The 'Security Digest' Archives (TM)

Archive: About | Browse | Search | Contributions | Feedback
Site: Help | Index | Search | Contact | Notices | Changes

ARCHIVE: TCP-IP Distribution List - Archives (1994)
DOCUMENT: TCP-IP Distribution List for April 1994 (632 messages, 332765 bytes)
SOURCE: http://securitydigest.org/exec/display?f=tcp-ip/archive/1994/04.txt&t=text/plain
NOTICE: securitydigest.org recognises the rights of all third-party works.

START OF DOCUMENT

-----------[000000][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 00:55:51 GMT
From:      resnick@cogsci.uiuc.edu (Pete Resnick)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not? UDP?

jmward@shavano.uccs.edu (Joel M. Ward) writes:

>2) Is UDP faster than a normal TCP socket?

Depends. UDP isn't reliable, so you don't get the confirmation packet
the other way, so it is often quicker (i.e. generates less traffic).
But, for instance, if you are going over a serial link which uses VJ
header compression for TCP, UDP goes through with full headers, so you
may actually end up with slower traffic than using TCP. Also, if you
are doing things like FTP with lots of data going over long hauls, TCP
retransmissions may be a good deal more efficent than having the
application do retransmits with UDP. So it depends on both application
and environment.

>3) some protocols are guaranteed to get packets thru and some aren't.
>under what conditions do they NOT get thruogh?

A router could be down, or miss a packet becuase it is overloaded. If
the former, you will hopefully get an ICMP unreachable message. If the
latter, it might go completely undetected.

pr
-- 
Pete Resnick             (...so what is a mojo, and why would one be rising?)
Graduate assistant - Philosophy Department, Gregory Hall, UIUC
System manager - Cognitive Science Group, Beckman Institute, UIUC
Internet: resnick@cogsci.uiuc.edu

-----------[000001][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 01:22:06 GMT
From:      garyh@sdsc.edu (Gary Hanyzewski)
To:        comp.protocols.tcp-ip
Subject:   Multicast IP : References please ?


	Hi all. I'm attempted to design/spec a piece of code to use
	Multicast IP. I've grabbed what little source code I could
	off the net, imm in particular, but I'm still at a loss as
	to how to code an application to use ipmcast communications.
	I would like to know what flags to set, what fcntl to invoke,
	what knobs to twiddle, etc. It would be nice if I could find
	a comprehensive information source, but at this point I guess
	I'd be satisfied with an outline or well commented source code.
	
	Any help?

	Thanks 
		
	Gary

---
////////////////////////////////////////////////////////////////////////
         Gary Hanyzewski                | "... There are more things in
         garyh@sdsc.edu      		|      heaven and earth than
                                        |      are dreamt of in your 	
     >>                       <<        |      philosophy ...."
////////////////////////////////////////////////////////////////////////


-----------[000002][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 10:29:59 -0500
From:      frederic@bingsun2.cc.binghamton.edu (Brian Frederick)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Re: VT100/TN3270 in one?

Hi there,
  If you are interested in both VT100 and 3270 emulation, I suggest
Rutger's CUTCP package.  It does include FTP.  Nice package.
=brian
-- 
Brian Frederick - Binghamton University - Computer Center - Network Analyst
frederic@bingvaxa - or - frederic@bingsuns.cc.binghamton.edu

-----------[000003][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 02:47:19 GMT
From:      yue@tuna.micro.umn.edu (Dongxiao Yue)
To:        comp.client-server,comp.protocols.tcp-ip,comp.unix.programmer,comp.unix.sys5.r4,navy.nswc.misc
Subject:   Re: How To Clea Or Free The Socket?

In <1994Mar28.171034.1350@relay.nswc.navy.mil> rchui@tethys.nswc.navy.mil (Raymond Chui) writes:
> problems with the client socket. When 1st time I get/put a file into the 
> socket with no problem at all. But when 2nd time I get/put a file into the
> socket, I got some garbage and append the command line into my 2nd file.
> This tell me after 1st data transfer, that socket still doesn't know EOF
> data file terminate, it still waiting there for more data, or tell the

For stream sockets:
Multiple write to the socket maybe concatenated;
A single write maybe broken into many deliveries.

I wrote a similar program a year ago, my orginal solution is let the
process sleep one second after finishing the data transfer, then send
reply. Recently, I modified this by using a more solid approach, 
that is send the size of the block each time before sending out the data,
the other side first read the size then read the 
promised amount(do you know how?),
the size info is 2byte, if send blocks of size 1024 the overhead is small.

At the end the process send size 0, then evrything is in sync, another
transfer cycle can begin.





-----------[000004][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 03:01:25 GMT
From:      yue@tuna.micro.umn.edu (Dongxiao Yue)
To:        comp.client-server,comp.protocols.tcp-ip,comp.unix.programmer,comp.unix.sys5.r4,navy.nswc.misc
Subject:   Re: How To Clea Or Free The Socket?

In <ESPEL.94Mar29235849@kayak.ens.fr> espel@ens.fr (Roger Espel Llima) writes:

>>>>>> On Mon, 28 Mar 1994 17:10:34 GMT, rchui@tethys.nswc.navy.mil (Raymond Chui) said:
 
>>> Hi, My Dare Net Dudes:
>>>  
>>> I am writting a simple socket client/server program with connection-oriented
>>> protocol in TCP. Please refers to W. Richard Stevens's book "Unix Network 
 
>FTP does open a new socket connection for every file (or directory listing)
>transfer! just do
 
>$ ftp
>ftp> debug
>Debugging on (debug=1)
>ftp> open somesite.somewhere.something
>     [log in]
>ftp> ls
>---> PORT 129,199,129,14,11,15
>	  ^^^^^^^^^^^^^^^^^^^^
 
>the ftp client opened a server socket at the address 129.199.129.14, port
>11*256+15, and is expecting the ftp server to send its data there.


>			Roger


Ftp open two TCP connections ( should not say sockets, since socket is
something only in 4.2BSD +... ?), one for data and one for exchange commands.

This is certainly more reliable(and simpler), but I do not know if it is
wasteful.

D. Yue



-----------[000005][next][prev][last][first]----------------------------------------------------
Date:      Fri, 01 Apr 94 16:10:13
From:      troutr@cloud.apana.org.au (Richard Trout)
To:        comp.protocols.tcp-ip
Subject:   Re: advice on ncsa, novel

 -=> Quoting Greg K Moeller to All <=-

 GKM> From: gkm@calvin.muug.mb.ca (Greg K Moeller)
 GKM> Newsgroups: comp.protocols.tcp-?ip.ibmpc
 GKM> Subject: Advice on NCSA, Novell's TCPIP, and token-ring.
 GKM> Message-ID: <Cn3p1A.3nv@calvin.muug.mb.ca>
 GKM> D?ate: Wed, 23 Mar 1994 04:41:33 GMT
 GKM> Organization: HOBBS BBSes and other wierd stuff
 GKM> Lines: 31

 GKM> Ok, first what w?e've got at work and what I'd like to do.

 GKM> We've got a token-ring network running Novell. I have to talk to some
 GKM> Uni?x boxes, which I do with Novell's tcpip, telapi, and TinyTerm.

 GKM> I can't change how tcp/ip is done at work, but what I'd li?ke to do is
 GKM> run NCSA, or some other telnet package which can do multiple sessions.

 GKM> The question is how?

 GKM> Here?'s what's loaded in autoexec.bat
 GKM> lsl
 GKM> token
 GKM> (odi is loaded here, can't remember the name of this one)
 GKM> bxn?et(Or is it xnetb?  Can't remember the name This seems to be IPX)
 GKM> tcpip
 GKM> telapi

Ok, firstly, this is what you actual?ly have:
lsl             The ODI 'manager' (Link Support Layer)
token           The Token Ring adapter driver or MLID in ODI ter?ms
ipxodi          Netware ipx protocol stack
bnetx           Netware burst-mode netx stack
tcpip           LAN WorkPlace for DO?S tcp/ip protocol stack
telapi          Telnet API for LW4DOS (NB: no longer needed for V4.1)

If you've done this, you now have? a telnet program which can do multiple
tcp/ip sessions!  You don't need NCSA!  The DOS program is tnvt220, and
the Windows prog?ram is Host Presenter.

However, if for some reason you WANT to use NCSA, you have to remove LAN
WorkPlace for DOS - I am not aw?are of an easy way to allow multiple
tcp/ip stacks to run on the same network adapter.  To run NCSA, I'm
pretty sure you have to? use the Packet Driver Specification instead of
ODI.  To do this you need to either:
1) Remove ODI entirely for a Packet Driver.?
2) Introduct a ODI to Packet Driver 'shim', eg. ODIPKT or PDETHER.  A
'shim' has the effect of making ODI appear as a Packet Dr?iver network
adapter (in this instance).
Either of these steps is usually quite do-able, but I won't elaborate
here, in case I'v?e misintpreted your original problem.

Regards, Richard.


--- Blue Wave/QWK v2.10
                                             ?

-----------[000006][next][prev][last][first]----------------------------------------------------
Date:      Fri, 01 Apr 94 16:10:14
From:      troutr@cloud.apana.org.au (Richard Trout)
To:        comp.protocols.tcp-ip
Subject:   Re: wfw 3.11 & mac client

 -=> Quoting Greg A. Bronzert to All <=-

 GAB> From: gbronzer@codon.nih.gov (Greg A. Bronzert)
 GAB> Newsgroups: comp.protocols?.tcp-ip.ibmpc
 GAB> Subject: WFW 3.11 & Mac Clients?  (not PhoneNet)
 GAB> Message-ID: <gbronzer.13.000A876F@codon.nih.gov>
 GAB?> Date: Wed, 23 Mar 1994 10:31:40
 GAB> Organization: Virginia Tech, Blacksburg, Virginia

 GAB> Hi all,
 GAB> Is there any soft?ware (other than PhoneNet PC)  that allows a
 GAB> windows for  Workgroups client to connect to a Mac on the network. 
 GAB> It'?s ok if the software  is for the Mac.  I'd just like to have
 GAB> something like an ftp Daemon for the  Mac that will work with? WFW, so
 GAB> my network can do easy "drag and drop"  transfers.
 GAB> We have a great need for this since we cut CD-ROMs for b?oth PC and
 GAB> Mac  using a Mac machine, and need an easy way to get loads of data
 GAB> from one  machine to the other.
 GAB>? Thanks a bunch for your help.

Hi Greg,
You're question is a very good one, and I haven't done this first hand.
However, I do u?nderstand that Microsoft are intending to allow
connectivity to their 'Remote Access Server' from Mac's.  This fits into
Microso?ft's 'Networking Strategy' which incorporates, Win NT Advancer
Server (previously LAN Manager); Win NT; and WFW.  What I'm getti?ng to
is that, with an NT Server host Remote Access Server, there is some
operatability, but I don't think it is peer-to-peer.

?There may be a LAN Manager package for the same purpose, but I really
don't know.  In any case, this may be a grander scheme tha?n you were
hoping.  Has anyone else got more suggestions?

Regards, Richard.


--- Blue Wave/QWK v2.10
                         ?

-----------[000007][next][prev][last][first]----------------------------------------------------
Date:      Fri, 01 Apr 94 16:10:15
From:      troutr@cloud.apana.org.au (Richard Trout)
To:        comp.protocols.tcp-ip
Subject:   Re: xwindows under win3.1

 -=> Quoting William Verthein to All <=-

 WV> From: wverthei@usr.com (William Verthein)
     Newsgroups: comp.protocols.tcp-ip.?ibmpc
 WV> Subject: XWindows under Win3.1
 WV> Summary: Looking for best XWindows package under Win 3.1
 WV> Keywords: XWindows ?IP
 WV> Message-ID: <wverthei.18.2D90817E@usr.com>
 WV> Date: Wed, 23 Mar 1994 17:50:54 GMT
 WV> Sender: news@usr.com (News Acco?unt)
 WV> Organization: U.S. Robotics Inc.

 WV> We recently acquired a SPARC10 running Solaris SunOS5.3 and we have a
 WV> numb?er  of X applications we would like to access across our IP network
 WV> on PCs running  Windows.

 WV> I'm looking for recommen?dations on the best XWindows package for
 WV> Windows 3.1.   We are running Novell's IP stack which I am pretty sure
 WV> is WIN?SOCK compatible.  We also use Novell's LAN workplace for IP
 WV> functions. 
 WV> Any help or horror stories would be appreciate?d.  Email to
 WV> wverthei@usr.com  please.

You should have no problems getting an XServer package for your PC.  If
your using ?LWP, then I'm sure you're running windows.  Hummingbird's
eXceed/W (for Windows) has been around for some time, and is supposedl?y
quite performant.  XVision is supposed to be quite good, and NCD have a
package they are pushing now, although I have absolute?ly no experience
with it.  All of these packages support LWP or Winsockets.  For
cheaper/shareware alternatives, I'm not that fa?miliar - I think they are
pretty scarce because most people work from the MIT/X11 libraries which
are licensed.

If you are comi?ng from a Sun background however, you may be disappointed
at the performance of XWindows on a pc.  You will certainly be wanting? a
Local Bus video, with Windows Acceleration.  Also, if you are using
high-resolution video applications, you will need appropr?iate support
from Windows/PC (although the above packages all 'simulate'
higher-resolution windows if required, you don't want t?o have to work in
those modes).

Regards, Richard.


--- Blue Wave/QWK v2.10
                                                   ?

-----------[000008][next][prev][last][first]----------------------------------------------------
Date:      Fri, 01 Apr 94 16:10:16
From:      troutr@cloud.apana.org.au (Richard Trout)
To:        comp.protocols.tcp-ip
Subject:   Multiple tcp/ip stacks

Hi all!

The problem we have is we use and need to use Novell LAN WorkPlace for
DOS (version 4.1).  In addition to this, we have? Cabletron's LANView for
Windows, which requires a packet-driver interface (LWP is a ODI
application).

We have got a pd shim fo?r ODI to allow the Cabletron application to run,
but certainly not concurrently with LAN Workplace.  Perhaps the two
application?s are competing to use tcp/ip stacks.

If this problem is because we are effectively trying to run two tcp-ip
stacks, does anyon?e know how we might remain with ODI and still get two
tcp-ip stacks to run?  (I am aware of pktmux, but because it is pd, I
don'?t think it will help juggle a ODI tcp/ip stack).

Any help appreciated!!!

Regards, Richard.

--- Blue Wave/QWK v2.10
          ?

-----------[000009][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 18:15:50 CST
From:      Rekha Raghu <U57604@uicvm.uic.edu>
To:        comp.protocols.tcp-ip
Subject:   Client/Server question

Hi!!
Normally, for the client to specify the server, as far as I there are
2 ways:
1: Hardcode the server name in the client program
2: Pass the Server's name as an arguement ..
However, I am in a situation where the client may not know the server name
until started. So it is not posssible for the client to open up connection
to the particular server, when started.

Is there any way of doing this where in you dont have to give the server
name in the client program?
But client can nondeterministically connect to a server?
I dont mind RPC's also, in case there is a way of doing this,
where in the server can broadcast on start, and have a domain name
server, which holds the server name, and then the client can contact
the name server.
I am not very sure how to implement this..

      Thanks in Advance
      Rekha..


-----------[000010][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 1994 00:16:54 -0700
From:      satel@xmission.com (Satel Corporation)
To:        comp.protocols.tcp-ip
Subject:   PPP for DOS & Windows - Where?

I am looking for a easy to install, configure and use PPP package to run
on my pc.  I would prefer it be for Windows.  (I would also prefer it to
be shareware) :)

If anyone knows of a site containing the like, or even a commercial package,
please let me know.

Thanx
Devan Green

current email: satel@xmission.com

-----------[000011][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 13:55:54 GMT
From:      ihl@ecs.soton.ac.uk (In Lee)
To:        comp.os.ms-windows.misc,comp.protocols.tcp-ip
Subject:   MS TCP-IP 32 beta and Vista eXceed

Hi,
I installed new MS TCP-IP 32 beta with no problem. However I have a few problems using it with Vista eXceedw. I can telnet but no Xstart. I get error message that it cannot start using the start method which is Rexec xterm etc. Once I telnet I can start xterm by starting Vista eXceed in the background. 

Also ftp gets hung if I start if after X session is started. Overall it appears that X session of Vista eXceed causes problem becasue I can use it with no problem unless X-session is started.

I will appreciate any help on this. Thanks.

-----------[000012][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 14:35:56 GMT
From:      carlson@Xylogics.COM (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not?  UDP?

In article <2nfi4i$enk@harpo.uccs.edu>, jmward@shavano.uccs.edu (Joel M. Ward) writes:
|> a few quiestions:
|> 
|> 1) It seems like Inet sockets aren't 8 bit clean.  Is this correct?  Is 
|> there are way to make emm that way?

No, that's not true.  They are most certainly 8 bit clean.

|> 2) Is UDP faster than a normal TCP socket?

No.  In fact, it's usually slower, since most major manufacturers are
actively optimizing the TCP data path, but ignoring UDP.

|> 3) somem protocols are guaranteed to get packets thru and some aren't.
|> under what conditions do they NOT get thruogh?

*Many*!  Buffer underruns and overflows, line noise, routing glitches,
and sometimes just because of bad karma.

|> 4) is there a way a server can send telnet a control signal to tell it
|> "Go into Line Mode"?

Yes.  IAC DO LINEMODE -- see RFC 1184.  It's complicated.

--
James Carlson <carlson@xylogics.com>            Tel:  +1 617 272 8140
Annex Software Support / Xylogics, Inc.               +1 800 225 3317
53 Third Avenue / Burlington MA  01803-4491     Fax:  +1 617 272 2618

-----------[000013][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 16:07:44 GMT
From:      bof@wg.saar.de (Patrick Schaaf)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not? UDP?

jmward@shavano.uccs.edu (Joel M. Ward) writes:

>1) It seems like Inet sockets aren't 8 bit clean.  Is this correct?

No.

>2) Is UDP faster than a normal TCP socket?
 
>3) somem protocols are guaranteed to get packets thru and some aren't.
>under what conditions do they NOT get thruogh?

IP packets can be dropped anywhere for any reason, and you won't even know it.
UDP is basically IP with end to end application addressing, so it loses, too.
TCP gives you the illusion of an errorless lossless bidirectional byte stream,
which is what you want in most cases.

>4) is there a way a server can send telnet a control signal to tell it
>"Go into Line Mode"?

I think there is one (the line mode negotiation is symmetric as far as
I remember, check the telnet RFCs to be sure), but the question is whether
the client will honor the request. telnet client sources are freely available,
so you could hack it in if it doesn't work.


Patrick

-----------[000014][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 17:25:15 GMT
From:      ktran@tethys.nswc.navy.mil (Ki Tran)
To:        comp.protocols.tcp-ip
Subject:   Overhead for TCP acknowledgements?

Hi,

     I am trying to estimate overhead due to TCP
     acknowledgements for the situation of say
     transferring a file from one workstation to
     another.  In this case, the receiving workstation
     would need to send acknowledgements to the sending
     workstation.

     How much traffic is involved in creating the 
     initial TCP connecting and in closing it at the
     end?  Can the connection be left open to await
     a future file transfer?

     What is a typical window size?  How does the 
     receiving workstation decide which datagrams to
     acknowledge?  My understanding is that they don't all
     need to be acknowledged.

     Thanks for your support.

     Kee Yu

-----------[000015][next][prev][last][first]----------------------------------------------------
Date:      Fri, 1 Apr 1994 17:59:32 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not?  UDP?

In article <2nhbgc$6ek@newhub.xylogics.com> carlson@xylogics.com writes:
>
>|> 2) Is UDP faster than a normal TCP socket?
>
>No.  In fact, it's usually slower, since most major manufacturers are
>actively optimizing the TCP data path, but ignoring UDP.

Or sometimes Yes.

The relative speed of UDP and TCP depends on many things.  Historically,
UDP was much faster than TCP, because the de facto standard implementation
of TCP and UDP (4.2 BSD and early 4.3BSD) had not had as much attention
paid to speed as to correctness, and because UDP is so trivial that little
optimizing is needed.

Since about 4.3BSD-tahoe, TCP is much faster than previous versions, and
can be faster than UDP.  The reason is that you often need to do a routing
lookup and other initial computations for each UDP transmission, while
the saved state of a TCP connection lets you amortize such overhead over
many packets.

UDP is still faster than TCP in special cases.  The most common one is
where you transmit faster than the wire can carry the bits.  For example,
most modern workstations will let you "transmit" much more than 10Mbit/sec
over an Ethernet.  The kernel just drops any packets that won't fit on
the wire.


Vernon Schryver    vjs@rhyolite.com

-----------[000016][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 19:18:26 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,vmsnet.networks.tcp-ip.misc
Subject:   Re: INFO WANTED ON ARP, RARP and PORTS

In article <2n74gu$hhj@ceres.king.ac.uk> cs_a206@ceres.king.ac.uk (Richard Smart) writes:
>I have been studying data communications and I am not sure on the 
>following areas:
>
>  i)  ARP requests (getting the internet address - broadcast)
>  ii) RARP requests (getting the ethernet address for a machine - to server)

You have ARP and RARP backward.  ARP is used to translate an IP address to
an ethernet address, and RARP is used to find out the current machine's IP
address.  Both normally use broadcasts.

>  iii) PORTS and Datagrams
>
>Does anyone have any information (including ftp site
>documents) that could outline the above areas ? 

I suggest you read a book such as "Internetworking with TCP/IP" by Comer,
which is a good introduction to the TCP/IP protocols.  Then for excuciating
details you could read "TCP/IP Illustrated" by Stevens.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000017][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 19:38:17 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Ports for Master and Slave Sockets

In article <1994Mar30.093717.61176@kuhub.cc.ukans.edu> anh@kuhub.cc.ukans.edu writes:
>When the server uses accept() it gets back a slave socket. What port does
>this slave socket use? The same port as the master socket? I don't want my
>"future" sockets to collide with this port.

[I've never heard the terms "master" and "slave" used in this context, but
I'll use them in this response for consistency.  More appropriate terms
would be the listening socket and the connected socket.]

The local port of the slave socket is the same as that of the master
socket.  This is necessary because the client will continue to send to the
port that it originally connected to.  There's no provision in the TCP
connection initiation protocol to assign a new port number.

The difference between the master and slave sockets is in the remote
address/port.  The master socket normally has no remote address or port,
meaning that it's willing to accept a connection from any client.  The
slave socket's remote address and port are those of the client that opened
this connection.  As a socket is defined by the tuple <local address, local
port, remote address, remote port>, every connected socket will be
different, even if some have the same local port.

You may be referring to Unix's design bug that prevents binding a master
socket to a local port if that port is being used by any other socket, even
if the other socket is a slave socket (are there any versions of Unix that
have fixed this?).  The solution to this is the SO_REUSEADDR socket option.
Set this option before binding, and the colliding slave sockets won't cause
a problem.

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000018][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 19:45:33 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Internet Providers with T3 access to NSFNet/CIXnet

In article <1994Mar30.050710.12657@rpslmc.edu> bwolfe@rpslmc.edu (Brian A. Wolfe) writes:
>Is anyone aware of any IP service providers with T3  access to the 
>NSFNet or CIXnet backbone besides ANS? 

NEARnet, the New England regional IP provider, has had a T3 connection for
about a year and a half.  (You didn't specify that you were looking for
nationwide providers.)
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000019][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 20:02:46 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Question - FTP TELNET requirement.

In article <CnHHAL.BtH@umassd.edu> rleary@UMASSD.EDU writes:
>   When FTP is implemented solely for use between programs - i.e., when 
>no user-terminal interface is provided - will some degree of TELNET 
>protocol support still be required? From my reading of RFC959 I get the 
>impression that some TELNET will still be required for the control 
>connection in any case.

Since the FTP server can't tell what kind of application -- interactive or
automated -- opened the connection, it will implement the same protocol in
all cases.  The control connection will still use NETASCII as its character
set, and the implementations must still recognize TELNET's special control
characters (e.g. IAC negotiations).

>   When there is no user-terminal interface in an FTP implementation is 
>there still any login activity of any kind? I.e., does the invoking 
>machine have to provide anything equivalent to username/password to 
>the remote host? I would assume not. If my assumption is correct, is it then 
>correct to say that the remote login aspects of TELNET are not employed 
>in the case that I describe - only other rudimentary aspects of TELNET 
>needed for the control connection.

The use of a username and password is optional in the FTP protocol.  If the
server has no need to authenticate clients, it need not require a login.
This is totally unrelated to the use of the TELNET protocol underneath
(logging in is *not* a feature of the TELNET protocol -- it's part of the
FTP protocol).

>  Would/could anyone use the remote login feature of TELNET in the situation 
>I describe - i.e., no user-terminal interface - to provide for some 
>kind of access control. I.e., the requesting machine must identify 
>itself - name/password - to the remote host? Would there be a good reason 
>to do this?

Even if an automated client is being used, you might want to make sure that
it's being used only by people who are permitted to access your files.  But
as I said above, there's no way for the server to know whether the client
is the special automated server or a random person using the "ftp" command
(have you ever been called on the phone by a recorded message, and not
realized for a few seconds that it was a machine calling you?).  If you
need passwords to protect against the random, you'll have to demand one of
the automated client as well.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000020][next][prev][last][first]----------------------------------------------------
Date:      1 Apr 1994 21:54:22 GMT
From:      karrer@ks.id.ethz.ch (Andreas Karrer)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not? UDP?

jmward@shavano.uccs.edu (Joel M. Ward) writes:

>1) It seems like Inet sockets aren't 8 bit clean.  Is this correct?  Is 
>there are way to make emm that way?

IP datagrams deliver data 8-bit clean, unless they are of the variety
described in RFC 1149, of course. Sockets are an implementation of IPC
endpoints, which may be IP connections. If your sockets alter the
contents of the data transmitted by truncating the eight bit, then it
is a strange implementation indeed.

>2) Is UDP faster than a normal TCP socket?

Depends. For things like sending a DNS or NTP server one packet and
waiting for one answer packet, UDP is the right choice. For
transferring large amounts of data, a connection-oriented protocol like
TCP -- retransmissions, sliding window -- is usually more appropriate.

>3) somem protocols are guaranteed to get packets thru and some aren't.
>under what conditions do they NOT get thruogh?

The key word used in TCP/IP is "reliable". TCP is reliable, IP and UDP
are not. In TCP, packets are acknowledged, arrive in order, and an
elaborate scheme is used to establish and clear a connection. This does
not mean that a TCP connection is guaranteed to survive network outages
-- there are timeouts and such -- but a program that uses a TCP
connection does not need to deal with them.
For UDP, there is no such guarantee. A program that uses UDP must take
care of dropped, duplicated packets, must worry about timeouts and the
like,


+-----------
  Andi Karrer, Communication Systems, ETH Zuerich, Switzerland
  karrer@bernina.ethz.ch    
  - CH stands for Calvin & Hobbs, the most influential Swiss religion
                                    <2n04i5$44j@hebron.connected.com>

-----------[000021][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 94 14:12:11 CST
From:      anh@kuhub.cc.ukans.edu
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not?  UDP?

In article <CnLDz9.2Jw@calcite.rhyolite.com>, vjs@calcite.rhyolite.com (Vernon Schryver) writes:
> In article <2nhbgc$6ek@newhub.xylogics.com> carlson@xylogics.com writes:
>>
>>|> 2) Is UDP faster than a normal TCP socket?
>>
>>No.  In fact, it's usually slower, since most major manufacturers are
>>actively optimizing the TCP data path, but ignoring UDP.
> 
> Or sometimes Yes.
> 
> The relative speed of UDP and TCP depends on many things.  Historically,
> UDP was much faster than TCP, because the de facto standard implementation
> of TCP and UDP (4.2 BSD and early 4.3BSD) had not had as much attention
> paid to speed as to correctness, and because UDP is so trivial that little
> optimizing is needed.
> 
> Since about 4.3BSD-tahoe, TCP is much faster than previous versions, and
> can be faster than UDP.  The reason is that you often need to do a routing
> lookup and other initial computations for each UDP transmission, while
> the saved state of a TCP connection lets you amortize such overhead over
> many packets.
> 
> UDP is still faster than TCP in special cases.  The most common one is
> where you transmit faster than the wire can carry the bits.  For example,
> most modern workstations will let you "transmit" much more than 10Mbit/sec
> over an Ethernet.  The kernel just drops any packets that won't fit on
> the wire.
> 
> 
> Vernon Schryver    vjs@rhyolite.com

So, if the server send small packets (around 48 bytes of data) every second
to a client. Then the communication protocol should be TCP ?

Everything being equal, and speed is the most important factor. Is there a thumbrule
on when to choose TCP over UDP with respect to packet size and how often the packets
are sent?

Anh

-----------[000022][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 1994 13:29:45 GMT
From:      comrade@uniwa.uwa.edu.au (Peter Cooper)
To:        comp.protocols.tcp-ip
Subject:   Virtual network device code?

I'm interested in finding a way to route IP packets to a process on my
machine in a similar way to (say) a SL/IP device might to a serial line.
ie. the process would appear to be an IP host to other machines on the
network, assuming routing was set up correctly.

I run a DECstation.  Is there anything like this around?

Thanks

Peter
--
Peter Cooper (comrade@gu.uwa.edu.au)           Computer geeks give better head.

-----------[000023][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 1994 16:57:54 GMT
From:      cliu@cs.sunysb.edu (Cheng-mean Liu)
To:        comp.protocols.tcp-ip
Subject:   Network performance evaluation(TCPIP)

Hi:

  I am a beginner in the field of network.  I am studying the TCPIP and
ATM API.  Although I have two sparc 10 which are connect to Ethernet and
ATM network, I don't have the feeling of how fast the these networks have.
Could you show me how to measure the network speed or send me some examples
to do some measurement?

  Thanks in advance.

Soccer




-----------[000024][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 1994 17:01:10 GMT
From:      lorenz@news.wu-wien.ac.at (Bernhard Lorenz)
To:        comp.protocols.tcp-ip
Subject:   Re: PPP for DOS & Windows - Where?

Satel Corporation (satel@xmission.com) wrote:
: I am looking for a easy to install, configure and use PPP package to run
: on my pc.  I would prefer it be for Windows.  (I would also prefer it to
: be shareware) :)
: current email: satel@xmission.com

this , indeed, applies to me as well, although i do not specifically
need it to be for windows. so, if anybody has got a clue where i
possibly might find it, please post into this group or email me.

thank you very much in advance,

+bl

--
===========================================================
Bernhard Lorenz                                  Fingwe@IRC
University of Economics and Business Administration
Department of Analytical Economics (VW5)
Augasse 2-6, 1090 Vienna, Austria.
email: lorenz@olymp.wu-wien.ac.at   phone: +43-1-31336-4579
===========================================================

-----------[000025][next][prev][last][first]----------------------------------------------------
Date:      Sat, 2 Apr 1994 17:01:18 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: routable protocols

In article <CnI92I.5xu@gfimda.uucp> rgs@gfimda.uucp (Robert G. Schaffrath) writes:
>>Hi, What makes a protocol routable?  I understand IP, appletalk, and
>>IPX/SPX are routable protocols but not NetBIEUS.  Why? Any books that
>>I can read more about it? Thanks.
>
>In a nutshell, what make a protocol routable is whether or not it maintains
>information somewhere inside it's packets indicating where it is coming from
>and where it is going to.  As you mentioned, IP, IPX/SPX are routable as well
>as DECnet and XNS.  Each of these protocols maintains information in the
>header which a "router" can examine and determine where it is destined.

But this definition would apply just as well to most protocols which must
be "bridged".  All LAN MAC header have a source and destination address
in their headers.  The major difference is that these are media specific
addresses.

>A
>protocol such as NetBIOS (NetBEUI for LAN Manager/Server) and DEC's LAT and
>LAST protocols, are essentially broadcast oriented in that they are not
>really targeted at any one machine from the standpoint of the protocol.  It
>is up to the target machine to listen and if it sees a datagram destined
>for it to act accordingly.  Some more sophisticated routers can do source
>level routing and examine the MAC frame addresses to help cut down on
>network congestion.  But I gather that would be another topic :)

Not true.  While LAN oriented protocols tend to make heavy use of broadcast
features, the bulk of the traffic is still usually station to station.

IMHO, the major distinction of a routable protocol, is that is was designed
to be run over a collection of potentially dissimilar link level technologies.
This usually also includes a structured address which carries some topological
information about the node's relationship to the network.  The structure may
be physical, administrative or a mix of both.

Bridged protocols usually only have media (link level) addresses to base
forwarding on.

Art

-----------[000026][next][prev][last][first]----------------------------------------------------
Date:      2 Apr 94 22:58:00
From:      billw@glare.cisco.com (William )
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not? UDP?

    Turning on linemode is actually turning off a TELNET feature (suppress go
    ahead) If a client does not honor that, it is _violating the RFC_ and
    deserves to lose, big time.

Heh.  I suppose after you've turned off 'supress go-aheads', your server
implementation is going to start sending go-aheads as well, huh?

Using "won't supress go-aheads" to mean "do line mode" is a hack implied by
a single line in an old telnet RFC, and isn't really part of the standard.

These days, there is a whole separate RFC on how to do "real" line mode.
It's supported by several server and client implementations, and has
advantages even on non-linemode systems.  (RFC1184)

BillW


-----------[000027][next][prev][last][first]----------------------------------------------------
Date:      Sat, 2 Apr 94 20:31:20 GMT
From:      ftlofaro@unlv.edu (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not? UDP?

In article <CnL8sx.F06@wg.saar.de> bof@wg.saar.de (Patrick Schaaf) writes:
>jmward@shavano.uccs.edu (Joel M. Ward) writes:
>
>>1) It seems like Inet sockets aren't 8 bit clean.  Is this correct?
>
>No.
>
>>2) Is UDP faster than a normal TCP socket?
 
>>3) somem protocols are guaranteed to get packets thru and some aren't.
>>under what conditions do they NOT get thruogh?
>
>IP packets can be dropped anywhere for any reason, and you won't even know it.
>UDP is basically IP with end to end application addressing, so it loses, too.
>TCP gives you the illusion of an errorless lossless bidirectional byte stream,
>which is what you want in most cases.
>
>>4) is there a way a server can send telnet a control signal to tell it
>>"Go into Line Mode"?
>
>I think there is one (the line mode negotiation is symmetric as far as
>I remember, check the telnet RFCs to be sure), but the question is whether
>the client will honor the request. telnet client sources are freely available,
>so you could hack it in if it doesn't work.
>
>
>Patrick

Turning on linemode is actually turning off a TELNET feature (suppress go ahead) 
If a client does not honor that, it is _violating the RFC_ and deserves to 
lose, big time.


-----------[000028][next][prev][last][first]----------------------------------------------------
Date:      Sun, 3 Apr 1994 00:54:44 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: 8 bit clean or not?  UDP?

In article <1994Apr2.141211.61402@kuhub.cc.ukans.edu> anh@kuhub.cc.ukans.edu writes:
> ...
>So, if the server send small packets (around 48 bytes of data) every second
>to a client. Then the communication protocol should be TCP ?
>
>Everything being equal, and speed is the most important factor. Is there
> a thumbrule
>on when to choose TCP over UDP with respect to packet size and how often
> the packets
>are sent?

48Bytes/sec is an incredibly low data rate.  I'd use TCP, and let the
system handle measuring the round trip time that is needed to set the
retransmission timers.  If you don't know what I'm talking about, then
TCP is the resoundingly right answer.

As I think someone use wrote in this thread, whether UDP or TCP is more
appropriate depends on the answers to the following (among other)
questions:
    1. is the data going to 1 destination instead of several (i.e. unicast
	vs. multicast)?
    2. is it a bad thing if data does not arrive at the destination?
    3. are both ends real computers with real TCP/IP software?
    4. is the application writer less than familiar with the many
	techniques of error detection and recovery that have been developed
	over the last 20 years?

Answers of "yes" urge TCP instead of UDP.  On the other hand, multicast
and TCP are not (currently) compatible so if you have many destinations,
UDP might be the only good answer.  If you are sending a steady stream of
data, such as video images, and you would rather continue after a lost
packet instead of stopping and waiting for retransmissions (e.g. video or
audio), then UDP is likely to be the better answer.  If one of the systems
is particularly dumb or has an usually slow or buggy TCP/IP software (e.g.
certain so called "easy to use" personal computers), many people choose
raw MAC packets instead of the far wiser and no more difficult faking of
UDP packets.


Vernon Schryver    vjs@rhyolite.com

-----------[000029][next][prev][last][first]----------------------------------------------------
Date:      3 Apr 1994 10:47:28 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Ping WORKS - Traceroute DOES NOT

In article <2mv280$aij@godel.denver.ssds.com> wjh@godel.denver.ssds.com (W. Joe Hewitt (Raleigh)) writes:
>	I have found a case where a machine will answer Pings
>and a traceroute fails.  Is this because the destination
>machine is not fully implementing the ICMP Ping?

Since ping works, the destination machine obviously *is* implementing ICMP
Echo -- the protocol that ping uses.  What it apparently doesn't implement
fully is UDP.  Traceroute sends a UDP packet to a random port, and expects
to get back an ICMP Port Unreachable.  It's possible that the port it uses
is actually being used by a server on the PC that doesn't send a reply, but
this is unlikely.  It's more likely that the PC doesn't send Port
Unreachables.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000030][next][prev][last][first]----------------------------------------------------
Date:      3 Apr 94 11:53:23 GMT
From:      bmg@numbat.cs.rmit.OZ.AU (Bernadette Garner)
To:        comp.protocols.tcp-ip
Subject:   finger protocol

This is probably commonly known and this is probably not the right news
group but are there log files kept on the local machine for who fingers you 
remotely? I hope this makes sense.

Please answer with e-mail to bmg@numbat.cs.rmit.oz.au since I seldom read news.

Bernadette
bmg@numbat.cs.rmit.oz.au

-----------[000031][next][prev][last][first]----------------------------------------------------
Date:      Sun,  3 Apr 94 22:09:00 -0600
From:      mark.stapleton@cld9.com (Mark Stapleton)
To:        comp.protocols.tcp-ip
Subject:   PPP for DOS & Windows -

SC>I am looking for a easy to install, configure and use PPP package to
SC>run on my pc.  I would prefer it be for Windows.  (I would also
SC>prefer it to be shareware) :)

Chameleon's got PPP. It's not shareware, but it does work in Windows, 
and it's easy to setup and use.

Mark

v?  

-----------[000032][next][prev][last][first]----------------------------------------------------
Date:      Sun, 3 Apr 1994 14:52:53 GMT
From:      smb@research.att.com (Steven Bellovin)
To:        comp.protocols.tcp-ip
Subject:   Re: Ping WORKS - Traceroute DOES NOT

In article <2nm6s0INNrtq@early-bird.think.com>, barmar@think.com (Barry Margolin) writes:
> In article <2mv280$aij@godel.denver.ssds.com> wjh@godel.denver.ssds.com (W. Joe Hewitt (Raleigh)) writes:
> >	I have found a case where a machine will answer Pings
> >and a traceroute fails.  Is this because the destination
> >machine is not fully implementing the ICMP Ping?
> 
> Since ping works, the destination machine obviously *is* implementing ICMP
> Echo -- the protocol that ping uses.  What it apparently doesn't implement
> fully is UDP.  Traceroute sends a UDP packet to a random port, and expects
> to get back an ICMP Port Unreachable.  It's possible that the port it uses
> is actually being used by a server on the PC that doesn't send a reply, but
> this is unlikely.  It's more likely that the PC doesn't send Port
> Unreachables.

Or that the destination machine is behind a firewall that will pass ICMP
echo packets, but not random UDP packets.

-----------[000033][next][prev][last][first]----------------------------------------------------
Date:      Sun, 3 Apr 1994 21:52:22
From:      jfoley@vt.edu (Joe Foley)
To:        comp.protocols.tcp-ip
Subject:   WinSock software recomendations

hello

i've just got slip working under win 3.1 using winsock, and i was wondering if 
anyone has any recomendations on good winsock compatible software.

thank you in advance

joe foley


-----------[000034][next][prev][last][first]----------------------------------------------------
Date:      Mon, 04 Apr 1994 07:44:20 -0800
From:      lehmann@globalvillag.com (Len Lehmann)
To:        comp.protocols.tcp-ip
Subject:   tcp/ip shareware for Windows?

Does anyone know about any low cost or shareware tcp/ip
for Windows?  I have already tried Trumpet, but have
encountered installation problems and would like to
try something else.  Thanks.

-----------[000035][next][prev][last][first]----------------------------------------------------
Date:      Sun, 3 Apr 1994 18:31:04 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   Re: Network performance evaluation(TCPIP)

cliu@cs.sunysb.edu (Cheng-mean Liu) writes:
>  I am a beginner in the field of network.  I am studying the TCPIP and
>ATM API.  Although I have two sparc 10 which are connect to Ethernet and
>ATM network, I don't have the feeling of how fast the these networks have.
>Could you show me how to measure the network speed or send me some examples
>to do some measurement?


There is application code for anonymous ftp on ftp.netcom.com in
~ftp/pub/gnn/.  The programs you are looking for are in bwmeas-0.3.tar.Z.

I hope this helps.

Later,
George


-- 
gnn@netcom.com

If people were more concerned with being reconciled than with being right, 
the world would be a better place.  -- Miss Manners

-----------[000036][next][prev][last][first]----------------------------------------------------
Date:      Sun, 3 Apr 1994 20:39:15 GMT
From:      sandyb@world.std.com (Sandy Bendremer)
To:        comp.protocols.tcp-ip
Subject:   Best ftpd for SunOS


What is considered the best replacement FTP Daemon for SunOS 4.1.3.

Of particular interest to me is detailed logging of all ftp activity.

Improved security and any other features are also appreciated.


Thanks for any help with this...

Sandy
sandyb@world.std.com

-----------[000037][next][prev][last][first]----------------------------------------------------
Date:      Sun, 03 Apr 1994 10:05:00 +1100
From:      Chris.Williams@f252.n620.z3.fidonet.org (Chris Williams)
To:        comp.protocols.tcp-ip
Subject:   Info Wanted On Arp, Rarp

Earlier Richard wrote...

>From: cs_a206@ceres.king.ac.uk (Richard Smart)
>Organization: Kingston University, Kingston-upon-Thames
 
>Hi,
 
>I have been studying data communications and I am not sure on the
>following areas:
 
> i) ARP requests (getting the internet address - broadcast)
> ii) RARP requests (getting the ethernet address for a machine - to se
> iii) PORTS and Datagrams
 
>Does anyone have any information (including ftp site
>documents) that could outline the above areas ?

The RFC collection has all this and more. Send email as shown:

To: rfc-info@ISI.EDU
Subject getting rfcs

help: ways_to_get_rfcs


Get the index... it ain't half useful!

RFC 1340 is the Assigned Numbers RFC (may have been suprseded)
RFC 826 ARP Specification
RFC 903 RARP Spec

All of the above may now be superceded.

Hope this helps
Chris W


-----------[000038][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 11:21:40 -0400
From:      jspencer@cass.ma02.bull.com (Joel Spencer)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   Tales from the Real World

Does anyone know if there is such a thing as an *illegal* host name? 

Last week, I visited a customer who had some routers named "router-0/1",
"router-0/2" etc etc etc.  This did not really fuss me until I abstracted 
a full snmp oid into a particular application.  Of course, the application
took the "/" character to be a delimiter in the oid.  Therefore, returning
a status indicating that i had requested an illegal/non-existent 
management object.  

Is it "illegal/inadvisable" to use the "/" char as part of a name?  Are
there any other characters to avoid?  Is this in an rfc anywhere?

Thanks, Joel.


-- 

=============================================================================
|	Joel Spencer				| "excusez-moi, je pense    |
|	Bull HN Worldwide Information Systems	|  ma bouchon d'essence     |
|	Billerica, Massachusetts USA		|  est mal..."              |
|	Email: J.Spencer@Bull.Com		|                           |
=============================================================================

-----------[000039][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 10:04:32
From:      clarkb@netstar.com (Clark Bremer)
To:        comp.protocols.tcp-ip
Subject:   Re: WinSock software recomendations

In article <jfoley.1.0015E01D@vt.edu> jfoley@vt.edu (Joe Foley) writes:
>From: jfoley@vt.edu (Joe Foley)
>Subject: WinSock software recomendations
>Date: Sun, 3 Apr 1994 21:52:22
 
>hello
 
>i've just got slip working under win 3.1 using winsock, and i was wondering if 
>anyone has any recomendations on good winsock compatible software.

I use the Trumpet News Reader (Shareware - check your local anonymous ftp site)

>thank you in advance

You're welcome.  CB.


===========================================================================
          _  _               Clark Bremer     clarkb@netstar.com
         /  /_)              Software Engineer, NetStar Inc.
         \_/__)              10250 Valley View Road  MPLS, MN 55344

-----------[000040][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 05:59:21 GMT
From:      jaesoo@cs18.cs.aukuni.ac.nz (Jaesoo Kim)
To:        comp.protocols.tcp-ip
Subject:   Re: docs for UNIX socket programming?

In article <16F8AD115.TCPMAINT@TWNMOE10.Edu.TW>, TCPMAINT@TWNMOE10.Edu.TW
wrote:
> 
> Hi!
>  
>     Is there any document on the net describing basic UNIX socket
> programming? Thanx.
>  
> /Liu
> 

I would appreciate it if you could send me the same information.

Regards,
J.Kim 

Jaesoo Kim                         |E-mail:jkim03@cs.auckland.ac.nz
Department of Computer Science     |       jaesoo@cs.auckland.ac.nz
The University of Auckland         |Tel.: +64 9 3737 599 (x 7265)
Private Bag 92019                  |Fax.: +64 9 3737 453
Auckland                           |No man is infalliable!
NEW ZEALAND                        |

-----------[000041][next][prev][last][first]----------------------------------------------------
Date:      Mon, 04 Apr 94 11:18:57
From:      troutr@cloud.apana.org.au (Richard Trout)
To:        comp.protocols.tcp-ip
Subject:   Re: Free tcp/ip

 -=> Quoting Bret A. Johnson to All <=-

 BAJ> In article <1994Mar24.150854.22947@jet.uk>, Paul Simmonds (ps@jet.uk)
 BAJ> write?s: >Microsoft have released their TCP/IP for DOS and Workgroups
 BAJ> for Windoes into >the Public Domain as FreeWare.
 >
 >It i?s available on ftp.microsoft.com, it comes with PING to show that
 >everything is working - and it does!

Can someone advise as ?to non-internet sites, or Australian sites
for this please?
--- Blue Wave/QWK v2.10
                                            ?

-----------[000042][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 13:46:25 GMT
From:      "Valentin A. Kaidalov" <root>
To:        comp.protocols.tcp-ip,comp.unix.sys5.r3
Subject:   TCP/IP & AX.25 facilities ?

I run ka9q NOS(v890421.1a) under ISC UNIX. It to work good,
but ... I have some problems and questions:
1. Where I can get doc for this stuff and more info ?
2. Can ISC TCP/IP "talk" (interaction) with net's TCP/IP ?
   (They operated parallel, but that is nonsense :) )
3. I need version for ISC UNIX like latest for DOS (if exist, and were
   can get it?).
4. Are you know of any solutions provide AX.25 facilities and allow
   switching IP and AX.25 packets and frames, except ka9q?
Thanks in advance.
  ----------------------
  Valery Rudich.
  rvv@rocket.kharkov.ua





-----------[000043][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 13:47:46 GMT
From:      unikofh@news.uni-c.dk (Ole Frendved Hansen)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   TCP applications over OSI transport?


In the field of mixed protocol stacks I recognize the need for running
TCP application protocols on top of OSI transport service.

I would like to know if specifications or implementations does exists?

I am doing a thesis at university, with mixed protocol stacks as subject.
Through this work I see a strong demand for the ability to use
TCP application protocols in combination with OSI protocols on the
lower layers.

What I am looking for is the same functionality (though mirrored) as outlined
in RFC 1006; ISO transport service on top of the TCP. So, what I want is a
module to put on top of OSI transport service, giving the service of TCP.

An example of use is the following stack:

	TCP Application				TCP Application

	>>TCP on ISO TS<<			>>TCP on ISO TS<<

	ISO TP4					ISO TP4

	ISO CLNS				ISO CLNS
	    |					    |
	    +---------------------------------------+

Please give answers to me in e-mail, I will summarize to the group.

Best regards,


Ole.Frendved.Hansen@uni-c.dk

-----------[000044][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 94 18:56:10
From:      billw@glare.cisco.com (William )
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

TCP over CLNP is in fact TUBA, one of the proposed "future IP" protocols.

BillW

-----------[000045][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 14:20:00 GMT
From:      cca04@cc.keele.ac.uk (Yufan Hu)
To:        comp.protocols.tcp-ip
Subject:   Two logical networks on a physical network, help for route needed

Hello,

I have a situation in which two logical networks (using two class C network
address) connected by one ethernet cable. There is a router on each of the
logical network. The problem is that the router on the network I am working
on does not seem to work very well. We are cut off from the outside world
frequently. As the router is not under my control, it is a bit anony when it
stops working outside office hour. But the router on the other network seems
always work. My question is that is there some way I can make use of this 
"better" router which is not on my logical network (but physically connected
directly)?

Any suggestion is appreciated.

Yufan.

-----------[000046][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 15:17:32 GMT
From:      ag257@FreeNet.Carleton.CA (Ross Macgillivray)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP applications over OSI transport?



The Internet community is currently working on a specification
called TUBA (Tcp/Udp with Bigger Addresses).  This may be of
interest to you.  However, the basic approach taken with TUBA
is to run TCP and UDP directly on top of ISO CLNP.
 
Ross MacGillivray
rossmac@aol.com

-----------[000047][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 16:33:24 GMT
From:      kottos@cti.gr (Kottos Kostas)
To:        comp.protocols.tcp-ip
Subject:   8-bit problem with telnet distribution


      Hi netters,

 I have a problem with the last release of telnet. I have compiled the telnet 
distribution on a vax system running Ultrix 4.2 . The telnetd server works fine,
it passes the 8-bit ( What I'm trying to do is to see and write greek, so the 8-bit is
essential). The problem is the telnet client. It doesn't seem to pass the 8-bit.
I've compiled the distribution on a Sparcstation 10 (SunOS 4.1.3 ) and it has the same problem. ( Server works fine, client doesn't seem to work with 8-bit ). I don't understand where the problem is. I've already used the commands :

         stty pass8   and
         setenv LC_TYPE iso_8859_1

but with no result. I think the problem is in some option in compilation of the client. If the server can understand 8-bit why doesn't the client ?

I would appreciate a good answer, either in this list or by e-mail


  You can write to: kottos@cti.gr

                                                      Kostas


-----------[000048][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 16:37:49 GMT
From:      mollett@lexmark.com (Vic Mollett)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP option 3

In article <CnqJAv.82I@lexmark.com> mollett@lexmark.com (Vic Mollett) writes:
>I just recently saw a DEC Alpha machine, running UN*X, send out a packet with
>TCP option 3 (03 03 00).  My sniffer didn't recognize it, so I got out the 
>RFC, and it doesn't mention it either.  Does anyone know what it is and why
>the DEC machine might send it out??  Thanks!!

Well, after some more digging, I was able to locate TCP option 3, it is the
Window Scale option and is defined in RFC1323, and, according to RFC1600, it
is in experimental status.  This brings up an interesting situation: how are 
older TCPs suppose to handle new TCP options?  RFC793 seems to suggest that
an option (kind as they put it) might have a length (like maximum segment size)
or might not have a length (like end of options list), but how can an old TCP
make a determination on a new option-kind??  I couldn't find anything that 
stated that the only options that are allowed to not have a length are noop
and eool.  The Window Scale option has a length, but what about (as of yet
unproposed) option 4, what if it doesn't have a length?  How do you skip it??



-- 
                                             /\    Vic Mollett
 These opinions are my own and do not       /  \   Lexmark International, Inc.
 necessarily reflect those of my employer.  \  /   mollett@lexmark.com
                                             \/  

-----------[000049][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 17:01:18 GMT
From:      west@mgmt3.ncsl.nist.gov (Jim West)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

In article <2np5q2$ri7@news.uni-c.dk> unikofh@news.uni-c.dk (Ole Frendved Hansen) writes:
>
>In the field of mixed protocol stacks I recognize the need for running
>TCP application protocols on top of OSI transport service.
>
>I would like to know if specifications or implementations does exists?
>
>I am doing a thesis at university, with mixed protocol stacks as subject.
>Through this work I see a strong demand for the ability to use
>TCP application protocols in combination with OSI protocols on the
>lower layers.
>
>What I am looking for is the same functionality (though mirrored) as outlined
>in RFC 1006; ISO transport service on top of the TCP. So, what I want is a
>module to put on top of OSI transport service, giving the service of TCP.
>
>An example of use is the following stack:
>
>	TCP Application				TCP Application
>
>	>>TCP on ISO TS<<			>>TCP on ISO TS<<
>
>	ISO TP4					ISO TP4
>
>	ISO CLNS				ISO CLNS
>	    |					    |
>	    +---------------------------------------+
>


Why not just implement TCP over ISO's CLNP.  CLNP is almost identical to IP in terms of
the network service offered.  The hard part is modifying the Name Service that
TCP application use to map names to NSAP addresses (versus the normal operation of
names to IP addresses).

Why does TCP need to sit over the ISO transport layer?  As I recall the reason
for put TP0 on top of TCP in RFC 1006 was that this was the easiest way to
implement the solution (all the programming could be done us user space,
and TP0 is simpler than TP4.  Since there is only one class of TCP, you can't
save any work in terms of protocol implementation, maybe you can keep from
doing UNIX kernel programming, but that's assuming you implement on a UNIX
system.

Jim


-- 
----------------------------------------------------------------------
Speaking only for myself, not necessarily for my employer.
west@mgmt3.ncsl.nist.gov

-----------[000050][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 00:15:41 -0400
From:      robb@epenviron.eapi.com (Rob Beckman)
To:        comp.protocols.tcp-ip
Subject:   XNTP is 1 hour off

We have xntpd running on our network and everything was working fine
until this weekend's time change and now we are 1 hour off.  Our
TIMEZONE file is set to EST5EDT which is right but we are still off.

When we were not using xntpd the time change worked fine but now it
doesn't.

Does anyone have any idea on how we can get our system back in sync
with the right time.

Thanks,
Rob Beckman
robb@eapi.com



-----------[000051][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 94 17:48:20 GMT
From:      john@qii.sialis.com (John Dassow)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Word 6.0a .doc saving problem under W4W 3.11


   When trying to save shared Word 6.0a files with the extention and mounted
with NCR's NFS product .doc onto our NCR tower we get an error messgage like
'File is locked, please disengage write protect tab'.  We are running Word 6.0a
under Word for Windows 3.11 and are using WinTCP+. This only happens for files
saved as '.doc', any other file format will save with no problems.  Does anyone
have any ideas of why this happens or how to correct it?

-- 

-----------[000052][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 00:32:23 -0400
From:      jl@panix.com (Jean-Louis Ecochard)
To:        comp.protocols.tcp-ip
Subject:   Large flat network

 I have been asked to comment on a network design of 500 Unix clients connected
to 200 servers on a flat high speed network (FDDI or CDDI). All segments are
bridged to eachother. There are 4 to 7 NIS domains and all machines need to
connect to all other machines. Most of the traffic is NFS and some broadcast 
information flowing through the network. I have received some concerns about the
flat architecture and the risk for broadcast storms. There has also been 
mention on the fact that all NIS broadcast traffic would be seen by other 
domain servers throughout the network. In addition there are other protocols
that may affect network performance (broadcast UDP). Some people have also
suggested that FDDI would not be the right solution for the problem and that
instead Ethernet (switched) and routers would insure proper operation and
management. I have been told that at this scale, a flat FDDI network could
easily colapse and would be very difficult to manage while a routed ethernet
network would allow better management.

I am interested in gathering opinions, experiences and information related to
the problems and management issues of the above mentionned network.
Thanks
-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ O ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jean-Louis Ecochard                    /_\                Voice   (212) 866-4908
Network Research Group, Corp.         (_v_)               Internet  jl@panix.com

-----------[000053][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 19:59:32 GMT
From:      subbu@cup.hp.com (Subbu Subramaniam)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP option 3

Vic Mollett (mollett@lexmark.com) wrote:
: In article <CnqJAv.82I@lexmark.com> mollett@lexmark.com (Vic Mollett) writes:
: >I just recently saw a DEC Alpha machine, running UN*X, send out a packet with
: >TCP option 3 (03 03 00).  My sniffer didn't recognize it, so I got out the 
: >RFC, and it doesn't mention it either.  Does anyone know what it is and why
: >the DEC machine might send it out??  Thanks!!
 
: Well, after some more digging, I was able to locate TCP option 3, it is the
: Window Scale option and is defined in RFC1323, and, according to RFC1600, it
: is in experimental status.  This brings up an interesting situation: how are 
: older TCPs suppose to handle new TCP options?  RFC793 seems to suggest that
: an option (kind as they put it) might have a length (like maximum segment size)
: or might not have a length (like end of options list), but how can an old TCP
: make a determination on a new option-kind??  I couldn't find anything that 
: stated that the only options that are allowed to not have a length are noop
: and eool.  The Window Scale option has a length, but what about (as of yet
: unproposed) option 4, what if it doesn't have a length?  How do you skip it??

From RFC 1122 (Requirements for Internet Hosts -- Communication Layers),p85:

   4.2.2.5  TCP Options: RFC-793 Section 3.1

     A TCP MUST be able to receive a TCP option in any segment.
     A TCP MUST ignore without error any TCP option it does not
     implement, assuming that the option has a length field (all
     TCP options defined in the future will have length fields).
     TCP MUST be prepared to handle an illegal option length
     (e.g., zero) without crashing; a suggested procedure is to
     reset the connection and log the reason.

-Subbu
subbu@cup.hp.com

-----------[000054][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 21:30:29 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Client/Server question

In article <94091.181550U57604@uicvm.uic.edu> Rekha Raghu <U57604@uicvm.uic.edu> writes:
>Hi!!
>Normally, for the client to specify the server, as far as I there are
>2 ways:
>1: Hardcode the server name in the client program
>2: Pass the Server's name as an arguement ..

3: Put the server name in a configuration file

>However, I am in a situation where the client may not know the server name
>until started. So it is not posssible for the client to open up connection
>to the particular server, when started.

I don't understand.  If "the client may not know the server name until
started", that implies that the client *does* know the server name after it
is started.  But maybe your English is not correct.

>Is there any way of doing this where in you dont have to give the server
>name in the client program?
>But client can nondeterministically connect to a server?

If the server is on the same network as the client, you can send a
broadcast, and the server can reply.  This is how Sun NIS works -- you find
a server by sending a broadcast RPC, and only servers that have ypserv
running will respond.

>I dont mind RPC's also, in case there is a way of doing this,
>where in the server can broadcast on start, and have a domain name
>server, which holds the server name, and then the client can contact
>the name server.

That would work, too.  And it should be pretty easy to implement with RPC.
But this means that you would have to have one of these name servers on
each subnet.

If you run NIS, you could put it in an NIS map.  Then use yp_match(3) or
yp_first(3) to look it up.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000055][next][prev][last][first]----------------------------------------------------
Date:      Mon, 4 Apr 1994 21:33:17 GMT
From:      parikh@enws237.eas.asu.edu (Ripple Parikh)
To:        comp.protocols.tcp-ip
Subject:   Need sample TCP/IP code


I need sample TCP/IP code.

Is there a FTP site from where I can download the code?

Thanks for your help.


-----------[000056][next][prev][last][first]----------------------------------------------------
Date:      4 Apr 1994 11:57:37 +1000
From:      tuc@lsupoz.apana.org.au (Scott Ellentuch)
To:        comp.sys.hp.hpux,comp.sys.hp.misc,comp.sys.sun.misc,comp.sys.protocols.misc,comp.protocols.tcp-ip
Subject:   Help needed with BOOTP/RARP

[ Article crossposted from comp.dcom.lans.ethernet,aus.computers.sun,comp.sys.sun.admin,comp.sys.hp ]
[ Author was Scott Ellentuch ]
[ Posted on 2 Apr 1994 18:31:38 +1000 ]

Hi all,

	Sorry to cross post all over the place, but I figured I'd give alot of
people the chance to hear me.  I'm trying to get a Sun 3/50M to ask for its
IP address via BOOTP to an HP/9000-G50.  I *THINK* I have the HP set correctly
to respond to a query (A bootpquery does give me what I want).  However, it
doesn't seem to do anything when the time comes.  The bootp daemon doesn't
even fire up.  I ran NETTL on the HP and this is what it sees :

=================================== ETHER ====================================
Source : 08-00-20-06-4c-bf [I] [Sun               ] TYPE: 0x8035              
Dest   : ff-ff-ff-ff-ff-ff [B] [                  ] TRACED LEN: 60
Date   : Sat Apr 02 16:44:33:27396 EST 1994
================================ RARP PROBE    (inbound -- [ICS]) ============
Source: 0.0.0.0       08-00-20-06-4c-bf
			Requesting: 0.0.0.0       

But doesn't seem to reply back.  Any help?

				Scott


-----------[000057][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 94 10:11:14
From:      billw@glare.cisco.com (William )
To:        comp.protocols.tcp-ip
Subject:   Re: ARAP Config Help !!!

    I am trying to configure a CS500 terminal server for Appletalk Remote
    Access at the moment and after doing the trivial config cannot get a
    connection to the server when attempting to dial in.

    We have a TACACS host to authenticate users so I don't think I need to
    worry about anything else ! What have I missed ???

If you want to use tacacs to authenticate ARA sessions, you will need
to change your CCL scripts.  Or did you mean that you can't even get a
terminal mode connection?

You should send your note to "tac@cisco.com", along with your serial
number and stuff...

BillW
cisco


-----------[000058][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 94 08:18:50 GMT
From:      rla@media03.UUCP (Raymond van der Laan)
To:        comp.protocols.tcp-ip
Subject:   binding & receiving broadcasts


When sending broadcast datagrams, why can't I bind the receiver's
socket to it's IP-address, and must I use INADDR_ANY?

   sock = socket (AF_INET, SOCK_DGRAM, 0);
   name.sin_family = AF_INET;

   name.sin_addr.s_addr = htonl(INADDR_ANY);        --> DG is received
   name.sin_addr.s_addr = htonl(<my IP address>);   --> DG is NOT received

   name.sin_port = htons(MYPORT);
   stat = bind(sock, (struct sockaddr *)&name, sizeof(name));

I send the DG as follows:

   sock = socket(AF_INET, SOCK_DGRAM, 0);
   on = 1;
   setsockopt(sock, SOL_SOCKET, SO_BROADCAST, &on, sizeof(on));
   name.sin_family = AF_INET;
   name.sin_addr.s_addr = htonl(<my IP address>);
   name.sin_port = htons(MYPORT+1);
   stat = bind(sock, (struct sockaddr *)&name, sizeof(name));

   remote.sin_family = AF_INET;
   remote.sin_port = htons(MYPORT);
   remote.sin_addr.s_addr = htonl(<my broadcast address>);
   stat = sendto (sock, TEXT, strlen(TEXT), 0,
                       (struct sockaddr *) &remote, sizeof (remote));
-- 
Raymond van der Laan           Email: sun4nl!media01!rla
Mediasystemen BV               Tel. : +31 23-319075            
Haarlem                        Fax  : +31 23-315210
The Netherlands                                   
                                    'There is no masterplan.
                                     This is what we do now.'
                                            - Stuart Adamson

-----------[000059][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 08:58:05 GMT
From:      pham@dsl.unit.no (BXD)
To:        comp.protocols.tcp-ip
Subject:   Re: wfw 3.11 & mac client

In article <1994Apr01.2974291815.wlm@cloud.apana.org.au> troutr@cloud.apana.org.au (Richard Trout) writes:
>From: troutr@cloud.apana.org.au (Richard Trout)
>Subject: Re: wfw 3.11 & mac client


> GAB> From: gbronzer@codon.nih.gov (Greg A. Bronzert)
 
> GAB> Hi all,
> GAB> Is there any soft?ware (other than PhoneNet PC)  that allows a
> GAB> windows for  Workgroups client to connect to a Mac on the network. 
> GAB> It'?s ok if the software  is for the Mac.  I'd just like to have
> GAB> something like an ftp Daemon for the  Mac that will work with? WFW, so
> GAB> my network can do easy "drag and drop"  transfers.
> GAB> We have a great need for this since we cut CD-ROMs for b?oth PC and
> GAB> Mac  using a Mac machine, and need an easy way to get loads of data
> GAB> from one  machine to the other.
> GAB>? Thanks a bunch for your help.


   Hi  

   Have you tried   MACLAN  Connect  Personal  from  MIRAMAR  SYSTEMS?  
Currently  I'm  using  MACLAN  v3.02   with  WFW 3.11 but  without  the  
network  part.  There is  a note of  how  to  run  MACLAN  with  WFW but I  
don't  have  time  yet.  After  installing  MACLAN  on  your  PC, it  can  
works  as  a  file server,  a  printer server  or  both  for  the  MAC.  That  
means  you  can  use your MAC  to  copying  files  to/from  your  PC  with   
MACLAN  installed.  If  there  were  a  laser printer connected to  the PC  
you  can also  use  that  printer from  the MAC.    

  There  also  are  a  product  which works exactly  vice versa.   

 Best regards. 

  Pham  

 

>Hi Greg,
>You're question is a very good one, and I haven't done this first hand.
>However, I do u?nderstand that Microsoft are intending to allow
>connectivity to their 'Remote Access Server' from Mac's.  This fits into
>Microso?ft's 'Networking Strategy' which incorporates, Win NT Advancer
>Server (previously LAN Manager); Win NT; and WFW.  What I'm getti?ng to
>is that, with an NT Server host Remote Access Server, there is some
>operatability, but I don't think it is peer-to-peer.
 
>?There may be a LAN Manager package for the same purpose, but I really
>don't know.  In any case, this may be a grander scheme tha?n you were
>hoping.  Has anyone else got more suggestions?
 
>Regards, Richard.


>--- Blue Wave/QWK v2.10
>                         ?


-----------[000060][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 94 10:59:44 GMT
From:      egger@slsvaat.sel.de (Jochen Egger Fa. SSW)
To:        comp.protocols.tcp-ip
Subject:   WANTED: Internet Provider in USA and Europe


Hello,

maybe this is not the perfect group for my question. Which is the right group?
No flames please.

I'm looking for a list of internet providers in the USA and Europe.
If You have such a list, please send it. If there is one to ftp,
please let me know, too. Even single email-adresses would be great.

Please reply by email.
Thank you.


--
-------------------------------+---------------------------------------------
Jochen Egger                   |
egger@lts.sel.alcatel.de      < >         intentionally left blank
+49 711 821-3578               |
-------------------------------+---------------------------------------------
ALCATEL-SEL Stuttgart Abt. US/ITC

-----------[000061][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 94 13:04:25 GMT
From:      jonesjg@dg-rtp.dg.com (Greg Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: Network performance evaluation(TCPIP)


In article <2nk86i$4je@newsserv.cs.sunysb.edu> cliu@cs.sunysb.edu (Cheng-mean Liu) writes:
> Hi:
> 
>   I am a beginner in the field of network.  I am studying the TCPIP and
> ATM API.  Although I have two sparc 10 which are connect to Ethernet and
> ATM network, I don't have the feeling of how fast the these networks have.
> Could you show me how to measure the network speed or send me some examples
> to do some measurement?
>   Thanks in advance.
> Soccer

Find yourself a copy of TTCP, it can send TCP/UDP traffic from memory to
memory on the 2 workstations.  This eliminates any disk I/O from your
benchmarks.  You can specify lots of different parameters, so you should
try all the options to see how they affect performance.

-- 
Greg Jones
jonesjg@dg-rtp.dg.com

-----------[000062][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 13:38:02 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: binding & receiving broadcasts

When you bind a non-wildcard IP address to a socket, that socket
will only receive IP datagrams whose destination IP address equals
the bound address.  This is how UDP and TCP work.  In your example
the destination IP address will be a broadcast address, but the
socket is bound to the interface's unicast IP address.  You can
bind the broadcast address to the socket for your example, you
don't have to bind the wildcard.

	Rich Stevens

-----------[000063][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 14:18:17 GMT
From:      ganzhorn@cisco.com (Charles Ganzhorn)
To:        comp.protocols.tcp-ip
Subject:   Re: Large flat network

In article <2nqpkn$f1d@panix.com>, jl@panix.com (Jean-Louis Ecochard)
wrote:

IMHO, the biggest problem of a bridged environment is the broadcast traffic
that is unnecessarily propogated to systems that don't care about it.  This
requires the node to deal with packets that it ultimately throws away.  In
a large enough environment ("large" is kind of slippery:  a few "noisy"
nodes can be just as bad as lots of nodes that might not talk as much) you
can lose a fair amount of system performance to the processing of useless
packets.  Routing minimizes as much as possible the amount of useless
traffic.  The whole point of routing is that you KNOW if the packet should
be propogated as opposed to bridging where you more of less GUESS if the
packet should be propogated.

As far as the choice of switched ethernet versus FDDI, the issues are
support, bandwidth and cost.  Fiber can be a bit of a pain to work with but
then there is always CDDI to deal with that.  After that, it's basically
dealers choice based on price for performance:  switched ethernet lets you
retain the ethernet adaptors you probably already have but then you still
have only a 10 Mb pipe and you might need more from a given station at a
given point in time.

A fairly typical approach would be to switch the clients, use FDDI to get
off the switch, put servers for those clients on the FDDI and then route
the FDDI's:  I'm assuming that you would have multiple client-server
"clusters".

Charles.

> 
>  I have been asked to comment on a network design of 500 Unix clients connected
> to 200 servers on a flat high speed network (FDDI or CDDI). All segments are
> bridged to eachother. There are 4 to 7 NIS domains and all machines need to
> connect to all other machines. Most of the traffic is NFS and some broadcast 
> information flowing through the network. I have received some concerns about the
> flat architecture and the risk for broadcast storms. There has also been 
> mention on the fact that all NIS broadcast traffic would be seen by other 
> domain servers throughout the network. In addition there are other protocols
> that may affect network performance (broadcast UDP). Some people have also
> suggested that FDDI would not be the right solution for the problem and that
> instead Ethernet (switched) and routers would insure proper operation and
> management. I have been told that at this scale, a flat FDDI network could
> easily colapse and would be very difficult to manage while a routed ethernet
> network would allow better management.
> 
> I am interested in gathering opinions, experiences and information related to
> the problems and management issues of the above mentionned network.
> Thanks
> -- 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ O ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Jean-Louis Ecochard                    /_\                Voice   (212) 866-4908
> Network Research Group, Corp.         (_v_)               Internet  jl@panix.com

--
Charles Ganzhorn                        Email:  ganzhorn@cisco.com
cisco Systems                           Phone:  612-368-8922
Chaska, MN                              FAX:    612-368-9977

-----------[000064][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 22:46:44 -0500
From:      lin@cs.purdue.edu (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   Re: Help implementing TCP/IP code?

In article <2nt3bj$6ee@darkstar.UCSC.EDU>, James G. Speth <speth@end.com> writes:
> I'm working on a TCP/IP implementation using the code from the book
> _Internetworking with TCP/IP_ Vol. 2 by Douglas Comer and David Stevens.  The
> code is great, and explained very well, but it uses some parts of the Xinu
> operating system that may or may not exist already in most unix systems.
> 
> Does someone out there have the code from this book working on a unix machine?
>  If so, would you mind sharing your replacements for Xinu's message passing,
> semaphores, and queueing functions?

   I don't know about Xinu-TCP running on a unix box. But, you can get 
the Xinu OS (for Sun3) and TCP/IP source code from the following anonymous 
ftp site:

   ftp.cs.purdue.edu:/pub/dls/xinu7.9.tar.Z

   Xinu-TCP uses a "process-oriented" approach to implement TCP/IP mainly
because Xinu is a multi-threaded kernel. Unix is not a multi-threaded kernel,
thus a "procedure-based approach" is more approate, IMHO.
 
John Lin (lin@cs.purdue.edu)

-----------[000065][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 94 15:27:11 GMT
From:      egn@athen.mch.sni.de (Emil Naepflein)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

In article <1994Apr4.174820.15737@qii.sialis.com> john@qii.sialis.com (John Dassow) writes:
>
>   When trying to save shared Word 6.0a files with the extention and mounted
>with NCR's NFS product .doc onto our NCR tower we get an error messgage like
>'File is locked, please disengage write protect tab'.  We are running Word 6.0a
>under Word for Windows 3.11 and are using WinTCP+. This only happens for files
>saved as '.doc', any other file format will save with no problems.  Does anyone
>have any ideas of why this happens or how to correct it?
>

This seems to be a problem of all AT&T System V based systems. The problem
is that WfW issues lock requests which use the full 32 bit address range
for the offset. This is allowed in the XNFS protocol, but causes problems
in the System V implementation of the lock manager. The System V lock manager
uses the fcntl(2) call to manage the locks, but it uses off_t to pass the
offset which is actually a signed 32-bit value. If a offset mit the 31st
bit set is used, the kernel interprets this value as negativ and returns
EINVAL. So all the System V implementations of the lock manager are not
XNFS compliant!

How to fix this problem?

1. Avoid mounting NFS volumes as shared.
   That should be possible, if documents are not *really* shared.
2. Fix problem in UNIX
   This can only be done by the UNIX vendor and its no easy task.
3. Fix cause for problem in WfW.
   There shouldn't be any reason for WFW issueing lock requests with
   this offsets. I assume this is a bug in WfW. But on the other side
   it is allowed to issue that kind of requests when using the XNFS protocol
   BTW, I traced the communication between a PC and a UNIX machine when
   saving a 6 KB word document and found out that around *1000* ethernet
   packets are exchanged. That's far from optimal and causes a lot of
   load for the server.
4. Look for an *original* SUN Lock manager.
   The original SUN lock manager does not use a system call. It manages
   the locks by itself.

I hope this helps. I'am also looking for an easy solution to this problem.

Emil


-----------[000066][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 16:17:18 GMT
From:      lparsons@world.std.com (Lee E Parsons)
To:        comp.protocols.tcp-ip
Subject:   KeepAlive not used for non-idle processes

We are seting up a database clinet/server envinroment that in the 
near future will have 200+ users.

I have noticed that if a user turns off his PC the server process
is not terminated. This is not a problem since the keepalive timer
will eventually catchup with him. 

However if the server process is actively accruing CPU time then 
the timer doesn't seem to be used. So if the client sends the
server a endless loop and then dies (or is rebooted), then the
server will sit and spin forever taking up as much CPU as it
can.

Is this behavior normal for the KeepAlive Timer? Can it be changed?

If not, any suggestions for tracking these processes down?

We are using Oracle under Ultrix for our test environment and
will be moving to AIX for production.
-- 
Regards, 

Lee E. Parsons                  		
Systems Oracle DBA	 			lparsons@world.std.com

-----------[000067][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 17:25:11 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Network performance evaluation(TCPIP)

In article <1994Apr5.130425.26285@dg-rtp.dg.com> jonesjg@dg-rtp.dg.com (Greg Jones) writes:
>In article <2nk86i$4je@newsserv.cs.sunysb.edu> cliu@cs.sunysb.edu (Cheng-mean Liu) writes:
>> 

...
>> Could you show me how to measure the network speed or send me some examples
>> to do some measurement?
 
>Find yourself a copy of TTCP, it can send TCP/UDP traffic from memory to
>memory on the 2 workstations.  This eliminates any disk I/O from your
>benchmarks.  You can specify lots of different parameters, so you should
>try all the options to see how they affect performance.

Look on sgi.com for ttcp.c, HP's (Rick Jones') netperf, and Cray's nettest.
All three meet the "network only, no disk" criterion, and each has
its unique strengths.

From the FTP logs, several other systems are mirroring what is on sgi.com,
so you should be able to find those benchmarks in many places.


Vernon Schryver    vjs@rhyolite.com

-----------[000068][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 17:36:17 GMT
From:      mansonw@kalpana.com (Manson Wong)
To:        comp.protocols.tcp-ip
Subject:   Re: BOOTP

I need to build a simple BOOT network with a PC based boot server to download boot image. BOOTP and TFTP will be used to download the boot image from the server to a booted target. Does anyone know where I can get or buy the server side software that support BOOTP and TFTP.



-----------[000069][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 18:18:36 GMT
From:      root@cti.gr (Operator)
To:        comp.protocols.tcp-ip
Subject:   8-bit problem with telnet still unsolved


 Hi netters, 

 I've already sent a message in this list, I received some answers and I thank all of
you for this. 
  I've compiled the latest telnet distribution on a Sparc10 running SunOS 4.1.3 ( I use
both telnet client and telnetd server). The telnetd server worked just fine ( it passed the eight bit) but the telnet client doesn't seem to work well.( it cuts the 8-bit).
The problem is not in the terminal settings. I know this because I 've tried the telnet
client packed with SunOS ( /usr/ucb/telnet) and it works just fine. Both the telnet
clients run from the same machine with the same terminal settings.
   Has anybody seen a similar problem ?

  Please respond in this list or send a mail-message at : kottos@cti.gr




-----------[000070][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 18:28:17 GMT
From:      john@iastate.edu (John Hascall)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP option 3

Vic Mollett <mollett@lexmark.com> wrote:
}>I just recently saw a DEC Alpha machine, running UN*X, send out a packet with
}>TCP option 3 (03 03 00).  My sniffer didn't recognize it,  ...
 
}  ... experimental status.  This brings up an interesting situation: how are 
}older TCPs suppose to handle new TCP options?  RFC793 seems to suggest that
}an option (kind as they put it) might have a length (like maximum segment size)
}or might not have a length (like end of options list), but how can an old TCP
}make a determination on a new option-kind??  I couldn't find anything that 
}stated that the only options that are allowed to not have a length are noop
}and eool.  The Window Scale option has a length, but what about (as of yet
}unproposed) option 4, what if it doesn't have a length?  How do you skip it??

   No TCP option can be length-less (besides noop and eool).

John
-- 
John Hascall              ``I can't tell you that Karla Homolka pled guilty''
Systems Software Engineer
Project Vincent
Iowa State University Computation Center  +  Ames, IA  50011  +  515/294-9551

-----------[000071][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 07:19:56 -0700
From:      phil@lykos.netpart.com (Phil Trubey)
To:        comp.protocols.tcp-ip
Subject:   Dynamic load sharing via a single IP address.

I have a situation where I have 2 machines offering an identical TCP 
service (clients can connect to either and get the same information/service).
Is there a way for clients to open a connection to a single IP address
and have the request go to either one of the machines via some 
combination of switching, routing, etc?  ie. I'd like to be able to
dynamically load share the service between these two machines - and
also allow all connections to go to one machine if the other is down.

If  this isn't possible by sending a request to a single IP address, any
suggestions on how I could load share in some other way?

Thanks,

______________________________________________________________________

 Phil Trubey                 | 
 NetPartners                 |
                             | Providing independent consulting in the    
 E-mail: phil@netpart.com    |   application of Internet technology        
 Phone:  714-759-1641        |                                             
 Fax:    714-644-0577        |
______________________________________________________________________
-- 
 

-----------[000072][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 20:30:23 GMT
From:      @udel.engr.sgi.com (Dave Crocker)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

In article <2np5q2$ri7@news.uni-c.dk>, unikofh@news.uni-c.dk (Ole Frendved
Hansen) writes:
|> 
|> In the field of mixed protocol stacks I recognize the need for running
|> TCP application protocols on top of OSI transport service.

I used to work for a professor named Dave Farber.  He has a rather
remarkable linguistic (actually deeper, I think, so that it's something in
his basic cognitive processing) style that leads to some interesting
distortions of language.  One that showed up in a book on Murphy's Laws
was that Necessity is the Mother of Strange Bedfellows.

Anyhow, he also once referred to a technology as filling a much-needed
gap.

There are many good and reasonable ideas.  The trick is to follow through
on the ones that are needed and useful.  While there has been discussion
of running TCP apps over OSI lower layers -- and there is no technical
impediment that I can think of -- the real test is whether there is a
compelling need for it.  Do you have some indication of market pressure
for this functionality?

Dave

-----------[000073][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 20:58:16 GMT
From:      clv2m@uvacs.cs.Virginia.EDU (Charles Viles)
To:        comp.protocols.tcp-ip
Subject:   TCP-IP performance references.

Forgive me if this is an FAQ...

I am looking for some references that document tcp-ip performance, in
particular, things like time to build a connection, bandwidth, time to
tear down the connection etc.

Any pointers to on-line documentation or the literature would be
appreciated. I'll summarize if there seems to be general interest
in the topic.

-- 
Charles Viles (clv2m@virginia.edu)
Department of Computer Science
University of Virginia
Charlottesville, VA 22902

-----------[000074][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 21:02:14 GMT
From:      sdong@scorpion.syr.edu (Sheng  Dong)
To:        comp.protocols.tcp-ip
Subject:   need your help!!!

Hi:

I am using the socketpair to setup two stream pipes. 
Can anyone kindly tell me how to bind in this condition. I did it this way,but
every time got error message for can not binding.

-----------------------------------------------------------------------------
    if( socketpair(AF_UNIX, SOCK_STREAM, 0,stm_pipe) <0)
       printf("\n server: can't open STREAM socket \n");
 
    bzero((char*) &serv_addr, sizeof(serv_addr));
    serv_addr.sin_family   = AF_UNIX;
    serv_addr.sin_addr.s_addr = htonl(INADDR_ANY);
    serv_addr.sin_port      = htons(SERV_TCP_PORT);
 
    unlink(stm_pipe[0]);
 
    if( bind(stm_pipe[0], (struct sockaddr *) &serv_addr, sizeof(serv_addr)) <0)
       printf("\n server: can't bind local address 1 \n");

    serv_addr.sin_port      = htons(SERV_TCP_PORT1);
 
    if( bind(stm_pipe[1], (struct sockaddr *) &serv_addr, sizeof(serv_addr)) <0)
       printf("\n server: can't bind local address \n");
 
------------------------------------------------------------------------------

help me out! I really need it.

send an email to sdong@rodan.syr.edu

thanks in advance


Victor Dong




-----------[000075][next][prev][last][first]----------------------------------------------------
Date:      05 Apr 1994 21:32:34 GMT
From:      gthaker@polyphony.sw.stratus.com (Gautam Thaker)
To:        comp.protocols.tcp-ip
Subject:   multiple interfaces on a machine and coping w/ failures


Suppose I have two machines with two ethernet interfaces each and
connect these two machines with two ethernet cables. Each machine thus
has two IP addresses. (call these addresses A and A' and B and B') If
I have a TCP/IP connection open from one machine to another (say from
A<->B, and if I then cut the ethernet cable that is actually carrying
this traffic, what all do I need to be sure that the traffic will be
rounted over the remaining ethernet with no loss of connection at the
TCP level?

Is this something I can achieve by just properly setting up routes?
Or is loss of connection unavoidable?

Any hints welcome.

Gautam Thaker, Isis Distributed Systems, Inc.
gthaker@isis.stratus.com, 609-767-4854, 609-767-4863 (fax)

-----------[000076][next][prev][last][first]----------------------------------------------------
Date:      Tue, 5 Apr 1994 21:51:45 GMT
From:      kardel@nessy.informatik.uni-erlangen.de (Frank Kardel)
To:        comp.protocols.tcp-ip
Subject:   Re: XNTP is 1 hour off

robb@epenviron.eapi.com (Rob Beckman) writes:

>We have xntpd running on our network and everything was working fine
>until this weekend's time change and now we are 1 hour off.  Our
>TIMEZONE file is set to EST5EDT which is right but we are still off.
 
>When we were not using xntpd the time change worked fine but now it
>doesn't.
 
>Does anyone have any idea on how we can get our system back in sync
>with the right time.

When you are using xntp your system will have the right time.
It is the time zone handling code that is usually wrong. The kernel
uses UTC (universal time coordinated). This time does not do any
daylight saving time switching. NTP distributes UTC. If you are off
by one hour chances are, that either your time zone table is misconfigured
or the time zone code is broken. It would help if you could mention your
system and OS version. There have been quite a few broken time zone library
versions out there. Misconfigurations and bugs are usually detected when
running NTP which distributes the correct version of UTC.

Frank Kardel (time@informatik.uni-erlangen.de)

-----------[000077][next][prev][last][first]----------------------------------------------------
Date:      Tue, 05 Apr 1994 22:55:33 GMT
From:      conklin@lilypond.win.net (Thomas R. Conklin)
To:        comp.protocols.tcp-ip
Subject:   Distinct TCP-IP -- Anybody using this?

I wanted to solicit some review from this group of Distinct
Corp's tcp-ip product.  My organization is considering
getting pretty involved with this package (winsock.dll) for
doing some development work (they seem to have some very
nice libraries for ftp and telnet).  We are also
considering liscensing their tcp-ip protocol stack for
Windows. 

Does anyone have some experience with Distinct's stuff
that they'd be willing to share?  How robust is the
software?  Are the ftp and telnet libraries workable,
useful?  Any notable limitations, areas of buggyness?  How
about Distinct's support?  What are the highlights and
lowlights of the package?

Thanks in Advance 


-- Rob Conklin (CWI)
Internet Address:    conklin@lilypond.win.net

-----------[000078][next][prev][last][first]----------------------------------------------------
Date:      5 Apr 1994 23:14:15 GMT
From:      wanning@skorpio.usask.ca (Mandy Zhu)
To:        comp.protocols.tcp-ip
Subject:   Re: Internet Map


In article <2n297s$j1q@tribune.usask.ca>, you write:
|> Hi Netfriends,
|> 
|> Could any of you point me to some ftp sites in which I can find
|> some maps of worldwide Internet connectivity? Any help will be
|> very appreciated.
|> 
|> My email address: wanning@cs.usask.ca
|> 
|> -Mandy

After I sent this message, I received some replies from some net friends.
Thanks. Since so many people want to share the answer, I like to post the
ftp site that has the maps. Try ftp to nis.nsf.net directory /maps.

Enjoy,

-Mandy                                             


-----------[000079][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 11:19:01 -0700
From:      guy@nova.netapp.com (Guy Harris)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

>4. Look for an *original* SUN Lock manager.
>   The original SUN lock manager does not use a system call. It manages
>   the locks by itself.

I've no idea what a "SUN Lock manager" is, as I know of no company named
"SUN" that puts out a lock manager.

The lock manager put out by Sun Microsystems (not Stanford University
Network Microsystems) in releases prior to SunOS 4.1 did, in fact,
manage the locks itself; it did not make any calls to pass those locks
down to the kernel.

As of SunOS 4.1, code to do locking on local files was put into the
kernel, in order to provide SV-style mandatory file and record locking
(although this is *NOT* a "System V implementation of the lock manager"
in the sense of having come from an AT&T SV release; Sun did it
themselves).  The lock manager was changed to use a call to pass locks
down to the kernel (the F_RSETLK and F_RSETLKW calls; see FCNTL(2V), and
do *NOT* use those calls in your own program - as the manual page says,
at least in later SunOS 4.1[.x] releases, those calls are there *solely*
for the use of the lock manager; they are *not* what a program that
wants to lock a file that happens to be on an NFS server should use -
such a program should use the same F_SETLK/F_SETLKW calls that it would
use on a local file).

You may, or may not, have any success trying to run a SunOS 4.0[.x] lock
manager, which would "manage the locks by itself", on a SunOS 4.1[.x]
system.  You probably will not have much luck at all trying to run it on
a SunOS 5.x system.

-----------[000080][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 01:30:27 GMT
From:      James G. Speth <speth@end.com>
To:        comp.protocols.tcp-ip
Subject:   Help implementing TCP/IP code?

I'm working on a TCP/IP implementation using the code from the book
_Internetworking with TCP/IP_ Vol. 2 by Douglas Comer and David Stevens.  The
code is great, and explained very well, but it uses some parts of the Xinu
operating system that may or may not exist already in most unix systems.

Does someone out there have the code from this book working on a unix machine?
 If so, would you mind sharing your replacements for Xinu's message passing,
semaphores, and queueing functions?

I'd also appreciate any advice for implementing these myself if no one's done
it already.

Thanks in advance.

Jim
_______________________________________________________________________________
james speth           email for pgp-compatible public key         speth@end.com
_______________________________________________________________________________
Our request is in keeping with the preference of the community,    -Post Office
as reflected by local ordinance.                                    Sign

-----------[000081][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 13:32:49 -0700
From:      guy@nova.netapp.com (Guy Harris)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

>You may, or may not, have any success trying to run a SunOS 4.0[.x] lock
>manager, which would "manage the locks by itself", on a SunOS 4.1[.x]
>system.

And even if you *do* get it to run, the 4.0[.x] lock managers had some bugs
fixed in later 4.1[.x] lock managers, so even if you *do* manage to run
the 4.0[.x] lock manager on a 4.1[.x] system, doing so may cause more
problems than it fixes....

-----------[000082][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 08:49:36 -0400
From:      philp@universe.digex.net (Phil Perucci)
To:        comp.protocols.tcp-ip,comp.os.ms-windows.misc
Subject:   Mosaic (Winsock) over Chameleon 4.0?

With WFW 3.11 (no microsoft TCP/IP) and ChameleonNFS 4.0 installed,
has anyone been able to run Mosaic?

A more general question, do the freeware/shareware Winsock apps run
OK over Chameleon 4 *IF* the "properties" of the Winsock app's icon
(working directory) point to the ChameleonNFS Winsock.DLL?

-- 
==============================================================================
 Phil Perucci             | "All postings are my own opinion - all comments
 Systems Integrator       |  are intended for research/educational purposes"
==============================================================================

-----------[000083][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 94 10:38:24 -0500
From:      kylesm@TIGGER.STCLOUD.MSUS.EDU (Kyle Smith)
To:        comp.protocols.tcp-ip
Subject:   AS/400 TCP/IP Problem

Hello all,

I am having a small problem with TCP/IP Connection Utilities/400 for 
the AS/400.  I am running version 2.3 along with OS/400 V2R3M0.  I
am trying to TELNET into a DEC host.  It seems, however, that the AS/400
is always negotiating terminal type VT52, even though I have the default
set to VT200B8 (8-bit VT200).  We have traced the communications and it is
definately negotiating VT52 not just timing out and defaulting to it.

Has anyone else expereinced similar problems?  

Thanks in advance.

Kyle
Kyle@pebbles.cmghp.HealthPartners.com
+----------------------------------------------------------------------------+
| Kyle J. Smith |   Amateur Radio N0NYS   |  KYLESM@TIGGER.STCLOUD.MSUS.EDU  |
 +---------------+-------------------------+----------------------------------+
|          A mind is like a parachute; it only functions when open.          |
 +----------------------------------------------------------------------------+
| In order to achieve the impossible, one must see the invisible - Nietzsche |
+----------------------------------------------------------------------------+


-----------[000084][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 08:44:25
From:      CLAY@tnsnoc2.sdsu.edu (Robert D. Clay)
To:        comp.protocols.tcp-ip
Subject:   Re: Dynamic load sharing via a single IP address.

In article <2nuget$qbl@lykos.netpart.com> phil@lykos.netpart.com (Phil Trubey) writes:
>Path: newshub.sdsu.edu!nic-nac.CSU.net!usc!howland.reston.ans.net!agate!library.ucla.edu!csulb.edu!csus.edu!netcom.com!netcomsv!lykos.netpart.com!lykos.netpart.com!not-for-mail
>From: phil@lykos.netpart.com (Phil Trubey)
>Newsgroups: comp.protocols.tcp-ip
>Subject: Dynamic load sharing via a single IP address.
>Date: 6 Apr 1994 07:19:56 -0700
>Organization: NetPartners, Newport Beach, CA
>Lines: 24
>Message-ID: <2nuget$qbl@lykos.netpart.com>
>NNTP-Posting-Host: localhost


>I have a situation where I have 2 machines offering an identical TCP 
>service (clients can connect to either and get the same information/service).
>Is there a way for clients to open a connection to a single IP address
>and have the request go to either one of the machines via some 
>combination of switching, routing, etc?  ie. I'd like to be able to
>dynamically load share the service between these two machines - and
>also allow all connections to go to one machine if the other is down.

DEC (yes DEC) offers a product that lets you do this on Unix boxes.  Here is 
the product description (I recently got it from DEC):
-----------------------------------------------------------------------------
LSF 1.1 Compute-Sharing Software For UNIX Farms

HIGHLIGHTS

o  Digital will distribute the Load Sharing Facility software that enables 
   Alpha AXP systems, along with other vendors' installed systems, to form 
   heterogeneous UNIX clusters or "farms"

o  Provides for NQS-like batch execution of compute intensive jobs with 
   powerful job control and flexible job queue configuration.

o  A complete application program interface allows application developers to 
   directly access information to and from the load-sharing scheduler

o  Available in mid-November 1993


In industries such as discrete manufacturing and defense, education and 
research, chemical, pharmaceutical, and oil and gas, three trends are 
occurring that are forcing changes to the numerically intensive paradigm.

One, management is demanding that departments eliminate supercomputers to 
save on capital and maintenance costs.  In a move called "downsizing", many 
supercomputer applications are being migrated to less expensive computing 
options.  Two, some workstations are overloaded by users performing compute 
intensive applications, while others remain essentially idle.  
Administrators are asked to purchase and upgrade equipment for specific 
applications even though the aggregate computing power on site far exceeds 
the total needed.  These administrators must find a way to harness these 
unused compute cycles and put them to good use.  Three, individual 
workstation users, who never had access to a supercomputer, are rapidly 
outgrowing the power of a single system as their compute intensive 
applications grow more complex.  They must "upsize" their workstation power, 
but a supercomputer is simply not an option.

The solution to each of these situations is to link all the available 
workstations and servers into clusters called "farms".  Customers have a 
strong need for software that can accomplish this link across systems from 
various manufacturers, and make the aggregate computing power available to 
all users on the network when they need it.

DIGITAL SOLUTION

Digital has teamed with Platform Computing Corporation to offer the Load 
Sharing Facility (LSF) cluster compute-share software for use on 
heterogeneous UNIX workstation clusters.  LSF 1.1 gives users transparent 
access to the full compute power of systems normally sitting unused on the 
network.  Interactive and batch jobs are automatically sent to the least 
loaded system in the cluster, eliminating situations where individual 
machines are compute-bound.  This software also provides transparent 
parallelization of common compute-intensive UNIX activities.  LSF 1.1 
supports DEC AXP OSF/1, IBM AIX on RS/6000s, HP-UX on HP9000s, SunOS on Sun 
SPARCstations, and Sun Solaris on Sun SPARCstations.

CUSTOMER BENEFITS

End users see the greatest benefit in reduced job completion time, 
especially for very large compute intensive jobs.  Now each user in the 
system, no matter what the workstation power, has access to a virtual 
supercomputer of computing power.  Administrators and managers benefit 
through better utilization of existing corporate computing resources, plus 
flexibility in system management and upgrades.  These benefits are currently 
being realized at LSF test sites containing thousands of networked 
computers.

PRICING/ORDERING INFORMATION

NOTE:  LSF software is offered under the terms and conditions of Platform 
       Computing Corporation.

Model No.               Description                          U.S. List

QB-2C2AA-WA             LSF Pkg O/A Kit                         $1,045
QM-2C2AA-BA             LSF Pkg O/A Lic agreement                  995
QB-2C2AA-WB             LSF Pkg U/R Kit                          1,045
QM-2C2AA-BB             LSF Pkg U/R Lic agreement                  995
QB-2C2AA-WC             LSF Pkg HP Kit                           1,045
QM-2C2AA-BC             LSF Pkg HP Lic agreement                   995
QB-2C2AA-WD             LSF Pkg IBM Kit                          1,045
QM-2C2AA-BD             LSF Pkg IBM Lic agreement                  995
QB-2C2AA-WE             LSF Pkg SUN Kit                          1,045
QM-2C2AA-BE             LSF Pkg SUN Lic agreement                  995
QB-2C2AA-WF             LSF Pkg SOL Kit                          1,045
QM-2C2AA-BF             LSF Pkg SOL Lic agreement                  995

----------
AIX and IBM are registered trademarks of International Business Machines 
 Corporation.
Sun is a registered trademarks of Sun Microsystems, Inc.
SPARCstation is a trademark of SPARC International, Inc.
SunOS and Solaris are trademarks of Sun Microsystems, Inc.
RS/6000 is a trademark of International Business Machines Corporation.
HP 9000 and HP-UX are trademarks of Hewlett Packard Corporation.
LSF and Platform Computing are registered trademarks of Platform Computing
Corp.
-----------------------------------------------------------------------------

Robert D. Clay                    San Diego State University
Data Communications Manager       Telecommunications and Network Services
619/594-7309 (Voice)              San Diego, CA 92182	
619/594-2912 (FAX)                Robert.Clay@sdsu.edu (Internet)

-----------[000085][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 11:40:59 -0400
From:      jkim@panix.com (Joachim Kim)
To:        comp.protocols.tcp-ip
Subject:   Re: Distinct TCP-IP -- Anybody using this?

I too would like information regarding Distinct... anybody using the VB tools also?


-----------[000086][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 12:54:04 GMT
From:      w-rolph@ds.mc.ti.com (Don Rolph)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

In article <1994Apr4.174820.15737@qii.sialis.com> john@qii.sialis.com (John Dassow) writes:
>From: john@qii.sialis.com (John Dassow)
>Subject: Word 6.0a .doc saving problem under W4W 3.11
>Date: 4 Apr 94 17:48:20 GMT


>   When trying to save shared Word 6.0a files with the extention and mounted
>with NCR's NFS product .doc onto our NCR tower we get an error messgage like
>'File is locked, please disengage write protect tab'.  We are running Word 6.0a
>under Word for Windows 3.11 and are using WinTCP+. This only happens for files
>saved as '.doc', any other file format will save with no problems.  Does anyone
>have any ideas of why this happens or how to correct it?
 
>-- 

Microsoft has changed the file locking with all new ole 2.0 compliant apps, 
of which Word 6.0a is one.  This requires server mods which are not presnet in 
many old servers.  Old server will then break just as you described.

You reinforce my hypothesis that nfs mounts will nearly all fail under ole 2.0 
apps.

Regards.
 
Don Rolph w-rolph@ds.mc.ti.com WD3 MS10-13 (508)-699-1263

-----------[000087][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 12:56:52 GMT
From:      w-rolph@ds.mc.ti.com (Don Rolph)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

In article <1994Apr5.152711.10409@athen.mch.sni.de> egn@athen.mch.sni.de (Emil Naepflein) writes:
>From: egn@athen.mch.sni.de (Emil Naepflein)
>Subject: Re: Word 6.0a .doc saving problem under W4W 3.11
>Date: Tue, 5 Apr 94 15:27:11 GMT
 
>In article <1994Apr4.174820.15737@qii.sialis.com> john@qii.sialis.com (John
>Dassow) writes:
>>
>>   When trying to save shared Word 6.0a files with the extention and mounted
>>with NCR's NFS product .doc onto our NCR tower we get an error messgage like
>>'File is locked, please disengage write protect tab'.  We are running Word 6.0a
>>under Word for Windows 3.11 and are using WinTCP+. This only happens for files
>>saved as '.doc', any other file format will save with no problems.  Does anyone
>>have any ideas of why this happens or how to correct it?
>>
 
>This seems to be a problem of all AT&T System V based systems. The problem
>is that WfW issues lock requests which use the full 32 bit address range
>for the offset. This is allowed in the XNFS protocol, but causes problems
>in the System V implementation of the lock manager. The System V lock manager
>uses the fcntl(2) call to manage the locks, but it uses off_t to pass the
>offset which is actually a signed 32-bit value. If a offset mit the 31st
>bit set is used, the kernel interprets this value as negativ and returns
>EINVAL. So all the System V implementations of the lock manager are not
>XNFS compliant!



>I hope this helps. I'am also looking for an easy solution to this problem.
 
>Emil

Microsoft has basically screwed us all.  This problem has cropped up with a 
Novell  3.11 server, my lanman/x servers, and your nfs server.  When microsoft 
changed the locking they killed virtually anything not based on lanman 2.0 or 
above.  The users dont really seem to care, however, since no one is 
complaining.  I guess we get what we deserve.


Regards.
 
Don Rolph w-rolph@ds.mc.ti.com WD3 MS10-13 (508)-699-1263

-----------[000088][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 13:19:08 GMT
From:      john@iastate.edu (John Hascall)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   Re: Tales from the Real World

Joel Spencer <jspencer@cass.ma02.bull.com> wrote:
}Does anyone know if there is such a thing as an *illegal* host name? 

   yes.

}Last week, I visited a customer who had some routers named "router-0/1",
}"router-0/2" etc etc etc.  This did not really fuss me until I abstracted 
}a full snmp oid into a particular application.  Of course, the application
}took the "/" character to be a delimiter in the oid.  Therefore, returning
}a status indicating that i had requested an illegal/non-existent 
}management object.  
}
}Is it "illegal/inadvisable" to use the "/" char as part of a name?  Are
}there any other characters to avoid?  Is this in an rfc anywhere?

   It is both illegeal and inadvisable.  The legal characters
   are outlined in RC952 as...

      1. A "name" (Net, Host, Gateway, or Domain name) is a text string up
      to 24 characters drawn from the alphabet (A-Z), digits (0-9), minus
      sign (-), and period (.).  Note that periods are only allowed when
      they serve to delimit components of "domain style names". 

   Case is not significant.  RFC1123 says hosts must now support names
   upto 63 characters -- although going over 31 still breaks some things :(

John
-- 
John Hascall              ``I can't tell you that Karla Homolka pled guilty''
Systems Software Engineer
Project Vincent
Iowa State University Computation Center  +  Ames, IA  50011  +  515/294-9551

-----------[000089][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 94 14:05:22 GMT
From:      gdlee@bsu-cs.bsu.edu (Gary D. Lee)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   tcp-ip software for windows and Mac


Sorry if I haven't got the correct groups here, but they are the
only ones I have access to.  
	Here at Ball State, we are looking for a tcp-ip communications
	suite for all our faculty, student, and staff users.  We are 
	currently using Pathway Access from TWG.  while it seems 
	to meet most of our needs at present, the customer support 
	and user interface leave much to be desired.  
		We are looking for a replacement package which meets
		the following requirements.

		1. Must provide Dec vt240 terminal emulation
		2. must include ftp
		3. must provide multiple tn3270 sessions.
		4. would prefer to have the same package for both 
		   Macintosh and Windows with nearly identical
		      interfaces.  This will make my job as 
			 primary campus support much simpler.

			 5. must co-exist with novell netware.

			 thanks for any and all help.  please respond
			 via email as I don't get to read news as
			 often as I would like.
-- 
		    Senior Microcomputer analyst
		    Ball State University
		     phone 317-285-1853
                    :  gdlee@bsu-cs.bsu.edu

-----------[000090][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 21:37:06 -0400
From:      robb@epenviron.eapi.com (Rob Beckman)
To:        comp.protocols.tcp-ip
Subject:   xntp between ver 2 & 3


We are running SCO UNix and using xntpd version 2.  We have another 
machine running SunOS with a public domain version of xntp that is
version 3.

Does anyone know what we have to do to get the Sun xntp to read a
time feed from the SCO xntp?  Is other words how do  we get a version 3
of xntp to read xntp version2 time feed.

Thanks,
Rob Beckman
robb@eapi.com



-----------[000091][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 15:31:18 GMT
From:      ced@bcstec.ca.boeing.com (Charles Derykus)
To:        comp.protocols.tcp-ip
Subject:   Can rdist pull?

I posted this in comp.unix.questions and got no
response so I decided to re-direct here.

  Does anyone know if rdist can be made to pull rather
  than push? 

  The FM doesn't specify a way but Stevens in Unix
  Network Programming drops a tantalizing tidbit on
  pg. 565:  
   
    "Also note that both the rcp and rdist programs 
     function as either client or server. They both
     do this by using undocumented command line ar-
     guments to tell the other copy of itself to act
     as a server instead of a client"


Any comments/advice appreciated.

---
-- 
Charles DeRykus				Internet:   ced@carios2.ca.boeing.com
Boeing Computer Services		UUCP:	    ...!uunet!bcstec!ced

-----------[000092][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 16:30:33 GMT
From:      jjj@cs.hh.ab.com (James J. Jankowski)
To:        comp.protocols.tcp-ip
Subject:   Re: Tcp-ip Stacks co-existing with IPX & Netbeu

Clayton,

 We are currently running WFW 3.11, Novell 4.x client and PCTCP 2.3 with
great success.  Using QEMM 7.03 we have 575k free conventional memory.
This is on Gateway 2000 486/66 PC's with SCSI CD-ROM's and the ATI
Graphics Ultra VLB.  This also includes a 40k driver for NFS access using
PCTCP.

 Use of Microsoft's Windows for Workgroups 3.11 resource kit was a great
help.  We used the section on setting up WFW, Novell 4.x and Microsoft
TCP/IP.  Instead of using the Microsoft TCP/IP (memory hog) we used
PCTCP from FTP software due to its ability to load into EMS memory.
The only other document I needed was the PCTCP release notes for WFW 3.11
compatibility.  We considered other TCP/IP products which use (much) less
conventional memory but we need TCP/IP and NFS access from DOS not
just Windows, not to mention the fact that we use PCTCP with some of our
own commercial products.


 -------------------------------------------------------------
|0|   James J. Jankowski      Network Analyst - Desktop Services           	|0|
|0|   Allen-Bradley Co.   747 Alpha Dr.   Highland Hts.,   Ohio   44143.   	|0|
|0|   E-mail:jjj@cs.hh.ab.com             Phone:+1(216)646-3933            	|0|
 -------------------------------------------------------------



-----------[000093][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 17:30:19 GMT
From:      frontline@cix.compulink.co.uk ("Frontline Distribution L")
To:        comp.protocols.tcp-ip
Subject:   POP Server Software

Looking for POP EMail software for a Sun Sparc2 running SunOS 4.1.1 or 
SCO UNIX.

Anyone know any good packages - freeby or commercial ?

-------------------------------------------------------------------
Chris Miles                 | EMail: cmiles@cix.compulink.co.uk
Frontline Tech Supp Centre  |        cmiles@frontline.co.uk
Hampshire House, Wade Road  | CIS  : 100065,621
Basingstoke, Hampshire      | Tel  : +44 (0)256 847220
ENGLAND, RG24 8PL           | Fax  : +44 (0)256 847900
-------------------------------------------------------------------


-----------[000094][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 18:43:08 GMT
From:      kerch@parc.xerox.com (Berry Kercheval)
To:        comp.protocols.tcp-ip
Subject:   Re: Free tcp/ip

> -=> Quoting Bret A. Johnson to All <=-
 
> BAJ> In article <1994Mar24.150854.22947@jet.uk>, Paul Simmonds (ps@jet.uk)
> BAJ> write?s: >Microsoft have released their TCP/IP for DOS and Workgroups
> BAJ> for Windoes into >the Public Domain as FreeWare.


Well, I looked, since it sounded unlikely.  At least the Window for
Workgroups is NOT public domain and NOT freeware.  (Public Domain has
a specific legal meaning, at least in the USA.  Essentially it would
mean that Microsoft had legally relinquished ALL RIGHTS to it and
anyone could do ANYTHING with it including reselling it as their own
(Kerch's TCP/IP for Windows, for instance).  FreeWare is typically
copyrighted to prevent this, although it is still free.)

This TCP/IP for WFW seems to be neither.  It's a free beta version but 
still covered with a license agreement.  I qoute from the README.1st file in 
ftp.microsoft.com:/Peropsys/WFW/tcpip/vxdbeta:


  [under 'how to install] 
  The license agreement will be displayed and you will be asked to accept
  the terms of the agreement. Once you have accepted the files will be
  expanded to the current working directory.

Sigh.

  --berry

 

--

Berry Kercheval :: kerch@parc.xerox.com 
"...start with Plan 9, which is free of sin..." -Mark V. Shaney


-----------[000095][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 19:05:03 GMT
From:      Genady@genady.RAD
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   SNMP Proxy Agent


Is anywere some matherials about
building SNMP Proxy Agent?
Any help will be accepted with pleasure.

========================
Gennady Yakubovich
genady@radmail.rad.co.il

-----------[000096][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 20:43:58 GMT
From:      lefkogt%scs@sc4199.usafa.af.mil (Gary Lefko)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP ?

Is there any place I can get info on protocols and how they work?  I am in a
Computer Systems Organization class and have to do a term paper.  Any help
would be appreciated.  TCP/IP info also appreciated...Thanks, again!

-----------[000097][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 20:49:40 GMT
From:      les@chinet.chinet.com (Leslie Mikesell)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

In article <2nshov$o54@tokyo.engr.sgi.com>,
Dave Crocker <@udel.engr.sgi.com> wrote:
>|> 
>|> In the field of mixed protocol stacks I recognize the need for running
>|> TCP application protocols on top of OSI transport service.
>
>While there has been discussion
>of running TCP apps over OSI lower layers -- and there is no technical
>impediment that I can think of -- the real test is whether there is a
>compelling need for it.

I can't think of many TCP apps that should be allowed to work without
being able to do a reverse lookup to see who is making a connection,
and all of the existing ones are going to depend on the underlying
IP addressing scheme.  I don't even know of a way to do this using
the TLI transport interface when the transport isn't IP (for example,
AT&T's osi-based StarGroup).  Is it possible? 

Les Mikesell
  les@chinet.com

-----------[000098][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 21:43:32 GMT
From:      adiwan@siacbbn.com (Arif Diwan (BBN))
To:        comp.protocols.tcp-ip
Subject:   Re: multiple interfaces on a machine and coping w/ failures

In article <GTHAKER.94Apr5173234@polyphony.sw.stratus.com>,
	gthaker@polyphony.sw.stratus.com (Gautam Thaker) writes:
 Suppose I have two machines with two ethernet interfaces each and
 connect these two machines with two ethernet cables. Each machine thus
 has two IP addresses. (call these addresses A and A' and B and B') If
 I have a TCP/IP connection open from one machine to another (say from
 A<->B, and if I then cut the ethernet cable that is actually carrying
 this traffic, what all do I need to be sure that the traffic will be
 rounted over the remaining ethernet with no loss of connection at the
 TCP level?

 Is this something I can achieve by just properly setting up routes?
 Or is loss of connection unavoidable?
...
<<
I have accomplished this by using virtual IP interfaces. Each of the
hosts should have a virtual IP interface. Establish TCP
connections between virtual IP interfaces then you use gated or
something similar to route between the real interfaces.

I know of two ways to create virtual interfaces:

1. Modify the if code to add lo1 then ifconfig lo1 with an IP
address with a 1 bit subnet!!! Note that the broadcast address &
the host address are the same in a one bit subnet. gated advertises
the 1 bit subnet. You can do it differently if you want.

2. Use MorningStar ppp to create a dynamic interface and use that
   as a virtual interface. Note you are not using ppp, just their
   dynamic interface device drivers (or you can write your own).

All this works on Suns under Sun OS 4.1.3. Other OSs can be made to
work similarly. I am also running OSPF with variable length subnet masks
to make optimal use of the available address space.

The end-systems (ESs) in my case are separated by a complicated internetwork
(WANs,LANs) but I can still rollover TCP connections between the real interfaces.


-- 

				-- Arif Diwan (adiwan@bbn.com)
						617-873-6274

-----------[000099][next][prev][last][first]----------------------------------------------------
Date:      Wed, 6 Apr 1994 22:02:49 GMT
From:      sjacobs@maroon.tc.umn.edu (Steve Jacobson)
To:        comp.protocols.tcp-ip,comp.os.ms-windows.misc
Subject:   Re: Mosaic (Winsock) over Chameleon 4.0?

In article <2nub50$lpe@universe.digex.net> philp@universe.digex.net (Phil Perucci) writes:

>With WFW 3.11 (no microsoft TCP/IP) and ChameleonNFS 4.0 installed,
>has anyone been able to run Mosaic?
 
>A more general question, do the freeware/shareware Winsock apps run
>OK over Chameleon 4 *IF* the "properties" of the Winsock app's icon
>(working directory) point to the ChameleonNFS Winsock.DLL?

Yes, Mosaic and other freeware apps run fine over Chameleon.  Cello does not, 
but apparently Cello has quite a few other problems.  Hopefully, Tom at LII 
will fix the problems Cello has with Chameleon.  

-----------[000100][next][prev][last][first]----------------------------------------------------
Date:      6 Apr 1994 20:48:06 +0200
From:      agulbra@nvg.unit.no (Arnt Gulbrandsen)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   Re: Tales from the Real World

In article <2npba4$afg@cass.ma02.bull.com>,
Joel Spencer <jspencer@cass.ma02.bull.com> wrote:
>Does anyone know if there is such a thing as an *illegal* host name? 

If I remember the RFCs correctly, each part of the fully qualified
domain name must consist entirely of letters, digits, "-" and "/",
and it's best to start with a letter, though RFC112[23] allows
digits as well as long as the name can't be mistaken for an IP
address.

I just saw a PTR record that said "${HOST}.${ORG}.no".  I wonder why
BIND even allows such names?  Not the autoritative server, of course
it must be hosed, but but why will my local name server accept,
cache and forward an RR that contains such trash?

--Arnt

-----------[000101][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 01:07:31 GMT
From:      fetrow@biostat.washington.edu (Dave Fetrow)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

In article <w-rolph.391.0007F2E1@ds.mc.ti.com> w-rolph@ds.mc.ti.com (Don Rolph) writes:
>
>Microsoft has basically screwed us all.  This problem has cropped up with a 
>Novell  3.11 server, my lanman/x servers, and your nfs server.  When microsoft 
>changed the locking they killed virtually anything not based on lanman 2.0 or 
>above.  The users dont really seem to care, however, since no one is 
>complaining.  I guess we get what we deserve.

 Actually, it appears even worse. According to todays InfoWorld
Cringely Rumours column: Several Microsoft Ole 2.0 Apps are sucking
sucking locks and not letting go. I don't think even lanman 2.0 can
save you from that.

-- 
   -dave fetrow-			


-----------[000102][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 01:12:40 GMT
From:      mleavitt@netcom.com (Michael R. Leavitt)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP Tutorial wanted


I have the unenviable task of trying to persuade--by means of a one-hour
briefing--some protocol-knowledge challenged, but otherwise very smart,
people of the nature and advantages of moving to a TCP/IP-based, routed
network. 

My questions, in order to save me the work of preparing such a briefing
(and reinventing what must be a much-invented wheel), are: has anyone
prepared such a briefing before and if so, would you be willing to share
it with me?  Alternatively, if anyone knows of a chapter or section in one
of the million books now availble on TCP/IP and the Internet, could you
send me a pointer to it?

If this is in a FAQ, I beg your pardon and humbly request a pointer to 
the appropriate archive.
-- 
Mike Leavitt<mleavitt@netcom.com>
Reston, Virginia

-----------[000103][next][prev][last][first]----------------------------------------------------
Date:      Thu, 07 Apr 94 08:13:01 PDT
From:      bwwhite@mmm.com
To:        comp.protocols.tcp-ip,comp.os.ms-windows.misc
Subject:   Re: Mosaic (Winsock) over Chameleon 4.0?


In article <2nub50$lpe@universe.digex.net>, <philp@universe.digex.net> writes:
> Path: 
dawn.mmm.com!umn.edu!news-feed-1.peachnet.edu!gatech!howland.reston.ans.net!new
s.intercon.com!news1.digex.net!digex.net!not-for-mail
> From: philp@universe.digex.net (Phil Perucci)
> Newsgroups: comp.protocols.tcp-ip,comp.os.ms-windows.misc
> Subject: Mosaic (Winsock) over Chameleon 4.0?
> Date: 6 Apr 1994 08:49:36 -0400
> Organization: Universal Access by Digital Express. 800-969-9090
> Lines: 12
> Message-ID: <2nub50$lpe@universe.digex.net>
> NNTP-Posting-Host: universe.digex.net
> Xref: dawn.mmm.com comp.protocols.tcp-ip:12891 comp.os.ms-windows.misc:26069
> 
> With WFW 3.11 (no microsoft TCP/IP) and ChameleonNFS 4.0 installed,
> has anyone been able to run Mosaic?
> 
> A more general question, do the freeware/shareware Winsock apps run
> OK over Chameleon 4 *IF* the "properties" of the Winsock app's icon
> (working directory) point to the ChameleonNFS Winsock.DLL?
> 
> -- 
> 
 ==============================================================================
>  Phil Perucci             | "All postings are my own opinion - all comments
>  Systems Integrator       |  are intended for research/educational purposes"
> 
==============================================================================

Phil,	

I've been able to run Mosaic using Chameleon 4.0 and WFWG 3.11.  I don't recall 
doing anything special to get it to work.  I know you need to install Chameleon 
*after* WFWG.

Also, I've run some other Winsock apps like ELM and Trumpet mail readers OK 
without having them point to the NETMANAG directory (where their WINSOCK.DLL is 
located).  I should mention that NETMANAG is in my path statement.

At this moment I'm trying the Microsoft beta VxD winsock and am running 
Netmanage's mail program that came with Chameleon 4.0.  I took advice from some 
responses to you (I think) in another group.  Rather than changing the working 
directories for a bunch of the Netmanage programs to point to the Microsoft 
Winsock rather than the Netmanage Winsock, I just renamed the Netmanage 
WINSOCK.DLL file to, e.g., WINSOCK.LLD.  Worked fine.

Later.


-----------[000104][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 03:25:22 GMT
From:      James G. Speth <speth@end.com>
To:        comp.protocols.tcp-ip
Subject:   Re: Help implementing TCP/IP code?

In article <2ntbb4INNl14@excalibur.cs.purdue.edu> John Chueng-Hsien Lin,
lin@cs.purdue.edu writes:
>    Xinu-TCP uses a "process-oriented" approach to implement TCP/IP mainly
> because Xinu is a multi-threaded kernel. Unix is not a multi-threaded kernel,
> thus a "procedure-based approach" is more approate, IMHO.

Could you by any chance point me toward some code that implements a procedure
based TCP/IP for a unix system?

Jim
_______________________________________________________________________________
james speth           email for pgp-compatible public key         speth@end.com
_______________________________________________________________________________
Our request is in keeping with the preference of the community,    -Post Office
as reflected by local ordinance.                                    Sign

-----------[000105][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 13:00:22 -0500
From:      lin@cs.purdue.edu (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   Re: Help implementing TCP/IP code?

In article <2nvuf2$mkt@darkstar.UCSC.EDU>, James G. Speth <speth@end.com> writes:
> Could you by any chance point me toward some code that implements a procedure
> based TCP/IP for a unix system?

  Sure, try the following anomynous ftp:

  ftp.uu.net:/systems/unix/bsd-sources/sys/netinet

John Lin (lin@cs.purdue.edu)

-----------[000106][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 16:35:22 -0700
From:      jdickson@nemesis.jpl.nasa.gov (Jeff Dickson)
To:        comp.protocols.tcp-ip
Subject:   Re: Timeout for socket read/write

In article <2o1uma$91e@phoebe.jpl.nasa.gov> pjs@euclid.jpl.nasa.gov writes:
>
>Hi.  I'm writing a TCP/IP client-server application with sockets
>and would like to introduce a timeout where I call read(2),
>since if the other end has gotten out of whack it may not be
>sending anything but still have the connection open.  I have looked
>in both of my W. Richard Stevens books without finding anything
>useful.  Can someone please tell me the best (er, make that
>simplest, I don't want to get into adaptive timeouts) way of
>coding a timeout on the read?  And might I also need a timeout
>when calling write(2) on a socket?
>

Normally system calls are restarted automatically when a signal is delivered,
but you can prevent this from happening. The man page for read briefly men-
tions something about this. Look at the man page for sigvec for a better under-
standing of how this is accomplished. The timeout can be arranged by using one
of the three interval timers. Each of these timers generate a signal when they
expire. See the man page for getitimer. 

Jeff S. Dickson				jdickson@vtol.jpl.nasa.gov








-----------[000107][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 06:55:18 GMT
From:      Cullhane Jason Gibbs <cullhane@medusa.itc.gu.edu.au>
To:        comp.protocols.tcp-ip
Subject:   Re: WinSock software recomendations

I use the following winsock compliant apps:


Eudora - email
Hgopher2.3 - gopher
WinTrumpet - Network news
QVTNet - Telnet, ftp, news, email
Mosaic
WinFTP, ftp ,etc , etc -- there's heaps out there.....


L8r

-----------[000108][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 14:43:40 -0400
From:      faldor@sol.cms.uncwil.edu (Sherman Brown)
To:        comp.protocols.tcp-ip
Subject:   http, where can I get a definition of this protocol?

	Curious, could someone point in the right path to getting a
formal definition of the htt-protocol used by mosaic.  And
possibly some examples in code.

thanks,
Faldor

-----------[000109][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 12:16:10 GMT
From:      mantor@ccsg.tau.ac.il (bercovici sam)
To:        comp.protocols.tcp-ip
Subject:   checking line status

Sorry if this is a FAQ.
I am currently involved in a project which does peer to peer message passing
using IP protocol. The networking will be done through routers.
Some areas of the network are connected using slow rate lines ~9600 baud.
I am interested to find books that are dealing with the followin problems:
1. discovery of connection fails as soon as possible over TCP/UDP.
2. ways to keep track of appearing and disappearing nodes.

please answer me through E-mail.
my E-mail address is mantor@ccsg.tau.ac.il
	Thanks

-----------[000110][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 12:40:26 GMT
From:      Richard_Lyons@ccmail.gsfc.nasa.gov (Richard Lyons)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP performance references.

In article <Cnt0x4.KKn@murdoch.acc.Virginia.EDU>, clv2m@uvacs.cs.Virginia.EDU (Charles Viles) says:
>
>Forgive me if this is an FAQ...
>
>I am looking for some references that document tcp-ip performance, in
>particular, things like time to build a connection, bandwidth, time to
>tear down the connection etc.
>
>Any pointers to on-line documentation or the literature would be
>appreciated. I'll summarize if there seems to be general interest
>in the topic.
>
>-- 
>Charles Viles (clv2m@virginia.edu)
>Department of Computer Science
>University of Virginia
>Charlottesville, VA 22902

If that list could be broadcasted, I'd like it to please!

Thanks much...

-----------[000111][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 13:50:26 GMT
From:      kurt@unirsvl.rsvl.unisys.com (Kurt Matthys )
To:        comp.protocols.tcp-ip
Subject:   Re: multiple interfaces on a machine and coping w/ failures

In <GTHAKER.94Apr5173234@polyphony.sw.stratus.com> gthaker@polyphony.sw.stratus.com (Gautam Thaker) writes:


>Suppose I have two machines with two ethernet interfaces each and
>connect these two machines with two ethernet cables. Each machine thus
>has two IP addresses. (call these addresses A and A' and B and B') If
>I have a TCP/IP connection open from one machine to another (say from
>A<->B, and if I then cut the ethernet cable that is actually carrying
>this traffic, what all do I need to be sure that the traffic will be
>rounted over the remaining ethernet with no loss of connection at the
>TCP level?
 
>Is this something I can achieve by just properly setting up routes?
>Or is loss of connection unavoidable?

Routing won't do it.  Moving to the remaining ethernet requires the use of
different addresses (i.e. A' and B') and that means a different connection.
This is something that just about has to be done at the application level.

Kurt Matthys
kurt@rsvl.unisys.com

-----------[000112][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 14:16:46 GMT
From:      lefkogt%scs@sc4199.usafa.af.mil (Gary Lefko)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP Tutorial wanted

In article <mleavittCnv7D4.6GD@netcom.com>, mleavitt@netcom.com (Michael R. Leavitt) says:
>I have the unenviable task of trying to persuade--by means of a one-hour
>briefing--some protocol-knowledge challenged
>
>
>
>
>
>
>
I would be interested in the same info, please!!!
I have a term paper on network protocols which includes same type of
briefing.






-----------[000113][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 14:28:19 GMT
From:      Dave Crocker <dcrocker@mordor.stanford.edu>
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

In article <CnqvA7.no@dove.nist.gov> Jim West,
west@mgmt3.ncsl.nist.gov writes:
>the network service offered.  The hard part is modifying the Name
 Service that
>TCP application use to map names to NSAP addresses (versus the normal
 operation of
>names to IP addresses).

Not such a hard problem.  As part of the TUBA effort, NSAP RR entries
for
the domain name system have been defined.  Adding RRs to the DNS
happens
rarely, but it does happen.  Modifying the apps.  Modifying the support
routines (gethostbyname, etc.) to work with NSAPs rather than IP
addresses
will take some work, but there's nothing risky in it.  All
straightforward.

Again, the real question is why to do it?

-----------[000114][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 14:48:07 GMT
From:      ps@jet.uk (Paul Simmonds)
To:        comp.protocols.tcp-ip
Subject:   Re: Free tcp/ip

>In <2nuvrs$17u@news.parc.xerox.com> kerch@parc.xerox.com (Berry Kercheval) writes:
 
>>> BAJ> In article <1994Mar24.150854.22947@jet.uk>, Paul Simmonds (ps@jet.uk)
>>> BAJ> write?s: >Microsoft have released their TCP/IP for DOS and Workgroups
>>> BAJ> for Windoes into >the Public Domain as FreeWare.
 
>>Well, I looked, since it sounded unlikely.  At least the Window for
>>Workgroups is NOT public domain and NOT freeware.  (Public Domain has
>>a specific legal meaning, at least in the USA.  Essentially it would
>>mean that Microsoft had legally relinquished ALL RIGHTS to it and
>>anyone could do ANYTHING with it including reselling it as their own
>>(Kerch's TCP/IP for Windows, for instance).  FreeWare is typically
>>copyrighted to prevent this, although it is still free.)
 
>>This TCP/IP for WFW seems to be neither.  It's a free beta version but 
>>still covered with a license agreement.  I qoute from the README.1st file in 
>>ftp.microsoft.com:/Peropsys/WFW/tcpip/vxdbeta:
 
>>  [under 'how to install] 
>>  The license agreement will be displayed and you will be asked to accept
>>  the terms of the agreement. Once you have accepted the files will be
>>  expanded to the current working directory.
 
>OK, since I have been inundated,  Microsoft released this with the following
>statement: "The software is available at no charge, and the licence is an
>unlimited user licence"  -- I translate that as FreeWare.
 
>Look in:  ftp.microsoft.com
 
>          /Advsys/MSClient/dos/35/*.*     MS DOS & windows 3.5in
>                          /dos/525/*.*    MS DOS & Windows 5.25in
>                          /WFW/WFWTCP.EXE Worgroups for Windows
 
>That's all I know, other than I have tested the DOS version and it works to 
>the extent I can ping our SUN boxes.  If anyone sucessfully gets any apps
>to run on top of the DOS version please let me know.  The Windows version
>comes with WINSOCK (as far as I understand) so should run with loads of stuff.
 
>Regards
 
>Paul Simmonds (ps@jet.uk)
- Disclaimer: Please note that the above is a personal view and should not 
  be construed as an official comment from the JET project.

-----------[000115][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 15:00:31 GMT
From:      Dave Crocker <dcrocker@mordor.stanford.edu>
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

In article <BILLW.94Apr4185610@glare.cisco.com> William,
billw@glare.cisco.com writes:
>TCP over CLNP is in fact TUBA, one of the proposed "future IP"
protocols.

The extent to which TUBA is, in fact, simply TCP over CLNP is a matter
of some debate.  The various IPng requirements are pushing proposals so
that all sorts of changes might be necessary.  Hence, to the extent
that
the original question to this list pertained to running Internet
applications
over vanilla, unmodified OSI-based stacks, TUBA might not prove to be
an
answer.

-----------[000116][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 94 20:21:38
From:      drw@zermelo.mit.edu (Dale R. Worley)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   Re: Tales from the Real World

In article <2nv056$4j9@trondviggo.nvg.unit.no> agulbra@nvg.unit.no (Arnt Gulbrandsen) writes:
   I just saw a PTR record that said "${HOST}.${ORG}.no".  I wonder why
   BIND even allows such names?  Not the autoritative server, of course
   it must be hosed, but but why will my local name server accept,
   cache and forward an RR that contains such trash?

Because generally programs should accept more than the rules allow.
In this case, it could process the record reasonably sanely, so it was
willing to do so.

Dale

Dale Worley		Dept. of Math., MIT		drw@math.mit.edu
--
A common anthropological dictum:  Every anthropologist gets the tribe
he deserves, and vice versa.	-- Adam S. Moskowitz

-----------[000117][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 94 20:25:48
From:      drw@zermelo.mit.edu (Dale R. Worley)
To:        comp.protocols.tcp-ip
Subject:   Re: XNTP is 1 hour off

robb@epenviron.eapi.com (Rob Beckman) writes:
   We have xntpd running on our network and everything was working fine
   until this weekend's time change and now we are 1 hour off.  Our
   TIMEZONE file is set to EST5EDT which is right but we are still off.

If you're not taking time from a network-wide NTP source, but rather a
local radio clock, it's possible that you have your radio clock
mis-configured.  The NTP code generally assumes that the radio clock
is configured to produce GMT timestamps.  If it is producing local
time, even though the clock is sending out the timezone correction it
is using, NTP assumes that the times are GMT.  This can cause your
system's clock to display bizarre behavior.

Learned about that the hard way,

Dale

Dale Worley		Dept. of Math., MIT		drw@math.mit.edu
--
Life in this society being, at best, an utter bore and no aspect of
society being at all relevant to women, there remains to civic-minded,
responsible, thrill-seeking females only to overthrow the government,
eliminate the money system, institute complete automation, and destroy
the male sex.	-- Valerie Solanas, radical feminist

-----------[000118][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 15:30:38 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Large flat network

In article <ganzhorn-050494084548@mojo.cisco.com> ganzhorn@cisco.com (Charles Ganzhorn) writes:
> ...
>A fairly typical approach would be to switch the clients, use FDDI to get
>off the switch, put servers for those clients on the FDDI and then route
>the FDDI's...

That sounds like it involves bridging between FDDI and Ethernet in the
switches, which is a potentially serious problem if the servers are real
FDDI machines.  Some FDDI-Ethernet switches are unable to correctly
translate FDDI frames to Ethernet packets.  This is commonly a problem
with 4KByte NFS/UDP/IP frames.

What you'd like is a box that functions as a bridge among Ethernets (i.e.
as a switch) and acts as a router for packets going between the Ethernets
and the FDDI link away from the group.  I have not heard of such a thing
yet.  The switch vendors have religious objections to understanding
anything except MAC packets and the router vendors don't have the right
hardware--yet, as far as I know.


Vernon Schryver    vjs@rhyolite.com

-----------[000119][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 16:27:57 GMT
From:      rmadhok@anl.liv.usw.com (Raghu Madhok)
To:        comp.protocols.tcp-ip
Subject:   Re: docs for UNIX socket programming?

In article <jaesoo-040494175734@jaesoo.cs.aukuni.ac.nz> jaesoo@cs18.cs.aukuni.ac.nz (Jaesoo Kim) writes:
 >In article <16F8AD115.TCPMAINT@TWNMOE10.Edu.TW>, TCPMAINT@TWNMOE10.Edu.TW
 >wrote:
 >> 
 >> Hi!
 >>  
 >>     Is there any document on the net describing basic UNIX socket
 >> programming? Thanx.
 >>  
 >> /Liu
 >> 
 >
 >I would appreciate it if you could send me the same information.
 >
 >Regards,
 >J.Kim 
 >
 >Jaesoo Kim                         |E-mail:jkim03@cs.auckland.ac.nz
 >Department of Computer Science     |       jaesoo@cs.auckland.ac.nz
 >The University of Auckland         |Tel.: +64 9 3737 599 (x 7265)
 >Private Bag 92019                  |Fax.: +64 9 3737 453
 >Auckland                           |No man is infalliable!
 >NEW ZEALAND                        |

I would strongly recommend the book "UNIX Network Programming" by Richard
Stevens.  It provides some excellent examples and covers the top thoroughly
from an application programming point of view.

...Raghu

comnetsv!anl!rmadhok

-----------[000120][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 17:56:24 +0000
From:      tkr@puffball.demon.co.uk (Tim Rylance)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP performance references.

The "Internet System Handbook" edited by Daniel C. Lynch and Marshall T. Rose,
Addison Wesley 1993, ISBN 0-201-56741-5 has an excellent 100-page chapter by
Jeffrey Mogul on "IP Network Performance".
--
Tim Rylance <tkr@puffball.demon.co.uk>

-----------[000121][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 18:46:54 GMT
From:      ez003145@othello.ucdavis.edu (Taylor Gautier)
To:        comp.protocols.tcp-ip
Subject:   SLIP as gateway

Since I have been attempting to do this with Linux, I have been posting 
over there.  But I have not succeeded in getting an answer.  Does anyone 
know if the following is possible w/out modify some programs to deal with 
it specifically?  

         point-to-point
Internet <----SLIP----> 128.120.250.xx Linux A         Linux B
128.120.250.100                          |                |
                                   122.122.122.1     122.122.122.2
                                         +----ethernet----+

My machine is Linux B.  I would like to be able to access the Internet 
directly.  So far, I have been able to login to Linux A, and then do 
whatever (ftp, telnet...), but cannot do whatever from my machine. 

As you can see, I gave just random addresses to our Linux machines, so I 
know that as of now, my packets get routed out, but never back in.  Is 
there a way to make the Linux A machine mask the packets to look like
it's own SLIP packets, and when they come back convert them back and send 
them to Linux B?

(I don't think it will be possible for me to get a certified IP address, 
because that would require UCD to have dynamic routing tables  -- since I 
could come in under any SLIP address --  and I just don't think that is 
going to happen)



-- Taylor Gautier   --  ez003145@hamlet.ucdavis.edu


-----------[000122][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 19:14:41 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Can rdist pull?

In article <CnuGG7.BzH@bcstec.ca.boeing.com> ced@bcstec.ca.boeing.com (Charles Derykus) writes:
>  Does anyone know if rdist can be made to pull rather
>  than push? 

No, it can't.  I've seen rdist-like programs that can posted to
comp.sources.unix, though.

>  The FM doesn't specify a way but Stevens in Unix
>  Network Programming drops a tantalizing tidbit on
>  pg. 565:  
>   
>    "Also note that both the rcp and rdist programs 
>     function as either client or server. They both
>     do this by using undocumented command line ar-
>     guments to tell the other copy of itself to act
>     as a server instead of a client"

This is totally unrelated to your question.  Rdist is implemented by having
the client (the pusher) perform the equivalent of "rsh host rdist -s" (I
don't know whether that's the actual special command line argument); rcp
works similarly.  This tells the remote rdist command to implement the
server side of the protocol rather than the client.  But it still expects
the client to tell it what to do, and it's still limited by the rdist
protocol; that protocol only includes sending files from the client to the
server, not vice versa.

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000123][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 20:55:22 GMT
From:      jeff@astph (Jeff Martin)
To:        comp.protocols.tcp-ip
Subject:   Optimizing TCP/IP Process Architecture

We are running Interactive UNIX Release 3 Version 3.2.  We are
developing a network database application using TCP/IP.

We have implemented a network database that creates a unique server
process for each client connection.  A master server waits at accept()
and fork()s after each connection.  Thus each client has its own server.

One server process could handle all the client connections using poll()
or select().  However we hoped to increase performance by increasing
concurrency.  We now have maximized concurrency along with a swamped
process table and limited scalability.

HOW CAN WE REDUCE THE NUMBER OF SERVER PROCESSES USING TCP/IP?  Preferably
allowing a tunable server/client ratio?  NOTE: Comer and Stevens suggest
in "Interworking with TCP/IP, Vol 3," that several server processes can
wait at accept() and establish a socket connection for each TCP/IP
transaction, which is way to costly.

Should you have any ideas or examples or be working on a similar
design we are very eager to talk with you.

Thanks, Jeff
-- 
Jeff Martin, dbms programmer,		Philadelphia Phillies
INET:	astph!jeff@attmail.com		Voice:	(814)234-8592x32
UUCP:	attmail.com!astph!jeff		FAX:	(814)234-1269

-----------[000124][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 21:24:17 GMT
From:      haynes@cats.ucsc.edu (James H. Haynes)
To:        comp.protocols.tcp-ip
Subject:   Re: Can rdist pull?


In article <CnuGG7.BzH@bcstec.ca.boeing.com>,
Charles Derykus <ced@bcstec.ca.boeing.com> wrote:
>
>  Does anyone know if rdist can be made to pull rather
>  than push? 

I don't know the answer for rdist, but you can get track from
athena-dist.mit.edu:/pub/ATHENA/misc/track.tar.Z    Track uses a
librarian-client model; or in other words it pulls rather than pushes.

-- 
haynes@cats.ucsc.edu
haynes@cats.bitnet

"Ya can talk all ya wanna, but it's dif'rent than it was!"
"No it aint!  But ya gotta know the territory!"
        Meredith Willson: "The Music Man"


-----------[000125][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 21:41:30 GMT
From:      pjs@euclid.JPL.NASA.GOV (Peter J. Scott)
To:        comp.protocols.tcp-ip,comp.unix.questions
Subject:   Timeout for socket read/write

Hi.  I'm writing a TCP/IP client-server application with sockets
and would like to introduce a timeout where I call read(2),
since if the other end has gotten out of whack it may not be
sending anything but still have the connection open.  I have looked
in both of my W. Richard Stevens books without finding anything
useful.  Can someone please tell me the best (er, make that
simplest, I don't want to get into adaptive timeouts) way of
coding a timeout on the read?  And might I also need a timeout
when calling write(2) on a socket?

-- 
This is news.  This is your       |    Peter Scott, NASA/JPL/Caltech
brain on news.  Any questions?    |    (pjs@euclid.jpl.nasa.gov)

-----------[000126][next][prev][last][first]----------------------------------------------------
Date:      Thu, 7 Apr 1994 21:44:27 GMT
From:      "George Stump" <gstump@twg.com>
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Re: TCP/IP software for DOS/Windows and MAc

 
=>	Here at Ball State, we are looking for a tcp-ip communications
=>	suite for all our faculty, student, and staff users.  We are 
=>	currently using Pathway Access from TWG.  while it seems 
=>	to meet most of our needs at present, the customer support 
=>	and user interface leave much to be desired.  
=>		We are looking for a replacement package which meets
=>		the following requirements.
=>
=>		1. Must provide Dec vt240 terminal emulation
=>		2. must include ftp
=>		3. must provide multiple tn3270 sessions.
=>		4. would prefer to have the same package for both 
=>		   Macintosh and Windows with nearly identical
=>		      interfaces.  This will make my job as 
=>			 primary campus support much simpler.
=>
=>			 5. must co-exist with novell netware.
=>
=>			 thanks for any and all help.  please respond
=>			 via email as I don't get to read news as
=>			 often as I would like.
=>-- 
=>		    Senior Microcomputer analyst
=>		    Ball State University
=>		     phone 317-285-1853
=>                    :  gdlee@bsu-cs.bsu.edu


Gary,

Are we talking about the same company?  If so the answer has been under your 
nose.  PathWay Access 3.0 provides ALL your stated requirements and more.  Many 
of the enhancements made to the latest version were interface related.  I would 
be very surprised if you felt the same with 3.0 at your desk.

The Wollongong Product Support manager and I are very interested in any details 
you can provide regarding unsatisfactory support.  While this impression is a 
rare one for us, no vendor's support can please everyone, neither will they get 
any better without detailed input from you.  I am biased, not because I work 
here, but because I have spent time listening in on their end of the phone with 
customers large and small, and was impressed with the amount of knowledge and 
patience that job takes.  Doing likewise would change anyone's view.

"If you're looking for a TCP/IP product that works the same way across PC and 
Macintosh platforms, with the same set of features, but is still well-integrated 
with platform-specific networking standards, Wollongong's PathWay Access is the 
better choice"  Network computing - Nov 93. (before 3.0) Need I say more?  

We value your business, tell us how we can better serve you.

George Stump
Wollongong
415 962-7134
gstump@twg.com

-----------[000127][next][prev][last][first]----------------------------------------------------
Date:      7 Apr 1994 23:42:26 GMT
From:      bpacker@hybrid.com ()
To:        comp.protocols.tcp-ip
Subject:   tcp checksum redundancy

 hello,

 i am  researching tcp checksum redundacny, i.e. tcp checksums redundant
 with mac crc for local lan connection. i have a couple of very basic
 questions...

 1. who has implemented this and written it up ? i glanced thru the
 latest rfc index and didn't see anything.

 2. i am assuming that all right minded tcp implementations require a
 checksum, i.e. they do not treat the checksum like udp does. thus, i am
 assuming that something like a bridge which was relaying packets between
 tcp endpoints on separate lans couldn't successfully peek into the
 tcp packet and decide to poke out the checksum. it seems that end to
 end agreement would be necessary....

 bummer for us would be pokers, spoofers, and hackers.

 thanks for help. 
 bp




---------
Robert Packer  
 Hybrid Networks, Inc    408/725-3261    bpacker@hybrid.com
 Result Oriented Systems 408/353-8191    Result_Oriented@cup.portal.com

-----------[000128][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 00:31:38 GMT
From:      mday@park.uvsc.edu (Matt Day)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   named performance problems

I'm running named 4.8 as a simple cache server on my system, and it
seems to resolve things a lot slower than other systems on the same
network.  Does anyone have any tips for speeding named up?  Do I have
to upgrade to a newer version perhaps?

Matt Day <mday@park.uvsc.edu>

-----------[000129][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 02:55:20 GMT
From:      grpjl@iastate.edu (Paul J Lustgraaf)
To:        comp.protocols.tcp-ip
Subject:   Re: Large flat network

In article <CnwB32.8qs@calcite.rhyolite.com>,
Vernon Schryver <vjs@calcite.rhyolite.com> wrote:
>In article <ganzhorn-050494084548@mojo.cisco.com> ganzhorn@cisco.com (Charles Ganzhorn) writes:
>> ...
>>A fairly typical approach would be to switch the clients, use FDDI to get
>>off the switch, put servers for those clients on the FDDI and then route
>>the FDDI's...
>
>That sounds like it involves bridging between FDDI and Ethernet in the
>switches, which is a potentially serious problem if the servers are real
>FDDI machines.  Some FDDI-Ethernet switches are unable to correctly
>translate FDDI frames to Ethernet packets.  This is commonly a problem
>with 4KByte NFS/UDP/IP frames.
>
>What you'd like is a box that functions as a bridge among Ethernets (i.e.
>as a switch) and acts as a router for packets going between the Ethernets
>and the FDDI link away from the group.  I have not heard of such a thing
>yet.  The switch vendors have religious objections to understanding
>anything except MAC packets and the router vendors don't have the right
>hardware--yet, as far as I know.

The Alantec Powerhub can do exactly that, as well as grouping any
arbitrary set of ports into any IP subset.  It's a good box.
But of course its not the cheapest switch out there.  But, hey, it's
cheaper than the major routers.  (:-)
-- 
Paul Lustgraaf             "It's easier to apologize than to get permission."
Network Specialist                                      Grace Hopper
Iowa State University Computation Center                    grpjl@iastate.edu
Ames, IA  50011                                                  515-294-0324

-----------[000130][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 94 11:24:25 -0500
From:      jauld@delphi.com
To:        comp.protocols.tcp-ip
Subject:   Reverse Telnet

I am looking for a RFC or any other documentation for reverse telnet.  If
anyone knows of such a beast please help.

-----------[000131][next][prev][last][first]----------------------------------------------------
Date:      08 Apr 1994 03:32:27 GMT
From:      nelson@crynwr.crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Dynamic load sharing via a single IP address.

In article <2nuget$qbl@lykos.netpart.com> phil@lykos.netpart.com (Phil Trubey) writes:

   I have a situation where I have 2 machines offering an identical
   TCP service (clients can connect to either and get the same
   information/service).  Is there a way for clients to open a
   connection to a single IP address and have the request go to either
   one of the machines via some combination of switching, routing,
   etc?  ie. I'd like to be able to dynamically load share the service
   between these two machines - and also allow all connections to go
   to one machine if the other is down.

   If  this isn't possible by sending a request to a single IP address, any
   suggestions on how I could load share in some other way?

The correct way to do this is with shuffle A records in bind.  It
swaps between different A records on each request for an A record.
You may need to update to the current version of bind for this to work
for you.

--
-russ <nelson@crynwr.com>      ftp.msen.com:pub/vendor/crynwr/crynwr.wav
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)    | Quakers do it in the light
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000132][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 03:56:50 GMT
From:      raj@acc.com (Richard Johnson)
To:        comp.protocols.tcp-ip
Subject:   Xsnmp (little network management program) available (v2.2)

I have only announced this program before on the comp.protocols.snmp newsgroup
but it sounds as if there may be some interest here as well, so...

"Xsnmp" is a small X Windows program which uses the MIT SNMP library to gather
information about your LAN and display the status graphically.  You can easily
add and remove systems on the display through a simple GUI.  You can do some
simple SNMP (version 1) queries on systems which support that.  It will show
you when a system isn't responding to "ping's" and when the last response was,
etc.  It uses SNMP to automatically discover additional interfaces on systems as
well as interface speed, etc.

The current version is located on ftp.acc.com under the pub/raj directory as
"xsnmp22.tar.Z".  (I started working on this program in my free time when I was
working at the local university.  ACC has been keeping me busy since then so I
haven't really had time to do many improvements to it since then.  I have code
already to do autodiscovery on a subnet and would like to make that available as
soon as possible, however the program seems to have a small memory leak and I'd
like to fix that as well.  I hope future improvements will include SNMP V2 as
well as support for reading MIB's and doing whatever queries the user wants on
them.)

Anyway, check it out if you're interested.

-- 
Richard Johnson
Senior Network Manager                               Internet:      raj@acc.com
Advanced Computer Communications                     UUCP:    ucbvax!ucivax!raj
315 Bollay Dr.                                       Phone: (805) 685-4455 x222
Santa Barbara, Ca. 93117          (Me?  Represent ACC?  You're kidding, right?)

-----------[000133][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 04:55:59 GMT
From:      matt@buc.edu.au (Matthew Pegg)
To:        comp.protocols.tcp-ip
Subject:   Info on the Internet..Help

I dont know if this is the right place to ask this question, but here goes anyway.

What I need is information about the Internet and its sub-networks.
Does anyone know where I can find such information on the net????
I'm really looking for overview type material, such as its size, history, who
maintains it, what services are available... That sort of stuff.

			Anything will do, and all help will be appreciated...

							Regards.. MATT.
		
___________________________________
-----------------------------------
Matthew Pegg

School of Mathematics and Computing
University of Ballarat, Vic.,Aust.

matt@nova.ballarat.edu.au        ;)
-----------------------------------



-----------[000134][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 05:11:24 GMT
From:      al012@un.seqeb.gov.au ( ANTHONY LEE)
To:        comp.protocols.tcp-ip
Subject:   tcpcon and nd


Dear all,

I have been trying to use the tcpcon and nd commands to set up a logger
service on a terminal server port.  It is working but under circumstances
the pty device cannot be reused.  So I would start with say /dev/ptyqe
and then after a couple of broken connections I will have to use ptyqa.
The rate I am going I will have to reboot my machin very soon.  What gives ?
Is there anyway of resetting the pty device ?

Thank you
Anthony

-- 
Anthony Lee				These are my opinions and not SEQEB.
SEQEB					
150 Charlotte Street			
Brisbane

-----------[000135][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 05:38:37 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: tcp checksum redundancy

In article <2o25p2$htp@hybrid.com> bpacker@hybrid.com () writes:
>
> i am  researching tcp checksum redundacny, i.e. tcp checksums redundant
> with mac crc for local lan connection. i have a couple of very basic
> questions...
>
> 1. who has implemented this and written it up ? i glanced thru the
> latest rfc index and didn't see anything.

There have been proposals about every 18 monts to omit the TCP checksum
on LAN's from someone new.  They are loudly decried, except when the
previous round was particulary loud, in which case the proposal is ignored.
There will never ever be an RFC that is other than "informational" that
suggests watering down the current end-to-end TCP checksum requirement.
It is likely that an informational draft proposing dropping the TCP
checksum would be followed with several deriding the idea as well as the
morals, ancestry, friends, and taste and good sense of its author.

This issue raises hackles, emotions, and lynch mobs.  

> 2. i am assuming that all right minded tcp implementations require a
> checksum, i.e. they do not treat the checksum like udp does. thus, i am
> assuming that something like a bridge which was relaying packets between
> tcp endpoints on separate lans couldn't successfully peek into the
> tcp packet and decide to poke out the checksum. it seems that end to
> end agreement would be necessary....

There is no reason for a bridge or a router to poke at the TCP or UDP
checksums.  Those checksums are generated and checked only on the end
systems.  (Well, I do know of a varient of SLIP that compressed TCP/IP
headers to 3 bytes including the framing byte, but I no longer recommend
it.)

Look at the TCP and UDP error counts on any larger server, and notice that
all of those errors would have been undetected data errors had the
checksums not been end-to-end.

Then talk to people who have suffered undetected data errors in NFS packets
from systems with UDP checksums turned off.

Then note that it is impossible for two IP systems to know whether they
are on the same wire (because of bridges, subnets, and proxy-ARP) and so
to know whether a MAC-level checksum is as end-to-end as the TCP or UDP
checksum, to know whether or not the TCP checksum is "redundant."

Then talk to people who have made TCP and UDP go 10's of MByte/sec with
not very fast hosts and with cheap hardware support for checksums.  (Ask
your friend Jan about TS's numbers.)  We're lucky that there's no reason
to cheat on the checksums.


Vernon Schryver    vjs@rhyolite.com

-----------[000136][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 12:18:54 -0400
From:      overbyte@acy1.digex.net (Overbyte)
To:        comp.protocols.tcp-ip,comp.os.ms-windows.misc
Subject:   Re: Mosaic (Winsock) over Chameleon 4.0?

In article <sjacobs.92.2DA33189@maroon.tc.umn.edu>,
Steve Jacobson <sjacobs@maroon.tc.umn.edu> wrote:
>In article <2nub50$lpe@universe.digex.net> philp@universe.digex.net (Phil Perucci) writes:
>
>>With WFW 3.11 (no microsoft TCP/IP) and ChameleonNFS 4.0 installed,
>>has anyone been able to run Mosaic?
 
>>A more general question, do the freeware/shareware Winsock apps run
>>OK over Chameleon 4 *IF* the "properties" of the Winsock app's icon
>>(working directory) point to the ChameleonNFS Winsock.DLL?
>
>Yes, Mosaic and other freeware apps run fine over Chameleon.  Cello does not, 
>but apparently Cello has quite a few other problems.  Hopefully, Tom at LII 
>will fix the problems Cello has with Chameleon.  

 Is Chameleon FTP'able? if so where at? 


-----------[000137][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 14:22:09 -0500
From:      news@wbm386 (News Administrator)
To:        comp.protocols.tcp-ip
Subject:   Networking questions

I don't know if this is the right place to start this so if not tell where to
go (which newsgroup that is ;).  I am currently part of a feasibility study
for a new product from our company.  We are trying to locate some information
on different networks and protocols, and I have been given the "dubious" honor
of trying to summarize everything into a short document.  If you have time,
I would like any information or experiences about performance, robustness,
hardware and software support, cost per system, and any limitations (distance,
repeaters, max nodes, etc) for any/all of the items in my list.  I would also
appreciate any information of good documentation or sources to same which you
can give me.  And finally, I would like any information or concerns about WAN
connectivity of these items.  NOTE: This system is not planned to have any
connection with the InterNet, so that type of compatibility is not a concern.

	 1) NFS
	 2) AppleShare
	 3) Portable Netware
	 4) Lan Manager
	 5) TCP/IP
	 6) Novell/IPX
	 7) FDDI
	 8) Ethernet
	 9) Token Ring
	10) Fiber Optics
	11) Wireless
	12) Token Bus(?)

Thank you for any help you can give me.

	brian


-----------[000138][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 13:04:37 -0400
From:      kchen@science.aaas.org (Ken Chen)
To:        comp.protocols.tcp-ip
Subject:   Slow FTP throughput

Hi folks!

I'm not sure if this is the correct place to post this, but here goes.  BTW,
I'd be very grateful if someone does direct me to a more appropriate 
newgroup.


As the subject states, our problem is with slow FTP throughput.  I average 
about 4 kbytes/s anywhere I ftp.  No matter if it's a site down the road at a 
different institution , or half way around the world, the fastest we've 
encountered is 5 kbytes/s.  Sickening isn't it?

Needless to say, we're looking for any suggestions on how to correct this 
situations.  I guess I should inform you of the setup before I go on:

56K leased line
Service through sprintlink
SPARCstation 10 (Only UNIX server in the building, takes care of telnet,ftp 
	and newsfeeds)
Cisco 3000
Connected to a backbone which consists of a few Novell servers and a VAX.
	
I've discussed the problem with people at sprint and they seem to think that
the problem is with our system.  They think the culprit is the newsfeed.  I
have some doubts about that, but I'm not exactly the UNIX whiz.  Could it be
a bottleneck at the router?  Or one of sprint's routers? 

They have recommended that we purchase another SPARCstation to seperate the 
news and the ftp.  This sound like a solution, but it sounds like a costly
solution.  I've also been informed that this problem occured even before
we started the news server.  

Any solutions?  Another 56k channel to make the total 128K?  Optimizing the 
kernel?  *Any* suggestions or references to literature would be ***GREATLY***
appreciated!

Thanks in advance!
Ken

PS - if anyone wants to try the connection, the anonymous ftp address for 
our server is science.aaas.org
 
 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
  Ken Chen                                |      Internet:     kchen@aaas.org
  Computer Specialist                     |      Phone:        (202) 326-7042
  American Association for the            |      Fax:          (202) 842-1711 
     Advancement of Science               | **Opinions stated may not 
  1333 H Street, NW, Washington, DC 20005 |   reflect that of the employer**
  -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

-----------[000139][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 07:00:55 GMT
From:      gu_jc5@uxmail.ust.hk (Steven Wong)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,vmsnet.networks.tcp-ip.misc
Subject:   INFO WANTED ON DEChub 900 and DEC' NETWATCH


Hi :

	Do anyone know any thing about the DEChub 900 and DEC' netwatch, is there a ftp site which I can ftp info on these two products ???

Thanks your help




-----------[000140][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 12:15:54
From:      jack@sfsu.edu (Jack E. Tse)
To:        comp.protocols.tcp-ip,comp.unix.questions
Subject:   Re: Traceroute for Unix SVR4.2 ?

In article <CnJE8v.MDv@oasis.icl.co.uk> ppg@oasis.icl.co.uk (Philippe Goujard) writes:
>Newsgroups: comp.protocols.tcp-ip,comp.unix.questions
>Path:
>csus.edu!csulb.edu!library.ucla.edu!agate!howland.reston.ans.net!pipex!Q.icl.co.
>uk!dsbc!iclbra!ppg
>From: ppg@oasis.icl.co.uk (Philippe Goujard)
>Subject: Traceroute for Unix SVR4.2 ?
>Message-ID: <CnJE8v.MDv@oasis.icl.co.uk>
>Organization: ICL, Bracknell, UK
>X-Newsreader: TIN [version 1.2 PL2]
>Date: Thu, 31 Mar 1994 16:10:06 GMT
>Lines: 12
>Xref: csus.edu comp.protocols.tcp-ip:19136 comp.unix.questions:45509


>Hello, I would like to install traceroute on my SVR4.2 Unix, however the
>only traceroute programs I found so far in FTP are very BSD/SunOS-ish and 
>require a tweak in the kernel. 
 
>Could any kind soul point me to a FTP which contains a traceroute that would 
>work with SVR4.2?
 
>Thanks in advance
 
>-- 
>-----------------------------------------------------------------------------
>Philippe GOUJARD                              Email : ppg@oasis.icl.co.uk


The executable file "/caltech/sol-fix/traceroute" at ftp.cco.caltech.edu works.

Jack
San Francisco State University


-----------[000141][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 94 18:14:01 -0500
From:      harvey@indyvax.iupui.edu (James Harvey)
To:        comp.protocols.tcp-ip,comp.os.ms-windows.misc
Subject:   Re: Mosaic (Winsock) over Chameleon 4.0?

In article <2o405e$ofn@acy1.digex.net>, overbyte@acy1.digex.net (Overbyte) writes:
	[chomp]
>
>  Is Chameleon FTP'able? if so where at?

Chameleon is a commercial product of NetManage, Inc.  You can FTP a demo
copy of Chameleon from ftp.netmanage.com (156.27.1.1).
--
James Harvey   harvey@iupui.edu   IUPUI OIT Technical Support/VMS/Unix/Networks
Disclaimer:  These are my own opinions.  I do not speak for Indiana University.
A bird in the hand is worth what it will bring.

-----------[000142][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 94 17:31:56
From:      billw@glare.cisco.com (William )
To:        comp.protocols.tcp-ip
Subject:   Re: Reverse Telnet

Except on terminal servers, where "telnet" lets a user on an async line
establish a network connection to a host, and "reverse telnet" lets a user
on a host establish a network connection to an async line on the terminal
server.

BillW


-----------[000143][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 94 13:14:08 GMT
From:      amolitor@blefscu.network.com (Andrew Molitor)
To:        comp.protocols.tcp-ip
Subject:   Re: Large flat network

In article <CnwB32.8qs@calcite.rhyolite.com>, vjs@calcite.rhyolite.com (Vernon Schryver) writes:
|> That sounds like it involves bridging between FDDI and Ethernet in the
|> switches, which is a potentially serious problem if the servers are real
|> FDDI machines.  Some FDDI-Ethernet switches are unable to correctly
|> translate FDDI frames to Ethernet packets.  This is commonly a problem
|> with 4KByte NFS/UDP/IP frames.

	Network Systems boxes do it fine, however. Bridged packets which
happen to be IP are fragmented per IP specs. If you *really* want a giant
bridged network mixing FDDI and ethernet, a mess of NSC gear plugged in
with bridging turned on ought to do it.

	Presumably other vendors can solve the problem as easily, I simply
don't have the hard data in hand to say with certainty that 'yes, it will
work, just plug it in and turn it on.'

	Andrew Molitor

|> Vernon Schryver    vjs@rhyolite.com

-----------[000144][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 15:36:58 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Timeout for socket read/write

> Hi.  I'm writing a TCP/IP client-server application with sockets
> and would like to introduce a timeout where I call read(2),
> ...
> Can someone please tell me the best (er, make that
> simplest, I don't want to get into adaptive timeouts) way of
> coding a timeout on the read?  And might I also need a timeout
> when calling write(2) on a socket?

There are 3 ways, the first IMHO being the simplest (but least portable).

1) Use the SO_RCVTIMEO and SO_SNDTIMEO socket options.  Unfortunately
   these aren't widespread yet (4.4BSD, AIX 3.2, and BSD/386 all support
   them; perhaps others?).

2) Use alarm() before the read to generate SIGALRM.  Better be aware of
   your system's signal handling semantics and if/how it restarts
   interrupted system calls.

3) Don't block in read() or write(), but in select() instead, and use
   select's built in timeout capability.

I'd probably recommend #3, for portability, writing functions named
something like read_timer() and write_timer() to hide the details.

	Rich Stevens

-----------[000145][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 94 15:37:20 GMT
From:      egger@slsvaat.sel.de (Jochen Egger Fa. SSW)
To:        comp.protocols.tcp-ip
Subject:   Anyone ever heared of this ?


Has anyone heared of a British provider
named bytex byteex or bypex or ... ?

If so, can you send me their address and phone
number?


Thanx,
--
-------------------------------+---------------------------------------------
Jochen Egger                   |
egger@lts.sel.alcatel.de      < >         intentionally left blank
+49 711 821-3578               |
-------------------------------+---------------------------------------------
ALCATEL-SEL Stuttgart Abt. US/ITC

-----------[000146][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 16:07:10 GMT
From:      carlson@Xylogics.COM (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: Reverse Telnet

In article <hO-M6Wh.jauld@delphi.com>, jauld@delphi.com writes:
|> I am looking for a RFC or any other documentation for reverse telnet.  If
|> anyone knows of such a beast please help.

You're not going to find an RFC for this -- it's just an implementation
detail.

A normal telnetd program listens for TCP connections to port 23 and
starts login sessions hanging off of ptys.  Reverse telnet is just a
telnetd which has been modified to wait for a pty to be opened by an
application and then make an active TCP connection to some destination.

--
James Carlson <carlson@xylogics.com>            Tel:  +1 617 272 8140
Annex Software Support / Xylogics, Inc.               +1 800 225 3317
53 Third Avenue / Burlington MA  01803-4491     Fax:  +1 617 272 2618

-----------[000147][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 17:15:34 GMT
From:      tan@ctt.bellcore.com (Yiwen Tan)
To:        comp.protocols.tcp-ip
Subject:   tcp/ip performance

Hi, netters,

	I did some TCP/IP throughput measurement between my pc and my SPARCstation.  I used FTP Software's PCTCP.  I got about 3.5 Mbps with
a large window size.  I think 3.5Mbps is not so bad.
	My problem is, for the application that I am developing, I need
to install two ethernet cards in a PC.  Just having an additional card
physically installed lower the tcp throughput to about 0.08Mbps.
	I used packet drivers for my Ethernet card (3com's 3c509).  Is this normanl? If not, any sulution ?  Thank you.

Yiwen
---------------------------
Disclaimer:  This is my opinion, not Bellcore's.

-----------[000148][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 19:52:50 GMT
From:      boris.burtin@sandiegoca.ncr.com (Boris Burtin)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Software-based network monitor?

I was wondering if there is a tool available that monitors TCP/IP packets sent 
over the net.  We have some client/server software packages that we need to 
monitor, and I'm trying to avoid having to buy extra hardware.  I'd appreciate 
any input I can get on this subject.

Thanks!

Boris

/------------------------------------------------------------------------\
| Boris Burtin			  | Client-Server Techology		 |
| boris.burtin@sandiegoca.ncr.com | AT&T Global Information Solutions	 |
 \------------------------------------------------------------------------/

-----------[000149][next][prev][last][first]----------------------------------------------------
Date:      8 Apr 1994 20:06:48 GMT
From:      bob@hobbes.dtcc.edu (Bob Rahe)
To:        comp.protocols.tcp-ip
Subject:   Re: Slow FTP throughput

In article <2o42r5$7p@science.aaas.org>,
Ken Chen <kchen@science.aaas.org> wrote:

>As the subject states, our problem is with slow FTP throughput.  I average 
>about 4 kbytes/s anywhere I ftp.  No matter if it's a site down the road at a 
>different institution , or half way around the world, the fastest we've 
>encountered is 5 kbytes/s.  Sickening isn't it?
 
>Needless to say, we're looking for any suggestions on how to correct this 
>situations.  I guess I should inform you of the setup before I go on:
 
>56K leased line

  Note that line is 56,000 >bits< per second.  That is 7K >bytes< per 
second flat out.  So with protocol overheades etc., running 4-5k bytes/s
with any TCP/IP protocol is about right.  If you need faster, you'll have
to up the line speed.
-- 
-------------------------------------------------------------------------
|Bob Rahe, Delaware Tech&Comm College | AIDS, Drugs, Abortion: -        |
|Computer Center, Dover, Delaware     |  - Don't liberals just kill you?|
|Internet: bob@hobbes.dtcc.edu        |Save whales; and kill babies?    |
-------------------------------------------------------------------------

-----------[000150][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 22:57:24 GMT
From:      ljdejan@pb2esac.com (Leo Dejan)
To:        comp.protocols.tcp-ip
Subject:   RPC specification files.


	Provided I'm in the correct group for the following question:

	I have been unable to create an RPC specification file that fits
	my particular server. I started with a working local call version
	and am attempting to convert it to a distributed one. Every example
	in every book that I could find they use a function for which there
	is only one parameter. My function happens to be called with two
	parameters. Hoping the following is enough to clarify.

	...in the client:

	CLIENT *CL;
	char lname[MAXCHAR], *Listings[MAXREC];

		if (RecCnt = searchdb_1 (lname, Listings, cl))
						   parm1---^  parm2--^     ^
										required---|

	What is the format of the procedure specification in the 
	.x file? I currently have:
				
				int SEARCHDB (string) = 1;

	Certainly, there is a simple answer?

	Thanks in advance for any help.

	Leo

-----------[000151][next][prev][last][first]----------------------------------------------------
Date:      Sat,  9 Apr 94 10:14:00 -0640
From:      john.will@dscmail.com (John Will)
To:        comp.protocols.tcp-ip
Subject:   Timeout for socket read/w


P >Hi.  I'm writing a TCP/IP client-server application with sockets
P >and would like to introduce a timeout where I call read(2),
P >since if the other end has gotten out of whack it may not be
P >sending anything but still have the connection open.  I have looked
P >in both of my W. Richard Stevens books without finding anything
P >useful.  Can someone please tell me the best (er, make that
P >simplest, I don't want to get into adaptive timeouts) way of
P >coding a timeout on the read?  And might I also need a timeout
P >when calling write(2) on a socket?

I'd look in the Stevens books as select(), seems that's the easiest
way to do it.

-----------[000152][next][prev][last][first]----------------------------------------------------
Date:      Fri, 8 Apr 1994 23:35:57 GMT
From:      subbu@cup.hp.com (Subbu Subramaniam)
To:        comp.protocols.tcp-ip
Subject:   TCP Embedded Trailer Checksum


Hi,

Monroe  Bridges and I (from HP) have put out an  internet-draft  on "TCP
Embedded Trailer Checksum" for review.

We believe that an extension to TCP to support  Trailer  checksums  will
enhance  the  performance  and  reduce  the price of  hardware  assisted
checksum  implementations,  especially  in links  such as ATM and  Fibre
Channel.

The use of trailer cheksums is negotiated via TCP options.  The extended
TCP  implementaions  will  still  inter-operate  with Host  Requirements
conformant TCP  implementations.

The Fibre Channel  Systems  Initiative  (FCSI)  Committee  has been very
supportive of this mechanism.

We would  appreciate  comments  from this group in this  regard.  One or
both  authors  subscribe  to this  group,  and are  listening  for  your
comments.

I am including the announcement along with this email.

Thanks,

-Subbu (subbu@cup.hp.com)

- --NextPart

A New Internet-Draft is available from the on-line Internet-Drafts 
directories.                                                               

       Title     : TCP Embedded Trailer Checksum                           
       Author(s) : M. Bridges, S. Subramaniam
       Filename  : draft-bridges-tcp-checksum-00.txt
       Pages     : 14
       Date      : 04/06/1994

This document describes the use of embedded trailer checksums in TCP, and 
proposes an extension to TCP to negotiate the use of embedded trailer 
checksums for a TCP connection.  Embedded trailer checksums reside in the 
last two octets of the segment's data field. The negotiation mechanism 
proposed in this document allows for new TCP implementations to exchange 
segments bearing embedded trailer checksums while still being able to 
inter-operate with TCP implementations that do not support them.      

The performance gain achieved by assisting software's checksumming of TCP 
segments with hardware has been clearly demonstrated in some existing TCP  
implementations.  We claim that the price and performance gains achieved by
providing hardware assists for Embedded Trailer Checksums are even greater.
These gains are particularly apparent when employed over segmenting links 
such as Fibre Channel and Asynchronous Transfer Mode (ATM).     

TCP implementations that do not use hardware checksum assists will not 
suffer any performance penalty by supprting Embedded Trailer Checksums.           

Internet-Drafts are available by anonymous FTP.  Login with the	
username "anonymous" and password "guest".  After logging in,
Type "cd internet-drafts".
     "get draft-bridges-tcp-checksum-00.txt".
 
Internet-Drafts directories are located at:	
	                                                
     o  US East Coast                            
        Address:  ds.internic.net (198.49.45.10)	
	                                                
     o  US West Coast                            
        Address:  ftp.isi.edu (128.9.0.32)  	
	                                                
     o  Pacific Rim                              
        Address:  munnari.oz.au (128.250.1.21)	
	                                                
     o  Europe                                   
        Address:  nic.nordu.net (192.36.148.17)	
	                                                
Internet-Drafts are also available by mail.	
	                                                
Send a message to:  mailserv@ds.internic.net. In the body type: 
     "FILE /internet-drafts/draft-bridges-tcp-checksum-00.txt".
							
For questions, please mail to Internet-Drafts@cnri.reston.va.us.
							

Below is the data which will enable a MIME compliant Mail Reader 
implementation to automatically retrieve the ASCII version
of the Internet-Draft.

- --NextPart
Content-Type: Multipart/Alternative; Boundary="OtherAccess"

- --OtherAccess
Content-Type:  Message/External-body;
        access-type="mail-server";
        server="mailserv@ds.internic.net"

Content-Type: text/plain
Content-ID: <19940406101509.I-D@CNRI.Reston.VA.US>

FILE /internet-drafts/draft-bridges-tcp-checksum-00.txt

- --OtherAccess
Content-Type:   Message/External-body;
        name="draft-bridges-tcp-checksum-00.txt";
        site="ds.internic.net";
        access-type="anon-ftp";
        directory="internet-drafts"

Content-Type: text/plain
Content-ID: <19940406101509.I-D@CNRI.Reston.VA.US>

- --OtherAccess--

- --NextPart--

-----------[000153][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 00:19:11 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP ?

In article <2nv6ue$sg2@usafa2.usafa.af.mil> lefkogt%scs@sc4199.usafa.af.mil (Gary Lefko) writes:
>Is there any place I can get info on protocols and how they work?  I am in a
>Computer Systems Organization class and have to do a term paper.  Any help
>would be appreciated.  TCP/IP info also appreciated...Thanks, again!

"Internetworking with TCP/IP, Vol.1" by Comer and Stevens.  "TCP/IP
Illustrated, Volume 1: The Protocols", by Stevens (no relation).  And the
RFCs themselves (both books have references sections that list the
interesting RFCs and tell how to get them).

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000154][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 11:41:51 -0400
From:      barr@pop.psu.edu (David Barr)
To:        comp.lang.perl,comp.protocols.tcp-ip
Subject:   ping in perl?

	I'm writing a program that needs to do lots and lots and
lots of pings.  Has anyone written a perl version of ping?

--Dave
-- 
"Oh that my words were now written!  Oh that they were printed in a book!"
- Job 19:23

-----------[000155][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 15:44:43 -0500
From:      lin@cs.purdue.edu (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   [Q] MTU of various non-Ethernet interfaces?

   Can someone tell me the maximum transmission unit (MTU) of various
non-Ethernet interfaces (e.g., Token-Ring, Frame-Relay, SMDS, etc.)?
I know the MTU of 10 Mb/s Ethernet and FORE Systems' 100 Mb/s SBA-200
ATM adapter is 1500 and 9188 octets, respectively.

Thanks,
John Lin (lin@cs.purdue.edu)

-----------[000156][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 08:19:02 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Two logical networks on a physical network, help for route needed

In article <2np7mg$2kv@gabriel.keele.ac.uk> cca04@potter.cc.keele.ac.uk writes:
>My question is that is there some way I can make use of this 
>"better" router which is not on my logical network (but physically connected
>directly)?

Most routers allow you to configure multiple addresses on the same physical
interface.  For instance, on a cisco router you would configure an address
on the other logical network as a "secondary address".
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000157][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 15:22:32 -0400
From:      dean@tbone.biol.scarolina.edu (Dean Pentcheff)
To:        comp.lang.perl,comp.protocols.tcp-ip
Subject:   Elementary DNS/naming question (getting a "*.org" name)

To the DNS gurus out there...  Two questions about getting high level
names in the DNS namespace.

1.  How does a non-profit organization (in this case, a professional
society for biologists) go about getting a high-level "*.org" name
(e.g. blahblah.org)?

2.  Can that name be aliased to a preexisting machine (for example, our
Sparcstation at tbone.biol.scarolina.edu) and moved elsewhere at a later
date?

Thanks, from the DNS illiterate...

-Dean
-- 
N. Dean Pentcheff 
Biological Sciences, Univ. of South Carolina, Columbia SC 29208 (803-777-3936)
Internet addresses: pentcheff@pascal.acm.org or dean@tbone.biol.scarolina.edu
WWW link:  <a href="http://tbone.biol.scarolina.edu/~dean/">home page</a>

-----------[000158][next][prev][last][first]----------------------------------------------------
Date:      Sat,  9 Apr 94 20:34:00 -0600
From:      mark.stapleton@cld9.com (Mark Stapleton)
To:        comp.protocols.tcp-ip
Subject:   Re: Mosaic (Winsock) ove

O> Is Chameleon FTP'able? if so where at?

NetManage, the publisher of Chameleon, has FTP available @

     ftp.netmanage.com

'Nuff said.

Mark

PQWK\Work\CLOUD9.MSG

-----------[000159][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 17:15:30 -0400
From:      bb08176@bingsuns.cc.binghamton.edu (Amritansh Raghav)
To:        comp.protocols.tcp-ip
Subject:   TCP-IP simulator


I was reading an article about the effects of mobility on
TCP-IP and was wondering if there are any simulators for TCP-IP
out there..i wanted to test some ideas re mobile communication 
but as i dont have access to a WaveLan like the big guns :-)
i shall have to lump it and do with a simulation. Now i'd
rather not write a full blown TCP-IP sim. package with all the
details, so was wondering (read as begging)
anyways please email to me
raghav

raghav@cs.binghamton.edu

-----------[000160][next][prev][last][first]----------------------------------------------------
Date:      Sat, 9 Apr 1994 11:05:31 GMT
From:      kardel@nessy.informatik.uni-erlangen.de (Frank Kardel)
To:        comp.protocols.tcp-ip
Subject:   Re: xntp between ver 2 & 3

robb@epenviron.eapi.com (Rob Beckman) writes:


>We are running SCO UNix and using xntpd version 2.  We have another 
>machine running SunOS with a public domain version of xntp that is
>version 3.
 
>Does anyone know what we have to do to get the Sun xntp to read a
>time feed from the SCO xntp?  Is other words how do  we get a version 3
>of xntp to read xntp version2 time feed.

put a "version 2" after the "server scobox" in ntp.conf.

Frank Kardel (time@informatik.uni-erlangen.de)

-----------[000161][next][prev][last][first]----------------------------------------------------
Date:      Sat, 9 Apr 1994 17:22:49 GMT
From:      itbwvb@puknet.puk.ac.za (Van Belkum, W)
To:        comp.protocols.tcp-ip
Subject:   Re: Distinct TCP-IP -- Anybody using this?

In article <128@lilypond.win.net> conklin@lilypond.win.net (Thomas R. Conklin) writes:

>I wanted to solicit some review from this group of Distinct
>Corp's tcp-ip product.  My organization is considering
>getting pretty involved with this package (winsock.dll) for
>doing some development work (they seem to have some very
>nice libraries for ftp and telnet).  We are also
>considering liscensing their tcp-ip protocol stack for
>Windows. 
 
>Does anyone have some experience with Distinct's stuff
>that they'd be willing to share?  How robust is the
>software?  Are the ftp and telnet libraries workable,
>useful?  Any notable limitations, areas of buggyness?  How
>about Distinct's support?  What are the highlights and
>lowlights of the package?
 
>Thanks in Advance 


>-- Rob Conklin (CWI)
>Internet Address:    conklin@lilypond.win.net
Hi 
Wellfleet uses this stack with their management software.
Try asking them !
Wilhelm van Belkum

-----------[000162][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 04:15:18 -0500
From:      karl@MCS.COM (Karl Denninger)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP "sniffer" software - freeware anywhere?

Hi!

I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
where I can find such a thing?  It would be EXTREMELY helpful to me right
now.

Thanks!

--
--
Karl Denninger (karl@MCS.Net)| MCSNet - Full Internet Connectivity (shell,
Modem: [+1 312 248-0900]     | PPP, SLIP, leased) in Chicago and 'burbs.
Voice/FAX: [+1 312 248-8649] | Email "info@mcs.com".  MCSNet is a CIX member.
Fan Friendly Internet Here   | WWW: http://www.mcs.net, gopher: gopher.mcs.net

-----------[000163][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 20:19:49 GMT
From:      schemers@leland.Stanford.EDU (Roland Schemers)
To:        comp.lang.perl,comp.protocols.tcp-ip
Subject:   Re: ping in perl?

In article <2o6ibv$7o8@bosnia.pop.psu.edu>,
David Barr <barr@pop.psu.edu> wrote:
>	I'm writing a program that needs to do lots and lots and
>lots of pings.  Has anyone written a perl version of ping?
>

I haven't written one in perl but I have written one in C that
is meant to be called from perl. You can get it from:

ftp://jessica.stanford.edu/pub/fping.shar

Given a list of hosts fping will ping them round-bobin until all the
hosts have replied or they timeout. You can easily ping a hundred hosts
in no time flat :-) There are plenty of options so you don't flood the
net too. The nice thing is you don't sit and wait for each host to
respond, the pings are received asynchronously.

Roland


-- 
Roland J. Schemers III            |    Networking Systems 
Systems Programmer                |    414 Sweet Hall  +1 (415) 723-6740 
Distributed Computing Operations  |    Stanford, CA 94305-3090
Stanford University               |    schemers@Slapshot.Stanford.EDU 

-----------[000164][next][prev][last][first]----------------------------------------------------
Date:      9 Apr 1994 21:05:05 GMT
From:      cricket@nsr.hp.com (Cricket Liu)
To:        comp.protocols.tcp-ip
Subject:   Re: Elementary DNS/naming question (getting a "*.org" name)

Dean Pentcheff (dean@tbone.biol.scarolina.edu) wrote:
: To the DNS gurus out there...  Two questions about getting high level
: names in the DNS namespace.
 
: 1.  How does a non-profit organization (in this case, a professional
: society for biologists) go about getting a high-level "*.org" name
: (e.g. blahblah.org)?

You start by FTP'ing a copy of /netinfo/domain-template.txt from
rs.internic.net.  Fill it out, then mail it to hostmaster@internic.net.

: 2.  Can that name be aliased to a preexisting machine (for example, our
: Sparcstation at tbone.biol.scarolina.edu) and moved elsewhere at a later
: date?

There's no *technical* reason the name can't be an alias, although if it
were an alias, it would preclude delegation of the domain.  I don't know
what the NIC's administrative policy is, though.  Even if the NIC's
administrative policy disallowed aliases, you could effectively create one
using address and mail exchanger records that point to
tbone.biol.scarolina.edu.

--
cricket

cricket@hp.com / Hewlett-Packard Professional Services / Englewood, CO

-----------[000165][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 05:19:22 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   Re: docs for UNIX socket programming?


I agree that the Stevens book is an excellent guide; one which you 
may already have, however, if you're a Unix shop, is in the docs
that describe network programming for your particular Unix/platform.

For example, the _Network Programming Guide_ that came with SunOS 4.1
systems is a great document -- covering not only sockets, but TLI, NIS
and NFS.  Oh yeah, RPC as well.  I'm not sure *what* comes with Solaris
2.x, however.

Richard 

rmadhok@anl.liv.usw.com (Raghu Madhok) writes:

>jaesoo@cs18.cs.aukuni.ac.nz (Jaesoo Kim) writes:
> >TCPMAINT@TWNMOE10.Edu.TW wrote:
> >> 
> >> [...socket programming guide available for free?...] 
> >
> > [...I'd like that, too...] 
> >
>
>I would strongly recommend the book "UNIX Network Programming" by Richard
>Stevens.  It provides some excellent examples and covers the top thoroughly
>from an application programming point of view.
>
>...Raghu
>
-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondance 			| installed base. 

-----------[000166][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 05:58:20 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   Re: 2 C's can't talk

Robert,

Since you've left out some information, I'll make some assumptions:

(1) SGI and PCs can talk to each other; 

(2) You've verified that the netmasks are correct if the PCs and SGI 
are on the same network as the Sun; 

(3) That if the PCs and SGI are on a different network from the
Sun, that you've established a route to that network from the
interface on the local wire (e.g., le0 on the Sun) -- i.e., that
you're not inadvertently sending to the local wire via the router. 

By "network" in the above, I mean either the same Class ? network ID
if you're not doing subnetting, or a subnet if you *are* subnetting
the Class ? network ID. 

THEN: 

I'd verify those assumptions, first, if I were you -- and then I'd
look at the concentrator.  I'm pretty fuzzy on this part, since I've
never had this problem (and, I'm not at work at the moment) -- but if 
you bring the Sun down using (SunOS 4.1 syntax) 'shutdown -h now' so 
that you get the 'ok ', you can look at 'help ethernet ' to run a test 
necessary to determine how the framing gets done on the Sun in order to 
operate correctly with certain concentrators.  I've never had the problem, 
personally, so I can't be any clearer than this.  'help ethernet ' will
explain it better.

If that doesn't cut it, I'd get a copy of tcpdump from the 'net (if you
can't find it, let me know and I'll send you an archie listing on it)
and examine what the blazes is happening on the local wire when the Sun
tries to ping the SGI and visa versa.  If you've got a Sniffer, then so
much the better.  If you're running Solaris 2.something on the Sun, then
you should have 'snoop' (or something like that -- sorry, I haven't yet
acquired the taste for Solaris 2.x), and you can examine packets that way. 

And, before I did any of this, I might try to put the Sun behind the
concentrator ( once I'd verified (1-3) above ), to see if that solved
the problem.  If it's the framing on the Sun, then what should happen
is that the Sun should either lose connectivity to everybody else, or
it should lose connectivity to the router/Internet (depending on where
within the concentrator the Sun and it are having a conflict.  If it's
at the concentrator port, then the Sun could not talk to anyone, for
example). 

Finally, I'd also look at the physical interface between the coax and
the concentrator.  I doubt that's the problem, but then again there
are some details missing from the sketch you've provided; you never
know, in that case. 

Send me some mail if you'd like. 

Good luck,

Richard 
rh310 

P.S., Please check for a possible extra dot in your localhost ID, below: 

robert@Sharenet.Edu (0000-Admin(0000)) writes:
>
>                                  internet
>                                     |
>[SUN]________coax_________________[ROUTER]_______[NOVELL-FILE-SERVER]
>                         |
>                    [10 baseT concentrator]
>                    |   |   |      |
>                   pc  pc  pc     SGI
>
>I can talk from SUN to internet.  I can talk to SUN from internet.
>I can talk from pc's or sgi to internet.  I can talk back to them
>from internet.  I CAN NOT TALK from pc or sgi to SUN, or from sun
>to pc or sgi.
>
>I have entries in routing table for both sun and sgi to 127.0.0.0.
                                                                  ^
                                                                  ^
                                          Typo?                   ^

>to local wire, and default via gateway.
>
>Any ideas?  thank you very much.
>
>-robert rowland  robert@cyclops.sharenet.edu


-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondance 			| installed base. 

-----------[000167][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 06:28:41 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   Re: Ping WORKS - Traceroute DOES NOT

Another possibility, since you haven't mentioned whether the destination
machine is on the other side of at least one router (and I'm assuming it
must be, otherwise why run traceroute? ;), is that IP packet filtering
is in place on at least one of the routers.  Since traceroute requires
a working UDP module at the far end, if the router is blocking UDP, or
only permitting specific UDP port numbers through (which is a common and
generally very smart thing to do these days), then you're out of luck.

An excellent reference to this is _TCP/IP Illustrated Volume 1_ by
Richard Stevens.  Much better, IMHO, than the Comer books.

Happy reading,

Richard
rh310

philp@universe.digex.net (Phil Perucci) writes:
>W. Joe Hewitt (Raleigh) <wjh@godel.denver.ssds.com> wrote:
>>
>>	I have found a case where a machine will answer Pings
>>and a traceroute fails.  Is this because the destination
>>machine is not fully implementing the ICMP Ping?
>>
>
>This happens...
>
>We have some old AT&T 3B2 systems which do the same thing.  Ping
>responds, but traceroute does not.  This is considered "normal" 
>operation here.  Don't know enough about TCP/IP to explain it 
>though...
-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondance 			| installed base. 

-----------[000168][next][prev][last][first]----------------------------------------------------
Date:      Sun, 10 Apr 1994 01:13:20 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Embedded Trailer Checksum

In article <Cnys7x.2DH@cup.hp.com> subbu@cup.hp.com (Subbu Subramaniam) writes:
> ...
>We believe that an extension to TCP to support  Trailer  checksums  will
>enhance  the  performance  and  reduce  the price of  hardware  assisted
>checksum  implementations,  especially  in links  such as ATM and  Fibre
>Channel.
> ...

I've heard of an HIPPI implementation that has some very cheap hardware
support for UDP and TCP checksums, gets respectible performance, and puts
the checksums in their normal location.  As most people are no doubt tired
of hearing, the FDDI implementions from the same vendor (but somewhat
different people) use inexpensive, puny 16MHz CPU's on the cards to compute
and check the TCP and UDP checksums as well as help with SMT and run the
DMA engines.

At low speeds (Ethernet and slower) checksums too cheap to do in the host
to worry about any hardware help.  At medium speed (e.g. FDDI), they're
easily done either in the host or with a little cheap hardware.  At
high speeds such as HIPPI, the extra adder to do checksums in hardware is
essentially free when you have the rest of the stuff necessary to run at
a respectible speed.

I think the idea of checksums in trailers died with XTP.  It just isn't
worth the trouble.  Are the people now talking about TCP-trailer checksums
familiar with BSD's TCP-PUP-trailers?  Have they had the pleasure of
implementing them?  I know litterally no one will say anything good about
them today, including people I presume were present at their birth.


Vernon Schryver    vjs@rhyolite.com

-----------[000169][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 06:55:51 GMT
From:      i897515@redgum.ucnv.edu.au (DERF)
To:        comp.protocols.tcp-ip
Subject:   IP Routing

Hello.

Can anyone recommend some readings on different IP Routing Protocols?
A list of different ones, past and present, would also be appreciated.

Thanks in advance.

Craig

i897515@redgum.ucnv.edu.au

-----------[000170][next][prev][last][first]----------------------------------------------------
Date:      Sun, 10 Apr 1994 13:21:04 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] MTU of various non-Ethernet interfaces?

>   Can someone tell me the maximum transmission unit (MTU) of various
> non-Ethernet interfaces (e.g., Token-Ring, Frame-Relay, SMDS, etc.)?

From RFC 1191:

   Plateau    MTU    Comments                      Reference
   ------     ---    --------                      ---------
              65535  Official maximum MTU          RFC 791
              65535  Hyperchannel                  RFC 1044
   65535
   32000             Just in case
              17914  16Mb IBM Token Ring           ref. [6]
   17914
              8166   IEEE 802.4                    RFC 1042
   8166
              4464   IEEE 802.5 (4Mb max)          RFC 1042
              4352   FDDI (Revised)                RFC 1188
   4352 (1%)
              2048   Wideband Network              RFC 907
              2002   IEEE 802.5 (4Mb recommended)  RFC 1042
   2002 (2%)
              1536   Exp. Ethernet Nets            RFC 895
              1500   Ethernet Networks             RFC 894
              1500   Point-to-Point (default)      RFC 1134
              1492   IEEE 802.3                    RFC 1042
   1492 (3%)
              1006   SLIP                          RFC 1055
              1006   ARPANET                       BBN 1822
   1006
              576    X.25 Networks                 RFC 877
              544    DEC IP Portal                 ref. [10]
              512    NETBIOS                       RFC 1088
              508    IEEE 802/Source-Rt Bridge     RFC 1042
              508    ARCNET                        RFC 1051
   508 (13%)
              296    Point-to-Point (low delay)    RFC 1144
   296
   68                Official minimum MTU          RFC 791

-----------[000171][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 21:09:21 -0400
From:      dcongdon@news.delphi.com (DCONGDON@DELPHI.COM)
To:        comp.protocols.tcp-ip
Subject:   HELP! with addressing

I am in desperate need of a clear explanation of a TCP/IP addressing
problem I am having.  I have a network which I am working on which 
consists of a large number of machines connected via token ring.  Some
of the machines are directly connected while others are remotely attached
via various means (T1, regular phone).  Essentially though, everyone is
on one single ring.  I have finished the installation of five machines
on the ring which communicate using the IP protocol.  I have assigned
class C addresses to these machines (for instance: 192.9.200.1) and
set their network mask to 255.255.255.0  These five machines consist
of one IBM RS/6000 and four PC's running Chameleon's TCP/IP.  The
PC's can login to the RS6000 via telnet and they also run X-Windows
software.  The RS6000 also communicates to an AS/400 via a 5250
version of telnet.  Everything works well up to this point.  The problem
arose when someone else attached a number of other machines to the
net (IBM ES/9000's) also running TCP/IP.  They addressed them using
the scheme 200.1.1.1 etc.  I am not able to reach these machines
from my five using PING or any other means.  I assume the problem
is caused by the machines not using a similar addressing scheme.
My familiarity with IP addressing is moderate but I do realize that
addresses which start with 200 are not class C (class B????).  What
issues must I address in order to get my machines to talk to the 
ES/9000's.  I cannot change the addresses of either group of machines
due to company politics among the groups which control the different
machines on the net.  Any kind soul who can explain in greater detail
the ramifications of addresses, subnets, etc. will be MUCH appreciated.
Please email me direct with any inputs you have.  Thanks in advance!

   ___________________________________________________________________
   Don Congdon                    | Those who scorn computer history
   Consultant                     | are those who really don't grasp
   Congdon ComputerTECH Services  | what is happening today and will
   dcongdon@delphi.com            | never really shape tomorrow.          
           
   -------------------------------------------------------------------




-----------[000172][next][prev][last][first]----------------------------------------------------
Date:      Sun, 10 Apr 1994 16:06:18 GMT
From:      stanb@netcom.com (Stan Brown)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP performance analasys tool ?


	Can someone point me to a source for a good TCP/IP performance
tool?  What I want to do is look at the total network load.

	Needs to run under UNIX (HP-UX) public domain preferred,
but I would consider a commercial product.

Thanks
Stan

-- 
Stan Brown     stanb@netcom                                          404-996-6955
Factory Automation Systems
Atlanta Ga.

-----------[000173][next][prev][last][first]----------------------------------------------------
Date:      Sun, 10 Apr 1994 21:45:29 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] MTU of various non-Ethernet interfaces?

In article <1994Apr10.132104.14245@noao.edu> rstevens@noao.edu (W. Richard Stevens) writes:
>>   Can someone tell me the maximum transmission unit (MTU) of various
>> non-Ethernet interfaces (e.g., Token-Ring, Frame-Relay, SMDS, etc.)?


>              17914  16Mb IBM Token Ring           ref. [6]

Don't use a default of more than about 4K for Token Ring in a product that
you'd like to ship to people who might complain about having to manually
reduce the MTU.  Bridges or some other arcane Token Ring stuff often do
not work with the obvious (i.e. biggest) choice for the IP MTU.

Or is it 4K that where they stop working?  I'f forgotten.

No, MTU path discovery would not help.


Vernon Schryver    vjs@rhyolite.com

-----------[000174][next][prev][last][first]----------------------------------------------------
Date:      Sun, 10 Apr 94 22:18:23 GMT
From:      ftlofaro@unlv.edu (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   Protection against ICMP bombers exist?

	I have a couple of questions about ICMP unreachable bombing:
(sending bogus ICMP unreachables to make people's TCP connection
disconnect, etc)

	Assume host A is connected to host B over the Internet, and a
hacker on host C wants to sent unreachables to A saying B is
unreachable and thus breaking the connection.  Assume all 3 are on
different nets (all connected to the Internet from different points)

1.	If host C forges its IP address and tries to send an ICMP to A
over the Internet, will it be stopped? Assume the worst case; i.e. the
person doing this is the site admin for the net host C is on and can
get it out the net's router that connects to the Internet. Wil the
backbone stop it?
2.	If host C does not forge its IP address, but uses its own IP
address, and tells A that B is unreachable (as if C was a router
between A and B, but it really is not) can A tell that C really is not
on the path between A and B?
3.	Are there any ways to protect against these?


-----------[000175][next][prev][last][first]----------------------------------------------------
Date:      10 Apr 1994 23:55:29 GMT
From:      pjs@euclid.JPL.NASA.GOV (Peter J. Scott)
To:        comp.protocols.tcp-ip
Subject:   Re: Timeout for socket read/write

In article <1994Apr8.153658.9903@noao.edu>, rstevens@noao.edu (W. Richard Stevens) writes:
> > Can someone please tell me the best (er, make that
> > simplest, I don't want to get into adaptive timeouts) way of
> > coding a timeout on the read?
> There are 3 ways, the first IMHO being the simplest (but least portable).
> 
> 3) Don't block in read() or write(), but in select() instead, and use
>    select's built in timeout capability.
> 
> I'd probably recommend #3, for portability, writing functions named
> something like read_timer() and write_timer() to hide the details.

So does this modification of the readn() in your network programming
book look reasonable?  It compiles and appears to run okay. 

static int readn (fd, ptr, nbytes)
register int fd;
register char *ptr;
register int nbytes;
{
  int nleft, nread;
  fd_set rset;
  static struct timeval tv = { TIMEOUT_SECONDS, 0 };

  FD_ZERO (&rset);
  FD_SET (fd, &rset);
  nleft = nbytes;
  while (nleft > 0)
  {
    nread = select (fd + 1, &rset, (fd_set *) NULL, (fd_set *) NULL, &tv);
    if (nread < 0)
      return nread;                                   /* error in select */
    else if (nread == 0)
    {
      errno = ETIME;                                  /* timeout */
      return -1;
    }
    nread = read (fd, ptr, nleft);                    /* Should not block now */
    if (nread < 0)
      return nread;                                   /* error */
    else if (nread == 0)
      break;                                          /* EOF */

    nleft -= nread;
    ptr += nread;
  }
 return nbytes - nleft;
}
 
-- 
This is news.  This is your       |    Peter Scott, NASA/JPL/Caltech
brain on news.  Any questions?    |    (pjs@euclid.jpl.nasa.gov)

-----------[000176][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 08:39:57 -0500
From:      temp@sneezy.cc.utexas.edu (Lee Thomas)
To:        comp.protocols.tcp-ip
Subject:   FTP PCTCP ---> HP1200C Jet Direct?

Hi,

I just bought an HP 1200C color printer with an ethernet card (Jet Direct).
I need to print to it from an IBM RS/6000 using TCP/IP and PC's with FTP's 
PCTCP software.  Does  anyone have any ideas or specific instructions on how
to get either system to print to the HP 1200C using TCP/IP?  Please E-mail
me if you have a solution or ideas.

Thanks!

-Lee

---------------------------------------------------------------------------
Lee Thomas                              Voice: (512)322-6386
Systems Analyst                           FAX: (512)322-6350
City of Austin Electric Utility      Internet: temp@ccwf.cc.utexas.edu
---------------------------------------------------------------------------

-----------[000177][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 11:08:13 -0600
From:      bjg@shaman.nexagen.com (Bryan Gonderinger)
To:        comp.protocols.tcp-ip
Subject:   lpd on VAX

Has anyone out there set up lpd on a VAX?  I've got lpd running, and
it accepts jobs, but always errors out on the VAX with "User <name>
has no permission to print on this host".  I've tried creating a proxy
entry for the user, but this does not work.  Any assistance would be
appreciated.

thanks,
Bryan Gonderinger

-----------[000178][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 10:32:20 -0500
From:      lin@cs.purdue.edu (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Embedded Trailer Checksum

In article <Co0rE8.7BM@calcite.rhyolite.com>, vjs@calcite.rhyolite.com (Vernon Schryver) writes:
> In article <Cnys7x.2DH@cup.hp.com> subbu@cup.hp.com (Subbu Subramaniam) writes:
> > ...
> >We believe that an extension to TCP to support  Trailer  checksums  will
> >enhance  the  performance  and  reduce  the price of  hardware  assisted
> >checksum  implementations,  especially  in links  such as ATM and  Fibre
> >Channel.
> > ...
> 
> [....] 
>
> I think the idea of checksums in trailers died with XTP.  It just isn't
> worth the trouble.  Are the people now talking about TCP-trailer checksums
> familiar with BSD's TCP-PUP-trailers?  Have they had the pleasure of
> implementing them?  I know litterally no one will say anything good about
> them today, including people I presume were present at their birth.

  If we put the checksum in the trailer, a harware can compute the checksum
as it is transmitting/receiving the segment, right?  I am just wondering
why this idea died with XTP?  And, why using protocol trailer is bad?
(I know ATM AAL5 uses trailer.) 

Thanks,
John Lin (lin@cs.purdue.edu)







-----------[000179][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 02:56:27 GMT
From:      crohrer@advtech.uswest.com (Chris Rohrer)
To:        comp.protocols.tcp-ip
Subject:   Hostroutes

Hi all,  I'm looking to understand the concept of hostroutes.  Is this a
simple, straightforward concept that can be explained in a few sentences or
do things get complicated fast?  Is this just an address translation
technique?  Anyone know where this concept is written about?  Are there
routers that do this?

Thanks for any light you can shed here on this preliminary query.

Chris Rohrer
crohrer@advtech.uswest.com

-----------[000180][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 02:59:52 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP performance analasys tool ?

stanb@netcom.com (Stan Brown) writes:
>	Can someone point me to a source for a good TCP/IP performance
>tool?  What I want to do is look at the total network load.
 
>	Needs to run under UNIX (HP-UX) public domain preferred,
>but I would consider a commercial product.

Look at anonymous ftp on netcom ~ftp/pub/gnn/bwmeas-0.3.tar.Z
It is a suite of programs for measureing UDP/TCP performance.

Later,
George

-- 
gnn@netcom.com

If people were more concerned with being reconciled than with being right, 
the world would be a better place.  -- Miss Manners

-----------[000181][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 94 03:44:35 GMT
From:      ftlofaro@unlv.edu (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   How to get ANCIENT rfcs?

Where can I find ancient RFCs?
I know they are way obsolete, but I am curious about the very early 
Internet (arpanet).

I could find rfc3, and a couple other single digit ones in the normal 
places, but I can not find rfc1 anywhere. Where could I find that and 
the rest of the first few rfcs?


-----------[000182][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 05:23:51 GMT
From:      smh@hakea.adept.oz.au (Stephen Hodgman)
To:        comp.protocols.tcp-ip
Subject:   TCP under Win 3.11 - Help Please

I would like to connect a Win 3.11 network to TCP.

Where does the TCP software come from?
Is is possible to configure printers in windows that are actually
printers on a Unix host. i.e. can I print using lpd etc ?

Any help appreciated
Thanks
----
Stephen Hodgman
Adept Software
CANBERRA, AUSTRALIA
smh@hakea.adept.oz.au
-- 


Stephen Hodgman                                      Tel: (06) 285 3460
                                                     Fax: (06) 285 3459

-----------[000183][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 16:00:23 -0500
From:      clancy@cs.utexas.edu (Daniel J. Clancy)
To:        comp.protocols.tcp-ip,comp.os.os2.networking.tcp-ip
Subject:   Good Resource to learn about TCP-IP, SLIP, IMAP and other protocols


I am just getting into running SLIP from home under windows.
I installed it using a package that was put together by the
university that I am at and it was very simple.

I am looking for some good references to gain a better
understanding of how the different protocols like SLIP TCP/IP,
IMAP, SMTP, etc relate to each other and what exactly they
are.

I would like to have  a better understanding of how my
system is working and how I can control it and change it.

Can anyone recommend any of the following:


   1.	A good book that I can buy that talks about all of this.

   2.	Internet resources that I can access that would have
	this information.


Thanks alot

Dan Clancy

-----------[000184][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 09:11:58 GMT
From:      hunen@brc.medtronic.com (Roger Hunen)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

In article <1994Apr10.221823.29510@unlv.edu> ftlofaro@unlv.edu (Frank Lofaro) writes:
#	I have a couple of questions about ICMP unreachable bombing:
#(sending bogus ICMP unreachables to make people's TCP connection
#disconnect, etc)
#
#	Assume host A is connected to host B over the Internet, and a
#hacker on host C wants to sent unreachables to A saying B is
#unreachable and thus breaking the connection.  Assume all 3 are on
#different nets (all connected to the Internet from different points)
#
#1.	If host C forges its IP address and tries to send an ICMP to A
#over the Internet, will it be stopped? Assume the worst case; i.e. the
#person doing this is the site admin for the net host C is on and can
#get it out the net's router that connects to the Internet. Wil the
#backbone stop it?

No & no.

#2.	If host C does not forge its IP address, but uses its own IP
#address, and tells A that B is unreachable (as if C was a router
#between A and B, but it really is not) can A tell that C really is not
#on the path between A and B?

It might (do a traceroute, which may take LONG), but it won't.

#3.	Are there any ways to protect against these?

Yes. Block all ICMP unreachables on your Internet router and rely on 
timeouts. Not really the greatest way, I must admit. TCP/IP was not designed
with any security in mind so we'll have to live with that. I'm afraid...

Regards,
-Roger

-----------[000185][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 09:13:09 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Hostroutes

In article <2oae8r$2bt@cherokee.advtech.uswest.com>
crohrer@advtech.uswest.com (Chris Rohrer) writes: 
    Hi all,  I'm looking to understand the concept of hostroutes.  Is this a
    simple, straightforward concept that can be explained in a few sentences or
    do things get complicated fast?  

It's pretty simple: it's just an entry in the routing table that only
applies to one host, instead of an entire network or subnet.

    Is this just an address translation
    technique?  Anyone know where this concept is written about?  Are there
    routers that do this?
    
Yes, lots of routers do this now (almost all?).  You can look at the OSPF
and RIP specifications for more information.

Tony

-----------[000186][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 09:14:19 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: IP Routing

In article <2o87tn$66k@sheoak.ucnv.edu.au> i897515@redgum.ucnv.edu.au
(DERF) writes: 

    Can anyone recommend some readings on different IP Routing Protocols?
    A list of different ones, past and present, would also be appreciated.
    
Suggest you get Interconnections by Radia Perlman.

Tony

-----------[000187][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 11:47:52 GMT
From:      Operators@datatek.csir.co.za (MVS Operators)
To:        comp.protocols.tcp-ip
Subject:   IP routing.

Hi all

At the risk of looking stupid I am asking for help. Can anyone help with
understanding IP routing. How it works, and how one can controll it. If 
there is any documentation out there I would really like to hear about it.
I would also like to get hold of any documentation on Firewall's. 

Thanks

Justin.

E-mail: justin@csir.co.za

-----------[000188][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 12:13:03 GMT
From:      klos@mv.mv.com (Patrick Klos)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In article <boris.burtin.74.000EE1E9@sandiegoca.ncr.com>,
Boris Burtin <boris.burtin@sandiegoca.ncr.com> wrote:
>I was wondering if there is a tool available that monitors TCP/IP packets sent 
>over the net.  We have some client/server software packages that we need to 
>monitor, and I'm trying to avoid having to buy extra hardware.  I'd appreciate 
>any input I can get on this subject.
>
>Thanks!
>
>Boris
>
>/------------------------------------------------------------------------\
>| Boris Burtin			  | Client-Server Techology		 |
>| boris.burtin@sandiegoca.ncr.com | AT&T Global Information Solutions	 |
>\------------------------------------------------------------------------/


We (Klos Technologies, Inc.) have a product called PacketView that goes for
$299 and can capture and decode network traffic on ethernet, token-ring and
ARCNET networks.  PacketView supports IPX/SPX/NCP, TCP/UDP/IP, VINES IP and a
few other protocols.  It has capture and display filtering, good symbolic
support, and allows you to write your own custom protocol decoders (should the
need arise).  A demo version of PacketView is available on our BBS by calling
(603) 429-0032 or via anonymous FTP at mv.mv.com:pub/users/klos/pvdemo.zip.

Feel free to call or email if you have any additional questions.

Sincerely,

Patrick Klos
-- 
============================================================================
    Patrick Klos                           Internet: klos@mv.mv.com
    Klos Technologies, Inc.                Voice: (603) 424-8300
    604 Daniel Webster Highway             FAX:   (603) 424-9300
    Merrimack, New Hampshire  03054        BBS:   (603) 429-0032
==========See us in Las Vegas at NETWORLD+INTEROP 94 booth 5290!============

-----------[000189][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 12:58:29 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: IP routing.

> At the risk of looking stupid I am asking for help. Can anyone help with
> understanding IP routing. How it works, and how one can controll it. If 
> there is any documentation out there I would really like to hear about it.

Any good introduction to TCP/IP should cover the basics of IP routing.
For details on the actual protocols and routing algorithms, check out

	%T Interconnections: Bridges and Routers
	%A R. Perlman
	%I Addison-Wesley
	%C Reading, Mass.
	%D 1992

> I would also like to get hold of any documentation on Firewall's. 

The following will likely become the standard reference on this, but
won't be out in the U.S. until the first week of May (I think).

	%T Firewalls and Internet Security: Repelling the Wily Hacker
	%A W. R. Cheswick
	%A S. M. Bellovin
	%I Addison-Wesley
	%C Reading, Mass.
	%D 1994

Rich Stevens

-----------[000190][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 14:14:17 GMT
From:      wilfred@zeus.wnc.nedlloyd.nl (Wilfred Mollenvanger)
To:        comp.protocols.tcp-ip
Subject:   Multiple Subnets on One Ethernet Segment?

Hi,

We are running out of ip addresses on one of our subnets. How should I
solve this? I'll have to use another subnet. Should I attach hosts with
addresses in the new subnet range on the same segment (possible?)
reconfiguring all existing hosts or should I use a router or
..... ?  What's the best solution performance- and administrative
wise?

E-mail prefered.

Thanx in advance,

Wilfred.

-----------[000191][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 14:55:25 GMT
From:      ww02ksrm@sbusol.rz.uni-sb.de (Ralf Muehlen)
To:        comp.mail.misc,comp.protocols.tcp-ip
Subject:   RFC-822: Format of first line in mail folder not covered by RFC-822?

Hi Netters,
on Sinix (the Unix clone from Siemens/SNI) version 2.1 (at least in this 
version, in newer ones maybe, too) the first line of each mail in the
mailbox folder is not in standard format (the line starting with From
without colon). They translated e.g. the names of months and days into 
German. (Bright idea, isn't it? ;-) So if one moves the mailbox to another 
system, elm and other MUAs can't read it.

Well, no big problem, I wrote a small perl script which converts these
lines back to standard. I just wonder if the format of these lines
is covered by a RFC. In RFC 822 (Standard for the format of ARPA Internet 
text messages) I haven't found anything appropriate. It seems as if only
the following lines were defined.

Is there another RFC or standard which covers this?

Ralf

--
Ralf Muehlen         | A computer terminal is not some clunky old television 
muehlen@rz.uni-sb.de | with a typewriter in front of it. It is an interface 
                     | where the mind and body can connect with the universe
                     | and move bits of it about.  
                     |                   -- Douglas Adams in Mostly Harmless

-----------[000192][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 94 16:05:23 GMT
From:      john@qii.sialis.com (John Dassow)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Re: Word 6.0a .doc saving problem under W4W 3.11

>
>Hi
>
>As has been mentioned here before there is a problem with PC-NFS v5.0x
>and the latest MSoft products saving to MustShared drives. The attached
>file is a fix from SunSelect in the states that is equivilent to 5.0b,
>but is called 'WordFix'. I have no more information, but it DOES solve
>this bug.
>
>Robert
>

[pcnfs.sys deleted]

  Thank you for providing this program, unfortunately we use Interdrive from
FTP Software as a client.  Is there a similar program for this client?

					-John Dassow

-- 

-----------[000193][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 94 16:37:10 GMT
From:      tbyrnes@ritz.mordor.com (Tim Byrnes)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

Boris Burtin (boris.burtin@sandiegoca.ncr.com) wrote:
: I was wondering if there is a tool available that monitors TCP/IP packets sent 
: over the net.  We have some client/server software packages that we need to 
: monitor, and I'm trying to avoid having to buy extra hardware.  I'd appreciate 
: any input I can get on this subject.
 
: Thanks!
 
: Boris
 
: /------------------------------------------------------------------------\
: | Boris Burtin			  | Client-Server Techology		 |
: | boris.burtin@sandiegoca.ncr.com | AT&T Global Information Solutions	 |
: \------------------------------------------------------------------------/

There is a freeware package out called ethload (I'm looking at ver 1.04)
that might be what your looking for.  From the documentation:

------ From Ethload 1.04 User's guide ------

...

In a TCP/IP network, ETHLOAD allows you to:
- see ARP table contents;
- see which host is sending (un)resolved ARP probes;
- see the IP host which is sending most of the IP, UDP or TCP packets;
- see what kind of protocols are in used (either TCP or UDP)
- see which is the mostly used telnet/rlogin server (or client);
- see the boot sequence with important BOOTP and TFTP events;
- see some characteristics of IP hosts (fragments size,MUT,IP 
retransmission, options used -- including source routing, ...)
- see main RFC 1001/1002 NetBIOS events and names;
- see the working of DNS;
- see important TCP events: start/stop of connections, ...

----End of copy from ETHLOAD 1.04 user guide -----


The ethload 1.04 software monitors more protocols than just tcp/ip.  I 
forget where I picked it up, but you could probably  find it via archie.  
If not, let me know.

Tim Byrnes
tbyrnes@mordor.com

-----------[000194][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 17:41:43 GMT
From:      pam@csn.org (Paul A. Morris)
To:        comp.protocols.tcp-ip
Subject:   subnetting

In an earlier posting on subnetting i ahd read that a subnet mask 
of 255.255.255.192 only gives 2 subnets of 62 hosts
a mask of 255.255.255.224 will give 6 subnets of 30 hosts
a mask of 255.255.255.240 will give 14 subnets of of 14 hosts
a mask of 255.255.255.248 will give 30 subnets of 6 hosts
etc

i understand the bit representation of 192, 224,240, and 248
however i do not understand where the number of subnets or the 
number of hosts is derived.  if someone could explain this in 
a more straightforward manner than most of the books i have 
looked at i would greatly appreciate it
thanks


-----------[000195][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 17:54:24 GMT
From:      asweeney@odin.sw.stratus.com (Andy Sweeney)
To:        comp.protocols.tcp-ip
Subject:   TCP Benchmarking


I am looking for any information on TCP/UDP benchmarking tools and/or
perfomance expectations.

I also trying to get a feel from what most people
use TCP/UDP for

i.e. single byte transactions such as TELNET 
     file transfer FTP (4k packets)
     transaction processing (256 byte packets)

Are there any existing performance numbers that I can 
compare my tests to?

Thanks
Andy

-----------[000196][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 18:00:50 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: Ping WORKS - Traceroute DOES NOT

In article <2o8kcp$k2f@bedrock.cs.umd.edu> reh@cs.umd.edu (Richard Huddleston) writes:
>Another possibility, since you haven't mentioned whether the destination
>machine is on the other side of at least one router (and I'm assuming it
>must be, otherwise why run traceroute? ;), is that IP packet filtering
>is in place on at least one of the routers.  Since traceroute requires
>a working UDP module at the far end, if the router is blocking UDP, or
>only permitting specific UDP port numbers through (which is a common and
>generally very smart thing to do these days), then you're out of luck.

Another possibility, is if the target machine does not generate ICMP Port Unreachable
responses.  If only a particular machine is behaving this way, and it has different
software, I'd suspect something in it's implementation.

Art


-----------[000197][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 18:07:29 GMT
From:      nelson@crynwr.crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

In article <1994Apr10.221823.29510@unlv.edu> ftlofaro@unlv.edu (Frank Lofaro) writes:

	   I have a couple of questions about ICMP unreachable bombing:
   (sending bogus ICMP unreachables to make people's TCP connection
   disconnect, etc)

Why would a TCP connection disconnect just because it got an ICMP
unreachable?  The only time a TCP connection should listen to ICMP
unreachable is when it's SYN is still un ACK'ed.

--
-russ <nelson@crynwr.com>      ftp.msen.com:pub/vendor/crynwr/crynwr.wav
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)    | Quakers do it in the light
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000198][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 94 18:09:29 GMT
From:      al137251@academ01.mty.itesm.mx (Joaquin Guillen Rodriguez)
To:        comp.protocols.tcp-ip
Subject:   Questions about the future of Internet

I'm doing an investigation about Internet. Could someone help me with the
following questions?

1. What is the future of Internet?

2. Will TCP/IP remain like the protocol of the internet?

3. I have heard talk about TUBA, Will this protocol be the new protocol
of the net?

4. Where can find information about this?

Thanks in advance
Joaquin Guillen Rodriguez
jguillen@socrates.mty.itesm.mx
al137251@academ03.mty.itesm.mx


-----------[000199][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 18:33:35 GMT
From:      henryh@well.sf.ca.us (Henry Hwong)
To:        comp.protocols.tcp-ip
Subject:   Many questions (sockets and protocol)

Forwarded from a person not on the net (yet)....

(in other words, I'm not quite responsible for the content :-))

-Henry (henryh@well.sf.ca.us)

------------------------------------------------------------------------
I have

A) three questions concerning C application coding using Berkeley sockets
with TCP transport and
B) a question on the choice of the appropriate protocol.

A.

1. Retransmission

Fact: TCP is reliable; it will retransmit MAX times upon non-receipt of ACK
      within TIMEOUT.

It is also stated (M. Santifaller "TCP/IP and NFS", p.129) that "if message
forwarding is unsuccessful, the RPC layer will be notified and may inform the
application program".

In contrast to these facts, having a TCP virtual circuit on a socket with
file descriptor 'sfd', one may code:

		n_writ = write( sfd, buf, n );

Consistent with delayed writes to regular files, the value returned into
n_writ will mean only "placed into kernel buffer", not "sent to destination".
No SIGPIPE will be raised, or error code returned even if the destination
application exited.

The transmission may still fail in at least two ways.  First, the actual
transmission may simply fail to elicit ACK upon MAX retransmissions, due to
whatever communication problems.

Second, it appears valid that even if the destination actively shut down its
side of the circuit (FIN flag), the sender can still transmit successfully.
Alternatively, if the application is not there anymore, the message will have
no place to go.  Can proper acknowledgment be forced by using the PSH flag?
But what, if anything, can the socket code say to force the PSH flag?  Is
this transmission successful or unsuccessful to TCP?


Question: when TCP eventually attempted and fled to send my 'n'  bytes, does
the Berkeley socket interface provide at C level a mechanism (such as the one
used by RPC), to find out that the transmission failed?  Or, being very
pragmatic, if application level reliable delivery is required, is there any
getting away from adding an application level ACK/NACK messaging?


2. Flow Control	

Fact:  TCP handles flow control by receive window size; ts value is in the
       TCP header.  When the receive buffer is filling up, a TCP algorithm
       reduces the window value.  A window size of 0 stops the source TCP. 

Question: This is the default behavior.  Right?  Must the socket interface
          say anything explicitly?  


3. Data formatting in a TCP messag

Worthy examples of read/recv & write/send (R. Stevens "UNIX Network
Programming") all show the transmitted data as a string.  Is there some
standard for how one does transmit structures, unions,.  ?

This must be an old problem: many people must have solved it by now...

(This question is not concerned with XDR coding, & ntoh and hton issues).


B.

For a medium to high transaction volume application, need some guidelines on
choosing between TCP sockets based code vs. UDP based RPC.  I understand that

  * for TCP the application will have to maintain record boundaries.  Also,
    depending on the resolution of Q.1 above, the application may have to do
    its own message ACK/NACK handling;

  * for UDP based RPC, I'll have to add to the session layer some code to
    handle retransmission and idempotence of communication.  Is there some
    "standard" protocol and/or code for this (kinda TFTP)?

When all the components above are in place how are the two solutions
different, as evidenced by

  * throughput and reliability;

  * effort required for code maintenance;

  * portability to heterogeneous platforms;



-----------[000200][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 01:15:29 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

ganzhorn@cisco.com (Charles Ganzhorn) writes:
>pam@csn.org (Paul A. Morris) wrote:
>> [...edited...] 
>> 255.255.255.192 only gives 2 subnets of 62 hosts
>> 255.255.255.224 will give 6 subnets of 30 hosts
>> 255.255.255.240 will give 14 subnets of of 14 hosts
>> 255.255.255.248 will give 30 subnets of 6 hosts
>> 
>> i understand the bit representation of 192, 224,240, and 248
>> however i do not understand where the number of subnets or the 
>> number of hosts is derived.  if someone could explain this in 
>> a more straightforward manner than most of the books i have 
>> looked at i would greatly appreciate it
>> thanks
 
>Given that you say you understand about the bit representations, then the
>missing piece of information that may be confusing you is this:  you can't
>use the node number which is all binary 0 or all binary 1. 

[...and so on...]

Paul: 

It's much, much simplier than this.  Once someone understands the bit
representations, they can completely stop using them and do it all in
2**n - 2, 2**m - 2, where n = the number of subnet bits and m = the 
remaining number of available bits.  Here's an illustration; it takes
about 10 minutes practice, and you'll never doodle out those 01101001s
on a napkin again.  You'll be able to do it in your head. 

For example, given a class C address:

8 total bits available, 2**8 = 256, minus 2 = 254 available host IDs.
The reason you subtract 2 is because of what Charles said: you can't
use all zeros or all ones in either the subnet mask or host ID fields.
This means that 00000000 and 11111111 can't be used -- but everything 
else is fair game.  Bear with me. 

Since you're subtracting two from whatever number you come up, then 
since 2**1 - 2 = 0 , then a one bit subnet is can't be used: it produces 
0 networks, which is exactly right ;). 

So, you start subnetting that class C address:

2 bit mask = 2**2 - 2 or two networks of 2**(8-2=6) - 2 or 62 hosts.
3 bit mask = 2**3 - 2 or six networks of 2**(8-3=5) - 2 or 30 hosts.
4 bit mask = 2**4 - 2 or 14 networks of 2**(8-4=4) - 2 or 14 hosts. 

The point is you're dealing with the number of available bits in the 
class X IP address: 24 for class A, 16 for class B, 8 for class C. 
n + m = total available bits, in other words. 

You also derive your networks by counting by 2**m:

C.C.C.32
C.C.C.64
C.C.C.96
C.C.C.128
C.C.C.160
C.C.C.192

And you derive your (class C) netmask by 256 - 2**m:

2 bits: 256 - 2**(8-2=6) = 256 - 64 = 192 
3 bits: 256 - 2**(8-3=5) = 256 - 32 = 224
4 bits: 256 - 2**(8-4=4) = 256 - 16 = 240 
etc. 

Ahh, someone will say, what about +/- 8 bit subnet boundaries?  So
what? is the answer; it still works.

Example: class B address, 10 bit subnet:

You have 16 total bits to play with, so n = 10 and m = 6.  You have
2**10 - 2 = 1022 networks and 2**6 - 2 = 62 hosts on each one.

Your networks still count up by 2**6 = 64, only in this case you'll
probably want the binary feature on a calculator to help you grab the
eight leftmost bits and stuff them into the third class B octet. 
In all honesty, I don't think many of us are going to be dividing
up many more class B addresses; the last one I personally did was
almost 2 years ago.  Those customers are pretty big, and there's 
no shame in pulling out a calculator in front of a big client ;). 

This is a great way to quickly determine how many subnet bits you'll
need on an ID, and if you're good with your 2**y calculations you
will impress the living stuffings out of people for some reason,
as you do this stuff in your head before their eyes.

More importantly, it's very difficult to make mistakes this way
unless you simply can't multiply and add. 

No offense meant, Charles; love those cisco routers! 

Richard 

By day:

Senior Network Analyst
I-NET Inc. Bethesda, MD 
reh@inet.com  

>Charles.
>--
>Charles Ganzhorn                        Email:  ganzhorn@cisco.com
>cisco Systems                           Phone:  612-368-8922
>Chaska, MN                              FAX:    612-368-9977


-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondance 			| installed base. 

-----------[000201][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 18:39:23 GMT
From:      martin@datacomm.ucc.okstate.edu (Martin McCormick)
To:        comp.protocols.tcp-ip
Subject:   RS6000 ARP Cache


	We are running "NV6000" on an RS6000 to manage our campus network.
is there a way to tune the size and expiration time of the ARP cache so
as to reduce the amount of ping traffic that the system is generating?  We
are getting complaints from some departments about this and I think they
have a point.

Martin McCormick WB5AGZ   Stillwater, OK
O.S.U. Computer Center Data Communications Group



-----------[000202][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 17:03:59 +0100
From:      tjc@ecs.soton.ac.uk (Tim Chown)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In <Co3GLs.HJA@mv.mv.com> klos@mv.mv.com (Patrick Klos) writes:

>We (Klos Technologies, Inc.) have a product called PacketView that goes for
>$299 and can capture and decode network traffic on ethernet, token-ring and
>...

An alternative is the PD sniffer for the PC called gopher, which is from
a University in Holland.  Good enough to see various packet types, filter 
them, look at them, show bar charts of this and that.  Very nice.

Tim

-----------[000203][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 18:55:04 GMT
From:      chris.resch@daytonoh.ncr.com
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In article <2obsdf$gtr@marr.ecs.soton.ac.uk> tjc@ecs.soton.ac.uk (Tim Chown) writes:
>In <Co3GLs.HJA@mv.mv.com> klos@mv.mv.com (Patrick Klos) writes:
>
>>We (Klos Technologies, Inc.) have a product called PacketView that goes for
>>$299 and can capture and decode network traffic on ethernet, token-ring and
>>...
>
>An alternative is the PD sniffer for the PC called gopher, which is from
>a University in Holland.  Good enough to see various packet types, filter 
>them, look at them, show bar charts of this and that.  Very nice.
>
>Tim

Where can I get this gopher sniffer software?

-----------[000204][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 19:33:40 GMT
From:      ganzhorn@cisco.com (Charles Ganzhorn)
To:        comp.protocols.tcp-ip
Subject:   Re: Large flat network

In article <CnwB32.8qs@calcite.rhyolite.com>, vjs@calcite.rhyolite.com
(Vernon Schryver) wrote:
> 
> In article <ganzhorn-050494084548@mojo.cisco.com> ganzhorn@cisco.com (Charles Ganzhorn) writes:
> > ...
> >A fairly typical approach would be to switch the clients, use FDDI to get
> >off the switch, put servers for those clients on the FDDI and then route
> >the FDDI's...
> 
> That sounds like it involves bridging between FDDI and Ethernet in the
> switches, which is a potentially serious problem if the servers are real
> FDDI machines.  Some FDDI-Ethernet switches are unable to correctly
> translate FDDI frames to Ethernet packets.  This is commonly a problem
> with 4KByte NFS/UDP/IP frames.

This is true:  you've got to set the MTU on the FDDI side correctly for
protocols which do not fragment.  For IP, cisco's Catalyst box supports
fragmentation as defined in RFC 791.

> 
> What you'd like is a box that functions as a bridge among Ethernets (i.e.
> as a switch) and acts as a router for packets going between the Ethernets
> and the FDDI link away from the group.  I have not heard of such a thing
> yet.  The switch vendors have religious objections to understanding
> anything except MAC packets and the router vendors don't have the right
> hardware--yet, as far as I know.

It's really not a matter of religion, it's a matter of price for
performance.  MAC layer switching is cheap compared to full blown routing
and it's easier considering the number of man-years it takes to build
routing software.  Also, it doesn't require re-subnetting and/or variable
length subnet masks.  From a strict layer interpretation, bridging from one
MAC layer to a different MAC layer is by definition a kludge but from an
addressing and price perspective, it is attractive.  It doesn't scale to
the enterprise (remember the broadcast domain discussion earlier in the
thread) but it is valid enough for the workgroup.

Charles.
--
Charles Ganzhorn                        Email:  ganzhorn@cisco.com
cisco Systems                           Phone:  612-368-8922
Chaska, MN                              FAX:    612-368-9977

-----------[000205][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 94 19:35:22 GMT
From:      a09878@giant.rsoft.bc.ca (Curt Sampson)
To:        comp.mail.misc,comp.protocols.tcp-ip
Subject:   Re: RFC-822: Format of first line in mail folder not covered by RFC-822?

ww02ksrm@sbusol.rz.uni-sb.de (Ralf Muehlen) writes:

>Well, no big problem, I wrote a small perl script which converts these [From_]
>lines back to standard. I just wonder if the format of these lines
>is covered by a RFC. In RFC 822 (Standard for the format of ARPA Internet 
>text messages) I haven't found anything appropriate. It seems as if only
>the following lines were defined.

They're not part of the RFC-822 standard, or any "standard" as far as
I know. They're just a convention used by Unix user mail agents and
local delivery systems.

cjs
--
Curt Sampson  a09878@giant.rsoft.bc.ca
Fluor Daniel		  604 691 5458
1444 Alberni Street
Vanouver, B.C., V6G 2Z4		 Non scholae sed vitae discimus. --Cicero

-----------[000206][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 19:57:05 GMT
From:      ganzhorn@cisco.com (Charles Ganzhorn)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

In article <Co3vtK.5M9@csn.org>, pam@csn.org (Paul A. Morris) wrote:
> 
> In an earlier posting on subnetting i ahd read that a subnet mask 
> of 255.255.255.192 only gives 2 subnets of 62 hosts
> a mask of 255.255.255.224 will give 6 subnets of 30 hosts
> a mask of 255.255.255.240 will give 14 subnets of of 14 hosts
> a mask of 255.255.255.248 will give 30 subnets of 6 hosts
> etc
> 
> i understand the bit representation of 192, 224,240, and 248
> however i do not understand where the number of subnets or the 
> number of hosts is derived.  if someone could explain this in 
> a more straightforward manner than most of the books i have 
> looked at i would greatly appreciate it
> thanks
To set the stage, the table above assumes a class C address.  I'm not sure
if that was a missing piece of information or not but there you go anyway.

Given that you say you understand about the bit representations, then the
missing piece of information that may be confusing you is this:  you can't
use the node number which is all binary 0 or all binary 1.  And the right
most bits of the subnet number (the bits that reside in the byte(s) which
would have belonged to the node number under a default mask) cannot be all
binary 0 or all binary 1.

For example, you didn't list the mask 255.255.255.128 in the table above. 
That's because there's only one bit in what should have been a node number
bit and the only two possible values are 0 and 1 thus making those subnet
numbers ineligible.  This results in zero useable subnets.

There are RFC's in the wind to deal with this but for the time being, this
is the nature of the beast.  And in fact under some configurations you can
in fact use subnet zero but given a heterogeneous environment, you'd have
to use some care to ensure that everyone could deal with it.

Charles.
--
Charles Ganzhorn                        Email:  ganzhorn@cisco.com
cisco Systems                           Phone:  612-368-8922
Chaska, MN                              FAX:    612-368-9977

-----------[000207][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 02:41:03 -0400
From:      radev@news.cs.columbia.edu (Dragomir R. Radev)
To:        comp.lang.c,comp.unix.questions,comp.protocols.tcp-ip
Subject:   QUESTION: clientless TELNET

I am looking for a reference on how to write a clientless TELNET
server. I want users to type "telnet xxx.yyy.zzz pppp" and
connect to my server which is waiting on port pppp for connections.
The reason I don't want clients is that I want to create a service
which people should want to use without having to compile
clients, etc.

Thanks

Drago


-- 
Dragomir R. Radev                                Graduate Research Assistant
Natural Language Processing Group          Columbia University CS Department
Office: (212) 939-7121        Lab: (212) 939-7108       Home: (212) 866-8548
          http://www.cs.columbia.edu/~radev/home.html

-----------[000208][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 02:42:43 -0400
From:      radev@news.cs.columbia.edu (Dragomir R. Radev)
To:        comp.lang.c,comp.unix.questions,comp.protocols.tcp-ip
Subject:   Re: QUESTION: clientless TELNET

In article <2odfpv$ruc@age.cs.columbia.edu>,
Dragomir R. Radev <radev@news.cs.columbia.edu> wrote:
>I am looking for a reference on how to write a clientless TELNET
>server. I want users to type "telnet xxx.yyy.zzz pppp" and
>connect to my server which is waiting on port pppp for connections.
>The reason I don't want clients is that I want to create a service
>which people should want to use without having to compile
>clients, etc.
>
>Thanks
>
>Drago

I forgot to say that I am using BSD sockets (SunOS 4.1.3.)

D.
-- 
Dragomir R. Radev                                Graduate Research Assistant
Natural Language Processing Group          Columbia University CS Department
Office: (212) 939-7121        Lab: (212) 939-7108       Home: (212) 866-8548
          http://www.cs.columbia.edu/~radev/home.html

-----------[000209][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 20:13:49 GMT
From:      leonard@telcom.arizona.edu (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?


In article <2o8g36$a56@Jupiter.mcs.com>, karl@MCS.COM (Karl Denninger) writes:
|Hi!
|
|I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
|where I can find such a thing?  It would be EXTREMELY helpful to me right
|now.
|
|Thanks!

You don't say what OS you're running.  For VMS, Andy Pavlin's 
ETHERMON program works very nicely.  It might be available
on gw.syr.ge.com.

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
  \ Don't lock yourself into open systems. /

-----------[000210][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 21:03:28 GMT
From:      pomeranz@imagen.com (Hal Pomeranz)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,comp.unix.admin,comp.unix.large
Subject:   BayLISA April Meeting: Paul Vixie


BayLISA will be holding it's monthly meeting on April 21, 1994.
Please note that we are at a new location, Synopsys. Please
see directions below.

This months speaker, Paul Vixie, sent the following abstract:

          Stolen Networks And How To Live With Them

  Since Class A networks have always been hard to get and Class B's
  have recently also become hard to get, a growing number of IP-using
  organizations are using Class A and B network numbers which officially
  belong to someone else.  This abuse creates some interesting problems
  for the Internet gateways to such organizations, especially if they
  want end-to-end IP connectivity from within their internal, illegal,
  unadvertised, ambiguous network all the way out to the real Internet.

  This talk will focus on an implementation of ``proxy forwarding'' done
  mostly via intercepted system calls and the ``hooks'' interface in the
  BIND 4.9.2 resolver.  The described system is similar to ``Socks'' but
  violates fewer invariants.  It is running in production on BSD/386,
  and should be portable to any BSD-like operating system whose Socket
  interface is implemented with system calls rather than library stubs.
  The code will shortly be released to the public.

We hope you can attend.

- -----

Notes: The BayLISA Summer Picnic is coming together, and we are
currently looking at late July as our target date.  Watch for further
announcements.

- -----
General Meeting Information:
 
        Date:   Thursday, Apr 21st, 7:30 PM
                (Third Thursday of every month)
                Please do not arrive before 7 PM.
 
        Place:  Synopsys
                Building C
                700 E. Middlefield Road
                Mountain View, CA 94043
                [SEE BELOW FOR MAP AND DIRECTIONS]
 
 
BayLISA Information:
 
The BayLISA group meets monthly to discuss topics of interest to systems
and network administrators.  The meetings are free and open to the public.
 
BayLISA Contact Information:
 
     For general information:     baylisa-info@baylisa.org
     BayLISA Board:               blw@baylisa.org
     Individual Board members:    first_lastname@baylisa.org
     FTP Service:                 ftp.baylisa.org:/BayLISA

     Non-electronic mail:         BayLISA
                                  PO Box 64369
                                  Sunnyvale, CA
                                  94088-4369

To join the BayLISA mailing list, send email to majordomo@baylisa.org with
a body message of help.

- -----
 
Synopsys is located at the corner of Highway 237 and Middlefield.
If you are going South on 101 you exit on Ellis street.  Turn right
towards Mountain View (not Moffet field).  Turn left onto Middlefield. 
Go past the light at Highway 237.  The next light will be Bernardo,
turn left into the parking lot.  Head left to Building C.
 
If you are going North on 101, exit on to Mountain View (I don't
remember if it's West or South 237, freeways here just don't make
sense).  Turn left at Middlefield.  It will be the second traffic
light.  The next light will be Bernardo, turn left into the parking
lot.  Head left to Building C.
 
Ascii diagram, not to scale :-)
 
 
                     Highway
                       101
                |       |
                | ellis \
                +--------\
                |         \
                |          \
- ---------+------+--------------------- Highway 237
                |M     |     \
                |i @   |maude \
        --------+d---  |       \
     bernardo   |d     |
                |l     |
                |f
                |i
                |e
                |l                @ = Synopsys
                |d

-----------[000211][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 19:24:03 +0100
From:      tjc@ecs.soton.ac.uk (Tim Chown)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In <2obsdf$gtr@marr.ecs.soton.ac.uk> tjc@ecs.soton.ac.uk (Tim Chown) writes:

>An alternative is the PD sniffer for the PC called gopher, which is from
>a University in Holland.  Good enough to see various packet types, filter 
>them, look at them, show bar charts of this and that.  Very nice.

Whoops, I (of course :) meant the gobbler package, my wires were crossed!

Cheers,
	Tim

-----------[000212][next][prev][last][first]----------------------------------------------------
Date:      11 Apr 1994 21:33:26 GMT
From:      hnn@charm.physics.ucsb.edu (Harry Nelson)
To:        comp.protocols.tcp-ip
Subject:   [q] Grand Junction Networks

  Hello, I am looking for phone/address of Grand Junction Networks, apparently
in Santa Clara.  Directory assistance there doesn't know them.  Please e-mail
reply to hnn@charm.physics.ucsb.edu.  Thanks you

-----------[000213][next][prev][last][first]----------------------------------------------------
Date:      Mon, 11 Apr 1994 21:54:46 GMT
From:      raj@cup.hp.com (Rick Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Benchmarking

Andy Sweeney (asweeney@odin.sw.stratus.com) wrote:
: I am looking for any information on TCP/UDP benchmarking tools and/or
: perfomance expectations.

I tend to use netperf (anon ftp from
col.hp.com:dist/networking/benchmarks/) but then, I'm a triffle biased...

rick jones

-----------[000214][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 00:17:08 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Benchmarking

asweeney@odin.sw.stratus.com (Andy Sweeney) writes:

>I am looking for any information on TCP/UDP benchmarking tools and/or
>perfomance expectations.

Performance depends on the underlying network.  If you are talking
about an Ethernet(tm) then TCP will give you about 800K/second
on a well tuned network.  Performance for UDP on a quiet
network is a bit higher because of the low overhead.  Once the
network is crowded with traffic performance for any pair of
communicating hosts is lowered.  

There are some tools for measuring performanc on ftp.netcom.com
in ~ftp/pub/gnn.

There have been several papers published showing results on 
Ethernet/FDDI as well as ATM networks.  

You might look at the following papers:

%z TechReport
%A J. C. Mogul
%T Observing TCP dynamics in real networks
%p DECWRL.
%R 92/2
%D 1992

%z TechReport
%A David R. Boggs
%A Jeffrey C. Mogul
%A Christopher A. Kent
%T Measured capacity of an Ethernet: myths and reality
%D Sep. 1988
%p DECWRL.
%R 88/4
%x Ethernet, a 10 Mbit/sec CSMA/CD netowrk, is one of the most
%x successful LAN technologies.  Considerable confusion exists as
%x to the actual capacity of an Ethernet, especially since some
%x theoretical studies have examined operating regimes that are not
%x characteristic of actual networks.  Based on measurements of
%x an actual implementation, we show that for a wide class of
%x applications, Ethernet is capable of carrying its nomial
%x bandwidth of useful traffic, and allocates the bandwidth fairly.

Both of these papers are available for anonymous FTP on gatekeeper.dec.com
in DEC/WRL/research-reports .


Hope this helps...

Later,
George

-- 
gnn@netcom.com

If people were more concerned with being reconciled than with being right, 
the world would be a better place.  -- Miss Manners

-----------[000215][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 02:07:17 GMT
From:      annadata@bashful.uhcc.Hawaii.Edu (Anil Annadata)
To:        comp.protocols.tcp-ip
Subject:   Re: Optimizing TCP/IP Process Architecture

In article <1994Apr7.205522.14976@astph> jeff@astph (Jeff Martin) writes:
>We are running Interactive UNIX Release 3 Version 3.2.  We are
>developing a network database application using TCP/IP.
>
>HOW CAN WE REDUCE THE NUMBER OF SERVER PROCESSES USING TCP/IP?  Preferably
>allowing a tunable server/client ratio?  NOTE: Comer and Stevens suggest
>in "Interworking with TCP/IP, Vol 3," that several server processes can
>wait at accept() and establish a socket connection for each TCP/IP
>transaction, which is way to costly.

	I am doing pretty much the same thing. You can have more than one client
per server depending upon the requirements of the client. In the server
process you can very effectively use the select() call for this. All this 
can be done while you are accepting the connections.


Anil

---------------------
email: annadata@uhunix.uhcc.hawaii.edu

-----------[000216][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 02:48:38 GMT
From:      fitz@wang.com (Tom Fitzgerald)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

pam@csn.org (Paul A. Morris) writes:

> In an earlier posting on subnetting i ahd read that a subnet mask 
> of 255.255.255.192 only gives 2 subnets of 62 hosts
 
> i understand the bit representation of 192, 224,240, and 248
> however i do not understand where the number of subnets or the 
> number of hosts is derived.  if someone could explain this in 
> a more straightforward manner than most of the books i have 
> looked at i would greatly appreciate it

If you subnet to n bits, 2<=n<=6, your netmask is 

255.255.255.X where X = (2^n - 1) * (2^(8-n))     ^ is exponentiation

This give you 2^n - 2 subnets of 2^(8-n) - 2 hosts per subnet.

-- 
Tom Fitzgerald   Wang Labs   Lowell MA, USA   1-508-967-5278   fitz@wang.com
Pardon me, I'm lost, can you direct me to the information superhighway?

-----------[000217][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 10:17 PDT
From:      adelman@tgv.com (Kenneth Adelman)
To:        comp.sys.dec,comp.protocols.tcp-ip
Subject:   Re: DECnet and IP

In article <1994Apr12.124609.24656@cti.wnc.nedlloyd.nl>, wilfred@zeus.wnc.nedlloyd.nl (Wilfred Mollenvanger) writes...
>Hi,
>
>We got a bunch of VAXs talking DECnet. We want to route as much
>traffic as possible using IP routing 'cause we are standardising to
>TCP/IP. Is it possible to run DECnet over IP or encapsulate it in IP?
>What are the problems if it is possible (performace, price $$) ?

    MultiNet, our TCP/IP for OpenVMS product, lets you do this.
Basically, you're creating a point-to-point DECnet link and circuit
that encapsulates DECnet packets in UDP datagrams (or optionally a TCP
vc).  Over an Ethernet, I recall measuring about a 15-30% performance
degradation for DECnet/UDP/Ethernet as compared to DECnet/Ethernet,
but this is going to be very sensitive to CPU speed (this was measured
on a MicroVAX II), so I would measure it in your environment.

    The other gotcha is that to be of any use (that is, to involve
more than two machines in your DECnet network) one of the machines
needs to be a DECnet router to connect you to other links (either
other DECnet over IP links, or the rest of your network).  Phase V
DECnet won't support OpenVMS machines as DECnet routers.

							Ken

-----------[000218][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 03:40:40 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

In article <NELSON.94Apr11140729@crynwr.crynwr.com> nelson@crynwr.crynwr.com (Russell Nelson) writes:
>Why would a TCP connection disconnect just because it got an ICMP
>unreachable?  The only time a TCP connection should listen to ICMP
>unreachable is when it's SYN is still un ACK'ed.

They shouldn't, but many do.  Why?  Because that's what the BSD implementors
did.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000219][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 03:41:34 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: Hostroutes

In article <2oae8r$2bt@cherokee.advtech.uswest.com> crohrer@advtech.uswest.com (Chris Rohrer) writes:
>Hi all,  I'm looking to understand the concept of hostroutes.  Is this a
>simple, straightforward concept that can be explained in a few sentences or
>do things get complicated fast?  Is this just an address translation
>technique?  Anyone know where this concept is written about?  Are there
>routers that do this?

IP routes specify the direction(s) to forward a packet to reach a member of a
group of machines with related IP addresses.  The group may be everyone not
covered by an explicit route (default route), everyone in the same IP network
number (network route), everyone in an IP subnet (subnet route), or a single,
specific IP address (host route).  Actually, there is an evolution away from
the net,subnet,host distinctions, toward defining IP routes in terms of an
IP destination with an associated netmask.  The netmask defines how much of
an address is used for routing purposes, and the remainder defines the set
of IP addresses being routed toward.  In this scheme, a host route is an IP
address with a netmask of all 1s.

Art


-----------[000220][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 03:49:29 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: How to get ANCIENT rfcs?

In article <1994Apr11.034435.7363@unlv.edu> ftlofaro@unlv.edu (Frank Lofaro) writes:
>Where can I find ancient RFCs?
>I know they are way obsolete, but I am curious about the very early 
>Internet (arpanet).
>
>I could find rfc3, and a couple other single digit ones in the normal 
>places, but I can not find rfc1 anywhere. Where could I find that and 
>the rest of the first few rfcs?

If I remember correctly, some of the very early RFCs were assigned RFC numbers,
but were never maintained in electronic form.  At one time the NIC would mail
paper copies, but it's not clear that they ever had copies of all documents
with RFC numbers.  I don't know if they will still send you a paper copy
if they have it and it's not available online. I don't even know if the paper
copies followed the NIC when it left SRI.  Only a few of the original ARPANET
developers are still involved with the Internet, and I expect even fewer of
them read this newsgroup.

Art


-----------[000221][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 03:57:27 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Many questions (sockets and protocol)

In article <2oc55v$3op@nkosi.well.com> henryh@well.sf.ca.us (Henry Hwong) writes:
>Second, it appears valid that even if the destination actively shut down its
>side of the circuit (FIN flag), the sender can still transmit successfully.

That's correct.  Each direction of a connection must be closed separately.
When the destination sends a FIN it only means that it won't send any more,
but it may still be willing to receive (and there are some protocols that
depend on this asymmetry).

>Alternatively, if the application is not there anymore, the message will have
>no place to go.  Can proper acknowledgment be forced by using the PSH flag?
>But what, if anything, can the socket code say to force the PSH flag?  Is
>this transmission successful or unsuccessful to TCP?

I think you can set the PSH flag by setting the TCP_NODELAY socket option.
However, this still doesn't make write() synchronous, it just means that
you'll learn of the error sooner.

>Question: when TCP eventually attempted and fled to send my 'n'  bytes, does
>the Berkeley socket interface provide at C level a mechanism (such as the one
>used by RPC), to find out that the transmission failed?  Or, being very
>pragmatic, if application level reliable delivery is required, is there any
>getting away from adding an application level ACK/NACK messaging?

After an error is detected, the next system call that tries to use the
socket will return an error code.  So you won't get an error from the
write() of the data that wasn't acknowledged, but a later operation.
>Fact:  TCP handles flow control by receive window size; ts value is in the
>       TCP header.  When the receive buffer is filling up, a TCP algorithm
>       reduces the window value.  A window size of 0 stops the source TCP. 
>
>Question: This is the default behavior.  Right?  Must the socket interface
>          say anything explicitly?  

It's handled automatically by the kernel.  The application doesn't have to
worry about this.

>3. Data formatting in a TCP messag
>
>Worthy examples of read/recv & write/send (R. Stevens "UNIX Network
>Programming") all show the transmitted data as a string.  Is there some
>standard for how one does transmit structures, unions,.  ?

No, there's no general standard.  It's specified by the application
protocol.

>This must be an old problem: many people must have solved it by now...

Yes, but they solve it in many different ways.

>(This question is not concerned with XDR coding, & ntoh and hton issues).

Those are some of the ways that it has been solved.  Another common
solution is ASN.1/BER.

It isn't really important *how* you specify it, as long as you remember to
specify it.  Otherwise you end up with a problem like the "talk" protocol.
The protocol was never specified in a machine-independent way, it was just
implemented by writing C structures; when the program was ported from Vaxes
to Suns the byte order got changed and the two systems couldn't talk to
each other.

However, using one of the common methods makes it easier for people to
implement your protocol, since they can use existing libraries.

>For a medium to high transaction volume application, need some guidelines on
>choosing between TCP sockets based code vs. UDP based RPC.  I understand that

Why not use TCP-based RPC, to get the best of both?
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000222][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 06:05:29 GMT
From:      annadata@curly.ics.Hawaii.Edu (Anil Annadata)
To:        comp.protocols.tcp-ip
Subject:   Re: Optimizing TCP/IP Process Architecture

>In article <1994Apr7.205522.14976@astph> jeff@astph (Jeff Martin) writes:
>>We are running Interactive UNIX Release 3 Version 3.2.  We are
>>developing a network database application using TCP/IP.
>>
>>HOW CAN WE REDUCE THE NUMBER OF SERVER PROCESSES USING TCP/IP?  Preferably
>>allowing a tunable server/client ratio?  NOTE: Comer and Stevens suggest
>>in "Interworking with TCP/IP, Vol 3," that several server processes can
>>wait at accept() and establish a socket connection for each TCP/IP
>

	But the best thing to do in this kind of applications is to use
UDP/IP. Check the source code of Talk. In that the server uses the UDP/IP
for checking talk requests and then back off by establishing a connection
between the two clients(TCP/IP). You can use something similar to 
this if it is possible.

>Anil
>
>---------------------
>email: annadata@uhunix.uhcc.hawaii.edu



-----------[000223][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 06:48:13 GMT
From:      John Matthews <matthews@mis.uswest.com>
To:        comp.protocols.tcp-ip
Subject:   socket(SOCK_RAW) & ICMP question

Does anyone know of a way to add additional ICMP input packet buffers
so that UNIX can queue multiple incoming ICMP packets?  I've already
tried using setsockopt() with the RCVBUF and verified that UNIX accepted
it, however UNIX is only sending the process one ICMP packet and appears
to be discarding the rest.  I am doing all this on SunOS 4.1.2 + 4.1.3.

I've re-written ping to utilize UDP ECHOs and create a sliding
window to accurately load test Frame-Relay and this works perfect.
I am now going back and attempting to get this same utility to use ICMP
as well, but I'm running into a problem where the process can miss ICMP
packets if it loses the CPU momentarily and more than one packet in
the sliding window queues up before the process can read it.  I easily
re-produce this by ^Zing the process while there are 16 packets in transit.
When I resume the process, 15 packets are dropped every single time.
With the UDP option, this all works great and it never drops a packet
even when loading an ethernet at 85-92% going towards a Cisco AGS+.

For those who will ask, I do plan on posting this utility to the net
once I can get this remaining issue resolved and clean up a few of the
options.
				John Matthews
				matthews@mis.uswest.com

-----------[000224][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 07:46:27 GMT
From:      hunen@brc.medtronic.com (Roger Hunen)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

In article <2od57oINNkm7@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
#In article <NELSON.94Apr11140729@crynwr.crynwr.com> nelson@crynwr.crynwr.com (Russell Nelson) writes:
#>Why would a TCP connection disconnect just because it got an ICMP
#>unreachable?  The only time a TCP connection should listen to ICMP
#>unreachable is when it's SYN is still un ACK'ed.
#
#They shouldn't, but many do.  Why?  Because that's what the BSD implementors
#did.

I think RFC 792 (ICMP) nor RFC 1122 (host requirements) supports this view, 
at least not for ICMP Port Unreachable:

RFC 1122, page 40:
  A transport protocol that has its own mechanism for notifying the sender
  that a port is unreachable (e.g., TCP, which sends RST segments) MUST
  nevertheless accept an ICMP Port Unreachable for the same purpose.

RFC 1122, page 104:
  Destination Unreachable -- codes 2-4
  These are hard error conditions, so TCP SHOULD abort the connection.

I couldn't find a phrase restricting response on ICMP Port Unreachable to 
connection startup. ICMP Host Unreachables are different though: they are 
soft errors. Implementations that abort a connection on a Host Unreachable, 
should be considered broken.

Regards,
-Roger

-----------[000225][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 07:56:43 GMT
From:      nike@indirect.com (Laurence Canter)
To:        comp.protocols.tcp-ip,mn.traffic
Subject:   Green Card Lottery- Final One?

Green Card Lottery 1994 May Be The Last One!
THE DEADLINE HAS BEEN ANNOUNCED.

The Green Card Lottery is a completely legal program giving away a 
certain annual allotment of Green Cards to persons born in certain 
countries. The lottery program was scheduled to continue on a 
permanent basis.  However, recently, Senator Alan J Simpson 
introduced a bill into the U. S. Congress which could end any future 
lotteries. THE 1994 LOTTERY IS SCHEDULED TO TAKE PLACE 
SOON, BUT IT MAY BE THE VERY LAST ONE.

PERSONS BORN IN MOST COUNTRIES QUALIFY, MANY FOR 
FIRST TIME.

The only countries NOT qualifying  are: Mexico; India; P.R. China; 
Taiwan, Philippines, North Korea, Canada, United Kingdom (except 
Northern Ireland), Jamaica, Domican Republic, El Salvador and 
Vietnam. 

Lottery registration will take place soon.  55,000 Green Cards will be 
given to those who register correctly.  NO JOB IS REQUIRED.

THERE IS A STRICT JUNE DEADLINE. THE TIME TO START IS 
NOW!!

For FREE information via Email, send request to
cslaw@indirect.com


-- 
*****************************************************************
Canter & Siegel, Immigration Attorneys
3333 E Camelback Road, Ste 250, Phoenix AZ  85018  USA
cslaw@indirect.com   telephone (602)661-3911  Fax (602) 451-7617

-----------[000226][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 15:24:52 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip,comp.dcom.sys.cisco,comp.dcom.sys.wellfleet
Subject:   RFC 950 is really ignored (was: subnetting)

(1) Notice that I've added comp.dcom.sys.{cisco,wellfleet} to the posting
list, and adjusted followups to comp.protocols.tcp-ip.  The points Vic
brings up are good ones, I think, but I question his statement that "no
useful software" distinguishes or uses the "this network" field.  And,
I provide a working counter-example. 

Vic seems to be saying: Yes, the RFC says it should be done like so,
but nobody really does it like that.  I'm hoping a couple of router 
software engineers will comment on it. 

mollett@lexmark.com (Vic Mollett) writes:
>reh@cs.umd.edu (Richard Huddleston) writes:

[...the rest deleted for irrelevancy to this specific thread...] 

>>Since you're subtracting two from whatever number you come up, then 
>>since 2**1 - 2 = 0 , then a one bit subnet is can't be used: it produces 
>>0 networks, which is exactly right ;). 
 
>This case is only true if you are concerned about accepting what RFC950
>terms "this network" and "all hosts."  I don't think I have ever seen a
>use of "this network" in any useful piece of software, it tends to suppose
>that a particular resource will have a particular address, which how do you
>determine that if you don't know what class your address is.  "All hosts"
>doesn't regularly get used, either.  Take the class A address of 9.0.0.0
>(I hope IBM doesn't mind if I borrow it for this demonstration) which I know
>has been assigned to various places around the world.  If someone sent a 
>packet to 9.255.255.255, it should, in theory, go to all of the machines in
>the 9.0.0.0 network.  I know for a fact that that will not happen, and I
>don't know of a case where it would even be desired.

Well, first, you're right: I do strive to comply with the RFC, and do what
it and the manuals suggest when configuring routers and IP hosts.  And I 
agree that unless you specifically tell a router to forward, that it will 
ignore a broadcast so poorly addressed as 9.255.255.255 would be -- unless 
the subnet mask on 9.0.0.0 were 255.0.0.0, which it isn't. 

>Both gentlemen have stated that a 1 bit subnet generates no networks, but if
>you subnet a class C address into 255.255.255.128, you _do_ get 2 usable net-
>works with 126 nodes apiece (0, 127, 128, and 255 are still reserved).

I just happen to have a cisco 4000 in my den at the moment, and I've tested
your assertion with IP address 195.195.195.65 netmask 255.255.255.128.

It fails, as I knew it would. 

I'll be polite at this point, and merely ask you to explain why that
is.  If you want to say that this is because cisco router software doesn't
meet your definition of "useful software" (see your comments above) then
I'll respectfully disagree and we'll call it a day. 
-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondance 			| installed base. 

-----------[000227][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 10:08:50 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Q: ICMP source quench and TCP ?

In article <zok.766141551@popp.ins.de> zok@popp.ins.de (Andreas Frackowiak)
writes: 
    The sender receives the source quenches (showed with netstat -s),
    but the TCP of the sender seems to work exactly as if the gateway
    silently discards packets without sending source quenches.
    
Correct.  Most BSD derived implementations do nothing with Source Quenches.
This is actually fine.  Source Quench has been deprecated by the IETF and
is going away.  Host implementations should use slow start instead.

    The behaviour is:

     - sending with maximum throughput
     - pause (timeout ?)
     - sending with maximum throughput
     - pause  (timeout ?)
    ... etc

... but this doesn't sound like it.
    
Tony

-----------[000228][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 10:39:06 GMT
From:      d.w.stevenson@strath.ac.uk (David Stevenson)
To:        comp.protocols.tcp-ip
Subject:   TCP Hardware (yes, Hardware)

Has anybody implemented TCP as a hardware engine, in order to acheive very high throughputs?

I'd be interested in research and/or products.

---
Dave Stevenson 				d.w.stevenson@strath.ac.uk
Computer Centre Communications		Tel : 44 41-552 4400 ext 3461
University of Strathclyde
Glasgow, Scotland, U.K. 


-----------[000229][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 11:30:34 GMT
From:      jgarven@mcl.cc.utexas.edu (James R. Garven)
To:        comp.protocols.tcp-ip
Subject:   Slip/PPP Access

Can anyone suggest a Slip or PPP Internet Access Provider 
that my sister-in-law can contact in the Portland, Oregon area?
She is really "pumped up" about Internet and has outgrown the Prodigy
"pond".  Any suggestions would be most appreciated.  For that matter,
perhaps someone knows of a service provider located somewhere else in the
U.S.A. that bundles in with its service a really good deal on long-distance
connections.

Thanks in advance.

Jim Garven

*=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=*
| James R. Garven                                                        
| ----------------------------------------------------------------------  
| Department of Finance, CBA 6.222   Voice:    (512) 471-6557            
| Graduate School of Business        Fax:      (512) 471-5073         
| University of Texas                Internet: jgarven@mcl.cc.utexas.edu   
| Austin, TX  78712, U.S.A.          BITNET:   jim.garven@utxvm.bitnet   
 *=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=*
| Moderator of RISKNet, a worldwide Internet mailing list/ftp/gopher/world
| wide web service devoted to risk and insurance issues:
|1. RISKNet Mailing List: Send subscription requests to jgarven@mcl.cc.utexas.edu
|2. World Wide Web Home Page: http://riskweb.bus.utexas.edu
*=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=*

-----------[000230][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 11:46:59 GMT
From:      etxmesa@eos.ericsson.se (Michael Salmon)
To:        comp.lang.c,comp.unix.questions,comp.protocols.tcp-ip
Subject:   Re: QUESTION: clientless TELNET

In article <2odfpv$ruc@age.cs.columbia.edu>
radev@news.cs.columbia.edu (Dragomir R. Radev) writes:
|> I am looking for a reference on how to write a clientless TELNET
|> server. I want users to type "telnet xxx.yyy.zzz pppp" and
|> connect to my server which is waiting on port pppp for connections.
|> The reason I don't want clients is that I want to create a service
|> which people should want to use without having to compile
|> clients, etc.

You have a client though, telnet. "Networking with TCP/IP" Volume III
by Comer and Stevens shows some servers suitable for use with telnet,
the only thing that I think you'll need to worry about is that you
can't use character 255 without duplicating it.

--

Michael Salmon

#include	<standard.disclaimer>
#include	<witty.saying>
#include	<fancy.pseudo.graphics>

Ericsson Telecom AB
Stockholm

-----------[000231][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 12:33:44 GMT
From:      mollett@lexmark.com (Vic Mollett)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

In article <2odaph$p0l@bedrock.cs.umd.edu> reh@cs.umd.edu (Richard Huddleston) writes:
>ganzhorn@cisco.com (Charles Ganzhorn) writes:
>>pam@csn.org (Paul A. Morris) wrote:
>>> [...edited...] 
>>> 255.255.255.192 only gives 2 subnets of 62 hosts
>>> 255.255.255.224 will give 6 subnets of 30 hosts
>>> 255.255.255.240 will give 14 subnets of of 14 hosts
>>> 255.255.255.248 will give 30 subnets of 6 hosts
>>> 
>>> i understand the bit representation of 192, 224,240, and 248
>>> however i do not understand where the number of subnets or the 
>>> number of hosts is derived.  if someone could explain this in 
>>> a more straightforward manner than most of the books i have 
>>> looked at i would greatly appreciate it
>>> thanks
>
>Since you're subtracting two from whatever number you come up, then 
>since 2**1 - 2 = 0 , then a one bit subnet is can't be used: it produces 
>0 networks, which is exactly right ;). 

This case is only true if you are concerned about accepting what RFC950
terms "this network" and "all hosts."  I don't think I have ever seen a
use of "this network" in any useful piece of software, it tends to suppose
that a particular resource will have a particular address, which how do you
determine that if you don't know what class your address is.  "All hosts"
doesn't regularly get used, either.  Take the class A address of 9.0.0.0
(I hope IBM doesn't mind if I borrow it for this demonstration) which I know
has been assigned to various places around the world.  If someone sent a 
packet to 9.255.255.255, it should, in theory, go to all of the machines in
the 9.0.0.0 network.  I know for a fact that that will not happen, and I
don't know of a case where it would even be desired.

Both gentlemen have stated that a 1 bit subnet generates no networks, but if
you subnet a class C address into 255.255.255.128, you _do_ get 2 usable net-
works with 126 nodes apiece (0, 127, 128, and 255 are still reserved).  I
suspect that many people are using subnets such as this because it is quite
simple, and meets virtually all of a users requirements.  Should it be used?
well that's a debate that probably can never be resolved.  Will it be used?
most definitely.

My own personal opinion is that class broadcast capabilities have lost
their usefullness as local networks become smaller and MANs and WANs become 
more diverse.  A local network broadcast is still useful, but I know I would 
have  people screaming at me if my equipment was send class broadcasts instead
of local net broadcasts.

>
>Richard 
>
>By day:
>
>Senior Network Analyst
>I-NET Inc. Bethesda, MD 
>reh@inet.com  
>
>>Charles.
>>--
>>Charles Ganzhorn                        Email:  ganzhorn@cisco.com
>>cisco Systems                           Phone:  612-368-8922
>>Chaska, MN                              FAX:    612-368-9977
>
>
>-- 
>Richard Huddleston rh310 		 		| Quality.
>Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
>Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
>Personal opinions/correspondance 			| installed base. 


-- 
                                             /\    Vic Mollett
 These opinions are my own and do not       /  \   Lexmark International, Inc.
 necessarily reflect those of my employer.  \  /   mollett@lexmark.com
                                             \/  

-----------[000232][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 12:35:36 GMT
From:      tbyrnes@ritz.mordor.com (Tim Byrnes)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

: >>...
: >
: >An alternative is the PD sniffer for the PC called gopher, which is from
: >a University in Holland.  Good enough to see various packet types, filter 
: >them, look at them, show bar charts of this and that.  Very nice.
: >
: >Tim
 
: Where can I get this gopher sniffer software?

Accroding to archie it can be gotten from:

dorm.rutgers.edu as
   /pub/msdos/wattcp/delft/gobbler.zip

or:

ftp.wustl.edu as
  /systems/ibmpc/umich.edu/communications/wattcp/delft/gobbler.zip



-----------[000233][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 12:46:09 GMT
From:      wilfred@zeus.wnc.nedlloyd.nl (Wilfred Mollenvanger)
To:        comp.sys.dec,comp.protocols.tcp-ip
Subject:   DECnet and IP

Hi,

We got a bunch of VAXs talking DECnet. We want to route as much
traffic as possible using IP routing 'cause we are standardising to
TCP/IP. Is it possible to run DECnet over IP or encapsulate it in IP?
What are the problems if it is possible (performace, price $$) ?

E-mail preferred.


Thanks in advance,

Wilfred Mollenvanger.

-----------[000234][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 94 22:03:51 -0500
From:      manohar@msus1.msus.edu
To:        comp.client-server,comp.protocols.tcp-ip,comp.protocols.nfs
Subject:   Stateless Servers and transport protocols

	I'm working on issues dealing with Statelessness in Servers, and from
what I've read, it seems that the main motivation for statelessness in
Client-server systems is protocol reliability. Statelessness simplifies
application protocol design as they need not bother about messages being
lost, delayed, duplicated etc. by the underlying network or system crashes 
server or client or network failures. That seems to be
the reason why Sun's NFS also tries to have its server as stateless as 
possible. A protocol like TCP would however guarantee
reliability and would serve to shield the application protocol from
network unreliability issues even with stateful servers.  
		
Does anyone have an idea of how widely unreliable transport protocols (
such as UDP) are used in Client-Server systems. It would be interesting 
to compare that with the current trend towards stateless servers.

Regards Ram

-----------[000235][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 20:52:02 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   Re: RFC 950 is really ignored (was: subnetting)

@cisco.com (Tony Li) writes:
>reh@cs.umd.edu (Richard Huddleston) writes: 
>
>    I just happen to have a cisco 4000 in my den at the moment, and I've tested
>    your assertion with IP address 195.195.195.65 netmask 255.255.255.128.
>    
>    It fails, as I knew it would. 
>    
>Sigh.  Here's the facts, folks:

"Sigh." ?  Did you read the previous posting?  The gentlemen stated that
one could use such a netmask on a class C address. 

>The RFCs (currently - they're being fixed) consider any subnet with all
>zero bits or all one bits in the subnet field to be verboten.  As a result,
>the subnet field must be two bits wide to have any useful subnets.
>
>Your example above fails because you've only defined one bit for the subnet
>field.

Yep.  I hope you don't think I was surprised.  I was testing the assertion
that you could define such a one-bit netmask, or that it was meaningful
given the current reading of the RFC.

I can't find *any* equipment that will allow me to set such a netmask, but
I'll have time to test it further tomorrow.

And, if one can philosophically define a single-bit subnet yet not get
the IP devices to use them, then that seems a pretty shallow point.  Here
I'm not referring to routers per se, but end systems. 

>In the future, the RFC's will be changing to permit all of this forbidden
>behavior.  

You wouldn't have any of the draft RFC numbers handy, would you?  I'd be
quite interested to see how the problem is going to be handled where a 
machine is determining which MAC address to use (i.e., is the destination
IP address on the local subnet, or do I send this packet to the gateway?)
when constructing to destination address of the frame.

>Tony
-- 
Richard Huddleston rh310 		 		| Quality.
Computer network weenie by day	:  reh@inet.com 	| Because today's hack 
Computer musician by night	:  reh@cs.umd.edu  	| is tomorrow's 
Personal opinions/correspondence 			| installed base. 

-----------[000236][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 14:33:47 GMT
From:      jmr@ibm1.nynexst.com ()
To:        comp.protocols.tcp-ip
Subject:   Looking for RFC distrib. site

I am looking for a few RFC documents:

	1. Is there a common distribution site privided by the ITF ?
	2. If not, can these RFC be purchased and from where ?

Any hint would be helpful. Thanks in advance.



-----------[000237][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 16:30:21 GMT
From:      jeff@zis.ziff.com (Jeff Macdonald)
To:        comp.protocols.tcp-ip
Subject:   new improved POP3?

About 2 months ago I stumbled upon a site that mentioned a new mail protocol 
was being developed that would improve upon the POP3 protocol. Basically this 
protocol would allow the mail to remain on the host, but the client would be 
able to read new/old mail from any location... 

Does anybody know where I can find this info?

I've tried Mosiac and the internet RFC site, but Mosaic bombeb....


thanks...





Jeff Macdonald
Ziff Information Services
Ziff Communications
10 President's Landing
Medford, MA  
jeff@zis.ziff.com

-----------[000238][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 13:11:34 +0200
From:      zok@popp.ins.de (Andreas Frackowiak)
To:        comp.protocols.tcp-ip
Subject:   Q: ICMP source quench and TCP ?


What is the correct usage and packet format of
ICMP-Source-Quench messages generated by a gateway ?

From RFC 792 (also read RFC 896) I understand a gateway
schould send an icmp-source-quench if he discards a packet
(caused of internal buffer fill up).

The Format of an ICMP-Source-Quench-Packet should be:
20 Bytes IP-Header
   8 Bytes ICMP-Header
     20 Bytes IP-Header of the discarded packet
      8 Bytes (first 64 Bits) data of the discarded packet

When I trace (with etherfind/tcpdump) a TCP connection over the 
gateway, the gateway sends ICMP (header looks correct, message
is generated for the right packet) to the sender.

The sender receives the source quenches (showed with netstat -s),
but the TCP of the sender seems to work exactly as if the gateway
silently discards packets without sending source quenches.

The behaviour is:
 - sending with maximum throughput
 - pause (timeout ?)
 - sending with maximum throughput
 - pause  (timeout ?)
... etc


There is no smooth adaption of data speed up to the
maximum bandwith of the pysical channel, and because of the
pauses the throughput is only 50-70% of the maximum throughput.

What is going wrong ?
Are there any newer RFCs about ICMP-Source Quenches and congestion control ?
Is the TCP of the sender broken (SunOS 4.1.3) ?

Andreas

-- 
Andreas Frackowiak					af@ins.de

-----------[000239][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 17:12:50 GMT
From:      nelson@crynwr.crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

In article <Co3vtK.5M9@csn.org> pam@csn.org (Paul A. Morris) writes:

   In an earlier posting on subnetting i ahd read that a subnet mask 
   of 255.255.255.192 only gives 2 subnets of 62 hosts

2^6 - 2 = 62

   a mask of 255.255.255.224 will give 6 subnets of 30 hosts

2^5 - 2 = 30

   a mask of 255.255.255.240 will give 14 subnets of of 14 hosts

2^4 - 2 = 14

   a mask of 255.255.255.248 will give 30 subnets of 6 hosts

2^3 - 2 = 6

   etc

   i understand the bit representation of 192, 224,240, and 248
   however i do not understand where the number of subnets or the 
   number of hosts is derived.  if someone could explain this in 
   a more straightforward manner than most of the books i have 
   looked at i would greatly appreciate it

You must leave two subnetworks unused -- the one with all zeroes, and
the one with all ones.  The one with all ones is the broadcast
address.  A packet with that destination address should be broadcast
on all the subnets (which is not to say that every router will be
configured to do that).  The one with all zeroes is the legacy
broadcast address.  BSD 4.2 used all zeroes, and had no way to change
that.  Newer systems have a way to configure the broadcast address in
case you have to be compatible with a BSD 4.2-derived system.

You might be able to use the subnet with a subnet of all zeroes, but
it's not advisable, because some software may have been written or
tested assuming that there will never be an all-zeroes subnet.

--
-russ <nelson@crynwr.com>      ftp.msen.com:pub/vendor/crynwr/crynwr.wav
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)    | Quakers do it in the light
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000240][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 13:56:43 +0200
From:      zok@popp.ins.de (Andreas Frackowiak)
To:        comp.protocols.tcp-ip
Subject:   255.255.255.25 via point-to-point link ?

Is it "legal" to send an ip-packet to the destination
address "255.255.255.255" via a point-to-point link or
it is forbidden in a RFC ?

Any pointer to the RFC number would be nice.

thanks
Andreas

-- 
Andreas Frackowiak					af@ins.de

-----------[000241][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 18:28:27 GMT
From:      minoo@khumbu.posix.tandem.com (Minoo Gupta)
To:        comp.protocols.tcp-ip
Subject:   How does OOB work ??

   I have a couple of questions regarding out of band data.
   The basic question is-- how does an application process discover
   that it has OOB data?


   Assume this scenario:
       One end of a socket sends 200 bytes, then 1 byte of OOB data,
       then 100 more bytes of normal data.

   Now the other end of the socket calls recv().
   Q1. I think that the recv() would complete OK and return 200 bytes,
       is that right?
   Now the receive queue has an OOB byte followed by 100 bytes of data.
   The other end calls recv() again. To get OOB data, the OOB flag must
   be set.
   Q2. Now what happens if the MSG_OOB flag is NOT set? Does the recv()
       return the next 100 bytes? Does it return 0 bytes? Does it return
       an error? What error? If it is a blocking socket, does it block?


   Now, let's start over and use select()--
       sender sends 200 bytes, 1 byte OOB, 100 bytes
   And, the receiving side calls select() for read/exception on this fd.
   Q3. Is this socket ready for both read and exception?

   Q4. If recv() is called and picks up the 200 bytes, and select() is
        called again, is the socket still ready on both read/exception
        or only exception?

   Q5. If recv() is called with MSG_OOB *before* the 200 bytes are read
       will it pull out the OOB byte "from the middle" of the sent data?
       Or would recv()/MSG_OOB fail until the 200 bytes "up front" is
       consumed? If the recv()/MSG_OOB can get the OOB data byte, will
       the next recv() get all 300 bytes of normal data?

   Now, let's start again and use OOB-InLine. Stevens in Unix Network
   Programming, says that the MSG_OOB flag is not needed to get OOB data
   for the inline case (p. 333). Sender sends 200 bytes, 1 byte OOB, and
   then 100 bytes.
   Now the receiving sides calls recv()--
   Q6. Does recv() return 301 bytes? (Is the OOB data a protocol between
       sender/recv-er, like "-1 in the data means OOB"?) Or do you need
       3 calls to get the data? How can the application know where the
       OOB data is? How does it even know to check for OOB data? Does it
       sit in a loop calling select() for read/exception?
               select for read/exception
               if exception then call ioctl(ATMARK)
                   calculate how much normal data
                   if no normal data in front of OOB then recv(OOBdata)
               if read then call recv(amount of normal data)
               go to top of loop

-----------[000242][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 20:18:38 GMT
From:      @udel.engr.sgi.com (Dave Crocker)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

Your note confused me on several counts:

In article <Cnuv6t.EE@chinet.chinet.com>, les@chinet.chinet.com (Leslie
Mikesell) writes:
|> I can't think of many TCP apps that should be allowed to work without
|> being able to do a reverse lookup to see who is making a connection,

In fact, the reliance on reverse lookup at a pseudo-security technique has
been roundly criticized.  The security hacks are quite clear and forceful
that the IP address must NOT be used as an authentication mechanism by
itself.

|> and all of the existing ones are going to depend on the underlying
|> IP addressing scheme.  I don't even know of a way to do this using

Of the major, standard protocols, only FTP carries the IP address at the
application.  There are some others and never remember the full list,
but it is fairly short.  It certainly is not most.

|> the TLI transport interface when the transport isn't IP (for example,
|> AT&T's osi-based StarGroup).  Is it possible? 

Since TLI is used for TCP access, also, I don't understand your
statement, here.

d/

-----------[000243][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 20:30:03 GMT
From:      prabha@ctron.com (V. Prabhakar)
To:        comp.protocols.tcp-ip
Subject:   IP addressing issue....

Hi,

    I am planning to use a PPP serial link as a backup line to a system,
which is already on the ethernet, with an IP address. Can I use the same IP
address while using the PPP link? If so, how can I switch back and forth
between the network channel and the PPP link? Any info on this will be very
helpful.

OR

Is it mandatory to use a seperate subnet for each of the interfaces (ether/
serial)? Can't I get away with seperate IP addresses instead of seperate
subnets? Ideally I would like to have a single IP address for both the
interfaces and dynamically switch back and forth.

Thanks,

-P


-----------[000244][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 22:02:24 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: RFC 950 is really ignored (was: subnetting)

In article <2oesi4$rik@bedrock.cs.umd.edu> reh@cs.umd.edu (Richard
Huddleston) writes: 

    Vic seems to be saying: Yes, the RFC says it should be done like so,
    but nobody really does it like that.  I'm hoping a couple of router 
    software engineers will comment on it. 
    
The routers do it more generally than the RFC allows.
    
    I just happen to have a cisco 4000 in my den at the moment, and I've tested
    your assertion with IP address 195.195.195.65 netmask 255.255.255.128.
    
    It fails, as I knew it would. 
    
    I'll be polite at this point, and merely ask you to explain why that
    is.  If you want to say that this is because cisco router software doesn't
    meet your definition of "useful software" (see your comments above) then
    I'll respectfully disagree and we'll call it a day. 

Sigh.  Here's the facts, folks:

The RFCs (currently - they're being fixed) consider any subnet with all
zero bits or all one bits in the subnet field to be verboten.  As a result,
the subnet field must be two bits wide to have any useful subnets.

Your example above fails because you've only defined one bit for the subnet
field.

cisco has some extensions here in that we will let you use the all-ones
subnet and (with a knob) the all-zeros subnet.  You should NOT use this
knob if you're running with classful IGPs.

In the future, the RFC's will be changing to permit all of this forbidden
behavior.  

Tony


-----------[000245][next][prev][last][first]----------------------------------------------------
Date:      12 Apr 1994 22:07:57 GMT
From:      henryh@well.sf.ca.us (Henry Hwong)
To:        comp.protocols.tcp-ip
Subject:   Neophyte Subnetting

What is the best way to subnet a Class C address? We wanted to use two
bits for subnetting (mask of FF.FF.FF.C0), hopefully yielding 4 subnets of
62 hosts (is that right?), but someone tested this out on a Novell server
acting as a router, and it wouldn't let us use 00 or 11 for the subnet bits.
The strange thing is that the TCPIP book from O'Reilly has this exact
scenario as a valid example. Help! What is the correct behavior?

-Henry (henryh@well.sf.ca.us)



-----------[000246][next][prev][last][first]----------------------------------------------------
Date:      Tue, 12 Apr 1994 23:13:19 GMT
From:      fitz@wang.com (Tom Fitzgerald)
To:        comp.protocols.tcp-ip,comp.protocols.iso
Subject:   Re: TCP applications over OSI transport?

les@chinet.chinet.com (Leslie Mikesell) writes:

> I can't think of many TCP apps that should be allowed to work without
> being able to do a reverse lookup to see who is making a connection,

Email, please, email.  Under no circumstances should email ever be
prevented from going through.

I once lived through a fiasco of trying to inform another postmaster that
his nameserver had eaten some bad NS records for my domain, but his mailer
wouldn't accept a connection from me because his nameserver had.....

You don't want to put anyone in this position.

-----------[000247][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 00:37:23 GMT
From:      rgs@gfimda.uucp (Robert G. Schaffrath)
To:        comp.protocols.tcp-ip
Subject:   Re: lpd on VAX


>Has anyone out there set up lpd on a VAX?  I've got lpd running, and
>it accepts jobs, but always errors out on the VAX with "User <name>
>has no permission to print on this host".  I've tried creating a proxy
>entry for the user, but this does not work.  Any assistance would be
>appreciated.
>
>thanks,
>Bryan Gonderinger

Who's TCP/IP package are you using on the VAX?  I am using TGV's Multinet
and have had no problems with the LPD functionality.  I have been using
it forward print requests to a high speed printer on another VAX system.
It is a heck of a lot easier than using DECnet!! :) 

******************************************************************************
* Robert G. Schaffrath, N2JTX   * Internet:   rgs%wpmax2%gfimda@uunet.uu.net *
* Systems Engineer              * CompuServe: 76330,1057                     *
* Maxwell House Coffee Company  * Phone:      914-335-2777                   *
* Kraft General Foods Corp.     * Slogan:     "ervice is ur mott"            *
******************************************************************************

-----------[000248][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 94 01:17:49 GMT
From:      jimmy_t@hnlv4.verifone.com
To:        comp.protocols.tcp-ip
Subject:   Looking for X.25 to Telnet gateway

We are looking for an X.25 to Telnet gateway box.
The only one I've found so far is the XYPLEX.  If you have
any suggestions for other vendors, please let me know.
 
Thanks.
 
Jim
+------------------------------------+--------------------------------------+
|  James H. Thompson                 |   jimmy_t@verifone.com    (Internet) |
|  VeriFone Inc.                     |   uunet!verifone!jimmy_t  (UUCP)     |
|  100 Kahelu Avenue                 |   808-623-2911            (Phone)    |
|  Mililani, HI 96789                |   808-625-3201            (FAX)      |
+------------------------------------+--------------------------------------+

-----------[000249][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 02:23:11 GMT
From:      starksm@genesis.mcs.com (Scott M. Stark)
To:        comp.protocols.tcp-ip
Subject:   gateway between en0 and slip0


Hi,
I have a computer that I use with dialup slip to access a local
service provider. I have just bought a second computer and
setup a local ethernet network and I would like the second host
to be able to access the external world through the slip0 interface
when it is active. Here's the host diagram:

en0 {203.202.200.2}                           en0 {203.202.200.1}
+-|---------------------+...................+-|----------------+
|  ironman.mcs.com    |                   | valkyrie.mcs.com  |
+------------------|----+                   +------------------+
                 slip0 {199.3.36.54}
                   |
                       {192.160.127.110}
+------------------|----+
| external net gateway  |
+------------------|----+

The routing tables I use on ironman that allow me to access the world
are:
Routing tables
Destination      Gateway            Flags     Refs     Use  Interface
127.0.0.1        127.0.0.1          UH          1      772  lo0
199.3.36.54      127.0.0.1          UH          0        0  lo0
192.160.127.110  199.3.36.54        UH          1       10  slip0
default          192.160.127.110    UG          1     6120  slip0
203.202.200      203.202.200.2      U          20    42785  en0

I added a default route from valkyrie to ironman:
Routing tables
Destination      Gateway            Flags     Refs     Use  Interface
localhost        localhost          UH          3     1669  lo0
default          ironman            UG          0       21  en0
203.202.200      valkyrie           U          26    54075  en0

Now if I ping a host on the external network from valkyrie it just
hangs:
	valkyrie.mcs.com:4# ping 192.160.127.82
	PING 192.160.127.82: 56 data bytes
traceroute shows:
	valkyrie.mcs.com:5# traceroute 192.160.127.82
	traceroute to 192.160.127.82 (192.160.127.82), 30 hops max,
	40 byte packets
	 1  ironman (203.202.200.2)  6 ms  3 ms  3 ms
	 2  * * *
	 3  * * *

While traceroute to the same host from ironman gives:
	ironman.mcs.com:17# traceroute 192.160.127.82
	traceroute to 192.160.127.82 (192.160.127.82), 30 hops max,
	38 byte packets
 	1  Chigate-1.Mcs.Net (199.3.35.101)  225 ms  187 ms  170 ms
 	2  MCSNet-AGS1.MCS.NET (199.3.35.100)  182 ms  183 ms  171 ms
	3  Mars (192.160.127.82)  183 ms  173 ms  177 ms

So, the packets get to ironman, but he does not know that they should
go out on the slip0 interface?  Any pointers to references or
suggestions would be appreciated.

Scott
--
Scott Stark
Stark Internation Software
4950 N. Marine Dr.
#102
Chicago, IL 60640
starksm@genesis.mcs.com	(NeXT mail accepted)

-----------[000250][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 94 02:53:09 GMT
From:      jkay@cs.ucsd.edu (Jon Kay)
To:        comp.protocols.tcp-ip
Subject:   Re: tcp checksum redundancy


	Checksum Redundancy Avoidance Frequently Asked Questions

Discussion about checksum redundancy avoidance (CRA) seems to arise
about once a quarter starting from when we published a Usenix paper on
the subject.  Thus, I have decided to write a FAQ on the subject and
simply post it whenever the subject arises, perhaps with the
occasional update.

NOTE: This FAQ details a slightly more advanced version of CRA than I
have previously described.  CRA has strategies for detecting bridges
and bad adapters now.

> What is checksum redundancy avoidance?

CRA is a technique for considerably raising the throughput of TCP/IP
implementations by permitting them to avoid doing checksumming in many
circumstances.  The basic observation is that both TCP/IP and LAN
hardware implement a checksum.  To a certain extent, the two are
redundant.  Specifically, they are redundant for packets that are
being sent over a single LAN, which is most packets.  Our design does
not checksum such packets; it does checksum routed or bridged packets,
since CRCs don't cover routers, and they don't cover some bridges.

> Where can I get more information than just this &^%&* FAQ??
> How can I run CRA?

A paper and implementations for Ultrix 4.2a and DEC OSF/1 1.3 are
available for anonymous FTP from the directory

ucsd.edu:pub/csl/fastnet

usenixw93.ps is the name of the paper; fetch the README to figure out
what you want in the way of implementations and for more information
about my work and other UCSD Computer Systems Lab work.

Incidentally, the CRA implemented described in the paper is an earlier
and less sophisticated version than is described in this FAQ.  We are
working on publishing the upgraded scheme more formally. 

> How is CRA different from turning off UDP checksums?

Just turning off UDP checksums means that all UDP transmissions, no
matter to where in the Internet, are unprotected.  Many WANs (notably
SLIP links) are not protected by CRCs.  So running NFS over a SLIP
link could spell disaster if you just turn off UDP checksums, whereas
CRA would detect this situation and automatically checksum packets.
Additionally, no router is protected by a CRC, and the chance of
corruption within a router goes up exponentially with the number of
hops a packet travels. 

Also, CRA applies to TCP just as much as to UDP.  Note that most data
is transferred across UDP (via NFS) these days.

> I'm looking at the output of netstat -s on my machine and it shows
> that checksum errors have been detected.  Doesn't that mean that CRA
> is a bad 
> idea?

There's no way of telling whether those errors occurred on your local
LAN or on a WAN or in a router.  It's my belief (and my experience
tends to somewhat back this up) that almost all checksum errors crop
up across WANs.

> What about end-to-end ness?  Aren't you sacrificing reliability by
> no longer checking memory-to-memory but rather adapter-to-adapter?

An small amount of reliability within the network subsystem, yes.
Reliability within the system as a whole, no.  Nobody checksums disk
activity or memory copies or program correctness to memory either
except in the most perfunctory way (LFS checksums one word per block,
probably less protection than the IP header checksum).  In a system
that is subject to errors from many influences, does it make sense to
demand complete reliability from the network subsystem, especially at
such a high cost as is imposed by data checksumming?

> What about whole bad designs of adapters like the DEQNA, that were a
> major factor in the decision to always checksum?

The new version will have statistical checksumming: checksumming every
tenth or twentieth packet.  The hope is that it will catch bad
adapters relatively quickly.  If too many errors are seen, the
implementation will print an error message and turn off CRA on that
interface.

This may result in higher reliabilities, since it would make bad
adapters easier to detect.

Remember that your SCSI interface can (and sometimes does) betray you
without any sign for years.  Yet you haven't yet raised a lynch mob to
hang the guy who wrote your filesystem.

> How do you tell that a host is on the same LAN?  What about bridges?

Whenever a host with this extension answers an ARP request, it also
transmits an IEEE spanning tree packet configured to look innocuous
(highest priority numbers, etc.).  Neither routers nor bridges will
forward such a packet.  Routers will simply ignore it (it isn't
IP/IPX/etc.), and bridges are compelled to interpret without
forwarding spanning tree packets.  If a host receives such a packet,
it can be sure that the sending host is on the same LAN segment.

This is different from the earlier version of the scheme and the
implementations, which both depend entirely upon IP routing
information.  The old scheme was usually good enough, but this new
scheme is safer. 

The new scheme still checks IP routing information so that
checksumming will be done in the case of multiple logical subnets on a
single piece of wire.  Sometimes packets travel through a router even
though they are on the same physical network.

> What about proxy ARP?  Can CRA detect that and "do the right thing"?

Yes.  It doesn't matter to CRA whether IP knows that a packet is
routed.  It only avoids checksumming if the host is actually on
the same piece of wire.

> What about repeaters?

The LAN CRC covers repeaters.  Repeaters don't recalculate the CRC, so
if any changes happen across the repeater, they show up as a CRC error
at the destination's adapter.

One problem with the earlier version of CRA is that multimedia bridges
do generally recalculate CRCs because of the change in header and lack
of an obvious associative property or subtraction operator in CRC
calculation.

> In a world where you can buy interfaces that do Internet
> checksumming, why do we need this scheme?

CRA is basically free.  No extra hardware required.  I have yet to see
a CHEAP interface that implemented Internet checksumming.  Admittedly,
it's easier than it might be for FDDI because those kinds of networks
so far seem to require an onboard processor for SMT basics and the
like.  However, SGI's checksumming FDDI board, for example, requires
an AMD 32000 32-bit RISC processor while DEC's FDDI board is able to
make do with the 16-bit MC68000.

> What mechanisms do you use to negotiate nonuse of checksums?

UDP already has a mechanism for this - just have the sender put a zero
in the checksum field.  We have the sender of UDP packets decide
whether CRA should be used and use that protocol.

For TCP we use Zweig and Partridge's TCP Alternate Checksum Option
(RFC1146).  The IANA has kindly granted us alternate checksum 3 to
be no checksum at all.

> Whose idea was it?

A number of people seem to have had the idea over the course of last
decade or so.  It's not the world's subtlest idea.  However, as far as
we know, Joe Pasquale and I are the first to actually implement it and
write it up and pursue it.

-----------[000251][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 06:14:34 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: 255.255.255.25 via point-to-point link ?

In article <zok.766151581@popp.ins.de> zok@popp.ins.de (Andreas Frackowiak)
writes: 
    Is it "legal" to send an ip-packet to the destination
    address "255.255.255.255" via a point-to-point link or
    it is forbidden in a RFC ?
    
Yes, that's legal.

Tony

-----------[000252][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 06:19:37 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: RFC 950 is really ignored (was: subnetting)

In article <2offni$sss@bedrock.cs.umd.edu> reh@cs.umd.edu (Richard
Huddleston) writes: 

    You wouldn't have any of the draft RFC numbers handy, would you?  I'd be
    quite interested to see how the problem is going to be handled where a 
    machine is determining which MAC address to use (i.e., is the destination
    IP address on the local subnet, or do I send this packet to the gateway?)
    when constructing to destination address of the frame.
    
Draft RFCs don't get numbers.  It won't look too different than today.
Basically a host needs to learn that there are a particular set of
addresses on the local cable.  These need not be contiguous, but for ease
of description, we probably want them to continue to be a power of two
range.  Of course, this doesn't preclude lists of host-specific-addresses.

Those in the list you ARP for.  Otherwise you follow your nose...

Tony

-----------[000253][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 94 07:31:18 GMT
From:      pb@idca.tds.philips.nl (Peter Brouwer)
To:        comp.protocols.tcp-ip
Subject:   Question usage of caching nameserver by other clients?

I have a problem with machines using NIS. Some architectures ( Solaris, HP-UX )
wants the nameserver and the NIS master to be the same machine.
In a large net with different NIS domains and a few central nameservers this is
a problem. As a quick fix I distribute the host file in the NIS domain. This is
not e real solution. I want to try to configure the NIS master as a caching
nameserver. Can clients use a caching nameserver as source for resolving names?
If so this would solve my problem.


--
Regards, Peter Brouwer                 /\_/\
pb@idca.tds.philips.nl                 (0 0)
-----------------------------------oOO--(_)--OOo----------------------------
# Digital Equipment Enterprise,        / WorkGroup Systems, Office Product Set
# DIGITAL : HLDEO1::BROUWER_P,829-4218 \ Dep Office Product Sets, P.O.Box 245,
# PHONE:[+31][0]55 43 ext 4218,fax 9133  7300AE Apeldoorn, The Netherlands.

-----------[000254][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 94 16:07:41 PDT
From:      Bruce White <bwwhite@mmm.com>
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.os.ms-windows.networking.windows,comp.os.ms-windows.networking.tcp-ip
Subject:   MS DOS TCPIP & WFWG & ...


Hel(p)lo:

I have been experimenting running Chameleon 4.0, MS TCPIP VxD 32 bit beta, and 
MS DOS TCPIP stack with WFWG 3.11.  I have tried all 3 with WFWG and have 
problems with each.

Chameleon lets me mount a network drive that is on a UNIX box, use Mosaic, 
NewtNEWS (or other news reader), run PC Eudora for Windows, and run eXceed X 
windows sessions.  However, there ain't no way I can figure out how to hook up 
to an HP 4si MX printer with a JetDirect card (configured for multiple network 
support--TCPIP and Ethertalk).  This printer hangs off UNIX machine that runs 
an LPD daemon (I hope I got all this terminology correct).  There is no LPD 
daemon running on the printer itself.

Microsoft's 32bit beta TCPIP makes some things nice and fast, but I seem to get 
too many random errors.  I also can't mount the network drive I want nor can I 
run an eXceed X windows session.  I didn't try printing.

Right now I'm running Microsoft's DOS TCPIP stack.  It works fine.  I can mount 
the drive I want and print to the printer I want.  But now PC Eudora won't run. 
 Nor can I use Netmanage's telnet program and whenever I exit their ftp 
program, Dr. Watson pops up with an error even though I got no GPF screen.  I 
can't run and X window with eXceed.  Mosaic (2.02a) occasionally freezes up.

I tried the new 32 bit Mosaic with both Microsoft products and it was slow and 
kludgy.  If I tried to adjust the window size, like as not, it would freeze up 
on me.

With all these combinations, I still get random reboots of my computer.  If I 
start WFWG with the /n switch to turn off network services, these random 
reboots go away almost completely (possibly completely, I can't remember).

Hardware:	Northgate Elegance 486-25, #9GXE level 12 ISA video card, 3C509 
ethernet card, Quantum 525 AT hard disk, CMS IDE controller, NEC MultiSync 5FGp 
monitor.

relevant Software:	DOS 6.2, WFWG 3.11, EMM386, smartdrv.

I can email you config.sys, autoexec.bat, system.ini, protocol.ini if you want 
to take a look at any of them.

Thanks for reading this and thanks in advance for any tips passed my way.  I've 
already been helped a lot by these news groups.  I'll pass along whatever 
helps.

BWW

-----------[000255][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 09:25:00 GMT
From:      gnb@bby.com.au (Gregory Bond)
To:        comp.protocols.tcp-ip
Subject:   What's the current SotA for reliable multicast/broadcast?

I'm looking into doing reliable multicast or broadcast for a project.
When I last looked into this (3-4 years ago) there was no definite
answer, but a number of research proposals.  Has the state of the art
changed in that time?

Any refs, pointers to commercial products (esp under Win NT), etc
gladly accepted.

Greg.

--
Gregory Bond <gnb@bby.com.au> Burdett Buckeridge & Young Ltd Melbourne Australia

Atilla The Hun's Maxim: If you're going to rape, pillage and burn, be sure to 
do things in that order.  -- P. J. Plauger, Programming On Purpose, p147


-----------[000256][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 19:36:02 -0500
From:      jgreco@brasil.moneng.mei.com (Joe Greco)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In comp.protocols.tcp-ip article <Co5CBD.1x5@ritz.mordor.com>, tbyrnes@ritz.mordor.com (Tim Byrnes) wrote:
:Accroding to archie it can be gotten from:
:
:dorm.rutgers.edu as
:   /pub/msdos/wattcp/delft/gobbler.zip
:
:or:
:
:ftp.wustl.edu as
:  /systems/ibmpc/umich.edu/communications/wattcp/delft/gobbler.zip

following up on comments at dorm.rutgers.edu led me to the home site:
dutepp0.et.tudelft.nl.

A comment there reads:

pub/Fergie        -->     dos SNMP ethernet Monitor and dos ethernet 
frame grabber (previously called Gobbler and Beholder)

which I am currently ftp'ing to examine...

(I'm looking for something to replace the "netwatch" program I currently
use)

Enjoy,

... Joe

-------------------------------------------------------------------------------
Joe Greco - The Data Capture Fellow (and UNIX/Network Hacker)      414/362-3617
Marquette Electronics, Inc. - Milwaukee, WI        jgreco@brasil.moneng.mei.com

-----------[000257][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 18:56:16
From:      jfoley@vt.edu (Joe Foley)
To:        comp.protocols.tcp-ip
Subject:   Finger client for windows

i am looking for a windows winsock compatible finger client that will let me 
keep a list of frequently fingered parties and machines

if anyone knows of suck a peice of software i would be gratefull if you would 
tell me where i can find it

joe foley

jfoley@vt.edu

-----------[000258][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 15:23:58 GMT
From:      nelson@crynwr.crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

In article <hunen.101.0@brc.medtronic.com> hunen@brc.medtronic.com (Roger Hunen) writes:
   In article <2od57oINNkm7@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
   #In article <NELSON.94Apr11140729@crynwr.crynwr.com> nelson@crynwr.crynwr.com (Russell Nelson) writes:
   #>Why would a TCP connection disconnect just because it got an ICMP
   #>unreachable?  The only time a TCP connection should listen to ICMP
   #>unreachable is when it's SYN is still un ACK'ed.
   #
   #They shouldn't, but many do.  Why?  Because that's what the BSD
   #implementors did.

   I couldn't find a phrase restricting response on ICMP Port Unreachable to 
   connection startup. ICMP Host Unreachables are different though: they are 
   soft errors. Implementations that abort a connection on a Host Unreachable, 
   should be considered broken.

Ahhh, I see.  Yes, Frank didn't specify which unreachable he was
worrying about, and I forgot that there are multiple types.  It seems,
then, that he has a definite worry.

--
-russ <nelson@crynwr.com>      ftp.msen.com:pub/vendor/crynwr/crynwr.wav
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)    | Quakers do it in the light
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000259][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 15:44:23 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Hardware (yes, Hardware)

In article <2odtoa$j5r@rockall.cc.strath.ac.uk> d.w.stevenson@strath.ac.uk writes:
>Has anybody implemented TCP as a hardware engine, in order to acheive very high throughputs?

(I assume a workstation or router doesn't count as hardware for this question.)

The closest I can think of was the "TCP-on-a-Chip" that Steve Holmgren did several
years ago, using an (Rockwell 6502 variant?) integrated single-chip CPU with RAM
and EPROM.  But that device only supported one or two connections and I don't recall
it really getting onto the market.  TCP (and presumably IP) would be a very complex
implementation for a pure VLSI implementation.  It would still need lots of RAM for
state and buffers.  And TCP and IP seem to continue to evolve, so casting them in
silicon may not be the wisest decision.

The late company Protocol Engines started off trying to build a silicon processor
for XTP.  Later they shifted toward building a specialized RISC engine for more
general protocol processing, but I don't recall silicon ever being cast (certainly
not in production).

Art


-----------[000260][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 94 15:48:21 GMT
From:      don@mars.dgrc.doc.ca (Donald McLachlan)
To:        comp.protocols.tcp-ip,comp.unix.programmer,comp.client-server
Subject:   problems programming tcp-ip server-application on Sun Unix

Hi all,

I am trying to write a small client server application on a Sun Sparcstation
using TCP/IP.

I want to set up a small server application which (at this point) simply
waits for a connection, reads data from the socket, writes the data back
to the socket, and closes the connection.  I have chosen to use fdopen
on the socket.



The first wrinkle is:

I use telnet to telnet to the TCP socket the server app is waiting at
(telnet host 2666).  If I comment out the code that echoes the data back,
all is as I expect.  That is telnet echoes back what I type in, and then
the connection gets shut down by the remote host.

The problem occurs when I have the server app echo data back by doing
an fputs to the stream I fdopen'ed on the socket.  In this case, I see
3 copies of the text I typed appear in the telnet session (1 echoed by
telnet, and 2 caused by the fputs).

Can this be reduced to 2 copies (with seeks or fflushes)?



The second problem is:

Even though I do explicit closes on the sockets, the server app cannot
be re-run (bind fails) until the previous connection times out.



The code follows below my signature.  Any help will be appreciated.

Thanks,
Don


---
Donald McLachlan			e-mail	donald.mclachlan@crc.doc.ca
Communications Research Centre / DRL	office	613-998-2845
3701 Carling Ave.,			lab	613-998-2423
Ottawa, Ontario				fax	613-990-7987
K2H 8S2


/*
	ivox_shout.c
*/

#include <stdio.h>
#include <sys/types.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netdb.h>
#include <string.h>

extern void exit();

#define SHOUT_PORT	2666
#define BUFLEN		80

main(argc, argv)
int argc;
char *argv[];
{
	char hostname[80];
	int sock;
	struct hostent *hp;
	struct sockaddr_in sin;

	if(argc > 2)					/* verify usage */
	{
		(void) fprintf(stderr,"usage: %s [timehost (client mode)]\n", argv[0]);
		exit(0);
	}

	if(gethostname(hostname, 80) != 0)		/* get local name */
	{
		perror("gethostname");
		exit(1);
	}

	if ((hp = gethostbyname(hostname)) == NULL)	/* get local addr */
	{
		perror("gethostbyname");
		exit(2);
	}

	if ((sock=socket(AF_INET, SOCK_STREAM, 0)) < 0)	/* connect to socket */
	{
		perror("socket");
		exit(3);
	}

	bzero((char *)&sin, (int)sizeof(sin));		/* setup address info */
	bcopy(hp->h_addr, (char *)&sin.sin_addr, hp->h_length);
	sin.sin_family = AF_INET;
	sin.sin_port = SHOUT_PORT;

	server(sock, &sin);				/* get request */
/*	
	if(shutdown(sock, 2) != 0)
		perror("shutdown(sock)");
*/
	if(close(sock) != 0)
		perror("close(sock)");

	return(0);
}

server(sock, sin)
int sock;
struct sockaddr_in *sin;
{
	char buf[BUFLEN];
	int sock2;
	FILE *sockfp;
							/* bind to socket */
	if(bind(sock, (struct sockaddr *)sin, sizeof(*sin)) < 0)
	{
		perror("bind");
		exit(20);
	}

	if(listen(sock, 1) < 0)
	{
		perror("server: listen()");
		exit(20);
	}

	if((sock2 = accept(sock, (struct sockaddr *)0, (int *)0)) == -1)
	{
		perror("server: accept()");
		exit(20);
	}

	sockfp = fdopen(sock2, "r+");
	if(sockfp == 0)
	{
		perror("fdopen()");
		exit(20);
	}

	fgets(buf, BUFLEN, sockfp);
	if(ferror(sockfp))
	{
		perror("fgets(sockfp)");
		clearerr(sockfp);
	}

	fputs(buf, stdout);

	fputs(buf, sockfp);
	if(ferror(sockfp))
	{
		perror("fputs(sockfp)");
		clearerr(sockfp);
	}

	fflush(sockfp);
/*
	if(shutdown(sock2, 2) != 0)
		perror("shutdown(sock2)");
*/
	if(close(sock2) != 0)
		perror("close(sock2)");
	
	if(fclose(sockfp) != 0)
		perror("fclose(sockfp)");
}



-----------[000261][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 16:00:36 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.lang.c,comp.unix.questions,comp.protocols.tcp-ip
Subject:   Re: QUESTION: clientless TELNET

In article <2oe1nj$22c@erinews.ericsson.se> etxmesa@eos.ericsson.se (Michael Salmon) writes:
]In article <2odfpv$ruc@age.cs.columbia.edu>
]radev@news.cs.columbia.edu (Dragomir R. Radev) writes:
]|> I am looking for a reference on how to write a clientless TELNET
]|> server. I want users to type "telnet xxx.yyy.zzz pppp" and
]|> connect to my server which is waiting on port pppp for connections.
]|> The reason I don't want clients is that I want to create a service
]|> which people should want to use without having to compile
]|> clients, etc.
]
]You have a client though, telnet. "Networking with TCP/IP" Volume III
]by Comer and Stevens shows some servers suitable for use with telnet,
]the only thing that I think you'll need to worry about is that you
]can't use character 255 without duplicating it.

The server also has to be prepared to handle the TELNET negotiations and
other special characters that the telnet program might send.  And it should
translate newline and carriage return into CR/LF and CR/NUL, to conform to
the TELNET specification.  See the TELNET RFC for details.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000262][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 16:16:30 GMT
From:      jayoung1@eos.ncsu.edu (JASON ADAM YOUNG)
To:        comp.protocols.tcp-ip
Subject:   Blocking Part of a Net


Background:

We currently have two seperate LANs within our building,  One is 
connected with the rest of the company WAN.  The other has been 
separated from the "main" LAN because we have a dedicated 56k line 
to one subcontractor and 3 dialup lines to other subKs and 
non-company machines cannot have access to corporate WAN (by a rather 
valid and very definitely set-in-stone corporate policy). The company 
WAN is behind an Internet FireWall, we can see out but no one else 
can see in (I'm not real familiar with firewalls, as the direction for 
that is outside our location, so if anyone can explain that one to me, 
I would be grateful)  As is usual and eventual with this situation LAN A 
needs to communicate with LAN B.

The Question:

What, if anything, can we do to keep the subKs on LAN B, connect LAN A to
LAN B, and keep them from seeing anything that they should not be seeing
(LAN A + WAN)  We need a method as secure as a FireWall, If there is
a solution what are the ballpark costs?  Product references, book 
references, and Vendor options are welcome.  

Thanks in advance,
Jason
-- 
#      Jason Adam Young               jayoung1@eos.ncsu.edu                 #
#                                     Jason.Young@gtegsc3.sprint.com        #
#      "If we weren't all crazy, we would go insane" -- Jimmy Buffett       #
#          Copyright 1994 Dark Side of Margaritaville Productions           #

-----------[000263][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 17:42:09 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   FAQ for this group?

Hi Folks,

	I've noticed over the past few weeks that we really need a FAQ
here.  Is there one that I'm not aware of?  If not I'm more than willing
to start one and post it monthly.

Later,
George



-- 
gnn@netcom.com

If people were more concerned with being reconciled than with being right, 
the world would be a better place.  -- Miss Manners

-----------[000264][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 94 23:48:33 EST
From:      nyseclug@rlgsc.com
To:        comp.os.vms,comp.osf.misc,comp.protocols.tcp-ip
Subject:   20-April-1994 NYSECLUG Meeting Announcement


                     The Security User Group


                Advances in UNIX System Security
                                &
                    TCP/IP Network Security
 
                   Wednesday, April 20, 1994
                     6:00 p.m. - 8:00 p.m.



                Advances in UNIX System Security
 
Numerous security incidents in recent years have, perhaps unfairly, labeled 
the UNIX system as unsecurable. Almost every UNIX vendor now provides 
systems with rich security features and many third party products are 
available to provide extra controls and alert the security administrator 
when security is breached or weakened. This presentation will review some 
of the basic and advanced security features provided by the UNIX system and 
discuss approaches to check deployed systems to ensure that they are 
installed and configured securely. Use of automated tools to assist in the 
security analysis process will be highlighted. The presentation will also 
review emerging security standards that will drive the security 
capabilities that vendors will be providing in the future.
 
Speaker:              David Bauer
                      Director - Security Product Line
                      Bellcore


                    TCP/IP Network Security
 
Over the past several years, the TCP/IP suite has emerged as the protocols 
of choice for providing enterprise-wide network and internetwork 
connectivity. In addition to interconnecting internal corporate networks 
and systems, TCP/IP also provides a gateway to a huge external world of 
interoperable on-line services. This external connectivity can potentially 
provide enhanced communications with clients, vendors, and telecommuting 
employees, as well as providing connectivity to the Internet. Connection to 
the Internet is becomming a business necessity. The openness of TCP/IP 
carries with it new risks and vulnerabilities. This talk will overview 
TCP/IP and the Internet and explore security issues associated with TCP/IP 
based networks.
 
Speaker:              Jim Giacopelli
                      Director - Secure Communications and Services
                      Bellcore

Location:             Coopers & Lybrand
                      1301 Avenue of the Americas (52nd Street)
                      Training Room 3 (2nd floor)
                      New York, New York

Host:                 Carol A. Siegel

If you plan to attend, please RSVP by Tuesday, April 19 to:

                                 Carol A. Siegel
                                Coopers & Lybrand
                                 +1 212 259 2931
                                nyseclug@rlgsc.com

RSVP is mandatory. Building security will not permit any person not on the 
attendee list to enter the building (This is the Security User Group after 
all!)
                         

-----------[000265][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 20:25:50 GMT
From:      blh@hpuerca.atl.hp.com (Bill Hassell)
To:        comp.sys.hp.hpux,comp.sys.hp.misc,comp.sys.sun.misc,comp.sys.protocols.misc,comp.protocols.tcp-ip
Subject:   Re: Help needed with BOOTP/RARP

Scott Ellentuch (tuc@lsupoz.apana.org.au) wrote:

: 	Sorry to cross post all over the place, but I figured I'd give alot of
: people the chance to hear me.  I'm trying to get a Sun 3/50M to ask for its
: IP address via BOOTP to an HP/9000-G50.  I *THINK* I have the HP set correctly
: to respond to a query (A bootpquery does give me what I want).  However, it
: doesn't seem to do anything when the time comes.  The bootp daemon doesn't
: even fire up.  I ran NETTL on the HP and this is what it sees :
 
: =================================== ETHER ====================================
: Source : 08-00-20-06-4c-bf [I] [Sun               ] TYPE: 0x8035              
: Dest   : ff-ff-ff-ff-ff-ff [B] [                  ] TRACED LEN: 60
: Date   : Sat Apr 02 16:44:33:27396 EST 1994
: ================================ RARP PROBE    (inbound -- [ICS]) ============
: Source: 0.0.0.0       08-00-20-06-4c-bf
: 			Requesting: 0.0.0.0       
 
: But doesn't seem to reply back.  Any help?

  Check /usr/adm/syslog on the hp-ux side.  If you don't see the Sun's
  address listed, either the Sun is not making it thru the network
  (routers and gateways will by default block all bootp requests), or
  bootp service has not been enabled in hp-ux.  For the former, put the
  Sun on the same physical LAN; for the latter, check /etc/services for
  the bootp entry.


--

                  ________________________________
         ________|         Bill Hassell           |________
         \       |     HP Customer Support        |       /
          \      |    blh@hpuerca.atl.hp.com      |      /
          /      |________________________________|      \
         /__________)                          (__________\

-----------[000266][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 22:10:18 GMT
From:      larry@onramp.net ()
To:        comp.protocols.tcp-ip
Subject:   X Windows for System Five Release 3 (SCO UNIX)


Does anybody know of an easy place to download X windows for system
5 release 3 (SCO UNIX)....Preferably both an X windows Client and 
Server package..???

Please Email to:

larry@onramp.net


Thanx,


Larry




-----------[000267][next][prev][last][first]----------------------------------------------------
Date:      Wed, 13 Apr 1994 22:12:13 GMT
From:      rich@kastle.com (Richard Krehbiel)
To:        comp.protocols.tcp-ip
Subject:   Re: What's the current SotA for reliable multicast/broadcast?

In article <GNB.94Apr13192500@baby.bby.com.au> gnb@bby.com.au (Gregory Bond) writes:

>   I'm looking into doing reliable multicast or broadcast for a project.
>   When I last looked into this (3-4 years ago) there was no definite
>   answer, but a number of research proposals.  Has the state of the art
>   changed in that time?
>   
>   Any refs, pointers to commercial products (esp under Win NT), etc
>   gladly accepted.

I'd also be very interested in any info about a reliable multicast
protocol.

We were developing a decentralized multicast protocol of our own,
based on udp, but have stopped work on it.  One reason is that we need
to support Windows 3.1 nodes, and Win just can't run the protocol we
designed because it can't meet the timing constraints.  The other
reason is that we're afraid that multiple-access network media like
Ethernet will be supplanted in the future by ATM and other protocols
where specific circuits must be established between points anyway, and
you can't hear needed traffic by just "listening".

Instead we're going to spec that systems must include a centralized
message server.
--
Richard Krehbiel          rich@kastle.com or richk@netcom.com
Nothing witty comes to mind...

-----------[000268][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 22:57:23 GMT
From:      kerch@parc.xerox.com (Berry Kercheval)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP Hardware (yes, Hardware)

art@acc.com (Art Berggreen) writes:

>The late company Protocol Engines started off trying to build a
>silicon processor for XTP.  Later they shifted toward building a
>specialized RISC engine for more general protocol processing, but I
>don't recall silicon ever being cast (certainly >not in production).

Protocol Engines silicon was never cast (I worked there until the very end...) 

Excelan did a "TCP/IP on a board" product once, too.  THe folks who
did that went on to found Kinetics...

  --berry
--

Berry Kercheval :: kerch@parc.xerox.com 
"...start with Plan 9, which is free of sin..." -Mark V. Shaney


-----------[000269][next][prev][last][first]----------------------------------------------------
Date:      13 Apr 1994 10:47:49 +0800
From:      ngps@nova.np.ac.sg (Ng Pheng Siong)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Software-based network monitor?

In article <Co3z7t.9GE@ranger.daytonoh.ncr.com>,
 <chris.resch@daytonoh.ncr.com> wrote:
>In article <2obsdf$gtr@marr.ecs.soton.ac.uk> tjc@ecs.soton.ac.uk (Tim Chown) writes:
>>An alternative is the PD sniffer for the PC called gopher, which is from
>>a University in Holland.  Good enough to see various packet types, filter 
>>them, look at them, show bar charts of this and that.  Very nice.
>
>Where can I get this gopher sniffer software?

Mayhaps he meant Beholder? Beholder is for Dos. Btng is for OS/2 and
Unix. Tricklet is SNMP addon for Perl. Cool stuff. Makefiles are a 
mess, tho.

Site is dnpap.et.tudelft.nl

- PS
-- 
Ng Pheng Siong * ngps@np.ac.sg * ngps@technet.sg
Computer Centre, Ngee Ann Polytechnic, Singapore

-----------[000270][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 02:14:58 GMT
From:      tclark@med.unc.edu (Thomas B. Clark III)
To:        comp.protocols.tcp-ip,comp.periphs
Subject:   trouble with SLIP and US Robotics 14.4 modem

I just "upgraded" my Gateway Telepath to a Telepath II--
which, according to Gateway, is a vanilla US Robotics
14.4/14.4 internal modem.  Unfortunately, it won't do SLIP.
I have (I think) configured it properly for hardware-only
flow control.  It runs through the init script just fine,
but once SLIP starts the modem transmits no more data.

I talked to Gateway's tech support (most of which was spent
trying to explain to them what SLIP is and why you can't type
directly into it) but did get one interesting tidbit--they
said they had heard that this modem "doesn't like" hardware
flow control.

I'd be interested to hear if anyone else has had trouble with
this modem--and, in particular, running SLIP.

-----------[000271][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 14:28:48 -0700
From:      phil@lykos.netpart.com (Phil Trubey)
To:        comp.protocols.tcp-ip
Subject:   Rough max # of TCPs on Unix?

I am trying to ballpark how many simultaneous active TCP connections
typical Unix boxes can handle.  I am looking at an application that
would require say 500-1000 active TCP connections - each of which 
would be busy sending/receiving data at a fairly low
average data rate (say 1 500 byte packet/second).  

I have no feel for just how large you can scale such an app on typical
to large Unix boxes.  Some possibilities for metrics - anyone know
how many active NNTP connections the big Internet USENET sites typically
carry?  Or how many active Telnet connections large Unix timesharing
systems handle?

Thanks for any info!
______________________________________________________________________

 Phil Trubey                 | 
 NetPartners                 |
                             | Providing independent consulting in the    
 E-mail: phil@netpart.com    |   application of Internet technology        
 Phone:  714-759-1641        |                                             
 Fax:    714-644-0577        |
______________________________________________________________________
-- 
 

-----------[000272][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 1994 03:01:23 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: What's the current SotA for reliable multicast/broadcast?

In article <RICH.94Apr13171213@rich.kastle.com> rich@kastle.com (Richard Krehbiel) writes:

> ...
>We were developing a decentralized multicast protocol of our own,
>based on udp, but have stopped work on it.  ...
 
>                                                  ...  The other
>reason is that we're afraid that multiple-access network media like
>Ethernet will be supplanted in the future by ATM and other protocols
>where specific circuits must be established between points anyway, and
>you can't hear needed traffic by just "listening".
> ...

That reason seems bogus to me.  One of the big controversies in ATM
standards groups is how to support multicast.  Not "whether to support
multicast," and not really "how", but "which way."

Look at multicast as a special case of broadcasting.  There are simply
too many applications that absolutely must have at least broadcast
facilities.  Once you accept broadcast, it's a modest step to do it right
and have multicast.

IP multicast is here to stay.


Vernon Schryver    vjs@rhyolite.com

-----------[000273][next][prev][last][first]----------------------------------------------------
Date:      14 APR 94 08:56:13 EST
From:      ciarfella@took.enet.dec.com (Paul Ciarfella)
To:        comp.protocols.tcp-ip
Subject:   Re: What's the current SotA for reliable multicast/broadcast?


In article <GNB.94Apr13192500@baby.bby.com.au>, gnb@bby.com.au (Gregory Bond) writes...
>I'm looking into doing reliable multicast or broadcast for a project.
>When I last looked into this (3-4 years ago) there was no definite
>answer, but a number of research proposals.  Has the state of the art
>changed in that time?
> 
>Any refs, pointers to commercial products (esp under Win NT), etc
>gladly accepted.
> 

As you said, there is plenty of ongoing research into the areas of
reliable multicast/broadcast and process group paradigms: Totem,
at the University of California, Santa Barbara; Transis, at the Hebrew
University of Jerusalem, and a myriad of others.

ISIS Distributed Systems (see comp.sys.isis) is the best-known commercially
available package for building reliable multicast and process group -based
distributed applications.

Paul C

    ----------------------------------------------------------------------  
    Paul Ciarfella                              ciarfella@took.lkg.dec.com
    Mobile & Wireless Networks			508 486-7712
    Digital Equipment Corporation               Littleton, MA   USA
    ----------------------------------------------------------------------  




-----------[000274][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 05:24:14 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: need your help!!!

In article <1994Apr5.210214.16189@newstand.syr.edu> sdong@scorpion.syr.edu (Sheng  Dong) writes:
>I am using the socketpair to setup two stream pipes. 
>Can anyone kindly tell me how to bind in this condition. I did it this way,but
>every time got error message for can not binding.

You can only bind a socket that isn't connected yet.  Socketpair creates
two sockets that are already connected to each other, so there's no need to
bind them.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000275][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 94 17:11:00 -0400
From:      john.childs@canrem.com (John Childs)
To:        comp.protocols.tcp-ip
Subject:   Winsock under Win-OS/2

Has anybody gotten Win-OS/2 to recognize winsock.  If so who's TCP/IP
product were you using?

John Childs

---
 * KWQ/2 1.2e * Two most common elements in the universe: Hydrogen & Stupidity.

-----------[000276][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 94 20:53:00 -0500
From:      Allan Eayrs <aeayrs@delphi.com>
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

Aaron Leonard <leonard@telcom.arizona.edu> writes:
 
>|I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
>|where I can find such a thing?  It would be EXTREMELY helpful to me right
>|now.
 
I'm running into the same situation.  If the software is written for a DOS or
DOS-able environment, I could sure use it.
 
Allan Eayrs
MKRail Corporation
Boise ID

-----------[000277][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 13:00:23 GMT
From:      comrade@uniwa.uwa.edu.au (Peter Cooper)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP redirector for unix

I remember reading about a driver written for some variety of unix that
permitted redirection of IP datagrams to a user process.  The impression
I had, was that the device was ifconfig'd and a route was added to the
routing tables, somewhat like the situation with SL/IP.

Something as nifty as this would be really useful to test how complete
and correct an implementation of something with pretty minimal IP
functionality was, without having to burn LOTS of EPROMs.

If anyone can direct me to the sources, I'd really appreciate it - it'd
be nice if the sources were written with a BSD-derived unix in mind, but
I'll try anything... in preference to the appalling alternative.

Peter
--
Peter Cooper (comrade@gu.uwa.edu.au)           Computer geeks give better head.

-----------[000278][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 1994 18:17:20
From:      jfisher@eagle.ndhm.gtegsc.com (John Fisher)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP FTP over X.25?

Hi,

I have a friend at work that needs to test out an x.25 interface in a couple 
Sparkstations.  He wants to connect 2 Sun spark-stations together with a 
standard RS232 cable running X.25  Then he wants to run TCP/IP on top of it 
and FTP files between the machines.  Is this possible??  I don't know anything 
about x.25 but I think IP can be routed over it, right??  

If it's possible does he have to purchase any software or get freeware to 
setup the x.25 stuff?

Any help or suggestions would be great.  Thanks in advance.

-John Fisher
jfisher@eagle.ndhm.gtegsc.com

-----------[000279][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 1994 13:43:17 GMT
From:      andrew@gmvt4.concordia.ca (Andrew Francis)
To:        comp.client-server,comp.protocols.tcp-ip,comp.protocols.nfs
Subject:   Re: Stateless Servers and transport protocols

In article <1994Apr12.220351.9774@msus1.msus.edu>
manohar@msus1.msus.edu writes:

>	I'm working on issues dealing with Statelessness in Servers, and from
>what I've read, it seems that the main motivation for statelessness in
>Client-server systems is protocol reliability.

   Actually the main reason one wants a stateless server is that it simplifies
the construction of transaction processors that has ACID properties (atomicity,
consistency, isolation, durablity). In the case of NFS, the designers felt
that it is easy to implement crash recovery, if the server does not need to 
have memory of prior transactions in order to implement the current transact-
ion (in short no state). 

>Statelessness simplifies application protocol design as they need not bother
>about messages being lost, delayed, duplicated etc. by the underlying 
>network or system crashes server or client or network failures. That seems
>to be>the reason why Sun's NFS also tries to have its server as stateless
>as possible. A protocol like TCP would however guarantee reliability and
>would serve to shield the application protocol from network unreliability
>issues even with stateful servers.  

   I believe that Sun picked UDP because of its low overhead in comparsion
to TCP. Since the client retransmits requests under it succeeds, and the
transaction model is stateless, and requests sizes are small (also I think
that requests for remote disk blocks appear like page faults to the client
processor), TCP does not add much functionality, since memory of the network
connection creates overhead and adds nothing to the transaction processing
robustness.  
   However there are good explanations about NFS' design in the O'Reilly
"NFS and NIS" book, and I believe, Sun's "Art of Distributed Applications."
Also consult the SunOS references for the real scoop.

--Andrew
-- 

-----------[000280][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 94 13:46:57 GMT
From:      chrisc@fir.canberra.edu.au (Chris Chlap)
To:        comp.protocols.tcp-ip
Subject:   Re: Need sample TCP/IP code

In <Cnr7vI.F5D@ennews.eas.asu.edu> parikh@enws237.eas.asu.edu (Ripple Parikh) writes:


>I need sample TCP/IP code.
 
>Is there a FTP site from where I can download the code?

An excellent book on this is Comer and Stevens "Internetworking with TCP/IP
Vol. 2" published by Prentice-Hall. The code in the book is available
from:

SUN Version: gwen.cs.purdue.edu in /pub/Xinu
PC Version: csc.canberra.edu.au in /pub/ise/xinu

The above site also contains recent manual pages.

Chris Chlap
University of Canberra, Australia

-----------[000281][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 94 13:53:26 GMT
From:      chrisc@fir.canberra.edu.au (Chris Chlap)
To:        comp.protocols.tcp-ip
Subject:   Re: BOOTP

In <2ns7ih$ntd@kalpana.kalpana.COM> mansonw@kalpana.com (Manson Wong) writes:

>I need to build a simple BOOT network with a PC based boot server to download boot image. BOOTP and TFTP will be used to download the boot image from the server to a booted target. Does anyone know where I can get or buy the server side software that support BOOTP and TFTP.

If you are a programmer, you might like to try PC Xinu 7.9. It has BOOTP
support and TFTP support, although you would need to write the server
yourself.
BOOTP sits on top of UDP and would be quite straightforward to write for
an OS like Xinu. You would be able to do alot of other things into the
bargain, like run lpr and similar.

The code is available from csc.canberra.edu.au in /pub/ise/xinu

Chris Chlap
University of Canberra, Australia

-----------[000282][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 94 14:03:21 GMT
From:      chrisc@fir.canberra.edu.au (Chris Chlap)
To:        comp.protocols.tcp-ip
Subject:   Re: Distinct TCP-IP -- Anybody using this?

In <128@lilypond.win.net> conklin@lilypond.win.net (Thomas R. Conklin) writes:

>I wanted to solicit some review from this group of Distinct
>Corp's tcp-ip product.  My organization is considering
>getting pretty involved with this package (winsock.dll) for
>doing some development work (they seem to have some very
>nice libraries for ftp and telnet).  We are also
>considering liscensing their tcp-ip protocol stack for
>Windows. 
 
>Does anyone have some experience with Distinct's stuff
>that they'd be willing to share?  How robust is the
>software?  Are the ftp and telnet libraries workable,
>useful?  Any notable limitations, areas of buggyness?  How
>about Distinct's support?  What are the highlights and
>lowlights of the package?

I used the Distinct Package for a while and was initially quite
impressed with it. It occasionally gave Windows Application Error
Messages and so I put it back in its box and never used it again.

I now use various Shareware Packages (WinFTP, Trumpet, Hgopher, Mosaic) on top
of Trumpet Winsock and have never had any problems with any of them.

I'm sure that Distinct would have provided Customer Support if I would
have taken the time to request it and follow up the problems, but I
think an A$800 software package should run properly without hours of
frustrating problem-searching by the customer.

Chris Chlap
University of Canberra, Australia
  

-----------[000283][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 14:07:35 GMT
From:      prabha@ctron.com (V. Prabhakar)
To:        comp.protocols.tcp-ip
Subject:   Re: Hostroutes

Can a host route be set to a host connected directly on the network? If so,
will it take effect? 

-Prabhakar

NB: I am under the impression the routing table will be consulted only for
routing datagrams that are not directly connected.

-----------[000284][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 1994 15:21:00 GMT
From:      clallo@stsci.edu (Christian Lallo)
To:        comp.protocols.tcp-ip
Subject:   MacTCP and ARA

Hello,

At work, we have several Macs that are hardwired to the internet via
ethernet connection to our ethernet backbone which goes out to the net.
 They are all running MacTCP 2.0.4.  One of those Macs is a Workgroup
Server which runs an Apple Remote Access multiport dial-in connection. 


I have been wanting to dial in from home and get 'real' internet
connection, like a SLIP connection, and I have been told that all I
need to do is to run MacTCP on the remote machine and make sure it has
a unique IP address.  Then when I connect via Remote Access, I will
have a true internet connection and I'll be able to run Gopher, Fetch,
Mosaic, etc. directly from my machine at home.

This sounds to good to be true, so before I go and bother my system
manager with an IP address request, I wanted to know if anyone has
gotten this type of connection working.

Thanks in advance,

Christian Lallo
Technical Director
Special Studies Office
Space Telescope Science Institute
410-516-5441
Internet: clallo@stsci.edu
AppleLink: christianL@applelink.apple.com
AOL: christianL@aol.com
NewtonMail: christian1@online.apple.com

-----------[000285][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 1994 17:17:23 GMT
From:      mre@teal.eng.sun.com (Mike Eisler)
To:        comp.client-server,comp.protocols.tcp-ip,comp.protocols.nfs
Cc:        
Subject:   Re: Stateless Servers and transport protocols

In article <1994Apr12.220351.9774@msus1.msus.edu>,
 <manohar@msus1.msus.edu> wrote:
>	I'm working on issues dealing with Statelessness in Servers, and from
>what I've read, it seems that the main motivation for statelessness in
>Client-server systems is protocol reliability. Statelessness simplifies
>application protocol design as they need not bother about messages being
>lost, delayed, duplicated etc. by the underlying network or system crashes 
       ^^^^^^^  ^^^^^^^^^^

Actually NFS servers are pretty much required to keep
track of duplicate requests for certain non-idempotent operations.
Chet Juscak's paper ``Improving the Performance and Correctness of an
NFS Server'', in the January 1989 USENIX proceedings explains this.
The NFS Version 3 specification also talks about this.

>server or client or network failures. That seems to be
>the reason why Sun's NFS also tries to have its server as stateless as 

The reason why the NFS server was conceived to be as stateless
as it is, is that with state, recovery becomes more complicated. Around
the time NFS was first released, there wasn't even agreement on what
the desired semantics should be in face of network partition,
server crash, or client crash. Compare what System V Release 3's
RFS file sharing system did with crash recovery versus NFS.

>possible. A protocol like TCP would however guarantee
>reliability and would serve to shield the application protocol from
>network unreliability issues even with stateful servers.  

It is possible for a client to send a request to the server, have the
server execute the procedure, and then before the server can respond,
the network partitions. This eventually breaks the TCP connection, and
the client has no choice to re-connect and retry the operation. If the
server is not doing duplicate checkings, a non-idempotent retry, such
as rename(), will fail, because the first one succeeded.

Still TCP does improve reliability, in that fewer requests get lost,
delayed, or duplicated. Initial experiences with the NFS/TCP
implementation here at SunSoft suggests that the hit-rate on the
duplicate request cache is reduced by a factor of between 10 and 100.
-- 
	-Mike Eisler
	mre@Eng.Sun.Com

-----------[000286][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 00:25 MST
From:      gavron@hearts.aces.com (Ehud Gavron)
To:        comp.protocols.tcp-ip
Subject:   Re: Hostroutes

In article <2ojin7INNk8b@dur-news.ctron.com>, prabha@ctron.com (V. Prabhakar) writes...
#Can a host route be set to a host connected directly on the network? If so,
#will it take effect? 

Yes.
Yes.

# 
#-Prabhakar
# 
#NB: I am under the impression the routing table will be consulted only for
#routing datagrams that are not directly connected.

The routing table is *always* consulted.  Packets to destinations that
are not forwarded via gateways are sent to lower layers for MAC resolution.

Ehud

--
Ehud Gavron        (EG76)
gavron@aces.com
Yow!  Am I having fun yet?

-----------[000287][next][prev][last][first]----------------------------------------------------
Date:      14 Apr 94 19:53
From:      hawat@hawat.abg.sub.org (Andreas Kellner)
To:        comp.protocols.tcp-ip
Subject:   Automatic Dial with KA9Q ?

Hello ,

One question: Is it possible to let ka9q automaticaly dial if i try
to establish a connetion over the modem interface ?

Background: I am running ka9q as a router on a dos machine.
The modem is connected to that machine as a interface (sl0).
From another computer (running OS/2 and TCP/IP) which is connected to
the dos machine with ethernet i want to login to my mailbox via tcp/ip.
Now i must first enter the "dial sl0 file.name" manualy on the dos machine
to establish the connection. Is there a way to do this automaticaly if a
packet is routed to sl0 ?


Thanks for help

Andy

--

                   Andreas Kellner (hawat@hawat.abg.sub.org)

                          APL II - Beyond the future                           

-----------[000288][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 08:37:12 -0700
From:      phil@lykos.netpart.com (Phil Trubey)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP FTP over X.25?

In article <jfisher.14.00124A86@eagle.ndhm.gtegsc.com>,
John Fisher <jfisher@eagle.ndhm.gtegsc.com> wrote:
>I have a friend at work that needs to test out an x.25 interface in a couple 
>Sparkstations.  He wants to connect 2 Sun spark-stations together with a 
>standard RS232 cable running X.25  Then he wants to run TCP/IP on top of it 
>and FTP files between the machines.  Is this possible??  I don't know anything 
>about x.25 but I think IP can be routed over it, right??  

Yes.

>If it's possible does he have to purchase any software or get freeware to 
>setup the x.25 stuff?

Sun has a product called SunLink X.25 that does exactly what you want - 
the Sunlink  product can work over the Sun's built in serial ports,
or over an Sbus high speed serial card.  Contact SunExpress or your
favorite Sun marketing org for more info.
-- 
 

-----------[000289][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 11:01:31 -0700
From:      lstowell@pyrnova.mis.pyramid.com (Lon Stowell)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP FTP over X.25?

In article <jfisher.14.00124A86@eagle.ndhm.gtegsc.com> jfisher@eagle.ndhm.gtegsc.com (John Fisher) writes:
>
>Hi,
>
>I have a friend at work that needs to test out an x.25 interface in a couple 
>Sparkstations.  He wants to connect 2 Sun spark-stations together with a 
>standard RS232 cable running X.25  Then he wants to run TCP/IP on top of it 
>and FTP files between the machines.  Is this possible??  I don't know anything 
>about x.25 but I think IP can be routed over it, right??  
>
    Yes.  There is even an RFC for this--which is a model of
    "incompleteness" in my non-humble opinion.  

    However the good folks at Perdue and a few other places have
    published white papers and such on how to do the address mapping
    from X.121 to IP and such. I dunno if there is any public domain
    code to do this, but there is certainly commercially available
    code.  I would be extremely surprised if you couldn't get it from
    your friendly Sun salesman.  

    But if your friend just wants to check out the Sun X.25, why not
    just use their X.25 socket interface?   


-----------[000290][next][prev][last][first]----------------------------------------------------
Date:      Thu, 14 Apr 1994 23:48:52 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Q: ICMP source quench and TCP ?

> Most BSD derived implementations do nothing with Source Quenches.

Not true.  The following is from Net/2, and I see it was in Net/1 and is
still in 4.4BSD.

    /*
     * When a source quench is received, close congestion window
     * to one segment.  We will gradually open it again as we proceed.
     */
    tcp_quench(inp)
            struct inpcb *inp;
    {
            struct tcpcb *tp = intotcpcb(inp);
     
            if (tp)
                    tp->snd_cwnd = tp->t_maxseg;
    }

UDP, however, does indeed ignore source quenches.

	Rich Stevens

-----------[000291][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 00:47:25 GMT
From:      debiso@westfield.win.net (Joe DeBiso)
To:        comp.protocols.tcp-ip
Subject:   Public domain or freeware TCP/IP for SCO

Can anyone point me to a public domain or freeware TCP/IP for SCO
unix??  I would like to put it on my clients machines and use a
SLIP or PPP connection to allow multiple logons on the same phone
line.  Any info would really help!

Thanx!

 

  ___           __      __
 (   >         /  )    /  )
  __/______/> /  / _  /--<  o _   __
 / /  (_) (__/__/_</_/___/_<_/_)_(_)
<_/Joe Debiso, Plantrol Systems, Ltd.
 

-----------[000292][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 00:48:37 GMT
From:      debiso@westfield.win.net (Joe DeBiso)
To:        comp.protocols.tcp-ip
Subject:   SLIP or PPP FAQ

Is there a SLIP or PPP faq?  Are there any good instructions for
setting up a SLIP or PPP connection?

Thanx!


 

  ___           __      __
 (   >         /  )    /  )
  __/______/> /  / _  /--<  o _   __
 / /  (_) (__/__/_</_/___/_<_/_)_(_)
<_/Joe Debiso, Plantrol Systems, Ltd.
 

-----------[000293][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 02:58:40 GMT
From:      dyrkacz@anlchm.chm.anl.gov (Gary Dyrkacz)
To:        comp.protocols.tcp-ip,comp.periphs
Subject:   Re: trouble with SLIP and US Robotics 14.4 modem

In article <2oi8v2$ea2@bigblue.oit.unc.edu> tclark@med.unc.edu (Thomas B. Clark III) writes:
>From: tclark@med.unc.edu (Thomas B. Clark III)
>Subject: trouble with SLIP and US Robotics 14.4 modem
>Date: 14 Apr 1994 02:14:58 GMT
>Keywords: slip modem
 
>I just "upgraded" my Gateway Telepath to a Telepath II--
>which, according to Gateway, is a vanilla US Robotics
>14.4/14.4 internal modem.  Unfortunately, it won't do SLIP.
>I have (I think) configured it properly for hardware-only
>flow control.  It runs through the init script just fine,
>but once SLIP starts the modem transmits no more data.
 
>... Gateway's tech support ...
>said they had heard that this modem "doesn't like" hardware
>flow control.
 
>I'd be interested to hear if anyone else has had trouble with
>this modem--and, in particular, running SLIP.

I am using  the Telepath II, (circa 1993) on my Gateway.   I think its the 
same modem you have.  The modem is running Tattum's TCPman with 
SLIP.   I have hardware flow control set and have had no problems.  Have you 
tried a manual login rather than an auto login?

Gary
Gary Dyrkacz
dyrkacz@anlchm.chm.anl.gov
Argonne National Laboratory
Chemistry Division, 9700 S Cass Ave, Argonne, IL, 60439


-----------[000294][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 04:51:13 GMT
From:      ibottema@alkaid.sce.carleton.ca (Ike Bottema)
To:        comp.protocols.tcp-ip
Subject:   Re: subnetting

mollett@lexmark.com (Vic Mollett) writes:

[binary calculations removed]

>This case is only true if you are concerned about accepting what RFC950
>terms "this network" and "all hosts."  I don't think I have ever seen a
>use of "this network" in any useful piece of software, it tends to suppose
>that a particular resource will have a particular address, which how do you
>determine that if you don't know what class your address is.  "All hosts"
>doesn't regularly get used, either.  Take the class A address of 9.0.0.0
>(I hope IBM doesn't mind if I borrow it for this demonstration) which I know
>has been assigned to various places around the world.  If someone sent a 
>packet to 9.255.255.255, it should, in theory, go to all of the machines in
>the 9.0.0.0 network.  I know for a fact that that will not happen, and I
>don't know of a case where it would even be desired.

OK, so what does happen to a packet sent to 9.255.255.255?  Not that I'd
want to try this in a Class A network, but I can think of applications
(BOOTP in particular) where a broadcast reaching a number of neighbouring
subnets would be useful.  I suspect that "directed broadcasting" is
employed.  Is the result therefore dependant on the router configuration,
the router manufacturer, or both?
>-- 
>                                             /\    Vic Mollett
> These opinions are my own and do not       /  \   Lexmark International, Inc.
> necessarily reflect those of my employer.  \  /   mollett@lexmark.com
>

Ike Bottema
                                             \/  

-----------[000295][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 07:58:14 GMT
From:      smckinty@sunicnc.France.Sun.COM (Steve McKinty - SunConnect ICNC)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP FTP over X.25?

In article <jfisher.14.00124A86@eagle.ndhm.gtegsc.com>, jfisher@eagle.ndhm.gtegsc.com (John Fisher) writes:
> Hi,
>
> I have a friend at work that needs to test out an x.25 interface in a couple
> Sparkstations.  He wants to connect 2 Sun spark-stations together with a
> standard RS232 cable running X.25  Then he wants to run TCP/IP on top of it
> and FTP files between the machines.  Is this possible??  I don't know anything
> about x.25 but I think IP can be routed over it, right??
>
> If it's possible does he have to purchase any software or get freeware to
> setup the x.25 stuff?


This is all perfectly possible with Sun's X.25 products. V7.0 for SunOS4.x
and V8.0 for Solaris 2.x.  Both products contain support for routing IP over
X.25.

I've never heard of any free X.25 implementations for SPARC though, outside
of the NOS-type programs used in the amateur radio world.

Steve



--
Steve McKinty
Sun Microsystems ICNC
38240 Meylan, France
email: smckinty@france.sun.com

-----------[000296][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 21:04:37 -0700
From:      skl@Connectivity.com (Samuel Lam)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for X.25 to Telnet gateway

In article <1994Apr13.141745.488@sfov1.verifone.com>,
 jimmy_t@hnlv4.verifone.com wrote:
>We are looking for an X.25 to Telnet gateway box.
>The only one I've found so far is the XYPLEX.  If you have
>any suggestions for other vendors, please let me know.

Cisco also has X.25/Telnet protocol converter options for their
router and terminal server products.

...Sam
-- 
<skl@Connectivity.com> -- Connectivity Technology Inc.

"Packet Driver, WinSock & TCP/IP CD-ROM" product information:
{gopher,www,ftp}.CDPublishing.com or <info@CDPublishing.com>


-----------[000297][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 10:46:06 GMT
From:      v932584@si.hhs.nl (Hoogsteder, P.M.)
To:        comp.protocols.tcp-ip
Subject:   Collecting traffic volume

I need to measure the amount of tcp/ip traffic over a SLIP connection,
running on a unix-box.  Is there software to do this (linux or Solaris),
or do I need a dedicated router?

Paul.

-------------------------------------------------------------------------------
|"Oh, and Zaphod?"                                         Paul M. Hoogsteder |
|"Er, yeah?"                                                                  |
|"If you ever find you need help again, you know, if       Sector Informatica |
|you're in trouble, need a hand out of a tight corner..."  Haagse Hogeschool  |
|"Yeah?"                                                   The Netherlands    |
|"Please don't hesitate to get lost."                      v932584@si.hhs.nl  |
-------------------------------------------------------------------------------


-----------[000298][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 14:40:05 GMT
From:      jmc@devjam1.ml.com (Network Communications/Performance Consultant)
To:        alt.computer.consultants,comp.protocols.snmp,comp.protocols.tcp-ip
Subject:   SNMP Network Management Tools

I am working on an evaluation of SNMP Network Management Tools.  I was wondering
if anyone else out there has prepared an analysis of these tools. The
Particular SNMP Network tools include HP/OV/Netmetrics/EASE, Concord Trakker, 
Cabletron Spectrum.  I have looked at others such as W&G, but for our needs we need
the combination of remote probes, extensive reporting capabilities, an SQL relational
database for reporting historical data, baselining etc.   It also has to run on 
multiple platforms, SUNs, Windows(NT, 3.1), OS/2, AIX etc.  We are using SAS for some
reports, MOTIF based tools and Windows/DOS based tools to do Network Performance reporting on a very large 20,000+ node plus network.  Its a big job but somebodies got
to do it.  

These are the ones that are mainly in the running.  I could use
an independent outside report if you have one.  I'll share my feature/list/results
with anyone else if they will share their's.  

Thanks in advance for your help. Be good and go get 'em Tiger!

-----------[000299][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 15:39:21 GMT
From:      jjm@cscns.com
To:        comp.protocols.tcp-ip
Subject:   Looking for few large network administrators



My name is John McDermott, and I am writing a course on TCP/IP Internetworking.
I'm looking for a few administrators of large internetworks who would be 
willing to grant me a short interview.  I'm particularly intrested in:
	* The types of problems you regularly experience
	* The tools (RMON boxes, SNIFFER, etc.) you use to solve problems
	* Your largest sources of confusion (my users all want 100Mbps, etc.)
	* Other troubleshooting, implementation or design issues which 
	  confront you.

If you would be willing to be interviewed for a few minutes on these topics
(either with or without attribution), please send e-mail to me at:
	jjm@lrntree.com
Thanks
--john

-----------[000300][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 16:29:28 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: tcp checksum redundancy

In article <65551@sdcc12.ucsd.edu> jkay@cs.ucsd.edu (Jon Kay) writes:

> ...
>> How do you tell that a host is on the same LAN?  What about bridges?
>
>Whenever a host with this extension answers an ARP request, it also
>transmits an IEEE spanning tree packet configured to look innocuous
>(highest priority numbers, etc.).  Neither routers nor bridges will
>forward such a packet.  Routers will simply ignore it (it isn't
>IP/IPX/etc.), and bridges are compelled to interpret without
>forwarding spanning tree packets.  If a host receives such a packet,
>it can be sure that the sending host is on the same LAN segment.

That assumes all bridges support spanning tree.
That assumption is false.


> ...
>       However, SGI's checksumming FDDI board, for example, requires
>an AMD 32000 32-bit RISC processor while DEC's FDDI board is able to
>make do with the 16-bit MC68000.

That is an intentionally false and misleading statement.
"32-bit RISC processor" is content free rhetoric in this context.

There is an AMD 29K on current SGI FDDI boards, but not just for
checksumming.  Those 29K's run at only 16MHz, slower than most PC's.  For
that matter, most 80386's and 80486's are at least as fast for computing
the checksum as a 16Mhz 29K.  Other SGI products with hardware support
for checksumming do not use CPU's to compute the checksum, and run a lot
faster than 100Mbit/sec.

The cost to compute the checksum in hardware is the cost of the functional
equivalent of a single 16-bit end-around-carry adder.


Vernon Schryver    vjs@rhyolite.com

-----------[000301][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 16:52:13 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: Hostroutes

In article <2ojin7INNk8b@dur-news.ctron.com> prabha@ctron.com (V. Prabhakar) writes:
>Can a host route be set to a host connected directly on the network? If so,
>will it take effect? 
>
>-Prabhakar
>
>NB: I am under the impression the routing table will be consulted only for
>routing datagrams that are not directly connected.

This probably depends on the TCP/IP implementation.  I haven't looked at recent
BSD Unix sources, but I think the older systems would allow you to do it.
In those systems, host routes were looked up first, and if no success, network
routes.  After a route was found, it was determined if the destination was
directly connected or via another hop.  But the end result would be the same
using the net route.  So why bother?

Art


-----------[000302][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 17:30:14 GMT
From:      cliffb@skeeter.bdb.com (Cliff Bedore)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

Allan Eayrs (aeayrs@delphi.com) wrote:
: Aaron Leonard <leonard@telcom.arizona.edu> writes:
:  
: >|I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
: >|where I can find such a thing?  It would be EXTREMELY helpful to me right
: >|now.
:  
: I'm running into the same situation.  If the software is written for a DOS or
: DOS-able environment, I could sure use it.
:  
: Allan Eayrs
: MKRail Corporation
: Boise ID


I think I saw a copy of it on ftp.digital.com under something like
/pub/micro/msdos/network


Cliff


-----------[000303][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 94 14:18:00 +0200
From:      hwm@mk.maus.ruhr.de (Hartmut W. Malzahn)
To:        comp.dcom.isdn,iijnet.dcom.isdn,sfnet.lists.tcp-isdn,tnn.dcom.isdn,bit.listserv.ibmtcp-l,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,relcom.tcpip
Subject:   UUCP via ISDN with TCP/IP under MS-DOS

Hello!

I have a very urgent problem at the moment, and I would be glad if one of you
would be able to help me solving it. I usually don't read this newsgroup, so
please reply by email to this posting. I've already browsed through all FAQs and
RFCs I could find that are in any way related to my problem, but I haven't found
anything at all there.

A few days ago, I installed a Teles.S0 ISDN board in my PC to reduce the cost of
my Internet access. The card's CAPI drivers and the Fossil driver work without
any difficulties, and I have been able to test it by calling several ISDN BBSs.
I have been told, however, that in order to use ISDN for my UUCP poll I would
have to establish a TCP/IP connection over ISDN to my host. Now the problem is
that I have no UUCICO program that is able to use a TCP/IP packet driver to send
or receive files. Even though it would be possible to establish an online TCP/IP
connection to my host, I have no idea how I could run an UUCICO poll on the PC
(with in that session. So what I am urgently looking for is either a UUCICO
program for DOS that is capable of transmitting files through a TCP/IP packet
driver, or alternatively an addition to the packet driver that enables me to use
the IP connection as if it were a direct connection to the host.

Maybe there is someone here who also uses a PC with DOS to transfer files via
ISDN with UUCP, and who can give me a hint how to do that!

I'd appreciate any help!

Cheers --- Hartmut
-- 
| Hartmut Malzahn | Internet:   hwm@mk.maus.ruhr.de             |
| Zimmerstrasse 9 |  ... Maus MK BBS in the beautiful Sauerland |
| 58638 Iserlohn  | CompuServe: 100012,1107                     |
| Germany         | FAX:        +49-2371-14491                  |

-----------[000304][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 18:17:54 GMT
From:      nrg@netcom.com (Ethan I. Miller)
To:        comp.protocols.tcp-ip
Subject:   netBSD info please

Greetings!

I'm in the early stages of trying to put my mac on the net. In addition to
running protocols like PPP and SLIP under the MacOS some have suggested
running netBSD *instead* However there seems to be precious little info on
netBSD. So if you have any I would greatly appreciate it. Too, any general
recommendations relating to putting a mac fully on the net, ie: domain name,
mail, news, uucp, etc. would also be greatly appreciated.

take it easy
ethan


-----------[000305][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 94 18:41:20 GMT
From:      stanley@Software.Mitel.COM (Stanley Douglass)
To:        comp.protocols.tcp-ip
Subject:   Limited number of sockets per process

I have established a server process in user space, in a UNIX (SunSoft ICS)
environment. For each remote client attached to this process via
TCP/IP sockets, this server process also opens a connection to a streams
mux to access desired services.

I am only able to attach 25 clients. Above that number, the server socket
simply stops receiving, and the existing 25 clients can also no longer
communicate with the server.

Since I have raise the tunable parameter NOFILES to 100, and each client
uses 2 file descriptors (one for the socket, one for connecting to the
mux), I would expect the maximum to be 48, not 25 clients.

Please email me if you have an idea as to why my limit is low, or if you
know of another tunable parameter which needs adjusting. Many thanks.
( Stan at stanley@Software.Mitel.COM )

ps. Is there any way at all to exceed the 100-limit on NOFILES? -SD

--
Stan Douglass				|   stanley@Software.Mitel.COM
Mitel Corporation, Kanata, Ontario	|    (613)592-2122 Ext.4262
"Death is swallowed up in victory" - Isaiah

-----------[000306][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 19:35:32 GMT
From:      rhaynal@netcom.com (Russ Haynal)
To:        comp.protocols.tcp-ip
Subject:   ?best: bootp vs novix

I am currently helping a 400+ organization get connected to the
internet.  They are running IPX/Novell 3.11.  We are going to install
Chameleon or Super TCp + some shareware (ie Mozaic, etc).
      
As we are trying to keep the installation and maintenance Simple (this is
for 400+ people) We are looking at BootP to dynamically assign IP numbers
and other "set-up" info.  We are also looking at two options called Novix
and catapult - which from what I understand:
          let the PC's continue to use IPX,  the IPX packets are caught
at the server and translated into IP packets for transmission to the
INternet.  Incoming IP pakets are then translated back to IPX and passed
along to the PC.  

has anyone done this sort of implementation?  Do the winsock applications 
work normally even with IPX at the PC connection?
are there any performance issues with the server doing the IPX-IP conversion?

thanks
Russ
rhaynal@netcom.com
russ@navigators.com




-----------[000307][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 19:41:42 GMT
From:      vmatt@netcom.com (Matthew S. Reeves)
To:        comp.protocols.tcp-ip
Subject:   dialup routers help !

I would like to link together workstations in several locations.

Are dialup routers the way to go ? I am using unix with sockets
and RPC calls. I don't want to pay for dedicated lines.

Recommendations of routers to buy are appreciated.

Matt


-----------[000308][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 19:53:18 GMT
From:      pensak@eplrx7.es.duPont.com (Dave Pensak)
To:        comp.protocols.tcp-ip
Subject:   Protocol conversion


I have a suite of messages which are coming into a LAN which is already
set up to receive tcp-ip packets except that these messages are tp4-ip.  I believe
that I can mix both on the same net since they are both ip, but will I
be able to route them on the same net ?  Are there special things that I will have
to do with my routers (Cisco) to make this possible ?

Is there any software out there which will let me convert between tcp
and tp4 (since I have to send the packets through a firewall which only
knows about tcp services) or should I consider encapsulating the tp4
packets inside tcp ?

The tp4 packets contain encrypted data and will be gobbled up by a
decryptor down inside a sub-LAN which has a secure network on the other
side.

Any pointers to papers, books, commercial products, code fragments, etc
will be very much appreciated.

-----------[000309][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 1994 21:04:50 GMT
From:      jeff@astph (Jeff Martin)
To:        comp.protocols.tcp-ip
Subject:   Multiple processes poll()ing on the same socket connection?

What will happen to the socket descriptors in the following situation?

	1. Client and server establish TCP/IP connection.
	2. Server passes this socket descriptor to several slave servers.
	3. Slave servers each poll() on the socket descriptor.
	4. Client process terminates or closes socket connection.

Will all slave servers be informed of an error on that socket descriptor?
Does each slave server have to explicitly close() the descriptor since
the client has closed it or has the descriptor entry already been removed?

Normally when a connection goes down the server recieves a 0 byte message
to indicate this.  However since several slave are all poll()ing on the
same socket connection, really only one can get the message.  Will the
other slave servers recieve a POLLERR after the first slave server closes
the socket connection?

I am also interested to discuss how to effectively coordinate multiple
processes poll()ing on the same socket connections with anybody who has
experience doing this.

Thanks, Jeff
-- 
Jeff Martin, dbms programmer,		Philadelphia Phillies
INET:	astph!jeff@attmail.com		Voice:	(814)234-8592x32
UUCP:	attmail.com!astph!jeff		FAX:	(814)234-1269

-----------[000310][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 21:30:45 GMT
From:      rick@seminole.b30.ingr.com (Rick Hopkins)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   NIS and NIS tech. papers

other tech papers created on NIS or NIS+?  Is there any public domain 
source code to NIS or NIS+ that I could review?

Thank you,

rick hopkins.
rehopkin@ingr.com

-----------[000311][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 22:38:09 GMT
From:      dsiebert@icaen.uiowa.edu (Doug Siebert)
To:        comp.protocols.tcp-ip
Subject:   Re: Rough max # of TCPs on Unix?

phil@lykos.netpart.com (Phil Trubey) writes:

>I am trying to ballpark how many simultaneous active TCP connections
>typical Unix boxes can handle.  I am looking at an application that
>would require say 500-1000 active TCP connections - each of which 
>would be busy sending/receiving data at a fairly low
>average data rate (say 1 500 byte packet/second).  
 
>I have no feel for just how large you can scale such an app on typical
>to large Unix boxes.  Some possibilities for metrics - anyone know
>how many active NNTP connections the big Internet USENET sites typically
>carry?  Or how many active Telnet connections large Unix timesharing
>systems handle?


I can give you one data point -- the ISCA BBS runs on an HP 9000/710 (i.e.
50MHz PA-RISC, about equal to the 715/50 which is HP's current low end
speed wise) and handles ~900 simultaneous users (plus > 300 more sometimes
queued up waiting to get in) and it runs fine.  I've seen incoming/outgoing
packet counts up to 1000/sec for each, the machine handles it fine, chugging
along at load averages of 40+ with response times so fast it generally isn't
possible to tell if you are one of 900 users or the only one.

As for how this compares to other makers, I can't easily guess.  I will say
I generally consider HP's the best out there, others will of course have
their own opinions.  But what you wish to achieve is certainly quite doable,
though with 1000 TCP connections of 1 500K packet per second you'd better not
be *doing* much with them, because the CPU will probably be pretty close to
getting maxed out handling that much traffic, even with the fastest machines
currently available.  And you might need to look into FDDI as well, since you
may get past what Ethernet can handle.


-- 
Douglas Siebert
dsiebert@isca.uiowa.edu

-----------[000312][next][prev][last][first]----------------------------------------------------
Date:      15 Apr 1994 22:55:02 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Protocol conversion

In article <1994Apr15.195318.26043@eplrx7.es.duPont.com>
pensak@eplrx7.es.duPont.com (Dave Pensak) writes: 
    
    I have a suite of messages which are coming into a LAN which is already
    set up to receive tcp-ip packets except that these messages are tp4-ip.
    I believe that I can mix both on the same net since they are both ip,
    but will I be able to route them on the same net ?  Are there special
    things that I will have to do with my routers (Cisco) to make this
    possible ?
    
Nope.  It's an IP packet...  there are some things that you won't be able
to do with TP4/IP, such as TCP port filtering.

    Is there any software out there which will let me convert between tcp
    and tp4 (since I have to send the packets through a firewall which only
    knows about tcp services) or should I consider encapsulating the tp4
    packets inside tcp ?
    
Seems unnecessary.

Tony

-----------[000313][next][prev][last][first]----------------------------------------------------
Date:      Sat, 16 Apr 1994 00:29:05 GMT
From:      dmausner@brauntech.com
To:        comp.protocols.tcp-ip,comp.periphs
Subject:   Re: trouble with SLIP and US Robotics 14.4 modem


In article <2oi8v2$ea2@bigblue.oit.unc.edu>, <tclark@med.unc.edu> writes:
> Originator: tclark@hudson
> 
> I just "upgraded" my Gateway Telepath to a Telepath II--
> which, according to Gateway, is a vanilla US Robotics
> 14.4/14.4 internal modem.  Unfortunately, it won't do SLIP.
> I have (I think) configured it properly for hardware-only
> flow control.  It runs through the init script just fine,
> but once SLIP starts the modem transmits no more data.
> [...]
> I'd be interested to hear if anyone else has had trouble with
> this modem--and, in particular, running SLIP.

i am using gateway2000 box with usrobotics v.32.bis (14.4) internal
modem over slip and clearly it works. it could be a config problem
in your initial AT string of course, or it could be a slip config
problem, or perhaps your client is asking the host the right questions
in order to receive data.

-----------[000314][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 00:50:58 GMT
From:      km@mathcs.emory.edu (Ken Mandelberg)
To:        comp.protocols.tcp-ip
Subject:   Subnetting a Class B

I think I understand how subnetting a Class C works. If you divide 
8 = s + h, you get (2^s -2) subnets with (2^h -2) hosts.

I thought it would work the same with class B, just partitioning 16
instead of 8. However, I've seen a chart that says that

    s=10 h=6 gives 508  subnets with 62 hosts
    s=11 h=5 gives 1524 subnets with 30 hosts.

This apparently comes by viewing a subnet number as a pair (s1,s2)
where s1 is the part from octet 3 and s2 is from octet 4, and requiring
neither s1 nor s2 take on the values 0 or 1.

Is the chart right?

---
Ken Mandelberg      | km@mathcs.emory.edu          PREFERRED
Emory University    | {rutgers,gatech}!emory!km    UUCP 
Dept of Math and CS | km@emory.bitnet              NON-DOMAIN BITNET  
Atlanta, GA 30322   | Phone: Voice (404) 727-7963, FAX 727-5611



-----------[000315][next][prev][last][first]----------------------------------------------------
Date:      Fri, 15 Apr 94 22:30:57 +0300
From:      pzz@tern.msk.su (Alexander E. Pevzner)
To:        comp.protocols.tcp-ip
Subject:   Re: Re: Optimizing TCP/IP Process Architecture

>>In article <1994Apr7.205522.14976@astph> jeff@astph (Jeff Martin) writes:
>>>We are running Interactive UNIX Release 3 Version 3.2.  We are
>>>developing a network database application using TCP/IP.
>>>
>>>HOW CAN WE REDUCE THE NUMBER OF SERVER PROCESSES USING TCP/IP?  Preferably
>>>allowing a tunable server/client ratio?  NOTE: Comer and Stevens suggest
>>>in "Interworking with TCP/IP, Vol 3," that several server processes can
>>>wait at accept() and establish a socket connection for each TCP/IP
>>
>
>	But the best thing to do in this kind of applications is to use
>UDP/IP. Check the source code of Talk. In that the server uses the UDP/IP
>for checking talk requests and then back off by establishing a connection
>between the two clients(TCP/IP). You can use something similar to 
>this if it is possible.

	The model is simple: you have one server process which
listen to tcp port. When it accepts a new incoming connection,
it forks. It's child inhirits the established connection
and can do it's work over it. That's why you always have 
as many server processes as clients plus one. BTW, there is
a standard daemon, inetd, which listens to _all_ standard
tcp service ports and spawns the specific server whith connection
established and ready to work. It's simply configurable
and you can work over it.


				Sincerely, 
				Alexander Pevzner.
				pzz@tern.msk.su

P.S.
	Sorry for bad english.


-----------[000316][next][prev][last][first]----------------------------------------------------
Date:      Sat, 16 Apr 1994 03:31:53 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: netBSD info please

In article <nrgCoBC5u.G3s@netcom.com> nrg@netcom.com (Ethan I. Miller) writes:
>Greetings!
>
>I'm in the early stages of trying to put my mac on the net. In addition to
>running protocols like PPP and SLIP under the MacOS some have suggested
>running netBSD *instead* However there seems to be precious little info on
>netBSD. ...

Look in comp.os.386bsd.announce.
I think one or both of the FreeBSD and NetBSD groups are talking about
68000 ports.


Vernon Schryver    vjs@rhyolite.com

-----------[000317][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 04:45:38 GMT
From:      gray@cac.washington.edu (Terry Gray)
To:        comp.protocols.tcp-ip
Subject:   Re: new improved POP3?

Jeff Macdonald <jeff@zis.ziff.com> wrote:

>About 2 months ago I stumbled upon a site that mentioned a new mail protocol
>was being developed that would improve upon the POP3 protocol. Basically this
>protocol would allow the mail to remain on the host, but the client would be
>able to read new/old mail from any location...

Jeff,
There is a revision of the POP3 specification in the works, but it does
not change the basic POP model, which is to fetch messages from a maildrop
to a (single) client machine.

There is another mail-related protocol that may be what you're looking
for... It's called "IMAP" for "Internet Message Access Protocol".
It is a functional superset of POP, and allows mail on a remote host
to be accessed/manipulated/stored from one or more client machines.

Recent versions of IMAP also allow for selective fetching MIME message
parts, handy if someone sends you a 40MB video attachment and you are
currently using a 9600bps slip link to look at your mail...

RFC-1176 describes "IMAP2", which has been around for quite awhile.  There
is an IETF working group developing its successor, IMAP4.  IMAP4 will also
have support for disconnected operation. See /mail/latest-imap-draft on
ftp.cac.washington.edu for details.

-teg
--
Terry Gray
 University of Washington
  gray@cac.washington.edu

-----------[000318][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 08:38:11 GMT
From:      robert@psy.uq.oz.au (Robert Dal Santo)
To:        comp.dcom.lans.ethernet,comp.dcom.lans,comp.protocols.tcp-ip
Subject:   Bridging/Routing Software on PCs

	I need some cheap routing/bridging solutions. I'm looking at
PCRoute/PCBrideg and possibly Drawbridge. All of these run on 386/486 PCs
and 2 or more ethernet cards.

	The one limitation I know of is that PCBridge/PCRoute can't cope
with 8k NFS packets. Drawbridge can (I think) cope with 8k packets.

	Can people with experience of these packages send me comments on them
(and any commercial packages that may exist). I need to decide if a 486
with 2 ethernet cards can ggive me the cheap routing I need. What are the
limitations of these packages? Whats the performance like? (throughput)

	Please E-mail replies as I may not see them in news.

	Thanks for any comments, pointers etc.

===========================================================================
Robert Dal Santo               Phone +61 7 365 6687
                               Fax   +61 7 365 4466
Department of Psychology,      
University of Queensland 4072,
AUSTRALIA. 		       

-----------[000319][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 17:12:03 -0500
From:      lin@cs.purdue.edu (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

 Check out RFC-1470: "FYI on a Network Management Tool Catalog:
 Tools for Monitoring and Debugging TCP/IP Internets and
 Interconnected Devices"

  Happy sniffing. 8^)

John Lin (lin@cs.purdue.edu)

-----------[000320][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 18:40:08 -0400
From:      bobf502@aol.com (BobF502)
To:        comp.protocols.tcp-ip
Subject:   Re: Books on Winsock or Sockets programming?

In article <Rob.Sadler.17.0008E9DC@AtlantaGA.NCR.COM>,
Rob.Sadler@AtlantaGA.NCR.COM (Robert F. Sadler) writes:

>> ...I was 
>> wondering if anyone out there can recommend example source code or books
 that 
>>explain or demo reading mail and news (SMTP, NNTP) over socket interfaces.

If you find anything, would appreciate your posting it here or letting me know.
I'm looking for the same thing.

Bob Fite, Houston TX
BobF502@aol.com
71220,2742@compuserve.com

-----------[000321][next][prev][last][first]----------------------------------------------------
Date:      Sat, 16 Apr 94 20:55:00 EST
From:      vgadmin@vgateway.com
To:        comp.protocols.tcp-ip
Subject:   TCP/IP Router


Hi..

   I was wondering if anyone could recommend any TCP/IP routers. What
my goal is is to be able to connect my Novell 3.11 network to the
internet, mainly to be able to offer more advanced internet services for
my BBS.

   The next problem is, how do I GET a connection to the internet...

If anyone can answer these questions, please let me know!

Brian
vgadmin@vgateway.com
Gateway Telecommunications

/-----------------------------------------------------------------------------\
| The Virtual Gateway * Florida's gateway to online excitement! (904)376-6601 |
| + Internet: vgateway.com  MajorNet: GAT  Mail info@vgateway.com for info! + |
 \-----------------------------------------------------------------------------/

-----------[000322][next][prev][last][first]----------------------------------------------------
Date:      16 Apr 1994 17:51:00 GMT
From:      dmag@engin.umich.edu (Daniel Demaggio )
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

>>|I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
>>|where I can find such a thing?  It would be EXTREMELY helpful to me right
>>|now.

There is a program called Fergie that includes a sniffer called "gobbler".
Search archie for Fergie or frgbin2.zip
				-=Dan=-


-- 
dmag@umich.edu | When laws are outlawed,      | Ono-Sendai: the best
Dangerous  Dan | only outlaws will have laws. | Sim Stim decks

-----------[000323][next][prev][last][first]----------------------------------------------------
Date:      Sat, 16 Apr 1994 20:56:45 GMT
From:      deano@areyes.com (Dean Carpenter)
To:        biz.sco.general,comp.protocols.ppp,comp.protocols.tcp-ip
Subject:   TCP errors in ppp after net382b

  Since applying the net382b patch, I've been seeing errors during ftp
transfers over ppp.  In the process of getting a couple of TLSs from
ftp.sco.com, I keep seeing about 5 or 6 of the following errors appear on
the current screen.

Notice: tcp sum: src 84936A06, sum 00000002
Notice: tcp sum: src 84936A06, sum 00000004
Notice: tcp sum: src 84936A06, sum 00000006

  The only difference seems to be in the last sum portion.  When I say 5
or 6, the total is more like 50, but in groups of 5 or 6.

  The connection is from a WorldBlazer on my end to unknown at the other
end.  Prior to applying net382b, there was never a problem.  This cuts
the transmission rate from 1.6kbps to around 0.7kbps, according to ftp.

  What's going on here ?

-- 
Dean Carpenter		uunet!areyes!deano		(203) 847-6003
Areyes, Inc.		deano@areyes.com

"No matter where you go, there you are"  sayeth Buckaroo 								 across the Eighth Dimension

-----------[000324][next][prev][last][first]----------------------------------------------------
Date:      17 Apr 1994 12:48:22 -0700
From:      craign@teleport.com (Craig R. Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Winsock under Win-OS/2

In <2orqg7$b7o@lastactionhero.rs.itd.umich.edu> sarr@citi.umich.edu (Sarr J. Blumson) writes:

>yes we've used IBM's Winsock for Win-OS/2 with IBM
>TCP/IP for
>OS/2.  Ran NCSA Mosaic for Wondows under it quite
>nicely.

Just off hand, does the IBM Winsock for Win-OS/2 support SLIP or PPP
connections? If so, is it simplistic to set it up, and above all, where
does one get this TCP/IP implementation?

Thanks for the info (and if this is in the FAQ *please* forgive me :-)


--

*  Craig R. Nelson       | Always looking for a better job. *
*  Located in Downtown   | If you need a Windows programmer *
*  Beaverton OR 97005    | I'm only a r)eply key away.      *

-----------[000325][next][prev][last][first]----------------------------------------------------
Date:      Sun, 17 Apr 1994 05:31:56 GMT
From:      spencer@equinox.unr.edu (Jason Spencer)
To:        comp.protocols.tcp-ip
Subject:   Need Help:  writing FTP client

I am writing a winsock ftp client, but I am building a dummy client in Unix 
first.  I have most everything going smoothly, but with 2 hosts (ftp.unr.edu 
& oak.oakland.edu), when I read the message which comes up after login 
(anonymous), I get about 3/4 the way through, and then my buffer appears to
run out.  I have tried to increase the buffer size w/ setsockopt, but it
doesn't change.  I get total byte counts of about 1050 and 1700 chars before
the buffer comes up short.  Here is the actual symptom:

after login, I send the password and read the reply.  "getreply()" consists
of an initial readline, then a "while the 4th char = '-', readline" loop.
Readline calls recv w/ MSGPEEK for 82 chars, and it looks for a '\n'.  
When it finds it, it remembers how many chars it has looked at so far, 
then calls recv again w/o MSG_PEEK, to clear the buffer.

The hosts send me about 25 lines before this happens:

n = recv(socket, buff, 82, MSG_PEEK) and n = 16

the 6th char is a \n (after the search, n=6), then recv(socket, buff, n, 0).

Then n=recv(socket, buff, 82, MSG_PEED) and n = 10, but the actual line
is 60+ chars long.
 
So for 2 consecutive PEEKS, I can see the buffer dwindling.  Most hosts 
are fine, but for some reason, these two kill me almost every time.  Actually,
I should mention that sometimes, ftp.unr.edu will work.

Anyway, after this happens, the code is not 230 (because the buffer now has
new chars in it (the continuation of the line) which are not a code), but
it continues to displays the lines perfectly fine.

ANY Ideas?  If anyone needs/wants more information, let me know.  Also, if
there is another place that this question would be better supported, let
me know.

Thanks in Advance.

Jason Spencer


-----------[000326][next][prev][last][first]----------------------------------------------------
Date:      Sun, 17 Apr 1994 15:37:36 GMT
From:      ucacjon@ucl.ac.uk (Jon Crowcroft)
To:        comp.protocols.tcp-ip
Subject:   Re: What's the current SotA for reliable multicast/broadcast?


|> ISIS Distributed Systems (see comp.sys.isis) is the best-known
|> commercially
|> available package for building reliable multicast and process group
|> -based
|> distributed applications.

however, a low cost ubiquitous reliable multicast transport in the
public
donmain would work wonders...

also nite that moist folks on the mbone are using things like wb, which
do application layer 
relaibility, as its often a better approach...in many applications, only
the highest
level can trade off between consistency and timeliness...

for trivcial things like software distribution, though,m you don't need
all the 
whizz bang causal stuff in ISIS, and there is some scope for someoene to
write a cute infinite
window UDP based FTP, with multicast congestion avoidance,m of
course...

how to do the last bit is a trade secrtet of ours, but if people ask
nicely, we 
will give it away:-)


-- 
jon crowcroft (hmmm...)

-----------[000327][next][prev][last][first]----------------------------------------------------
Date:      Sun, 17 Apr 94 15:58:15 GMT
From:      Geoffc@ozinkl.pc.my (Geoff Collins)
To:        comp.protocols.tcp-ip
Subject:   Encyption of data between nodes across the net

A hyperthetical question (which may have a later practical use)

Consider.. The Internet, and a corporation which has Internet connection
on one side of the world, and an office of same corp, on the other side
of the world, with another Internet connection thru a third-party
Internet access provider.

This hyperthetical corp would like to communicate across the net, as an
alternative to a primative uucp over X.25 in use now(at 2400bps).

Both sides use Cisco routers, and Unix systems.

Without extensive changes to all TCP stacks on all workstations, and
machines connected on each side, is there an easy way to implement some
form of encryption, to ensure that proprietary data and other
intelectual property passed between the two sites is protected, as the
data passes across the net, where is might be intercepted by others?

This would also need to be specific, in that encryption was ALWAYS used
and ONLY used when communicating to this corps systems, while still
providing non-encrypted communication for net access to non-corporate
systems.

I would assume that some form of gateway/firewall would be required. Is
there such a beast, or are there other alternatives?

Thanx in advance.

--geoff
geoffc@ozinkl.pc.my

-----------[000328][next][prev][last][first]----------------------------------------------------
Date:      17 Apr 1994 17:09:27 GMT
From:      sarr@citi.umich.edu (Sarr J. Blumson)
To:        comp.protocols.tcp-ip
Subject:   Re: Winsock under Win-OS/2

In article <60.16811.4471.0N19AAD5@canrem.com>,
john.childs@canrem.com (John Childs) writes:
|> Has anybody gotten Win-OS/2 to recognize
|> winsock.  If so who's TCP/IP
|> product were you using?
|> 
|> John Childs
|> 
I'm not completely sure I'm answering the right
question, but
yes we've used IBM's Winsock for Win-OS/2 with IBM
TCP/IP for
OS/2.  Ran NCSA Mosaic for Wondows under it quite
nicely.

-- 
--------
Sarr Blumson                         sarr@citi.umich.edu
voice: +1 313 764 0253               home: +1 313 665 9591
CITI, University of Michigan, 519 W William, Ann Arbor, MI 48103-4943

-----------[000329][next][prev][last][first]----------------------------------------------------
Date:      17 Apr 1994 22:19:37 GMT
From:      billy@utdallas.edu (Billy Barron)
To:        comp.protocols.tcp-ip
Subject:   Re: Promposed new commands for FTP servers

peter.lewis@info.curtin.edu.au (Peter N Lewis) writes:

>FTP Structured List command - Peter N Lewis, Feb 1994.
 
>Note: I don't expect anyone to pay any attention to this document.  It's
>better solution than the LIST parsing method described seperately, but it
>is unlikely to ever be implemented.
>
As you probably remember from the past, I disliked your other method.
The new commands documented here were much better.  Personally, I believe
that this is more likely to be implemented than your other idea.  Nobody
likes to change stuff, but this is independent.

May I recommend two things?  First, try to get an implementation into
the WU-FTPD code.  Second, get this published as an experimental RFC.

--
Billy Barron,  Network Services Manager, Univ of Texas at Dallas
billy@utdallas.edu 

-----------[000330][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 02:52:09 GMT
From:      hpa@ahab.eecs.nwu.edu (H. Peter Anvin N9ITP)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP redirector for unix

Followup to:  <2ojep7$2qj@styx.uwa.edu.au>
By author:    comrade@uniwa.uwa.edu.au (Peter Cooper)
In newsgroup: comp.protocols.tcp-ip
>
> I remember reading about a driver written for some variety of unix that
> permitted redirection of IP datagrams to a user process.  The impression
> I had, was that the device was ifconfig'd and a route was added to the
> routing tables, somewhat like the situation with SL/IP.
> 
> Something as nifty as this would be really useful to test how complete
> and correct an implementation of something with pretty minimal IP
> functionality was, without having to burn LOTS of EPROMs.
> 
> If anyone can direct me to the sources, I'd really appreciate it - it'd
> be nice if the sources were written with a BSD-derived unix in mind, but
> I'll try anything... in preference to the appalling alternative.
> 

Check out your implementation of SLIP.  My own experience with it is
limited to Linux (a freeware UNIX clone for i386 class machines), but
on that OS SLIP channels are allocated by the OS (as ifconfig
devices), but by default not bound to any tty.  Conventionally, of
course, they would be bound to a serial device, but it is perfectly
legal to bind them to the slave end of a pseudo-tty.  If you have your
user process monitor the master end, you should be able to access the
IP datagrams in the form of a byte stream (and send datagrams the same
way, of course).

Note that this is probably quite inefficient, and if it is speed
you're looking for, this is not it.  However, it appears to solve your
problem if all you want is a testbed.

Linux is available from sunsite.unc.edu, tsx-11.mit.edu and many other
sites. 

	/hpa
-- 
INTERNET: hpa@nwu.edu               FINGER/TALK: hpa@ahab.eecs.nwu.edu
IBM MAIL: I0050052 at IBMMAIL       HAM RADIO:   N9ITP or SM4TKN
FIDONET:  1:115/511 or 1:115/512    STORMNET:    181:294/101
Linux: It is not too late to turn back from the Gates of hell

-----------[000331][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 02:53:31 GMT
From:      hpa@ahab.eecs.nwu.edu (H. Peter Anvin N9ITP)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP FTP over X.25?

Followup to:  <jfisher.14.00124A86@eagle.ndhm.gtegsc.com>
By author:    jfisher@eagle.ndhm.gtegsc.com (John Fisher)
In newsgroup: comp.protocols.tcp-ip
>
> Hi,
> 
> I have a friend at work that needs to test out an x.25 interface in a couple 
> Sparkstations.  He wants to connect 2 Sun spark-stations together with a 
> standard RS232 cable running X.25  Then he wants to run TCP/IP on top of it 
> and FTP files between the machines.  Is this possible??  I don't
> know anything  
> about x.25 but I think IP can be routed over it, right??  

What's wrong with good ol' SLIP?

	/hpa

-- 
INTERNET: hpa@nwu.edu               FINGER/TALK: hpa@ahab.eecs.nwu.edu
IBM MAIL: I0050052 at IBMMAIL       HAM RADIO:   N9ITP or SM4TKN
FIDONET:  1:115/511 or 1:115/512    STORMNET:    181:294/101
Linux: It is not too late to turn back from the Gates of hell

-----------[000332][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 08:56:02
From:      andre@wmhmis1.wmh.iupui.edu (Andre' J. Huff)
To:        comp.protocols.tcp-ip
Subject:   Work Groups for Windows and TCP/IP



I am in need of help on getting Work Groups for Windows working correctly.  I 
can get everything on my side working great but I can't run Mosaic, Cello, 
FTP, TELNET or any other programe using TCPMAN.EXE.  I get an error stating 
need frame class of 1 or 6.  Does anyone know how to change the frame class of 
work groups?  Please respond to my address for I don't always search the news.


Thanks
Andre'

Andre@wmhmis1.wmh.iupui.edu

Andre@vax1.iupui.edu

-----------[000333][next][prev][last][first]----------------------------------------------------
Date:      Sun, 17 Apr 1994 19:05:47 +0800
From:      peter.lewis@info.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   Proposed FTP LIST Specification

FTP LIST Specification - Peter N Lewis, Feb 1994.

Note: For those purists out there, I will also post a proposed new command
for servers that would address this problem in a far more sensible manner,
except for the fact that it's unlikely to ever be adopted.

Target Audience: 
FTP server authors and FTP client authors who wish to parse the LIST
command to display information in a local (non-english or graphical)
manner.

Premise:
There is no way to change existing servers in a reasonable amount of time.

For a long time the FTP LIST output has been unspecified and yet there are
more and more programs each year that rely on this format in one way or
another.  To try to combat this problem I am proposing a specification of
a LIST format that is easily decoded and very close to the most common
current format.

Please note that I am offering up this specification in the hope of
increasing compatibility amongst servers and clients but no one is under
any obligation to pay any attention to this document.

It is necessary to distinguish between the output format and the method of
decoding the format since we wish to allow all decoders to interpret as
many of the current versions of the LIST format, while trying to reduce
the variation in the output.

The output should be something like this:

-rw-------  1 peter         848 Dec 14 11:22 00README.txt
or
- whatever you feel like 848 Dec 14 11:22 00README.txt

This specification only defines the first character, and the information
(size/date/name) at the end of the line.  The infoirmation between is at
the discretion of the server author, although you need to be careful to
avoid displaying something that will match the date format.

The first character should be one of [-dl].  "-" means a file, "d" means a
directory and "l" means a link.  In many cases it may be preferable to
resolve the link and display it simply as a file or directory, since the
client otherwise has no way to tell which it is except by trying to change
directory into it (or by using the -p switch, see later).  Currently the
only way I've been able to find to determine if a link is a file or
directory is to check if it has an extension (ie terminates in a dot
followed by 1 to 3 characters) (this is obviously less than an ideal
solution)

The rest of the line should be found by matching
<space><3-letter-month-code><space><any><digit><space><digit|space><digit>
(regexp: / ([A-Za-z][A-Za-z][A-Za-z]) .[0-9] [0-9 ][0-9]/ where $1 is a
valid english month) and using the earliest such match (since that string
may be part of the file name.  The month name match should be case
insensitive, and should match one of the twelve English three-letter month
abbreviations: Jan, Feb, Mar, Apr, May, Jun, Jul, Aug, Sep, Oct, Nov, Dec.
Servers should always use the English version.  Clients should of course
display the date in the local language and format where possible.  Some
servers may currently use French month names, so you might like to also
accept them (though it would be better to have the servers change to
English).

Once you have found that, you can determine the file size by looking at
the number immediately before that match location.  This number may be
missing or invalid for directories, links, or other "file types" (and thus
should be ignored except for files).

The date format is either:

  MMM DD hh:mm
OR
  MMM DD  YYYY
OR
  MMM DD YYYY 

In the first case, the year is taken to be the most recent past occurence
of the date.  Servers should choose to switch over well before this
becomes close to allow for variations in the local time (not to mention
user confusion).

And the file name follows the date and continues to the end of the line.
If the file name is a link, it may include a pointer to the original, in
which case it is in the form "name -> link".  This is really very bad,
since that is also a perfectly valid file name under unix and other
systems (especially the Macintosh).  Servers should not do this where
possible (resolve the link, correct the filesize and display as a file or
directory is in general a better plan).

The information between the first character and the size (for files) or
date (for other types) should be ignored.  Servers may put any system
dependent information here (although as noted about, you should avoid
displaying anything that looks like a date).

When decoding, it is important to note that many implementations include a
line at the start like "total <number>".  Clients should ignore any lines
that don't match the described format.  If no matches are found, and more
than a few lines are read, then the server is probably non-conformant, so
you might consider using the NLST command.  Also, listings may include the
special files "." and "..".  I have no idea what you do with them or how
you tell if they are special.  Servers should not display them where
possible.

Servers should try to support the important unix "ls" switches:

-p - add / on the end of directories, especially important for links.
-F - add /, *, @, = to the end of various files.
-l - in the long format described above (ie LIST should ignore it)
-R - recursive listing (some mirroring software relies on this)

Site maintainers should consider the problem of determining the type of a
link and try to maintain the original name or at least the original
extensions (this will also help users as well as client software).

Caveates:

A better sollution would be to implement a structured list command, but
this is not practical (even if this were done, it would be years before a
reasonable percentage of servers were converted, and thus this document
remains necessary).

Although it looks unix-centric, and english-centric, the intention is the
exact opposite.  By standardizing the list display, we allow clients to
display dates in the local language, and listings in a local and/or
graphical format.

BNF:

<list> ::= <line> <list>
         | <comment> <list>
         | <null>

<line> ::= <type-char> <stuff> <size> SPC <date> SPC <longname> <crlf>

<type-char> ::= "-" | "d" | "l"

<stuff> ::= <ascii-char> <stuff>
          | <null>

<size> ::= <digit> <size>
         | <null>

<date> ::= <month> SPC <day> SPC <time>
         | <month> SPC <day> SPC SPC <year>
         | <month> SPC <day> SPC <year> SPC

<month> ::= "Jan" | "Feb" | "Mar" | "Apr" | "May" | "Jun" | "Jul"
          | "Aug" | "Sep" | "Oct" | "Nov" | "Dec"

<day> ::= <digit> <digit>
        |   SPC   <digit>

<time> ::= <digit> <digit> ":" <digit> <digit>
         |   SPC   <digit> ":" <digit> <digit>

<year> ::= <digit> <digit> <digit> <digit>

<longname> ::= <name> SPC "->" SPC <name>
             | <name>

<name> ::= <ascii-char> <name>
         | <ascii-char>

<comment> ::= <stuff> <crlf>

<crlf> ::= CR LF

<null> ::=

Notes:
Month names are case insensitive.

<ascii-char> is any character from 1 to 255 of any ISO 8859-1, excluding
CR and LF.

Names in capital letters and without angle bracket ("<>") quotes are names
of ASCII control characters.

The ambiguity between <comment> and <line> must always be resolved in
favor of <line>.

There is also an ambiguity between the first and second forms of
<longname>.  Resolution of this ambiguity is up to the programmer, but
always prefering the first form is likely to give reasonable results.  It
may be possible to use the FTP NLST command to obtain disambiguating
information.

More ambiguity: Should <stuff> ever contain a legal <date>, all bets are off!

Author:

Peter N Lewis <peter.lewis@info.curtin.edu.au>
10 Earlston way,
Booragoon, 6154, WA,
AUSTRALIA

Contributors:

Quinn <quinn@cs.uwa.edu.au>
James W. Matthews <James.W.Matthews@Dartmouth.EDU>
Stephen Trier<trier@ins.cwru.edu>

Obviously, any errors in this document are my own!
_______________________________________________________________________
Peter N Lewis <peter.lewis@info.curtin.edu.au>       Ph: +61 9 368 2055

-----------[000334][next][prev][last][first]----------------------------------------------------
Date:      Sun, 17 Apr 1994 19:09:36 +0800
From:      peter.lewis@info.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   Promposed new commands for FTP servers

FTP Structured List command - Peter N Lewis, Feb 1994.

Note: I don't expect anyone to pay any attention to this document.  It's
better solution than the LIST parsing method described seperately, but it
is unlikely to ever be implemented.

This document defines two proposed FTP commands to work with structured
lists in an attempt to make a machine-readable LIST format.  This is
necessary to give people a viable alternative to attemptint to parse the
LIST command (a method for which is documented elsewhere).

The two commands are

XMRA <attribute-list>
XMRL <path>

The first specifies the attributes the client wishes the server to list
for each file.  The latter is the equivalent of the LIST/NLST command, and
asks the server to display the specified files/directories in the format
documented here with attributes as listed in the XMRA command.

<attribute-list> is simply a comma seperated list of attribute names.  An
attribute name is a sequence of ascii seven bit characters from the 32 to
126 excluding comma and space.  Current attribute names should include
only characters from the set [A-Z,a-z,0-9,_,-,+,.] starting with a letter
to allow for future expansion.  Attribute names are case insensitive. 
This document will attempt to describe all the standard attributes. 
System specific attributes should be in the format
<system-name>-<attribute> (where <system-name> is the System Name as
defined in the Assigned Numbers document).  Attributes starting with X
(including both Xattribute and UNIX-Xattribute) are reserved for
experimental or local attributes.  A server can respond to a system
specific attribute even if it is not actually on the system.  In general,
generic attributes should be used whereever possible, and clients MUST be
able to back off to the defined minimal set of attributes.

The XMRA command should reply with either:
200 Everything's cool
502 Command not implemented
504 <attribute-list> command not implemented for that parameter.
(<attribute-list> is a subset of the parameters to the XMRA command, and
may be either a single attribute or all attributes that the server does
not understand)

The XMRL command works in the same way as the LIST and NLST command (and
SHOULD list the same files, with the exception that the special files like
 and .. MUST NOT be listed).  The returned list should consist of one
line per file/directory which is a comma seperated list.  Characters in
each entry may be escaped using the MIME = method (ie, = followed by two
hex character is translated into that character).  Characters SHOULD come
from the ISO-8859-1 character set (which SHOULD be the same characterset
you use in the filename parameters (LIST, STOR, etc)).  Character 0-31,
127, 255, space, comma (,) and equals (=) MUST be escaped, other
characters SHOULD NOT be escaped.  The comma seperated entries in the list
are the requested attributes (in the same order) as specified in the XMRA
command (which MUST preceed it).

The XMRL command should reply with any response the LIST command can give.

Mandatory Attributes:

[ These need a lot of work!]

NAME: File name
TYPE: "file" or "dir" or ???
LINK: 0 or 1 ???
SIZE: Approximate size of the file in bytes, empty string for directories)
PERM: [g-][p-][c-][d-][r-] (case insensitive) where dash (-) means "can't"
and letter means "can", the characters are (in this order):
g - Can get (RETR) the file (or files inside this directory for a directory)
  p - Can put (STOR) over this file (or in this directory for a directory)
  c - Can cd (CWD) into this directory
  d - Can delete (DELE/RMD) this file or directory
  r - Can rename (RNFR/RNTO) this file or directory
DATE: Modification date or closest approximation (create date?). 
Formatted as 19940227114609 (or what is an ISO date format???)

Other Generic Attributes:
MOD_DATE
CREATE_DATE
ACCESS_DATE
BACKUP_DATE

System Specific Attributes:

Macintosh:
MACOS-FTYPE: File type (4-letter code)
MACOS-FCREATOR: File creator
MACOS-FLAGS: Finder flags
MACOS-DIRID: DirID (empty string for files)
MACOS-DLEN: Data fork length
MACOS-RLEN: Resource fork length
MACOS-PARID: Parent DirID

Unix:
UNIX-MODE: 9-letter mode? or octal number?
UNIX-FTYPE: 1-letter mode? or "file" output?
UNIX-LINKS: Link count
UNIX-UID: number
UNIX-GID: number
UNIX-UNAME: name
UNIX-GNAME: name

VMS:

Authors:
Peter N Lewis <peter.lewis@info.curtin.edu.au>
Robert J Minich <minich@a.cs.okstate.edu>

Contributers:
Steve Fosdick <stevef@aom.bt.co.uk>
_______________________________________________________________________
Peter N Lewis <peter.lewis@info.curtin.edu.au>       Ph: +61 9 368 2055

-----------[000335][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 94 09:19:05 GMT
From:      avalon@cairo.anu.edu.au (Darren Reed)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

nelson@crynwr.crynwr.com (Russell Nelson) writes:

>In article <1994Apr10.221823.29510@unlv.edu> ftlofaro@unlv.edu (Frank Lofaro) writes:
 
>	   I have a couple of questions about ICMP unreachable bombing:
>   (sending bogus ICMP unreachables to make people's TCP connection
>   disconnect, etc)
 
>Why would a TCP connection disconnect just because it got an ICMP
>unreachable?  The only time a TCP connection should listen to ICMP
>unreachable is when it's SYN is still un ACK'ed.

RTFRFC.

The host spec. requires that they treat various ICMP datagrams as the same
as TCP's RST.

The problem is it is easier to correctly forge an ICMP message than a TCP
RST.


-----------[000336][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 09:22:56 +0000
From:      paulr@motiv.demon.co.uk ("Paul B. Richardson")
To:        comp.protocols.tcp-ip,comp.mail.mime,demon.ip.support
Subject:   Quickmail from CE Software ??

I have been told about a cross platform mail product called Quickmail
by CE Software in the States.  Does anyone have contact details for
this company?  Also does anyone have any experience with either:
a. encapsulating the mail in IP for Internet transmission AND/OR
b. using a Quickmail <-> SMTP gateway.

Thankyou in anticipation of your replies.

-- 
 --------------------------------------------------------------------
| Paul Richardson            | Motiv Systems Ltd                     |
| PaulR@Motiv.demon.co.uk    | Internet and Open Systems Consultancy |
| Tel: +44 223 576318        | 22 Hills Road                         |
| Fax: +44 223 576319        | Cambridge CB2 1JP                     |
 --------------------------------------------------------------------

-----------[000337][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 11:23:59 GMT
From:      jardarso@dhhalden.no (JARDAR SUNDE OLSEN)
To:        comp.protocols.tcp-ip
Subject:   Where is the FAQ?

Where can I find a FAQ on TCP/IP communication?

I am specialy interested in pointers to get me going on TCP/IP programming 
in C++ on HPUX machine.





Thanks,

Jardar Sunde Olsen
jardarso@dhhalden.no


-----------[000338][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 12:01:19 GMT
From:      jbasara@vienna.itd.sterling.com (Jim Basara)
To:        comp.protocols.tcp-ip
Subject:   Which RFC defines ICMP?

Can someone please tell me which RFC defines ICMP?

thanks,

--
-------------------------------------------------------------------------------
Jim Basara
     jbasara@Sterling.COM   



-----------[000339][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 13:50:21 GMT
From:      carlson@Xylogics.COM (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: Which RFC defines ICMP?

In article <2otsqf$32f@ssw.vienna.itd.sterling.com>, jbasara@vienna.itd.sterling.com (Jim Basara) writes:
|> Can someone please tell me which RFC defines ICMP?

792.

--
James Carlson <carlson@xylogics.com>            Tel:  +1 617 272 8140
Annex Software Support / Xylogics, Inc.               +1 800 225 3317
53 Third Avenue / Burlington MA  01803-4491     Fax:  +1 617 272 2618

-----------[000340][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 14:22:54 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

> > Why would a TCP connection disconnect just because it got an ICMP
> > unreachable?  The only time a TCP connection should listen to ICMP
> > unreachable is when it's SYN is still un ACK'ed.
>
> RTFRFC.
> The host spec. requires that they treat various ICMP datagrams as the same
> as TCP's RST.

I think there's a misunderstanding here about which unreachable codes
do what to the connection.  I think the original poster was referring to
the common network unreachable and host unreachable, both of which are
considered soft errors by RFC 1122 (along with source route failed) and
TCP MUST NOT abort the connection when it receives these.  Then the latter
posting is referring to the other unreachables (protocol and port unreachable,
fragmentation required and DF set) which the HRRFC says are hard errors and
should abort the connection.

So the answer depends on the unreachable code, and getting back to the
original question, on whether the implementation is broken or not (i.e.,
the old 4.2BSD bug that let a host unreachable abort an existing connection).
Also, come to think of it, the fragmentation one shouldn't be a hard error if
you're doing path mtu discovery.

	Rich Stevens

-----------[000341][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 14:28:48 GMT
From:      jeff@zis.ziff.com (Jeff Macdonald)
To:        comp.protocols.tcp-ip
Subject:   Re: new improved POP3?

Thanks for all who have replied. IMAP was indeed what I was looking for.

In article <2onqhi$5do@news.u.washington.edu> gray@cac.washington.edu (Terry 
Gray) writes:>From: gray@cac.washington.edu (Terry Gray)
>Subject: Re: new improved POP3?
>Date: 16 Apr 1994 04:45:38 GMT
>Keywords: POP3 mail, IMAP
 
>Jeff Macdonald <jeff@zis.ziff.com> wrote:
 
>>About 2 months ago I stumbled upon a site that mentioned a new mail protocol 
>>was being developed that would improve upon the POP3 protocol. Basically this 
>>protocol would allow the mail to remain on the host, but the client would be 
>>able to read new/old mail from any location... 
 
>Jeff,
>There is a revision of the POP3 specification in the works, but it does
>not change the basic POP model, which is to fetch messages from a maildrop
>to a (single) client machine. 
 
>There is another mail-related protocol that may be what you're looking
>for... It's called "IMAP" for "Internet Message Access Protocol".
>It is a functional superset of POP, and allows mail on a remote host
>to be accessed/manipulated/stored from one or more client machines.
 
>Recent versions of IMAP also allow for selective fetching MIME message
>parts, handy if someone sends you a 40MB video attachment and you are
>currently using a 9600bps slip link to look at your mail... 
 
>RFC-1176 describes "IMAP2", which has been around for quite awhile.  There
>is an IETF working group developing its successor, IMAP4.  IMAP4 will also
>have support for disconnected operation. See /mail/latest-imap-draft on
>ftp.cac.washington.edu for details. 
 
>-teg
>--
>Terry Gray
> University of Washington
>  gray@cac.washington.edu




Jeff Macdonald
Ziff Information Services
Ziff Communications
10 President's Landing
Medford, MA  
jeff@zis.ziff.com

-----------[000342][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 14:41:31 GMT
From:      barnett@grymoire.crd.ge.com (Bruce Barnett)
To:        comp.protocols.tcp-ip,comp.sys.hp.hpux
Subject:   netstat -i input packet errors?


	I have some HP/UX systems (Running 9.0[123]) that has a large
number of input packet errors on Ethernet.  I am referring to the
"Ierrs" field in the netstat -i command.

	Does anyone have some suggestions on how I can find the cause,
and eliminate the problem? If I had a LAN analyzer, what would I look
for?  
--
Bruce Barnett <barnett@crd.ge.com> uunet!crdras!barnett

-----------[000343][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 15:07:34 GMT
From:      jmc@devjam1.ml.com (J. Clairmont)
To:        comp.protocols.tcp-ip
Subject:   Re: Network Management Platform Comparisons

I am working on an evaluation of SNMP Network Management Tools.  I was wondering
if anyone else out there has prepared an analysis of these tools. The
particular SNMP Network tools include HP/OV/Netmetrics/EASE, Concord Trakker, 
Cabletron Spectrum.  I have looked at others such as W&G, but for our needs we need
the combination of remote probes, extensive reporting capabilities, an SQL relational
database for reporting historical data, baselining etc.   It also has to run on 
multiple platforms, SUNs, Windows(NT, 3.1), OS/2, AIX etc.  We are using SAS for some
reports, MOTIF based tools and Windows/DOS based tools to do Network Performance reporting on a very large 20,000+ node plus network.  Its a big job but somebodies got
a do it.  
 
These are the ones that are mainly in the running.  I could use
an independent outside report if you have one or know of one on the net. 
I'll share my feature/list/results with anyone else who wishes thanks. 
 
Thanks in advance.

---
-- 
   |                 Jan Clairmont Road Warrior(Scars on the Back to Prove it!)
   |  __   _         email:   jmc@devjam1.ml.com
 \_|_(_(__/ )        (212) 236-2405


-----------[000344][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 15:53:32 GMT
From:      paulf@panic.Eng.Sun.COM (Paul Fronberg [CONTRACTOR])
To:        ba.seminars,ba.internet,alt.internet.services,alt.bbs.internet,comp.security.misc,comp.unix.questions,comp.unix.admin,comp.protocols.tcp-ip,comp.dcom.servers,comp.sys.sgi.admin,comp.unix.solaris,comp.sys.hp.hpux,comp.security.unix
Subject:   SVNet meeting April 20th: Firewalls and Internet Security


              SVNet Meeting  Wednesday, April 20, 1994, 7:30pm
	           -- FREE and OPEN TO THE PUBLIC. --




WHAT:     Firewalls and Internet Insecurity

     In today's world of hackers and crackers, a corporation's enterprise 
     network needs a secure gateway (or "firewall") to protect its systems 
     from the Internet while at the same time providing interconnectivity.  
     Too much security and it becomes unusable or bypassed; too little and 
     it is easily usurped. We will be discussing current firewall alternatives
     hardware and software, including bastion hosts and DMZ networks.
     
WHO:      Geoff Mulligan, SunSoft

     Geoff Mulligan is a Member of the Technical Staff in the Internet 
     Engineering group at SunSoft. He works on the TCP/IP kernel and ultiity 
     software in Solaris 2, Sun's operating system. He also works on 
     disconnected e-mail processing, and network security firewalls and 
     encryption.  Prior to joining Sun, Geoff worked at Digital's Network 
     Systems Laboratory where is was project lead on the DEC SEAL (a firewall) 
     product, and Director of Education.  Before working at Digital, he spent

     11 years in the Air Force working at the Pentagon on networking and 
     network security (an oxymoron) issues throughout the Air Force and 
     building campus wide networks and teaching computer science at the Air
     Force Academy. Geoff received is M.Sc in 1988 from the University of 
     Denver and B.Sc.  in 1979 from the United States Air Force Academy.
     
WHEN:    Wednesday, April 20, 1994 at 7:30pm
     Please note that we will now be meeting on the 3rd Wednesday of the month!


WHERE:  Sun Microsystems Bldg 6, 2750 Coast Avenue, Mountain View
      Coast Ave appears to be just a driveway next to Bldg 5 on Garcia Ave 
      between Amphitheatre Pkwy and San Antonio, so don't get confused.

FOR MORE INFORMATION: please call either Paul Fronberg at (415) 366-6403 
      or Ralph Barker at (408) 559-6202.

SVNet is a UNIX  and open systems user group supported by member dues 
      and donations.


             SVNet Meetings are FREE and OPEN TO THE PUBLIC.

UNIX is a registered trademark of Unix System Laboratories, Novell or X/Open 
                     (few people are sure anymore)



-----------[000345][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 16:01:53 GMT
From:      sarr@citi.umich.edu (Sarr J. Blumson)
To:        comp.protocols.tcp-ip
Subject:   Re: Winsock under Win-OS/2

In article <2os3q6$q26@linda.teleport.com>,
craign@teleport.com (Craig R. Nelson) writes:
|> 
|> Just off hand, does the IBM Winsock for
|> Win-OS/2 support SLIP or PPP 
|> connections? If so, is it simplistic to set it
|> up, and above all, where 
|> does one get this TCP/IP implementation?
|> 
|> Thanks for the info (and if this is in the FAQ
|> *please* forgive me :-)
|> 
SLIP at least, I don't remember seeing anything
about PPP but I wasn't
looking.  Almost all of the setup was in TCP/IP
(the winsock layer is just a
call translator) and was pretty straightforward. 
It's an IBM product, an
add on for IBM's TCP/IP for OS/2, but you may have
to hunt a little for a
dustributer that has a clue what you're talking
about.  I don't have the
product number any more.

If this group had a FAQ, this might be in it.  You might find more details
in comp.os.os2.networking, though.

-- 
--------
Sarr Blumson                         sarr@citi.umich.edu
voice: +1 313 764 0253               home: +1 313 665 9591
CITI, University of Michigan, 519 W William, Ann Arbor, MI 48103-4943

-----------[000346][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 17:46:39 GMT
From:      raj@cup.hp.com (Rick Jones)
To:        comp.protocols.tcp-ip,comp.sys.hp.hpux
Subject:   Re: netstat -i input packet errors?

Bruce Barnett (barnett@grymoire.crd.ge.com) wrote:
: 	I have some HP/UX systems (Running 9.0[123]) that has a large
: number of input packet errors on Ethernet.  I am referring to the
: 	Does anyone have some suggestions on how I can find the cause,
: and eliminate the problem? If I had a LAN analyzer, what would I look

Well, you might start with landiag. That program will give you more
detail on what sort of errors are happening...

rick

here is an example...the Unknown Protocol field is often the result of
Netware broadcasts and such. When we receive a packet on a SAP we do
not have bound to the driver (ie we aren't listening for that
protocol), the Unknown Protocol field will be incremented.

$ landiag


           LOCAL AREA NETWORK ONLINE DIAGNOSTIC, Version 1.0
                       Mon,Apr 18,1994     10:46:31

               Copyright 1985 Hewlett Packard Company.
                       All rights are reserved.

Test Selection mode.

        lan      = LAN Interface Diagnostic
        menu     = Display this menu
        quit     = Terminate the Diagnostic
        terse    = Do not display command menu
        verbose  = Display command menu

Enter command: lan

LAN Interface test mode. LAN Interface device file = /dev/lan0

        clear    = Clear statistics registers
        display  = Display LAN Interface status and statistics registers
        end      = End LAN Interface Diagnostic, return to Test Selection
        menu     = Display this menu
        name     = Name of the LAN Interface device file
        quit     = Terminate the Diagnostic, return to shell
        reset    = Reset LAN Interface to execute its selftest

Enter command: name
Enter LAN Interface device file name. Currently /dev/lan0: /dev/lan0

LAN Interface test mode. LAN Interface device file = /dev/lan0

        clear    = Clear statistics registers
        display  = Display LAN Interface status and statistics registers
        end      = End LAN Interface Diagnostic, return to Test Selection
        menu     = Display this menu
        name     = Name of the LAN Interface device file
        quit     = Terminate the Diagnostic, return to shell
        reset    = Reset LAN Interface to execute its selftest

Enter command: dis

                      LAN INTERFACE STATUS DISPLAY
                       Mon,Apr 18,1994     10:46:44

Device file                     = /dev/lan0
Lu number                       = 0
Current state                   = active
LAN station address, hex        = 0x080009786398
Number of multicast addresses   = 4
Frames received                 = 37419860
Frames transmitted              = 22488649
Undelivered received frames     = 26
Untransmitted frames            = 0
CRC errors received             = 0
Transmit collisions             = 66582
One transmit collision          = 31114
More transmit collisions        = 35468
Excess retries                  = 0
Deferred transmissions          = 482267
Carrier lost when transmitting  = 0


Press return to continue 

No heartbeat after transmission = 0
Frame alignment errors          = 0
Late transmit collisions        = 0
Frames lost                     = 8
Unknown protocol                = 9368958
Bad control field               = 0
IEEE 802.3 XID packets          = 0
IEEE 802.3 TEST packets         = 0
Unable to respond TEST/XID pkts = 0

LAN Interface test mode. LAN Interface device file = /dev/lan0

        clear    = Clear statistics registers
        display  = Display LAN Interface status and statistics registers
        end      = End LAN Interface Diagnostic, return to Test Selection
        menu     = Display this menu
        name     = Name of the LAN Interface device file
        quit     = Terminate the Diagnostic, return to shell
        reset    = Reset LAN Interface to execute its selftest

Enter command: quit
Diagnostic terminated by operator.
$ 

-----------[000347][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 22:08:58 EDT
From:      Ed Wolfe <CEW108@psuvm.psu.edu>
To:        comp.protocols.tcp-ip
Subject:   TCP/IP

I'm looking for information on TCP/IP in general.  I could use any information
you could provide me.  Thanx in advance.

-----------[000348][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 00:49:59 -0400
From:      dbadrak@info.census.gov (Don Badrak)
To:        comp.protocols.tcp-ip,comp.mail.mime
Subject:   Re: Quickmail from CE Software ??

In article <766660976snz@motiv.demon.co.uk>,
Paul B. Richardson <PaulR@motiv.demon.co.uk> wrote:
>I have been told about a cross platform mail product called Quickmail
>by CE Software in the States.  Does anyone have contact details for
>this company?  Also does anyone have any experience with either:
>a. encapsulating the mail in IP for Internet transmission AND/OR
>b. using a Quickmail <-> SMTP gateway.

Paul,

CE Software is based in Iowa, USA.  You can get information by sending
mail to ce_info%cedsm@uunet.uu.net.  I have a phone number here, but I'm
not sure if it's correct (try the mail first, I guess).  515.224.4534 (in
USA, so I guess that +1 515 224 4534, right?)

Star*Nine makes a QuickMail/SMTP gateway that we use and works great.  They
are located in California.  Phone +1 510 649 4949 (this may be the support
number, though).  E-mail info@starnine.com (I think, this is off the top of
my head).

I have about 6 months experience with this product (the smtp gateway), and over
a year with QuickMail itself.  QM has a version that runs on DOS and on
Windows.  They use file based servers (either Novell running Netware for
Macintosh to get the AFP stuff, or an AppleShare Server).  We've had some
difficulties with the file-based stuff.

Anything else you'd like to know?

Don


-- 
--
Don.Badrak@Census.GOV
U.S. Bureau of the Census, Geography Division
Suitland, MD

-----------[000349][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 18:32:21 GMT
From:      evansmp@mb48025.aston.ac.uk (Mark Evans)
To:        comp.protocols.tcp-ip
Subject:   Re: Protection against ICMP bombers exist?

Roger Hunen (hunen@brc.medtronic.com) wrote:

: Yes. Block all ICMP unreachables on your Internet router and rely on 
: timeouts. Not really the greatest way, I must admit. TCP/IP was not designed
: with any security in mind so we'll have to live with that. I'm afraid...

Alturnativly fix your TCP so that it does NOT drop established connections
except if it receives a valid FIN or it times out.

-----------[000350][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 94 18:36:25 GMT
From:      kenlee@iitmax.iit.edu (Kenneth H. Greenlee)
To:        comp.dcom.lans.ethernet,comp.dcom.lans,comp.protocols.tcp-ip
Subject:   Re: Bridging/Routing Software on PCs

In article <2oo85j$p6u@dingo.cc.uq.oz.au> robert@psy.uq.oz.au (Robert Dal Santo) writes:
[stuff]
>
>	The one limitation I know of is that PCBridge/PCRoute can't cope
>with 8k NFS packets. Drawbridge can (I think) cope with 8k packets.
>
As far as I can tell from the documentation, this is not technically
true.  The documentation was written originally for use with the WD8003E
cards which have a 6.5K buffer.  Other cards don't have this limitation,
eg WD8013EBT.
>
>	Can people with experience of these packages send me comments on them
>(and any commercial packages that may exist). I need to decide if a 486
>with 2 ethernet cards can ggive me the cheap routing I need. What are the
>limitations of these packages? Whats the performance like? (throughput)
>
I have had trouble setting PCRoute up with a 286 and two WD8013EBT cards.
These cards should be natively supported by the package, ie, no need for
drivers, so I'm not sure where I'm failing.  Once again, according to the
documentation, an AT class machine using 16-bit cards should handle any
load.
>
>	Please E-mail replies as I may not see them in news.
>
I am emailing him.  As for me, I would like to see a dialog start regarding
these packages.  I like tremendouly the idea of a public domain solution
to routing problems (partly out of nece$$ity), but headaches I don't want.
>
>	Thanks for any comments, pointers etc.
>
>===========================================================================
>Robert Dal Santo               Phone +61 7 365 6687
>                               Fax   +61 7 365 4466
>Department of Psychology,      
>University of Queensland 4072,
>AUSTRALIA. 		       

Ken
-- 
Kenneth H. Greenlee * idgreenlee@id.iit.edu
Manager, Design Processes Laboratory * Institute of Design
Illinois Institute of Technology * Chicago * Illinois * USA

-----------[000351][next][prev][last][first]----------------------------------------------------
Date:      Mon, 18 Apr 1994 18:37:09 GMT
From:      evansmp@mb48025.aston.ac.uk (Mark Evans)
To:        comp.protocols.tcp-ip
Subject:   Re: 255.255.255.25 via point-to-point link ?

Tony Li (tli@cisco.com) wrote:
: In article <zok.766151581@popp.ins.de> zok@popp.ins.de (Andreas Frackowiak)
: writes: 
:     Is it "legal" to send an ip-packet to the destination
:     address "255.255.255.255" via a point-to-point link or
:     it is forbidden in a RFC ?
:     
: Yes, that's legal.

But make sure the other end treats it as a broadcast. e.g. never sends
and ICMP error if there is no protocol or port which wants it.

-----------[000352][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 19:21:45 GMT
From:      jdsmith@novell.com (Doug Smith)
To:        comp.protocols.tcp-ip
Subject:   Setting up a SL/IP Server

I am trying to configure my OS/2 2.1 machine as a SL/IP server.  Can
anyone point me to a FAQ or some other document that may give me some
details about SL/IP?  Specifically, I want to have a machine at home
dial into my computer at work (wihch has an ethernet connection to the
Internet).  I don't have another network number to use for SL/IP so I 
want to use proxy arp.

Thanks in advance for any pointers you can give me.


---------------------------------------------------------------------
Doug Smith                               Internet: jdsmith@novell.com
Novell, INC                                 phone: (801) 429-7324
122 E. 1700 S.
C-11-1
Provo, UT 84606


-----------[000353][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 22:13:31 GMT
From:      rddietz@blackmail (Richard D Dietz)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.os.ms-windows.networking.windows,comp.os.ms-windows.networking.tcp-ip
Subject:   Question about UNIX end of WINSLIP/SLIP/PPP

Hi everyone.  

I've got a question I hope _someone_ out there can help me with.  
Basically, what it amounts to is that I've got a 486 and 14.4kb 
modem and want to get a WINSOCK/SLIP/PPP connection to our Suns 
at the University.

Here's my problem:  I've downloaded everything I need on the PC
end and believe I can set it up ok myself, but the network I want
to connect to does not offer SLIP or PPP.  I have located (via archie)
and downloaded UNIX source for both, but each requires "root"
access to install (which I do not have).  I do have an account on 
the network, so I was wondering if anyone knew of some kind of
hack or work-around that would allow me to run PPP (or maybe even
SLIP) from my user account without any installing at the system
level?

On a related topic: I have been using the UWWIN program to log in and
have multiple terms for several months now.  The only problem with it 
is that it lacks up/downloading in the version I have (1.04).  All the
file are dated 1992, so does anyone out there know of a more recent 
version?  Although I doubt it, does anyone know if there exists a
WINSOCK that will run over a UW connection?  Anyone willing to write 
one?!? 8^)

Thanks for your time,
Dick Dietz (rddietz@eng.uiowa.edu)

-----------[000354][next][prev][last][first]----------------------------------------------------
Date:      18 Apr 1994 22:57:14 GMT
From:      miyaguch@igate1.hac.com (Darryl Miyaguchi)
To:        comp.protocols.tcp-ip
Subject:   Re: Slip/PPP Access

In article <2oe0oq$dr2@geraldo.cc.utexas.edu>, jgarven@mcl.cc.utexas.edu (James R. Garven) says:
>
>Can anyone suggest a Slip or PPP Internet Access Provider 
>that my sister-in-law can contact in the Portland, Oregon area?
>She is really "pumped up" about Internet and has outgrown the Prodigy
>"pond".  Any suggestions would be most appreciated.  For that matter,
>perhaps someone knows of a service provider located somewhere else in the
>U.S.A. that bundles in with its service a really good deal on long-distance
>connections.
>
>Thanks in advance.
>
>Jim Garven

Try Netcom
email:  info@netcom.com
voice:  800-501-8649, 408-554-8649
fax:    408-241-9145
ftp more info:  ftp.netcom.com:/pub/netcom/

Also Teleport:
email:  info@teleport.com
voice:  503-223-4245
ftp more info:  teleport.com:/about


For the latest Public Dialup Internet Access List, sending email
containing the phrase "Send PDIAL" to "info-deli-server@netcom.com".

Good luck   -- Darryl Miyaguchi  miyaguch@igate1.hac.com

-----------[000355][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 07:25:11 -0500
From:      georgiad@cperi.forth.gr (Danis Georgiadis)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.sys.mac.comm
Subject:   Problem with PCRoute & Farallon PC card - need help

Hello netters,

For about a week now, we have been trying to setup a router to the Internet
using PCroute. Any kind of help or advice will be much appreciated.
Please email your responces to <georgiad@cperi.forth.gr> - our news feed
is unstable.

-----------------------------------------------------------------------
We are trying to connect our localtalk network (Macs) to a Unix machine using
SLIP over a leased-line. Up to a point we made it, but the connection isn't
functional yet. What makes things really weird is that the link is capable
of carrying some traffic, but seems to discard the rest:

  We are able to ping all relevant machines (UNIX host, PC-Router, Macs) in all
possible combinations.
  Using MacTCP Watcher we can also verify that UDP packets go through the link
fine (by sending data to UDP's echo port on the UNIX host, through PCRouter).
In addition to that, having specified a debugging level of 8h, we are logging
PCRouter's operation on the remote UNIX host, over syslogd's local0 facility!

(Localtalk Lan) <----> (PC Router) <-------> (UNIX host)

Anyway, we were unable to use any TCP-based applications successfullly. From
what we can tell, when we initiate traffic from a Mac to UNIX (over the
PCrouter), the packets originating from the Mac are routed to UNIX (the 
PCRouter modem's LEDs indicate traffic) and UNIX responds back. If the protocol
is UDP, all works as noted above. Also 'ping' will work, passing packets with
500bytes of data at full speed just fine. 

But if the protocol is TCP, the returned packets seem to dissapear between the
PCRouter and Macs: the UNIX host knows about the connections (which is visible
with 'netstat') but eventually times-out.
Some degree of communication is present though, because if we try to connect to
from UNIX to an arbitrary port on Mac, which has no server running, we'll get
a "Connection refused" message immediately. Anyway, if we start 'fingerd'
on the Mac, and invoke 'finger' from UNIX to the Mac, it'll act as if all is OK
but no data is exchanged and the command times out.

Of course, the logfile of PCroute indicates no problems. Both sides of the
SLIP interface use the same MTU, subnetting works, the Router finds out
about the IP addresses of the Localtalk Macs over ARP, and it receives RIP
updates from its UNIX peer (logging all these to the *remote* UNIX host, to
which the Macs are unable to connect!) It all looks as if TCP packets are
dropped between the PCrouter and the Macs ...

Router Setup:
   Software:
   1) PhoneNet PC stack s/w: Link Support Layer (LSL), Driver (PHNETPC3),
      ATALK (Appletalk for Localtalk) and COMPAT (compatibility).
   2) PCRoute v2.24 compiled without any src-code modifications apart from
      interface declarations (DECLARE.INC) with TASM 2.0
   3) 2 Interfaces: AppleTalk and SLIP
	
   Hardware:
   1) 486/50 (temporarily ;-)
   2) Farallon Phonenet PC III card (brand new)
   3) Serial card (8250) running at 9600bps to a 2400bps/MNP modem using
      hardware flow control.

The UNIX System is running MorningStar PPP and has multiple configurations just
like ours (only the IP addresses differ)

On our Macs we use the latest TCP/IP software: MacTCP 2.04, latest NCSA Telnet.
Sofar we have succeeded working with a multitude of TCP/IP applications, over
dialup SLIP though, using MacPPP/VersaSLIP/InterSLIP etc.

We would appreciate any help on this problem, especially regarding the support
of our LocalTalk card in PCRoute.

Danis Georgiadis,
Escape Information Services.
georgiad@cperi.forth.gr


-----------[000356][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 10:19:14 -0400
From:      jwmanly@unix.amherst.edu (John W. Manly)
To:        comp.protocols.tcp-ip
Subject:   NTP algorithm specification?

Hi there.  I was wondering if anyone can point me to an article or an RFC
that discusses the algorithm that underlies the NTP protocol.  The RFC on
NTP itself (I forget the number) seems to only talk about the protocol being
used for passing time around, not for actually doing the synchronization by
taking things like network delay into account, particularly when using 
multiple timesources.

If anyone has any suggested reading, please let me know.

-- John W. Manly  <JWMANLY@AMHERST.EDU> (System Manager -- Amherst College)


-----------[000357][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 03:55:09 GMT
From:      elufker@hns.com (Ed Lufker)
To:        comp.protocols.tcp-ip
Subject:   telnet help please

Hi all:

	Can anyone tell me why telnet will work for the root account and
not any other accounts. Is there a way to see what files telnet hits while
its executing or does someone know what files telnet uses when it executes.
I have a hunch that its a permissions problem.

thanks in advance
eddie lufker
elufker@oscar.hns.com


-----------[000358][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 04:47:30 GMT
From:      ibottema@alkaid.sce.carleton.ca (Ike Bottema)
To:        comp.protocols.tcp-ip
Subject:   Re: Which RFC defines ICMP?

carlson@Xylogics.COM (James Carlson) writes:

>In article <2otsqf$32f@ssw.vienna.itd.sterling.com>, jbasara@vienna.itd.sterling.com (Jim Basara) writes:
>|> Can someone please tell me which RFC defines ICMP?
 
>792.

And 791, 950, 919, 922, 1112  = STD 5

Enjoy!

Ike Bottema


-----------[000359][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 10:12:15 +0000
From:      proyse@peeras.demon.co.uk (Phil Royse)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP under Win 3.11 - Help Please

In article <1994Apr11.052351.18958@hakea.adept.oz.au> smh@hakea.adept.oz.au writes:

>I would like to connect a Win 3.11 network to TCP.

Do you mean Windows for Workgroups 3.11?

>Where does the TCP software come from?

Anonymous ftp it from ftp.microsoft.com  (you'll have to poke around
to find which directory.) The file is called WFWTCP.EXE, and is a 
self-exploding PKZip executable, of about 400k).

I downloaded it last week, read the README.1ST file and  configured
into my two Windows for Workgroup PCs.  

All it provides is the Microsoft TCP/IP stack (to replace NETBEUI
under WfWG 3.11's NETBIOS), plus sockets, name resolution and ping.   
The NETBIOS naming works fine, and both my WfWG computernames were 
retained without any reconfig.
All the other WfWG 3.11 functions stayed the same & worked fine.

It was very easy to configure and also made the necessary changes
to the AUTOEXEC.BAT & SYSTEM.INI files.

I haven't done any formal performance tests, but I would swear that
WfWG seems to work **faster** over TCP/IP than NETBEUI.  (This is
based on 2 486sx's on the same wire, in the same room, sharing
files and one laser printer.)

>Is is possible to configure printers in windows that are actually
>printers on a Unix host. i.e. can I print using lpd etc ?

Not directly from WfWG 3.11, because all that MS's TCP/IP gives you 
is the transport to support the WfWG's NETBIOS network functions.  
Therefore, any printers need to be seen by NETBIOS as "MS network 
printers" and recognised by WfWG.

I believe that the MS TCP/IP stack will support any higher layer
TCP/UDP applications via its sockets API, therefore if you have
a sockets-conformant LPS applications, you may be able to configure it to
work over the MS TCP/IP sockets i/f.

But others may give you better and more detailed advice.

>Any help appreciated
>Thanks


-- 

Phil Royse     Comms Consultant  |  PRA Consulting Ltd.
TUDOR HOUSE                      |
12 Woodside Road, Purley
Surrey  CR8 4LN   (UK)        Tel: (+44) 81 645-986

-----------[000360][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 10:14:32 GMT
From:      jenwen@pdd.iii.org.tw (Jenwen)
To:        comp.protocols.tcp-ip
Subject:   Re: new improved POP3?

Jeff Macdonald (jeff@zis.ziff.com) wrote:
: About 2 months ago I stumbled upon a site that mentioned a new mail protocol 
: was being developed that would improve upon the POP3 protocol. Basically this 
: protocol would allow the mail to remain on the host, but the client would be 
: able to read new/old mail from any location... 

As I know, we can RETR mail but DELE mail in POP3 so that the mail
can remain in host. And TOP n command can retrieve top n line of each
mail. We can use top n line to see mail subject/from...etc to 
perform selective mail retrieve. Above is my exeperience. Comments
appeciated!

Jenwen,Chien-Wen Huang.
Institute for Information Industry,Taiwan.

-----------[000361][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 11:43:21 GMT
From:      schaumann@oms2.usmc.mil (David K Schaumann)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

Guys-

I have a friend of mine that is a GUI programmer, and is writing a GUI based network sniffer.  But it might come into a problem
for the needs of most people.  The primary objective is to get is monitoring Vines IP, and I don't think that it will help most of 
you.  I am trying to push him to make it applicable to TCP/IP.  If and when this gets written, I will post to comp.protocols.tcp-ip 
and let you know where to find it.  I am sure that there is someone out there that has already done this, and if there is I would like
to see it.

Thanks,

Dave

David Schaumann
Okinawa  Japan
InterNet: schaumann@okr-smtp1.usmc.mil

-----------[000362][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 13:00:06 GMT
From:      s900387@minyos.xx.rmit.EDU.AU (Craig Macbride)
To:        biz.sco.general,comp.protocols.tcp-ip,comp.sys.amiga,comp.sys.amiga.programmer
Subject:   rwhod warped with an Amiga

At work, we have a machine running SCO Unix and another running SCO Xenix
on a network. (There are also the inevitable MeSsyDOS and Novell toys.)

Now, these SCO machines both run rwhod happily and will keep the information
up to date as appropriate and have done so for months.

Recently, someone (speaking of toys) connected an Amiga to the network
running some TCP/IP package which does some strange things. When the Xenix
machine sends out a udp who message, the Amaga sends some sort of broadcast
in response! After this broadcast, the Xenix machine continues to operate
normally, but the Unix machine's rwhod refuses to process any more incoming
who packets. (It still generates the outgoing ones as usual.)

Does anyone have any idea why the Amiga might be responding to a who message?
Can I just put it down to bad programming, along with the lack of ANSI or
vt100 arrow keys and function keys under telnet and the lack of an "lcd"
command in ftp?!

The real curiosity to me is why the Amiga reacts to the Xenix machine's
who packets, but seems to ignore the ones from the Unix machine and why the
Unix rwhod takes the Amiga's broadcast so personally that it ceases to listen
to any producer of user information (including itself) thereafter.

-- 
 _--_|\		Craig Macbride	<craig@rmit.edu.au>
/      \
\_.--.*/	How many Anglican priests does it take to change a lightbulb?
      v		... "Change"? What's change?

-----------[000363][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 94 18:00:21 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Re: Books on Winsock or Sockets programming?


RS>Rob.Sadler@AtlantaGA.NCR.COM (Robert F. Sadler) writes:
RS>>> ...I was
RS>>> wondering if anyone out there can recommend example source code or books
RS>that
RS>>>explain or demo reading mail and news (SMTP, NNTP) over socket interfaces.

BF>I'm looking for the same thing.
BF>Bob Fite, Houston TX
BF>BobF502@aol.com
BF>71220,2742@compuserve.com

I don't know where such source code might be available for anonymous
FTP (I have the source to BSD/386 on CD-ROM myself), but here are some
pieces of the puzzle:

The WINSOCK spec can be found by:  ftp SunSite.UNC.EDU
                                   cd /pub/micro/pc-stuff/ms-windows/winsock
                                   They have both ASCII text and PostScript
                                   forms (nice!)

For Berkeley Sockets
programming in general:  UNIX Network Programming, W.R.Stevens, chapter 6

SMTP: TCP/IP Illustrated, Volume 1, W.R.Stevens, chapter 25

NNTP: RFC's 977 (NNTP), 850 (USENET), and 822 (messaging)
      (IOW, I still haven't found a good explanation for NNTP
      other than the RFC's themselves.  But try RFC 977, it's
      not so bad.)

For RFC's:  ftp ds.internic.net
            cd /rfc
            get rfcnnn.txt

-- Bob Stein

   Galacticomm, Inc.                            (305) 583-5990 (voice)
   4101 SW 47 Ave.                              (305) 583-7846 (fax)
   Ft Lauderdale, FL 33314                      (305) 583-7808 (BBS)

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000364][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 94 18:00:21 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Proposed FTP LIST Specification


I have been trying to think of a way for an intelligent FTP client to
parse the directory and file information at an FTP server site for
better presentation to users.  The client should be able to handle an
ever-increasing propotion of sites, asymtotically approaching 100%.

It makes sense to have a pool of machine-readable LIST format
recognition and decoding specifications, available to all, that many can
contribute to (ideally FTP server authors), until servers can be made to
universally support a more machine-readable variant of the LIST command.
Peter Lewis seems to have gone a long way toward this, if I understand
his papers correctly.

I'm not sure if you've addressed this elsewhere, Peter, but how exactly
do clients become fluent with all the formats for the LIST output of ftp
servers?  This is a logistical question.  Two possibilities:

       1.  A publicly available file specifies them all.  Clients
           run with the latest version of the file at hand.

       2.  Individual server sites offer their format specs in
           the "ftplist" or "/etc/ftplist" files (clients look in
           both places).

These all assume a custom or robot client, but no changes to server
programming.  Is this the purpose you had in mind with your "FTP LIST
Specification", or have I missed the point?

-- Bob Stein

   Galacticomm, Inc.                            (305) 583-5990 (voice)
   4101 SW 47 Ave.                              (305) 583-7846 (fax)
   Ft Lauderdale, FL 33314                      (305) 583-7808 (BBS)

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000365][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 13:34:08 GMT
From:      tor@spacetec.no (Tor Arntsen)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for X.25 to Telnet gateway

In article 4md@vanbc.wimsey.com, skl@Connectivity.com (Samuel Lam) writes:
>In article <1994Apr13.141745.488@sfov1.verifone.com>,
> jimmy_t@hnlv4.verifone.com wrote:
>>We are looking for an X.25 to Telnet gateway box.
>>The only one I've found so far is the XYPLEX.  If you have
>>any suggestions for other vendors, please let me know.
>
>Cisco also has X.25/Telnet protocol converter options for their
>router and terminal server products.

And the MegaPAC from Satelcom (UK) Ltd.
You can configure it with a number of Ethernet and X.25 boards and other interfaces.

Tor




-----------[000366][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 13:49:18 GMT
From:      barnett@grymoire.crd.ge.com (Bruce Barnett)
To:        comp.protocols.tcp-ip,comp.sys.hp.hpux,comp.dcom.lans.ethernet
Subject:   Re: netstat -i input packet errors?

(I added the ethernet newsgroup)

I am trying to isolate the cause of lan netstat -i ierrs values. (HP/UX)

I don't have HP sources, but I looked at the BSD sources.
According to those, the AMD 7990 LANCE chip will increment network input
errors on two conditions:

		Runt Packets (length <= 0)
		Bad Packets (the RMD1 word has LE_ERR set. This is 0x4000)

Can someone explain to me what conditions can cause the "bad packet"
to be set? How do I find out what system is causing this problem?
Some have suggested the network may not be properly terminated.
How would I measure this?
--
Bruce Barnett <barnett@crd.ge.com> uunet!crdras!barnett

-----------[000367][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 13:50:43 GMT
From:      lchutson@netcom.com (Lawrence Hutson)
To:        comp.protocols.tcp-ip
Subject:   HP-UX, Am I there yet ?

Is this the "Best" place for disscussions on HP-UX ? I noticed the comments 
on this group were for HP products. Sounded kinda proprietary ;-).

Thanks

Lawrence Hutson, Consultant
On Assignment @ GTE Network Operations


-----------[000368][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 15:54:00 GMT
From:      mgleason@cse.unl.edu (Mike Gleason)
To:        comp.protocols.tcp-ip
Subject:   Re: Promposed new commands for FTP servers

peter.lewis@info.curtin.edu.au (Peter N Lewis) writes:

|FTP Structured List command - Peter N Lewis, Feb 1994.
 
|Note: I don't expect anyone to pay any attention to this document.  It's
|better solution than the LIST parsing method described seperately, but it
|is unlikely to ever be implemented.

As someone else suggested, let's get an RFC going.

|XMRA <attribute-list>
|XMRL <path>
 
|The XMRL command works in the same way as the LIST and NLST command (and
|SHOULD list the same files, with the exception that the special files like
| and .. MUST NOT be listed).

How about excluding some other special files too. Like those bloody
.FSP* files.  The attr list should include an option to print everything,
though.  I never liked the default method that typical ftpd's do, which
is "ls -a".  

|The returned list should consist of one line per file/directory which is
|a comma seperated list. Characters in each entry may be escaped using the
|MIME = method (ie, = followed by two hex character is translated into that
|character).

Personally, I don't really care for that.  The Unix/C world is the
big fish, so I think we should use their traditional backslash escaping
method, with:

\,    escape a comma
\n    escape a newline
\\    escape a backslash
\010  escape a character with 10 octal
\x10  escape a character with 10 hex

|LINK: 0 or 1 ???

How about LINK-TO: "item link points to" or empty string

|SIZE: Approximate size of the file in bytes, empty string for directories)

This has always been tricky.  The 'SIZE' command will give you back the
number of bytes it will take given the current transfer mode.  For ASCII
mode, this could very well mean the transferred size would be larger
or smaller that the actual amount.  It would seem no big deal, but I
believe we've discussed this before how it would be a pain in the ass
for sending Macintosh BinHex files.

|SIZE: Approximate size of the file in bytes, empty string for directories)

Maybe instead of empty str for dirs, have it be the number of items in
the directory.  It's always nice to know in advance how many things you
have.

|PERM: [g-][p-][c-][d-][r-] (case insensitive) where dash (-) means "can't"
|and letter means "can", the characters are (in this order):
|g - Can get (RETR) the file (or files inside this directory for a directory)
|  p - Can put (STOR) over this file (or in this directory for a directory)
|  c - Can cd (CWD) into this directory
|  d - Can delete (DELE/RMD) this file or directory
|  r - Can rename (RNFR/RNTO) this file or directory
 
|DATE: Modification date or closest approximation (create date?). 
|Formatted as 19940227114609 (or what is an ISO date format???)

The current format from 'MDTM' suits me.

Hmmm... the more general specs cover this.  I can't decide if I like
parsing the straight-forward "drwxr-x---"'s or would prefer straight
integers.

But, if we're going to get an RFC going, I'd like to suggest a 'FIND'
command.  This is long overdue!  Such a command, if supported, could
take a regular expression (if no special chars, then a case-insensitive
substring search) and have the remote host either grep it's "ls-lR.Z"
file, or actually search the file system.  Matches would be returned
in the format specified by XMRA.

--
===== Mike Gleason <mgleason@cse.unl.edu> ================= Go Huskers! =====
Current version of NcFTP is 1.7.3, and is available from cse.unl.edu, in the
/pub/mgleason/ncftp directory.  Public permissions are turned on daily at 5pm
CST;  To decode the gzip'd tar file, do "gunzip -c ncftp.tgz | tar xvof -". 

-----------[000369][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 16:07:09 GMT
From:      andy@max.uis.com (Oz)
To:        comp.protocols.tcp-ip,comp.mail.mime,demon.ip.support
Subject:   Re: Quickmail from CE Software ??

paulr@motiv.demon.co.uk ("Paul B. Richardson") writes:

>I have been told about a cross platform mail product called Quickmail
>by CE Software in the States.  Does anyone have contact details for
>this company?  Also does anyone have any experience with either:
>a. encapsulating the mail in IP for Internet transmission AND/OR
>b. using a Quickmail <-> SMTP gateway.
 
>Thankyou in anticipation of your replies.
 
>-- 
> --------------------------------------------------------------------
>| Paul Richardson            | Motiv Systems Ltd                     |
>| PaulR@Motiv.demon.co.uk    | Internet and Open Systems Consultancy |
>| Tel: +44 223 576318        | 22 Hills Road                         |
>| Fax: +44 223 576319        | Cambridge CB2 1JP                     |
> --------------------------------------------------------------------

CE Software is located in West Des Moines, IA.  The primary
number is (515) 224-1995.

        --andy

-----------[000370][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 00:08:38 -0400
From:      stewartc@cs.nyu.edu (C. A. Stewart)
To:        misc.legal.moderated,misc.legal,comp.protocols.tcp-ip
Subject:   Re: Green Card Update


Well, today the "Green Card Incident" made the New York Times.  It seems
that Mr. Canter is PLANNING TO WRITE A BOOK on....


		H O W   T O  A D V E R T I Z E   O N   T H E


			I N T E R N E T  !!!


I kid you not.   I get the feeling that current usenet members might 
want to form a more formal organization, with, say a charter and rules
for membership?  Any corporate lawyers out there willing to donate their 
services to draw up a charter with by-laws?  Anyone want to sort out the
rats nest of .mil, .org, .com, .edu, .se, .no, .fr, etc. etc. etc.?

I've been lurking on the usenet -> arpanet -> NSFnet -> 'usenet' since 1984.  
I DON'T want the internet to be turned into just another taxpayer-subsidized
junk-mail service, like the US postal service.  Maybe we could take a CS, 
approach, and just require that all ads be marked as such in the header.  

Oh!  Require an "ad" bit on IP packet header for easy filtering! Good idea! 

If individual news feeds (and gateways, and intermediate sites everywhere)
simply filter out the ads, well, maybe that's their legal perogative.  The 
forum might be public, but each individual newsfeed is privately (or publically)
owned... and each individual newsfeed can, if they want, gladly transmit ads 
while charging exhorbitant prices to the advertizer.  

It should be a price per packet!  Since a message gets split up into many 
packets, getting a whole advertizment through could become virtually 
impossible if enough intermediate machines are refusing to carry them.  
Retransmits everywhere, broadcast storms further enfuriating sysadmins!  
Drive the price way up since every dropped packet will be billed for by 
at least a few intermediate machines.  And then again on retransmit.  

They'll need a really big bucket for all of those dropped packets!

Cheryl




-----------[000371][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 19:15:57 +0000
From:      daveh@dhcs.demon.co.uk (Dave Hodgkinson)
To:        comp.protocols.tcp-ip
Subject:   RTT estimation - Berkeley cop out?


I'm looking after a TCP/IP written in 68k assembler, and I'm 
being bugged by the following question regarding round-trip
time estimation:

What is the effect of only performing RTT estimation on
one un-acked segment at a time?

I can't remember Jacobson, Karn mentioning it, I vaguely remember
Comer keeping a queue of times for each outgoing segment, but the
clincher is in Stevens where (p303!) he claims that Berkeley-
derived code only times one segment at a time. If a timer is 
already running when a new segment goes out, then no new
timer is started.

This has the effect that with a 32k window and 1500 byte packets,
only one in twenty segments might get timed. Or is it that 
frequent reception of ACKs (at least one for every two
segments according to rfc1122) means that we might hit one
in three?

Any offers?

Cheers,

-- 
Dave - Just another freakin' hacker...

-----------[000372][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 19:20:52 GMT
From:      apollo1@netcom.com (Doug A. Chan)
To:        comp.protocols.tcp-ip,biz.sco.general
Subject:   Hung rcp process preventing other rcp/rcmd from working?

I've noticed a strange problem on a platform running SCO Unix 3.2.4.2
with TCP v1.2.1.  If I try a rcp or rcmd (also known as rsh on some
other varieties of Unix) from another machine onto this platform, it
would just hang.  I found that there was an outgoing rcp that was
"stuck" (stayed on the system for several hours) on this platform 
and as soon as I killed it off, everything would return to normal.

-I already checked streams usage- no overflows/failures.
-Rlogin and telnet still worked even when rcp/rcmd failed.

Anyone have some ideas as to why this is happening?

-Doug
apollo1@netcom.com
apollo@world.std.com


-----------[000373][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 19:38:25 GMT
From:      raj@cup.hp.com (Rick Jones)
To:        comp.protocols.tcp-ip,comp.sys.hp.hpux,comp.dcom.lans.ethernet
Subject:   Re: netstat -i input packet errors?

Bruce Barnett (barnett@grymoire.crd.ge.com) wrote:
: I don't have HP sources, but I looked at the BSD sources.
: According to those, the AMD 7990 LANCE chip will increment network input
: errors on two conditions:

HP 700's use the Intel 82596...for both built-in and EISA Ethernet
connects.

rick jones

-----------[000374][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 20:50:51 GMT
From:      skibo@florida.wpd.sgi.com (Thomas Skibo)
To:        comp.protocols.tcp-ip
Subject:   Re: RTT estimation - Berkeley cop out?

In article <766782957snz@dhcs.demon.co.uk>, daveh@dhcs.demon.co.uk (Dave Hodgkinson) writes:
|> 
|> I'm looking after a TCP/IP written in 68k assembler, and I'm 
|> being bugged by the following question regarding round-trip
|> time estimation:
|> 
|> What is the effect of only performing RTT estimation on
|> one un-acked segment at a time?

You might want to take a look at RFC 1323 (TCP Extensions for High
Performance).  It discusses RTT estimations on connections that can
have many oustanding segments.  It also introduces a time-stamp option
to help achieve better RTT estimations.



-- 
---
Thomas Skibo		Media Systems Division
skibo@sgi.com		Silicon Graphics, Inc.

-----------[000375][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 21:40:03 GMT
From:      rockbttm@hebron.connected.com (Jeff Marrison)
To:        comp.protocols.tcp-ip
Subject:   Subnet Masks?????

Hi... I have written my own TCP/IP packet driver, and have based most everything
On RFCs... I can list the numbers if needed, for there are several that were
needed to create it, but I am currently unable to establish any kind of
connection... I believe all my fields for both the IP header and the
TCP header are correct, but I have seen several messages relating to
subnet masks, and I have not implemented such a thing. Need I mask
both the source/destination fields in the IP header? This could be my
problem if so, and if that is indeed the case, is there an RFC that
specifies the mask vals and how to calculate it? In the IP rfc, it
states the checking of the high order bits of the address to determine
the "class" of the address, A-D and E for extended addressing... but
it states nothing to the effect that I have to MASK it... 
A

Any help would greatly be appreciated. 


-----------[000376][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 94 22:41:57 GMT
From:      havens@CS.ColoState.EDU (Mark Havens)
To:        comp.protocols.tcp-ip
Subject:   FTP's PC/TCP DEVKIT code


I'm working with FTP Software's PC/TCP Development Kit, and I'm trying
to get a client/server process under MSDOS running. I need the client
to be a TSR, run in STREAM mode, and maintain a connection with the server.
Does anyone have experience using this product. I am utilizing tech support,
but they're kinda slow. Any suggestions are appreciated. Thanks in advance.

Mark

********************************************************************************
Mark & Tricia Havens     | _   /|  -ACK!    | havens@mozart.cs.colostate.edu   
                         | \'o.O'  -PFHHHT! | havens@wapa.gov
                         | =(___)= -COUGH!  | "Nobody gets me. I'm the wind,   
                         |    U    -ACK!    |  baby." - Tom Servo             
********************************************************************************

-----------[000377][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 22:55:37 GMT
From:      mcn@lanl.gov (Michael Neuman)
To:        comp.protocols.tcp-ip
Subject:   IP traffic statistics?

Are there any publicly accessible IP traffic stastitics? I'm interested
in a breakdown by protocol and port.

-Mike
mcn@lanl.gov

-----------[000378][next][prev][last][first]----------------------------------------------------
Date:      Tue, 19 Apr 1994 09:58:52 +0800
From:      peter.lewis@info.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   Re: Promposed new commands for FTP servers

In article <2osclp$s8a@news.utdallas.edu>, billy@utdallas.edu (Billy
Barron) wrote:

>As you probably remember from the past, I disliked your other method.
>The new commands documented here were much better.  Personally, I believe
>that this is more likely to be implemented than your other idea.  Nobody
>likes to change stuff, but this is independent.

Sure, but the other proposal is mostly aimed at people writing clients and
servers today, rather than existing servers (although I'd love to see the
NetWare server revved slightly to comply).

>May I recommend two things?  First, try to get an implementation into
>the WU-FTPD code.  Second, get this published as an experimental RFC.

Feel free.  I don't have the time to do this, and even if I did I would
make the changes to my own Mac FTP server, not some unix one.  And I
certainly don't have the time, skill, or patience to try to get something
thru the RFC process.

The base of the proposal is there, if people with sufficient ideology wish
to try to move it to an RFC state, then by all means, go for it.
   Peter.
_______________________________________________________________________
Peter N Lewis <peter.lewis@info.curtin.edu.au>       Ph: +61 9 368 2055

-----------[000379][next][prev][last][first]----------------------------------------------------
Date:      19 Apr 1994 22:46:04 +0100
From:      nreadwin@micrognosis.co.uk (Neil Readwin)
To:        comp.protocols.tcp-ip
Subject:   Re: Encyption of data between nodes across the net

In article <766598295.222snx@ozinkl.pc.my> Geoffc@Ozinkl.pc.my writes:>
>is there an easy way to implement some
>form of encryption, to ensure that proprietary data and other
>intelectual property passed between the two sites is protected, as the
>data passes across the net, where is might be intercepted by others?

Yes. Tunnel IP over TCP :-) Some PPP products allow this (eg
MorningStar). There are also commercial hardware products designed to
solve this problem. This came up on the firewalls list last week and
the names mentioned were Semaphore, UUnet (the LAN Guardian?) and the
commercial version of the KarlBridge (ie not the freeware version).

>This would also need to be specific, in that encryption was ALWAYS used
>and ONLY used when communicating to this corps systems, while still
>providing non-encrypted communication for net access to non-corporate
>systems.

Imagine it like this ... on internal net 1 you have an IP route to
internal net 2 which points to your box running PPP (or the
special-purpose hardware). That box encrypts the packets and sends them
over a TCP circuit to the remote site. A similar box at the remote site
decrypts them and pumps them out onto the local network.  The same
thing happens for packets coming back the other way. Either way, it only
affects packets bound for the other site.

Depending on how your firewall is configured you may need to open a small
hole for this (ie allow incoming connections from a pre-defined source to a
pre-defined destination with a fixed port number). Neil.
-- 
 nreadwin@london.micrognosis.com       Phone: +1 718 273 8234
 Anything is a cause for sorrow that my mind or body has made

-----------[000380][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 03:10:31 GMT
From:      jgs@hadron.merit.edu (John G. Scudder)
To:        comp.protocols.tcp-ip
Subject:   Re: IP traffic statistics?

In article <2p1nh9$34n@newshost.lanl.gov>, Michael Neuman <mcn@lanl.gov> wrote:
>Are there any publicly accessible IP traffic stastitics? I'm interested
>in a breakdown by protocol and port.

Try gopher://nic.merit.edu:7043/00/nsfnet/statistics

--John Scudder
  Merit/NSFNET

-----------[000381][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 03:16:36 GMT
From:      drukker@mundo.eco.utexas.edu (David M. Drukker)
To:        comp.protocols.tcp-ip
Subject:   Please Repost Word 6.0a PC-NFS .doc

	 I have precisely that problem with saving Word 6.0a documents from
within PC-NFS.  Could you please repost that document that you referred to,
or even mail it to me.  Many Thank yous,
DAVID M. Drukker
drukker@mundo.eco.utexas.edu

-----------[000382][next][prev][last][first]----------------------------------------------------
Date:      Wed, 20 Apr 1994 03:23:39 GMT
From:      sngmedia@world.std.com
To:        comp.protocols.tcp-ip
Subject:   MARKET SURVEY

We are a small, dedicated group of people desperately trying to get
a video production company off the ground.

Recently, we secured enough financial backing to produce several
videos on the subject of personal computing.

Because we don't have the financial resources to perform a
traditional market analysis, we considered several alternatives and
concluded that we should consult the people of the Internet, who
are technically proficient and tend to be enthusiastic personal
computer users.  

We would like you to indicate which three of the below described
videos would appeal to you most.

Please note that each video synopsis is numbered.  When you are
ready to send your response, please specify the numbers
corresponding to your three video selections, and use commas to
separate the numbers.

For example, if you choose selections four, eight, and eleven, the
body of your letter would simply be:

4,8,11

and nothing more.

When you send back your response, please specify "MARKET SURVEY" in
the subject field.  Then, direct your response back to
sngmedia@world.std.com.

Please realize that this is not an advertisement, or solicitation. 
We simply would like to know your response.  You will not be
contacted further.

Thank you in advance for your help and cooperation.

Tom and Randi Fecker
sngmedia@world.std.com.

Please select three videos that interest you most:

1. The Internet Pilot's Guide
A plain English guide to getting around on the Internet by veteran
users.  How to use Internet resources to send/receive e-mail, and
to search for documents and files relating to a particular subject. 
How to use LISTSERVs, GOPHER, ARCHIE, WWW and WAIS.  How to use FTP
and Telnet to access remote computers for file transfers, remote
logins, etc.  Also, learning how to observe Internet's
"netiquette."  And how to access multimedia Mosaic bulletin boards.

2. How to Make Money at Home With Your PC I -- Successful PC-Based
Businesses
Meet five successful entrepreneurs who started their own home-based
businesses using a PC.  Includes businesses that provide typing
services, mailing list management, horoscopes, bookkeeping, and
even a software developer who does not know how to program -- he
hires other people to implement his software designs.  Includes
ideas for two dozen home-based businesses you can start using your
PC.

3. How to Make Money at Home With Your PC II: Portfolio Management.
Whether you manage your own portfolio for investment gains, or you
take on clients for a consulting fee, a PC can give you powerful
tools for portfolio management.  Meet three successful home-based
portfolio managers who use their PCs to maximize profits.  They'll
inspire you, and share some tricks of the trade that can help you
land clients and get started.  We'll also show you how to determine
which software applications will meet your needs, and which online
services will satisfy your requirements.

4. How to Make Money at Home With Your PC III: Telecommuting
Learn how to use your PC to telecommute to companies across America
that are looking for specialized consultants with skills like
bookkeeping, software design and programming, editorial and script
writing, technical proofreading, graphic design, copywriting, and
more.  Includes interviews with three successful telecommuters who
use their PCs, faxes, modems and phones to serve distant clients,
and information about a telecommuting referral service that could
help you find employers.

5. How to Upgrade Your Hard Drive, Step-By-Step
From choosing your new hard drive, through ordering, installing and
testing it.  This plain English video shows you how to do the job
correctly, one simple step at a time.  Using industry-wide
standards, we'll show you what all hard drives have in common. 
Includes how to determine if your controller will support a
particular hard drive.  How to remove the old drive, and install,
format and test the new drive.  Includes a discussion on the
differences between the various standards.

6. How to Design Your Own PC
You don't have to be an engineer to design and build the PC of your
dreams.  This plain English video shows you how to figure out the
PC design that is best for you, how to specify components, how to
make sure they'll work together, and where to buy them.  You'll end
up with a top quality system that will save you money.

7. How to Build Your Own PC
Once you've designed your PC, we'll show you how to build it.  The
actual process will take you only a few hours.  Using an easy-to-
understand method, we'll show you how to inspect, install and test
components.  Includes tips and tricks from computer production
experts.  The technical skills can be easily mastered by almost
anyone, and you probably already own most of the tools you would
need.
8. How to Increase Your Computer's Memory
This plain English video shows you how to determine whether your
computer memory can be increased, and how to do the job correctly,
one step at a time.  You'll learn about industry-wide standards for
memory, how to configure additional RAM and cache, how and where to
buy RAM chips, and three ways to eliminate low-quality RAM chips.
Covers all phases of the process from opening your computer, to
testing your memory.  Includes discussions on how to ensure your
DOS set-up is able to access all available memory, and how to use
various memory management software applications.

9. How to Use MS-Windows 3.1
This powerful graphical user interface can help you work smarter
and faster, but the manual and the online tutorial that come with
Windows leave many questions unanswered.  This plain English, step-
by-step video will show you how to install Windows on your computer
and set it up to get optimum performance.

10.  How to Find a Job in the Information Age
A PC can give you an incredible advantage when you're searching for
a new job -- or even a new career.  But you have to know just how
it can help you.  In this video, an experienced employment
counselor will show you how to tap the power in your PC to find job
leads, create a winning resume and customized cover letters, tap
into databases and find bulletin boards that will lead you to job
openings, and use online services to research potential employers.

11. How to Install a Sound Card in Your Computer
Here's how to add incredible stereo sound to your computer with
step-by-step help. In plain English, you'll learn how to determine
if your computer can support a sound card, how and where to buy a
high-quality sound card.  How to open your computer, and install
and test the sound card.

12. How to Install a CD-ROM Drive in Your Computer
Using simple tools, this plain English video shows you how to
install a CD-ROM Drive in your computer.  You'll learn how to make
sure your computer can support a CD-ROM drive -- and what to do if
it can't.  Covers internal vs. external drives, how and where to
buy a high quality CD-ROM drive, what you need to know about
differing industry standards, preparing the drive bay, testing and
trouble-shooting.  Covers SCSI and IDE.

13. How to Fix the Most Common Computer Problems
Your computer serviceman may not want you to know this, but all you
need is the know-how you'll get from this video, simple tools, and
easily-obtainable diagnostic software -- and you can fix most
common problems you'll ever encounter with a PC.

14. What to Do When a Virus Strikes Your Computer
Viruses can come from almost anywhere: a colleague or friend's
disks, a network, a bulletin board, even commercial software.  If
you ignore the first warning signs, a virus can wipe out your data
and permanently damage your computer's memory.  In plain English,
this video will tell you how to scan disks for viruses, how to
check downloaded files from bulletin boards, how to set up a virus
prevention program for your home or office computer, and how and
where to buy the best anti-virus software.  We'll also cover the
pros and cons of the antivirus software in DOS 6.X and Windows 3.X,
how to use antivirus software, and more.

15. How Your PC Works: Inside the Case
Here's a down-to-earth explanation of how your PC actually
processes information, and what really goes on inside the case. 
You'll get a guided tour of the insides of a PC, learn about how
the various components work and how they communicate with each
other, and get a clear explanation of technical terms.  A must for
anyone who wants to really understand how to program, use and
repair a PC.

16. How to Create Config.Sys, Autoexec.Bat and Batch Files
These basic files can make it much easier to use your computer --
or cause incredible headaches if they are not written properly for
your particular software and peripherals.  Now you don't have to be
at the mercy of murky tech manuals, because we'll show you how to
create files that work for your system -- step-by-step, in plain
English.  You'll learn how to write, modify and test Autoexec.Bat
and Config.Sys files; and how to create batch files.

17. How to Add a Modem or Faxmodem to Your Computer
Here's the easy way to add a modem or faxmodem to your computer,
with step-by-step guidance from this plain English video.  You'll
learn how to determine if your computer can support a modem or
faxmodem, and what to do if it can't, how to choose and buy the
best modem or faxmodem, how to open your computer, and install the
modem or faxmodem, how to test it, how to quickly eliminate common
problems, and how to set your modem or faxmodem correctly.

18. How to Make Money at Home With Your Computer
The information age is opening up incredible new opportunities for
PC owners to make undreamed of money, using skills and knowledge
you may already have!  Here's inside information on the ten most
promising telecommuting jobs and 12 small businesses you can run
right from your home, using your PC.  Includes profiles of PC
owners who are actually running PC-based home businesses.

19. The Super-Highway Roadmap
This is your guide to where to go and what to see.  You can make
incredible contacts and gather powerful, valuable information on
the Internet, but the problem is that most people can't begin to
imagine the potential of something that seems so abstract.  This
plain English video will introduce you to the Internet, and make
these opportunities concrete.  Includes interviews with 7 people
who did the impossible by gathering information and making contacts
on the Internet.

20. How to Upgrade and Repair Your PC I: Basic
This is the video your repairman doesn't want you to know about! 
Since the components of most PCs are highly modular, PC repair is
easier than you think.  Just pinpoint the problem component, unplug
it, remove a few screws, replace it, and presto! You're in business
again.  This step-by-step video shows you how to pinpoint problems
and replace your PC's components, using ordinary household tools.

21. How to Upgrade and Repair Your PC II: Multimedia
Here's how to save big money on a PC with all the latest multimedia
peripherals.  You learn how to determine if your PC can be
upgraded, how to upgrade your video card and bus, and how to add a
CD-ROM drive, sound card, video accelerator, and more.  Presented
in plain English.  The procedures you'll learn require ordinary
household tools -- nothing fancy!

22. Plain English Guide to DOS 6+.
The powerful sub-programs buried deep within DOS 6.0 and higher can
help you work smarter and faster, but the manual and the online
tutorial that come with DOS leave many questions unanswered. This
plain English, step-by-step video will show you how to install DOS
on your computer and set it up to get optimum performance. In
addition to DOS commands, you'll learn how to use the shell,
defragmentation, scan and antivirus programs that come with DOS.

23. Home Financial Management on a PC.
Your computer can help you create and manage a budget, keep track
of your credit card accounts, handle your tax deductions, and
reconcile your bank accounts.  But that's not all!  You can also
determine whether you should pay down your mortgage, finance a new
car or pay cash, buy or rent your home, and how much you'll need
for retirement.  The financial information your computer can give
you might mean the difference between just getting by and a very
comfortable lifestyle -- if you ask the right questions and use
your PC to develop a financial strategy.

24. The Online Bulletin Board Guide
Bulletin boards can be the on-ramps to the Information Super
Highway -- if you know how to access and use them.  This step-by-
step guide shows you how to find bulletin boards, set-up your
modem, log on, find out what they have to offer, find bulletin
board users who share your interests, search for information, and
upload and download files.


Thank you.

-----------[000383][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 03:26:40 GMT
From:      jgs@hadron.merit.edu (John G. Scudder)
To:        comp.protocols.tcp-ip
Subject:   Re: IP traffic statistics?

(talking to yourself again, John?)

In article <2p26f7$rsn@lastactionhero.rs.itd.umich.edu>,
John G. Scudder <jgs@hadron.merit.edu> (that's me) wrote:
>In article <2p1nh9$34n@newshost.lanl.gov>, Michael Neuman <mcn@lanl.gov> wrote:
>>Are there any publicly accessible IP traffic stastitics? I'm interested
>>in a breakdown by protocol and port.
>
>Try gopher://nic.merit.edu:7043/00/nsfnet/statistics

Hmm, just tried this myself (I had just clipped the URL from Mosaic
after navigating there by hand) and for some reason it doesn't work for
me.  The following will definitely work:

gopher://nis.nsf.net   --or--   ftp://nis.nsf.net
<choose NSFNET> 
<choose statistics>

--John Scudder

-----------[000384][next][prev][last][first]----------------------------------------------------
Date:      Wed, 20 Apr 1994 09:16:04 EST
From:      Ray Hunter <RHUNTER@ESOC.BITNET>
To:        comp.protocols.tcp-ip
Subject:   Is an ALL 1s UDP bcast on a DEC VAX possible?

OK ip gurus, here's a question for you.

We have some home brew code on site that performs UDP broadcasting to
distribute information on server processes (the detail isn't important)

It has been running quite happliy for years on a subnet mask of 255.255.255.0
(Class B + 8 bits subnet) & subnet directed broadcasts of net.subnet.255.

We are now looking to restructure our network & would like to use all 1s
broadcast absolutely everywhere, and let our Ciscos sort out any cross-
segment broadcasting. This is because we want to run more than on logical
network on a physical segment.

My programmer chums are telling me that although they set up all ones
bcast WITHIN the machine (DEC VAX running VMS 5.5? & DEC UCX), the
broadcast appears on the net as net.1s (131.176.255.255)

Can other people confirm this behaviour (is the VAX UCX broken?) or
give us some configuration tip to overcome this?

Any help appreciated.

Ray Hunter
rhunter@esoc.bitnet

-----------[000385][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 94 09:46:09
From:      hafner@informatik.tu-muenchen.de (Walter Hafner)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,de.comm.internet,comp.protocols.ppp
Subject:   help with ka9q setup needed

Hello out there!

I need a little bit help with my ka9q installation. It seems, I can't get
ka9q to dial out!

My configuration: 386DX/16 (Yes!), MD-DOS 5.0 with 4DOS 5.0D, QEMM 5.01
and a couple of TSRs. The ka9q version is that of the e920603.zip file.
My modem is a ZyXEL 1496E with the 5.01 firmware.

The ka9q setup is as follows:

ftpusers:

: hafner <password> / 7
: anonymous * /ka9q/public 3

domain.txt:

: augusta.de             IN A 193.175.25.17
: forwiss.tu-muenchen.de IN A 131.159.128.1

autoexec.net:

: # logging from the start
: log netlog
: 
: # local name
: hostname dartpub.augusta.de
: 
: # local IP address
: ip address 193.175.25.87
: 
: # IP datagramm time-to-live in hops
: ip ttl 30
: 
: # raise TCP roundtrip time (10 sec.)
: tcp irtt 10000
: 
: # TCP max. segment size.
: tcp mss 1024
: 
: # AT type
: isat on
: 
: # net setup
: # asy - ioaddr - vector - slip - iface - bufsize - mtu - speed - crv
: # (Defaults for COM1)
: # crv: CTS/RTS detect. - RLSD carrier detect. - Header compression
: attach asy 0x3f8 4 slip sl0 1024 256 19200 crv
: 
: # How far to send broadcasts?
: ifconfig sl0 broadcast 255.255.255.0
: # Interface <-> Name (not necessary)
: # ifconfig sl0 ipaddress dartpub.augusta.de
: # We've got a class-C net
: ifconfig sl0 netmask 255.255.255.0
: 
: # Routing (default is sufficient)
: route add default sl0 augusta.de
: # route add 193.175.25/24 sl0 augusta.de
: 
: # Domain Name Server. From domain.txt
: #domain add forwiss.tu-muenchen.de
: domain add augusta.de
: 
: # starting Servers (allow remote operations)
: start echo
: start finger
: start ftp
: start remote
: start telnet
: start ttylink
: 
: # dial!
: # dialer sl0 dialer

My dialer file is that of the docu ... I fiddled around with almost every
command ...

The (quite old) docu says about the 'dialer' command:

dialer <iface> [<dialer-file> [<seconds> [<tests> [<hostid>]]]]

The online-help says:

dialer <iface> <timeout> [<raise script> <lower script>]

From the docu I think I understand the parameters. But what do the
parameters of the online-help mean? And what format is the right one? In
doubt I tend to believe the online-help.

However - which format I use - the modem does simply noting. I tried to
send an init-string but got no response ('wait "OK"'). From the 'asystat'
command I KNOW I configured and attached my COM1 the right way: When I
switch my modem on and off, 'asystat' recognizes the change in the
RTS/CTS and the DTR/DSR.

But the modem simply doesn't dial out!

I don't have any clue. Am I a looney, who is simply too stupid? Didn't I
understand the manual correctly? Is the example dialer-file of the docu
wrong? Any hints? Could some kind person send me a commented working
dialer-file?

Thanks in advance,

-Walter

PS: Followup to comp.protocols.tcp-ip.ibmpc
--
Walter Hafner__________ hafner@forwiss.tu-muenchen.de
FORWISS Muenchen,  Forschungsgruppe Kognitive Systeme
Raum O-134,                        Tel: 089/48095-220
_____________________________________________________

-----------[000386][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 94 09:57:25 GMT
From:      jraja@snakemail.hut.fi (Jarno Tapio Rajahalme)
To:        biz.sco.general,comp.protocols.tcp-ip,comp.sys.amiga,comp.sys.amiga.programmer
Subject:   Re: rwhod warped with an Amiga

In article <2p0kkm$8os@aggedor.rmit.EDU.AU> s900387@minyos.xx.rmit.EDU.AU (Craig Macbride) writes:

>At work, we have a machine running SCO Unix and another running SCO Xenix
>on a network. (There are also the inevitable MeSsyDOS and Novell toys.)
 
>Now, these SCO machines both run rwhod happily and will keep the information
>up to date as appropriate and have done so for months.
 
>Recently, someone (speaking of toys) connected an Amiga to the network
>running some TCP/IP package which does some strange things. When the Xenix
>machine sends out a udp who message, the Amaga sends some sort of broadcast
>in response! After this broadcast, the Xenix machine continues to operate
>normally, but the Unix machine's rwhod refuses to process any more incoming
>who packets. (It still generates the outgoing ones as usual.)

Just if the TCP/IP package happens to be AmiTCP/IP. AmiTCP (the
protocol stack) uses the BSD net/2 protocol code, and thus doesn't
send ICMP replies to broadcasted udp packets. So, if the Amiga sends
something back, there must be some daemon installed (by whoever did
install the AmiTCP/IP on your Amiga in the first place). On the Amiga,
you might check the file AmiTCP:db/inetd.conf to see if there is
something strange there.

>Does anyone have any idea why the Amiga might be responding to a who message?

My guess is that someone has installed a poorly written rwhod
(AmiTCP/IP distribution doesn't have a rwhod), or then someone has
messed with the AmiTCP:db/services (and mixed port numbers).

>Can I just put it down to bad programming, along with the lack of ANSI or
>vt100 arrow keys and function keys under telnet and the lack of an "lcd"
>command in ftp?!

Maybe, but please note that the ftp and telnet clients you are
referring to are not written by the AmiTCP Group, but by some other
individuals (as you can see by reading the AmiTCP:COPYRIGHTS file).
They are included in the AmiTCP binary distribution because there was
nothing better available at that time. The current AmiTCP version
(2.3) has the NcFTP, which you might want to use instead of the "ftp".

>The real curiosity to me is why the Amiga reacts to the Xenix machine's
>who packets, but seems to ignore the ones from the Unix machine and why the
>Unix rwhod takes the Amiga's broadcast so personally that it ceases to listen
>to any producer of user information (including itself) thereafter.

Maybe you should update the rwhod of your Unix box? (Hmm. why don't
you mention the "bad programming" here?)

	Jarno
-- 
--------------PLEASE NOTE THE NEW ADDRESS AND TELEPHONE NUMBER!--------------
| Address: Jarno Rajahalme            | EMail:                              |
|          Servin Maijan tie 12 G 102 |   Jarno.Rajahalme@hut.fi            |
|          02150  ESPOO, FINLAND      | tel: +358 0 468 3111                |
-----------------------------------------------------------------------------



-----------[000387][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 17:57:32 -0400
From:      moe@netcom.com (Morris Bisted)
To:        misc.legal.moderated,misc.legal,comp.protocols.tcp-ip
Subject:   Re: Green Card Update

In article <2p29s6$qhg@panix.com>, C. A. Stewart <stewartc@cs.nyu.edu> wrote:
>
>Well, today the "Green Card Incident" made the New York Times.  It seems
>that Mr. Canter is PLANNING TO WRITE A BOOK on....
>
>
>		H O W   T O  A D V E R T I Z E   O N   T H E
>
>
>			I N T E R N E T  !!!


Net folks,

This is an informational posting stemming from 3 phone calls totalling 
approximately 120 minutes to Canter & Siegal(C&S).  There is some information
in this post that tends to muddy the waters on the issue of complete
and sole responsibility for a breach of net.etiquette.  There is only
editorializing in one place which will be prefaced with m>.  

As there have been threats made on their lives and property, I will
not publish the name of the person to whom I talked.  However if you
have questions, I will pass on your information to them, and they can
contact you should you and they mutually desire it. 

Note: I, in no way, condone posting to every news group. I condemn
	threatening murder, mayhem, and arson in response .

Note 2: there are an abundance of threads that don't speak directly
to this issue.  I did not want this posting to get "lost in the sauce", so
"your indulgence please"  if this offends your sensibilities.


1.  C&S state regarding the mass posting...

	They approached 3 internet providers (Winnet, Pipeline,
	and Internet Direct) and met with them regarding a mass posting.    
	Apparently one internet provider, netcom, suggested they not
	do it.

	The response was (paraphrased)... We really don't care what you
	post.  You can do anything that you like as long as it doesn't
	hurt our company.

	C&S point out, that there was no content or breadth response
	raised.  The companies only had objection if it hurt their business.  

2.  C&S state regarding the response to the mass posting.
	They would not mass post again.  People commenting on the
	utility of the post, but concerned about its mass nature did not
	collectively agree on what was an appropriate dispersal. 

	There seems to be no consensus on where it was appropriate
	to post.  Some people say post in:
	1) alt.visa.*
	2) alt.visa.* and soc.culture.(country)
	3) alt.visa.* and soc.culture.(country) and relcom.*

3.  C&S state that the posting threatening individual net users with a suit for
"obvious libel attempted conspiracy" reposted in "alt.visa.us" (excerpted
below) is a FORGERY and  was not written or sent by (Laurence Canter) or C&S.

Begin Excerpt:
#
#[...]
#Newsgroups: alt.visa.us
#Subject: Re: Bycott Canter & Siegel !!!!
#Date: 14 Apr 1994 09:46:14 -0500
#[...]
#Message-ID: <sib1.766333053@Ra.MsState.Edu>
#Keywords: Canter & Siegel
#[...]
#-------------------------------------------------------------------------
#Mailer: WinNET Mail, v2.04
#Reply-To: cslaw@lcanter.win.net (Laurence A Canter)
#Date: Thu, 14 Apr 1994 06:59:52
#Subject: Re: Bycott Canter & Siegel !!!!
#
#Thank you so much for providing your name to your obvious libel and
#attempted conspiracy.  It is about time you people learn that
#there are no laws prohibiting advertising on the net, but there are
#plenty of laws, both state and federal, criminal and civil, that
#prohibit defamation, interference with business, junk fax, obscene
#phone calls, etc.
#
#[other stuff deleted]
#[...]

End Excerpt:

4. C&S state that the responses have been ...
	a, Death threats		
	b, Their house should be burned down.
	c, A hot poker should be stuck up her ass
	d, Obscene phone calls
	e, Fax bombing
	f, Mail bombing (email sending of large files)

m> "Death threats", "burning houses down", and "sticking hot pokers up her ass" 
m> are personal violent attacks of one person on another stemming from 
m> a C&S posting about immigration.  I am stretched to find one reasonable
m> explanation of these threats.  These responses are barbaric and although
m> I personally disagree with C&S's actions, I can in no way condone or
m> give any sympathy to people that respond in this way.  A posting that
m> can be deleted with 1 keys stroke or 5000 keystrokes, does not warrant
m> death, or loss of property, or "hot pokers's up someone's ass".  It does
m> not even warrant the threat of such action.  Think about it.  The activities
m> comtemplated by the folks making these threats are murder, arson, and
m> mayhem.  
m>
m> Regarding the Obscene phone calls.  This type of response, ripe with
m> vulgar language and intended to scare and intimidate, again resulted
m> from an inane posting.  This cowardly  response is not quite as 
m> anti-social as murder, but still is quite uncivilized and juvenile.
m>
m> I would hope that the usenet community would join me in condemming
m> any further response in this realm.  There was no violence nor
m> threat of violence visited upon anyone reading the C&S postings.  The
m> response of violence or threat of violence is unacceptable.  For those
m> that were impacted monetarily there are other appropriate responses.
m>
m> In addition, the internet is used by a small minority of U.S. citizens.
m> Comments on this issue and others issues have been in the New York times,
m> NewsWeek and other national media.  The masses as a whole after seeing
m> this activity, will have more evidence to show the lawless, uncivilized
m> nature of this medium.  The Feds answer, if the responses to the
m> C&S posting are found to be representative of the users as a whole is
m> going to be regulation.  I don't think that the U.S. or the 
m> internet community benefits if there is excessive regulation stemming 
m> from the activities of an irresponsible minority.

Morris (moe@netcom.com)



-----------[000388][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 11:56:21 GMT
From:      jfjr@mbunix.mitre.org (Freedman)
To:        comp.protocols.tcp-ip
Subject:   Re: Encyption of data between nodes across the net

In article <766598295.222snx@ozinkl.pc.my> Geoffc@Ozinkl.pc.my writes:
>A hyperthetical question (which may have a later practical use)
>
>Consider.. The Internet, and a corporation which has Internet connection
>on one side of the world, and an office of same corp, on the other side
>of the world, with another Internet connection thru a third-party
>Internet access provider.
>
>This hyperthetical corp would like to communicate across the net, as an
>alternative to a primative uucp over X.25 in use now(at 2400bps).
>
>Both sides use Cisco routers, and Unix systems.
>
>Without extensive changes to all TCP stacks on all workstations, and
>machines connected on each side, is there an easy way to implement some
>form of encryption, to ensure that proprietary data and other
>intelectual property passed between the two sites is protected, as the
>data passes across the net, where is might be intercepted by others?





Check out swIPe, a network layer security protocol that is in the
public domain, or close to it. It was developed by J. Ioannidis at
Columbia, P. Karn at Qualcom and M. Blaze at Bell. 

I believe there are some proprietary protocols too look at NETLOCK by
HP (although I am not entirely sure of that)

 We have done some work wit swIPe. I think it will do what you wan.



                           Jerry Freedman,Jr 





-----------[000389][next][prev][last][first]----------------------------------------------------
Date:      Wed, 20 Apr 1994 12:30:24 GMT
From:      sheersc@cwi.nl (Simon Heerschap)
To:        comp.protocols.tcp-ip
Subject:   HTTP

I have to write a HTML server during my stage and now I'm looking for information
about the HTTP protocol with some example request en respons messages. I would
like to know how a connection is being setup.

e-mail : sheersc@cwi.nl


-----------[000390][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 21:12:24 -0400
From:      alight@panix.com (Alan Light)
To:        misc.legal.moderated,misc.legal,comp.protocols.tcp-ip
Subject:   Re: Green Card Update

moe@netcom.com (Morris Bisted) writes:

>2.  C&S state regarding the response to the mass posting.
>	They would not mass post again.  People commenting on the
>	utility of the post, but concerned about its mass nature did not
>	collectively agree on what was an appropriate dispersal. 
 
>	There seems to be no consensus on where it was appropriate
>	to post.  Some people say post in:
>	1) alt.visa.*
>	2) alt.visa.* and soc.culture.(country)
>	3) alt.visa.* and soc.culture.(country) and relcom.*

Total BS.

This is a direct contradiction of what Mr. Canter stated in the NYT.
He was TOTALLY unrepentent regarding the mass posting. He also stated
that he would "definitly" do it again, and furthermore, suggests that
he plans to instruct others on how to do the same.

Also, am I to infer from the above that the reason that the mass posting
occurred as because they couldn't decide on what group(s) to post it to?

Come on, that is an insult to  the intelligence of everyone reading this.




-- 
/-\ |_ /-\ |\|  Alan Light  alight@panix.com
(finger alight@panix.com for PGP public key)

 


-----------[000391][next][prev][last][first]----------------------------------------------------
Date:      Wed, 20 Apr 94 13:34:42 BST
From:      walmsley@VNET.IBM.COM (Adrian Walmsley)
To:        comp.protocols.tcp-ip
Subject:   bootptab entries

I'm trying to use a bootp server on an AIX system to distribute
IP addresses and PC/TCP authentication keys to PCs running
FTP Systems' PC/TCP.
The PCs have fibre ethernet cards which are connected via a bridge
to an fddi backbone which is connected to the AIX server.
Question: how should I set the :ht tag in bootptab?
 Should :ht=ether:? Or should it define the fddi adapter?
          If so, what's the setting of :ht for an fddi card?
When I set :ht=ether:,
I can see by starting the bootp daemon in debug mode that it is
receiving the bootp requests from the PCs but its response never
arrives at the pc.

Adrian Walmsley            (+44)-256-344698
IBM UK Ltd                  walmsley@vnet.ibm.com
AL 15, PO Box 32, Basingstoke RG21 1EJ, UK

-----------[000392][next][prev][last][first]----------------------------------------------------
Date:      Wed, 20 Apr 1994 16:08:39 GMT
From:      aab@cichlid.com (Andy Burgess)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   56K leased line VS V.FAST or has anyone measured latency lately?

Here is an excerpt from one of the FAQs from a local internet provider. 
I've asked a few questions at the end.

> Q4: "What's so great about 56K, when it's only twice as fast as V.FAST
> modems?"
> 
> A: Bytes-per-second throughput is only half the answer. For relatively
> simple "download" situations, flat-out speed transferring bulk data
> is the answer. However on the Internet interactivity is the norm,
> and client/server software puts great demands on command/response
> type activity. The limiting factor is not just bytes/sec, but
> *latency* -- ie. how much delay is there?
> 
> The best modems on the market now have latencies between 100 and
> 300 milliseconds, with most v.32bis models around 250-280, and
> v.FAST around 150. This is the lag you see on an otherwise idling
> system while typing in telnet. Latency determines the overall
> response of the network, not bulk throughput. For short packets,
> which are common in interactive applications like remote logins,
> email, WWW, Gopher, or Usenet news, your network's latency determines
> how responsive the network feels to you -- how much delay you see
> whenever you ask for something.
> 
> Latency on a 56K link is about 10 milliseconds (1/100th of a second),
> virtually unnoticable, and doesn't vary much with load (though
> "ping" times will of course stretch out on a busy system). Modem
> latencies vary highly with load, and the amount of instantaneous
> bi-directional data, since modems are software-based. Latency on
> a T1 link is about 3 milliseconds, as measured informally.


Has anyone measured or seen published measurements of 14.4Kb or 28.8 Kb 
modem latency? 

Does the latency of different modem brands really vary this much 
(100-300 msec)?

What is a good test of latency? I can ping my internet provider (8 byte 
packets) in 150msec. Can anyone report better results? Is ping a good 
test? (BTW my modem is Intel 144/144E, my machine is Sun SS1+ SunOS 
4.1.3, the other modem is a US Robotics Sportster, the other machine is 
486DLC40 NetBSD and there is a Livingston portmaster at the other end 
too)

I've used telnet via PPP between two machines connected to the internet 
(total of 4 14.4Kb modems involved: two to my internet provider, a few 
T1 lines from CA to OR, and two 14.4Kb modems to the other site) and 
interactive performance does really suck. Would 28.8Kb help latency? 
What if only (say) my end went to 28.8? Would running two phone lines in 
parallel help interactive performance (assume the packets get routed so 
that both lines are equally loaded)?

Is this routing assumption valid? Is it hard to make common TCP/IP 
implementations share multiple SLIP/PPP lines?

Thanks for any information you can provide.

-- 
Andrew A. Burgess                    SLIP internet connections
aab@cichlid.com                      UUCP email and newsfeeds
                                     Santa Cruz CA and vicinity
                                     Email info@cichlid.com for info

-----------[000393][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 19:37:58 GMT
From:      willeva@falcon.st.usm.edu (Evan Willet)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.os.ms-windows.networking.windows,comp.os.ms-windows.networking.tcp-ip
Subject:   Re: Question about UNIX end of WINSLIP/SLIP/PPP

Richard D Dietz (rddietz@blackmail) wrote:


: Here's my problem:  I've downloaded everything I need on the PC
: end and believe I can set it up ok myself, but the network I want
: to connect to does not offer SLIP or PPP.  I have located (via archie)
: and downloaded UNIX source for both, but each requires "root"
: access to install (which I do not have).  I do have an account on 
: the network, so I was wondering if anyone knew of some kind of
: hack or work-around that would allow me to run PPP (or maybe even
: SLIP) from my user account without any installing at the system
: level?

There is no way to get a slip link without cooperation from the system 
administrator of the UNIX machine on the internet.

There is however something called term, which is almost slip, except 
it does not require root access nor a registered ip address.  Term 
uses sockets for all clients such as telnet, ftp, mosaic etc.

Unfortunately, term only runs on unix machines, and I don't think a
port is comming to MS-windows any time soon.  There is a free unix 
for PC's called linux(see sunsite.unc.edu:/pub/Linux/docs for info on 
linux).  Try out linux, it's the best thing since transitors replaced 
vacuum tubes.

Ev

-----------[000394][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 20:10:11 GMT
From:      km@mathcs.emory.edu (Ken Mandelberg)
To:        comp.protocols.tcp-ip,com.dcom.lans
Subject:   Optivity and Subnets

We are about to add a Class B address which will be subnetted
with a 10 bit subnet and and 6 bit host field (netmask 0xffffffc0),
This should give use 2^10 -2 = 1022 subnets. Our network service
group is planning on how to divide up the subnets and has concluded
that only 508 of them are actually usable.

They make heavy use of Synoptics Optivity product and have been told
that Optivity won't work properly with any subnet of the form
XXXXXXXX00 or XXXXXXXX11 (ie the 2 bit piece that intersects the last
octet is 00 or 11). 

Frankly, I find this hard to believe. Can anyone shed some light on this.

---
Ken Mandelberg      | km@mathcs.emory.edu          PREFERRED
Emory University    | {rutgers,gatech}!emory!km    UUCP 
Dept of Math and CS | km@emory.bitnet              NON-DOMAIN BITNET  
Atlanta, GA 30322   | Phone: Voice (404) 727-7963, FAX 727-5611



-----------[000395][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 94 20:33:25 GMT
From:      terry@spcvxb.spc.edu (Terry Kennedy, Operations Mgr.)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   Re: 56K leased line VS V.FAST or has anyone measured latency lately?

In article <CoKFIG.AEq@cichlid.com>, aab@cichlid.com (Andy Burgess) writes:
> Has anyone measured or seen published measurements of 14.4Kb or 28.8 Kb 
> modem latency? 

  I've done extensive testing of a few configurations. For example, using
TeleBit NetBlazer PN2's with Microcom DeskPorte FAST modems (28.8 DCE,
115.2 DTE), the times reported with ping are about 100ms on an unloaded
link. Note that DEC's DECserver 700 uses an 80ms timer to decide when to
send packets from the terminal server to the host, so 100ms isn't unreason-
able for interactive use.

> Does the latency of different modem brands really vary this much 
> (100-300 msec)?

  It depends on whether they're doing compression, encountering errors, etc.
It also depends on the processing power of the modem. For example, a modem
with an 8031 microcontroller will usually be slower than one with an 80C186
general-purpose CPU. You can tune a lot of this by playing with the modem's
settings - turning off compression makes less work for the modem, and if your
data is already compressed you won't get much benefit from modem-based com-
pression anyway.

> I've used telnet via PPP between two machines connected to the internet 
> (total of 4 14.4Kb modems involved: two to my internet provider, a few 
> T1 lines from CA to OR, and two 14.4Kb modems to the other site) and 
> interactive performance does really suck.

  Are you sure the problem is entirely in the PPP link segments and not out
on the backbone? Have you tried CSLIP? (Some implementations are better at
SLIP/CSLIP than PPP, performance-wise). Also, if the system on the other end
is doing a FTP transfer or something similar, performance will be bad even
on a 56Kb line (I've observed ping times of over 4000ms on a 56Kb line).

> Would 28.8Kb help latency? 

  It depends on where the bottleneck is. Since a given packet will go out
in half the time (28.8 vs. 14.4), that will help. However, if the delay is
from retransmissions or congestion on the backbone, it won't help.

> What if only (say) my end went to 28.8?

  Unless the modem you were replacing was woefully under-powered, CPU-wise,
this shouldn't make any change.

> Would running two phone lines in 
> parallel help interactive performance (assume the packets get routed so 
> that both lines are equally loaded)?
> Is this routing assumption valid? Is it hard to make common TCP/IP 
> implementations share multiple SLIP/PPP lines?

  Telebit NetBlazers do this for PPP lines. They can also bring up additional
circuits only when needed to save phone charges.

	Terry Kennedy		  Operations Manager, Academic Computing
	terry@spcvxa.spc.edu	  St. Peter's College, Jersey City, NJ USA
        +1 201 915 9381 (voice)   +1 201 435-3662 (FAX)

-----------[000396][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 20:54:33 GMT
From:      timur@seas.gwu.edu (Timur Tabi)
To:        comp.protocols.tcp-ip
Subject:   What is NetBIOS?

Subject says it all.  I know it has something to do with peer-to-peer,
since IBM's TCP/IP for OS/2 says that it's NetBIOS provides p2p service.
-- 

------------------------------------------------------------------ Timur Tabi
Contributing Editor for "OS/2 Monthly"        Internet:    timur@seas.gwu.edu
                                              Fidonet: Timur Tabi @ 1:109/347

-----------[000397][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 22:58:54 GMT
From:      kwia4000@bronto.zrz.TU-Berlin.DE (Manfred Kwiatkowski)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   Re: 56K leased line VS V.FAST or has anyone measured latency lately?

In article <CoKFIG.AEq@cichlid.com>, aab@cichlid.com (Andy Burgess) writes:
> 
> [..]
> Does the latency of different modem brands really vary this much 
> (100-300 msec)?

Yes. Analog modems need enough horsepower to handle the LAPM protocol
and sophisticated algorithms to decide on the packetizing strategy.
*Bargins* often lack one or both and thus exhibit an avg. 60% the
performance of name brand modems. Actually an asynchronous link is
3 consecutive links. E.g.:

-----  57.6k  -----          14.4k         ----    57.6k    -----
|    +-------+     +----------------------+    +---------- +    |
-----         -----                        ----             -----
host          modem                        modem            host

So you start out with 3 times the delay you encounter on a 
synchronous link of similar speed. The tranmitting modem (if capable) 
will packetize on the fly thus minimizing the first delay to (at best) 
2 bytes. Nevertheless the longer the LAPM-packet gets, the longer the 
third delay gets as the receiving modem has to wait for the checksum. 
Sending short packets instead means increased overhead (less effective 
bandwidth) and more need of processing power for sustained throughput. 
Increasing the DTE-DCE speed will do any good only if modem and host 
are capable of simultaneous processing (not just buffer) the data at 
higher speeds. So the actual delay of a specific byte varies with the 
packet-mix sent in conjunction with the modem characteristics. 
Any delay times published should therefore taken with a grain of salt. 
> 
> What is a good test of latency? I can ping my internet provider (8 byte 
> packets) in 150msec. Can anyone report better results? Is ping a good 
> test? (BTW my modem is Intel 144/144E, my machine is Sun SS1+ SunOS 
> 4.1.3, the other modem is a US Robotics Sportster, the other machine is 
> 486DLC40 NetBSD and there is a Livingston portmaster at the other end 
> too)

The ping delay of a 56k link is some 35ms. If I take the above example 
I would consider anything from 4-6 times this delay as satisfactory. 
If in addition I get wire speed with ftp on compressed files and no 
type-ahead effect on an idle link I would not bother.
> 
> I've used telnet via PPP between two machines connected to the internet 
> (total of 4 14.4Kb modems involved: two to my internet provider, a few 
> T1 lines from CA to OR, and two 14.4Kb modems to the other site) and 
> interactive performance does really suck. Would 28.8Kb help latency? 
> What if only (say) my end went to 28.8? Would running two phone lines in 
> parallel help interactive performance (assume the packets get routed so 
> that both lines are equally loaded)?

Any low speed (< T1) link will suck without priority queueing (either 
on length or by peeping into the protocol field). An interactive packet 
will have at least one TCP window full of bulk-transfer data in front 
of it. This is usually more than the 300ms delay considered tolerable...

-- 
Manfred Kwiatkowski         kwiatkowski@zrz.tu-berlin.de

-----------[000398][next][prev][last][first]----------------------------------------------------
Date:      20 Apr 1994 23:12:29 -0000
From:      sunborn@news.delphi.com (SUNBORN@DELPHI.COM)
To:        comp.protocols.tcp-ip
Subject:   Class B address with class c subnet mask, Why ?

I ran across an installation where they have a class B address but are 
using class C addressses.  Is there any good reason for this ?


-----------[000399][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 11:19:19 -0700
From:      lelina@kaiwan.com (Rico Lelina)
To:        comp.protocols.tcp-ip
Subject:   Multicast support in WinSock

Will it happen?  When?

-------------------------------
Rico Lelina (lelina@kaiwan.com)
"Blueberry bagels on Monday."

-----------[000400][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 01:06:40 GMT
From:      jna@bu.edu (John Adams)
To:        comp.protocols.tcp-ip
Subject:   Novell and IPX

If this is the wrong group, my apologies.

Has anyone had experience in tunnelling IPX packets, coming from a novell
server into IP for transmission via PPP ? 

I'm trying to find a way to connect a remote novell client (some pc in
the middle of nowhere) via PPP to our local (PC) novell network. 

The network has both IP machines (unix boxes) and IPX machines (novell
servers and clients) on the same wire. We accomplish this via 
ODIPKT (Clarkson Packet Ethernet Drivers)... 

I Just don't know how to move IPX over a PPP link! 

thanks!
		-john

-----------[000401][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 09:30:30 -0500
From:      ghelmer@alpha.dsu.edu (Guy Helmer)
To:        comp.protocols.tcp-ip
Subject:   Access controls (blocking particular ports) on Proteon routers?

Apologies if I'm using an incorrect newsgroup for this query.

I'm running what I believe is version 11 of Proteon's software on a 4100+ 
router, and I'd like to block TCP/IP access to specific ports.  For 
example, I'd like to block all access to NFS, rlogin, and rsh.

Alas, the documentation appears unclear, confusing, and ambiguous.  It
appears I need to "set access-control" to "exclusive", but the "add
access-control" configuration stumps me; for example, to block rlogin, do
I enter: 

add access-control E 0.0.0.0 0.0.0.0 128.128.0.0 255.255.0.0 6 6 0 513

to block all packets from any TCP port on any IP addresses to TCP port 
513 in network 128.128 (this number to be replaced by my actual network 
number)?

and for NFS:

add access-control E 0.0.0.0 0.0.0.0 128.128.0.0 255.255.0.0 17 17 0 2049

The documentation says the syntax of the command is:

add access-control <type> <IP-source> <source-mask> <IP-dest> <dest-mask> \
	[<protocol> <protocol>] [<port> <port>]

I'm most unclear about the <type> parameter (an example shows it being 
"I", which I assumed stood for "Inclusive") and whether a <port> 
value of 0 matches all ports.

Any help at all would be appreciated.

-- 
Guy Helmer, Dakota State University Computing Services - ghelmer@alpha.dsu.edu

-----------[000402][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 14:14:31 -0700
From:      craign@teleport.com (Craig R. Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Books on Winsock or Sockets programming?

In <21621MEFBKBASHMMHAS@gcomm.com> stein@gcomm.com writes:

>NNTP: RFC's 977 (NNTP), 850 (USENET), and 822 (messaging)
>      (IOW, I still haven't found a good explanation for NNTP
>      other than the RFC's themselves.  But try RFC 977, it's
>      not so bad.)

Bob,

For what it's worth RFC 850(USENET) has be "obsoleted" by RFC 1036, at 
least it is where *I'm* getting my RFC's. Still, the finger is pointing 
in the right direction. The 977 RFC is good for people writing their nntp 
client/server programs. I'm in the process of implementing an nntp news 
reader for Windows using BC++ 4.0 and I would be lost without it.


-- 

*  Craig R. Nelson       | Always looking for a better job. *
*  Located in Downtown   | If you need a Windows programmer *
*  Beaverton OR 97005    | I'm only a r)eply key away.      *

-----------[000403][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 05:45:41 GMT
From:      fetrow@biostat.washington.edu (Dave Fetrow)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Need configuration files for an archive

 ftp.biostat.washington.edu's anonymous ftp archive of hints, tips
and configuration files fell prey to a flakey tape drive and a large
phone book. The phone book has been moved and the tape drive replaced.

 We're taking the opportunity to try and organize it this time and,
to that end, could really use:

	autoexec.bat
	config.sys
	 ...whatever extra config files
	 ...comments about what you are running

 to set up a new, larger and organized collection of working configurations
which either can, or can not, have your email address attached to them (the
default being: email address attached). The idea is to spare others the 
agonies we went through.

 Please send such examples to: ftp@biostat.washington.edu

 In the meantime; what could be restored is in various subdirectories of:

 ftp://ftp.biostat.washington.edu/ibmpc/network

 ..which is mostly a collection of pointers to get the real answers.



-----------[000404][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 06:21:55 GMT
From:      george@ache.mad.adelaide.edu.au (George Travan)
To:        comp.protocols.tcp-ip
Subject:   Bootp



I'm having trouble getting bootp working. Infact I thought I'd try the bootp
test program 'bootptest', but it reports that it 'Can not get my ethernet
address'! 
Both programs do not seem to work.
Anyway of testing why pc's and macs can't  get their IP addresess from my bootp
server. I am sure I have set the thing up correctly, but I am now at a loss.

/Geo
-- 
                  George Travan "A Rebel Without A Clue!"                   
                  E-mail: george@ache.mad.adelaide.edu.au                       
     Phone: (Australia) +61 08 228 5968   Fax: (Australia) +61 08 224 0464  
        University of Adelaide,G.P.O Box 498 Adelaide, AUSTRALIA 5001       

-----------[000405][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 15:00:16 -0400
From:      timur@seas.gwu.edu (Timur Tabi)
To:        comp.protocols.tcp-ip
Subject:   REPOST: What is NetBIOS?

Sorry to post this twice, but I really need an answer quick, and no one's
responded to me yet.

>Subject says it all.  I know it has something to do with peer-to-peer,
>since IBM's TCP/IP for OS/2 says that it's NetBIOS provides p2p service.
-- 

------------------------------------------------------------------ Timur Tabi
Contributing Editor for "OS/2 Monthly"        Internet:    timur@seas.gwu.edu
                                              Fidonet: Timur Tabi @ 1:109/347

-----------[000406][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 94 09:02:00 GMT
From:      bryan.mawhinney@aecibbs.proxima.alt.za (Bryan Mawhinney)
To:        comp.protocols.tcp-ip
Subject:   Multiple subnet masks?

Dear TCP/IP gurus,

I thought I had a fair understanding of subnetting until I came across
the following example - two ethernets linked by two Wellfleet routers
connected by a serial line:

EthernetA               EthernetB
|     255.255.255.192   |
|                       |
Router1-----------------Router2
255.255.255.252

Each LAN has a subnet mask of 255.255.255.192, but the mask for the
serial 'net' is 255.255.255.252.  This seems somewhat confusing to me.
What are the advantages/disadvantages of this approach?  What are the
implications for routing?

Is it possible to configure the routers so that the serial link is
transparent (has no IP addresses)?  Would this make management or
monitoring of the serial links difficult?

Any and all opinions will be appreciated.  Email is probably best, and
I'll summarize if the replies prove interesting.

Thanks,
Bryan

--
Bryan Mawhinney
mawhinne@beastie.cs.und.ac.za


-----------[000407][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 1994 09:46:12 GMT
From:      ian@unipalm.co.uk (Ian Phillipps)
To:        comp.protocols.tcp-ip
Subject:   Is 'Congestion Avoidance and Control' online anywhere?


This paper by Van Jacobson is cited in RFC 1122 [TCP:7] as a "MUST",
but is itself not an RFC, nor can I find it on line anywhere.

It apparently is the only description of the "slow start" algorithm.

It was published as ACM SIGCOMM-88, August 1988.  Does ACM still own
the copyright and restrict its existence to paper copies?  I've looked
in Veronica, and on the ACM server (but that only covers CACM).

Ian
-- 
Ian Phillipps. Tech support manager, Unipalm. News admin, pipex. Internic: IP4
Many system managers claim that holes in an NNTP stream are more valuable than
the data. Van Jacobson, RFC 1144


-----------[000408][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 94 11:39:22 GMT
From:      TCPMAINT@TWNMOE10.Edu.TW
To:        comp.protocols.tcp-ip
Subject:   too many routes problem

Hi!
 
    We operate a national academic network and we have a few dozen
class B network numbers and hundreds of class Cs. We choose OSPF as the
backbone routing protocol and the routers work fine with those routes.
But some of the low-end routers in the customer's site seem overwhelmed
by too many routes available in our network and the exchange of routing
imformation makes their low speed telephone link even more conjested.
 
    Is there any way to solve this problem? Any document on the net
discussing this? I can't find any RFC dealing with this problem.
Any suggestions will be very much appreciate.
 
 
/Leo
 
 
 

-----------[000409][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 1994 12:21:37 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Is 'Congestion Avoidance and Control' online anywhere?

> This paper by Van Jacobson is cited in RFC 1122 [TCP:7] as a "MUST",
> but is itself not an RFC, nor can I find it on line anywhere.
> It apparently is the only description of the "slow start" algorithm.

From the Bibliography of my recent "TCP/IP Illustrated" (Addison-Wesley,
1994):

    A classic paper describing the slow start and congestion avoidance
    algorithms for TCP.  A PostScript copy of this paper is available via
    anonymous FTP from the host ftp.ee.lbl.gov in the file congavoid.ps.Z.

	Rich Stevens

-----------[000410][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 14:02:57 GMT
From:      nguyen@Phuong.raleigh.ibm.com (Phuong Thanh Nguyen)
To:        comp.protocols.tcp-ip
Subject:   Re: too many routes problem

In article <16F9FA3EA.TCPMAINT@TWNMOE10.Edu.TW>, TCPMAINT@TWNMOE10.Edu.TW writes:
|> Hi!
|>  
|>     We operate a national academic network and we have a few dozen
|> class B network numbers and hundreds of class Cs. We choose OSPF as the
|> backbone routing protocol and the routers work fine with those routes.
|> But some of the low-end routers in the customer's site seem overwhelmed
|> by too many routes available in our network and the exchange of routing
|> imformation makes their low speed telephone link even more conjested.
|>  
|>     Is there any way to solve this problem? Any document on the net
|> discussing this? I can't find any RFC dealing with this problem.
|> Any suggestions will be very much appreciate.
|>  
|>  
|> /Leo
|>  
|>  
|>  

Use sumary route would solve this problem.

Phuong Nguyen
-- 
--------------------------------------------------------------------------
| Disclaimer: My opinions and IBM's are two different entities           |
| Hope you have a "phuongtastic" day      PHUONG@RTP.VNET.IBM.COM        |
--------------------------------------------------------------------------

-----------[000411][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 14:08:23 GMT
From:      nguyen@Phuong.raleigh.ibm.com (Phuong Thanh Nguyen)
To:        comp.protocols.tcp-ip
Subject:   Re: Class B address with class c subnet mask, Why ?

In article <2p4cst$fkq@news.delphi.com>, sunborn@news.delphi.com (SUNBORN@DELPHI.COM) writes:
|> I ran across an installation where they have a class B address but are 
|> using class C addressses.  Is there any good reason for this ?
                 ^^^^^^^^^^
Is it a typo? The subject says "class c subnet mask". Anyway, the word
subnet says it all. A typical mask for class B would be 255.255.0.0
but since they want to have many subnetworks then they use subnet masks
i.e. 255.255.248.0 or 255.255.240.0 etc.

Phuong Nguyen

-- 
--------------------------------------------------------------------------
| Disclaimer: My opinions and IBM's are two different entities           |
| Hope you have a "phuongtastic" day      PHUONG@RTP.VNET.IBM.COM        |
--------------------------------------------------------------------------

-----------[000412][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 14:45:01 GMT
From:      smd@hrt213.brooks.af.mil (Sten Drescher)
To:        comp.protocols.tcp-ip,alt.security
Subject:   Re: Encyption of data between nodes across the net

In article <2p1jet$8kk@moria.london.micrognosis.com>, nreadwin@micrognosis.co.uk (Neil Readwin) writes:

 Readwin> In article <766598295.222snx@ozinkl.pc.my> Geoffc@Ozinkl.pc.my
 Readwin> writes:>
 >> is there an easy way to implement some form of encryption, to ensure
 >> that proprietary data and other intelectual property passed between
 >> the two sites is protected, as the data passes across the net, where
 >> is might be intercepted by others?

 Readwin> Yes. Tunnel IP over TCP :-) Some PPP products allow this (eg
 Readwin> MorningStar). There are also commercial hardware products
 Readwin> designed to solve this problem. This came up on the firewalls
 Readwin> list last week and the names mentioned were Semaphore, UUnet
 Readwin> (the LAN Guardian?) and the commercial version of the
 Readwin> KarlBridge (ie not the freeware version).

	With the recent packet sniffing attacks on the net, has the idea
of using public key/private key encryption of TCP and UDP packets been
discussed?  I don't want to trigger a long discussion if it's been
hacked to death already, but if the discussion's been archived, I'd
appreciate a pointer to it.
--
Sten Drescher			2709 13th St #1248
smd@floyd.brooks.af.mil		Brooks AFB, TX 78235
#include <disclaimer.h>
 We don't have an administration policy on long-run fiscal policy at the
moment.
	- Alice Rivlin, Deputy Director of the Office of Management and
Budget, quoted in _The Washington Post_


-----------[000413][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 15:57:33 GMT
From:      leonard@telcom.arizona.edu (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: too many routes problem


In article <2p612h$10e2@locutus.rchland.ibm.com>, nguyen@Phuong.raleigh.ibm.com (Phuong Thanh Nguyen) writes:
|In article <16F9FA3EA.TCPMAINT@TWNMOE10.Edu.TW>, TCPMAINT@TWNMOE10.Edu.TW writes:
||> Hi!
||>  
||>     We operate a national academic network and we have a few dozen
||> class B network numbers and hundreds of class Cs. We choose OSPF as the
||> backbone routing protocol and the routers work fine with those routes.
||> But some of the low-end routers in the customer's site seem overwhelmed
||> by too many routes available in our network and the exchange of routing
||> imformation makes their low speed telephone link even more conjested.
||>  
||>     Is there any way to solve this problem? Any document on the net
||> discussing this? I can't find any RFC dealing with this problem.
||> Any suggestions will be very much appreciate.
 
|Use sumary route would solve this problem.

If there is only one path to your stub nets, then you don't need
to run a routing protocol over the phone line at all.  On the
backboneward router, just set a static route to the stub net,
which you redistribute into OSPF for dissemination to the backbone.
At the remote site, just enter static default routes pointing
backboneward.

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
  \ Don't lock yourself into open systems. /

-----------[000414][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 16:59:04 GMT
From:      padgett@tccslr.dnet.orl.mmc.com (padgett peterson)
To:        comp.protocols.tcp-ip,alt.security
Subject:   Re: Encyption of data between nodes across the net

>From: smd@hrt213.brooks.af.mil (Sten Drescher)
>Subject: Re: Encyption of data between nodes across the net
>
>	With the recent packet sniffing attacks on the net, has the idea
>of using public key/private key encryption of TCP and UDP packets been
>discussed?  I don't want to trigger a long discussion if it's been
>hacked to death already, but if the discussion's been archived, I'd
>appreciate a pointer to it.

Would expect a public key method to be more difficult than some sort of
synchronous encryption layer placed at the packet mux level. On a PC,
ODIPKT might be a good place. If it were I, one-time-password tokens
would provide a source for session seeds. This would take care of the
case of known links (ftp or telnet).

For one-time communications, I suspect that off-line encryption using
PGP/ViaCrypt, SecureExchange, or some other system would be easier
to use.

Security by Obscurity (e.g. using non-standard packet types) just does
not work against a determined attacker.
					Warmly,
						Padgett

-----------[000415][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 94 17:41:07 +0200
From:      p_moore@roam.agfa.be (Paul Moore)
To:        comp.protocols.tcp-ip
Subject:   Is there a sepearate newsgroup for SNMP issues?

 
Hello,

I would like to post a message concerning SNMP.

Is there a separate SNMP newsgroup, or should I direct my
enqiory to the TCP/IP newsgroup?

Please email, as I don't have a news feed at this site.

Thanks in advance

Paul

----
Paul Moore               Voice: +32 3 444 6315
Agfa-Gevaert EBS
Mechelsesteenweg 432     Email: p_moore@roam.agfa.be
2650 Edegem
BELGIUM


----- End Included Message -----


-----------[000416][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 1994 21:01:15 GMT
From:      mwallace@sur3ax.ess.harris.com (Mark Wallace)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   MIL-STD for SNMP?

Question for a colleague:

Does anyone know if there is a MIL-STD (military standard)
number associated with SNMP?

Please email responses :-)

-- 
Mark Wallace
Harris Corporation, Electronic Systems, Melbourne, Florida, USA
mwallace@sur3ax.ess.harris.com

-----------[000417][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 21:14:42 GMT
From:      ganzhorn@cisco.com (Charles Ganzhorn)
To:        comp.protocols.tcp-ip
Subject:   Re: Optivity and Subnets

In article <2p4273$g3p@emory.mathcs.emory.edu>, km@mathcs.emory.edu (Ken
Mandelberg) wrote:
> 
> We are about to add a Class B address which will be subnetted
> with a 10 bit subnet and and 6 bit host field (netmask 0xffffffc0),
> This should give use 2^10 -2 = 1022 subnets. Our network service
> group is planning on how to divide up the subnets and has concluded
> that only 508 of them are actually usable.
> 
> They make heavy use of Synoptics Optivity product and have been told
> that Optivity won't work properly with any subnet of the form
> XXXXXXXX00 or XXXXXXXX11 (ie the 2 bit piece that intersects the last
> octet is 00 or 11). 
> 
> Frankly, I find this hard to believe. Can anyone shed some light on this.

They might have confused this with the class C situation where, with a two
bit subnet mask, those subnets WOULD be illegal.  For any class network
number where the default mask for that class has been extended by n bits,
the subnets represented by setting those n bits to all zero or all ones are
not usable.  

In the case of all zeros, the summary route for that subnet can be confused
with the summary route for the entire network.  In the case of all ones,
the subnet broadcast can be mistaken for a network wide broadcast.

Charles.
--
Charles Ganzhorn                        Email:  ganzhorn@cisco.com
cisco Systems                           Phone:  612-368-8922
Chaska, MN                              FAX:    612-368-9977

-----------[000418][next][prev][last][first]----------------------------------------------------
Date:      21 Apr 1994 21:44:30 GMT
From:      steve@eps.rain.com (Steve Kornreich)
To:        comp.protocols.tcp-ip
Subject:   slip and routing

I have a small network at my home which includes a NeXT computer running  
slip, and a Mac and PC all on my loacl net.. I would like to be able to  
use the Mac and PC sometimes for telnet, ftp, etc using my NeXT as a  
gateway. My NeXT works fine, but when I try to use my pc or mac I can only  
get to my local net and the remote local net.. I cannot go any farther..

Here is my routing table on my NeXT running slip

eps:6# netstat -rn
Routing tables
Destination      Gateway            Flags     Refs     Use  Interface
127.0.0.1        127.0.0.1          UH          2      291  lo0
199.2.100.11     127.0.0.1          UH          0        0  lo0
199.2.96.35      199.2.100.11       UH          0        0  pni0
default          199.2.96.35        UG          1      777  pni0
199.2.192        199.2.192.1        U           4      853  en0

199.2.192.1 NeXT
199.2.192.2 PC
199.2.192.3 Mac
Slip 199.2.100.11
Slip remote 199.2.96.35

Any ideas....

Thanks

--
Steven Kornreich
steve@eps.rain.com

-----------[000419][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 1994 23:33:06 GMT
From:      NADEL@litc.lockheed.com (Ron Nadel)
To:        comp.protocols.tcp-ip
Subject:   AFS

I'm trying to find out where I would get AFS software.  Like most sites, we 
use NFS, but find it such a poor performer regardless of media, and I 
understand that AFS (Andrews File System?) is supposed to be much better.  Can 
someone tell me something about it, and where to go to get it?

Thanks very much,

Ron

-----------[000420][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 1994 23:44:24 GMT
From:      alastair@cadence.com (Alastair Young)
To:        comp.protocols.tcp-ip
Subject:   Re: Encyption of data between nodes across the net

In article <padgett.4.0@tccslr.dnet.orl.mmc.com>,
padgett@tccslr.dnet.orl.mmc.com (padgett peterson) wrote:
> 
> >From: smd@hrt213.brooks.af.mil (Sten Drescher)
> >Subject: Re: Encyption of data between nodes across the net
> >
> >	With the recent packet sniffing attacks on the net, has the idea
> >of using public key/private key encryption of TCP and UDP packets been
> >discussed?  I don't want to trigger a long discussion if it's been
> >hacked to death already, but if the discussion's been archived, I'd
> >appreciate a pointer to it.
> 
> Would expect a public key method to be more difficult than some sort of
> synchronous encryption layer placed at the packet mux level. On a PC,
> ODIPKT might be a good place. If it were I, one-time-password tokens
> would provide a source for session seeds. This would take care of the
> case of known links (ftp or telnet). 
> 
> For one-time communications, I suspect that off-line encryption using
> PGP/ViaCrypt, SecureExchange, or some other system would be easier
> to use.
> 
> Security by Obscurity (e.g. using non-standard packet types) just does
> not work against a determined attacker.
> 					Warmly,
> 						Padgett

The hardware is available.

Semaphore Network Communications (408) 980-7750

They sell boxes which encrypt the packet contents selectively by
destination. The encryption session key is negotiated using RSA public key
technology. Session keys can be reset at any time.

There is also a SunOS 4.1.x software implementation of a similar type of
thing from Hughes Aircraft called NetLock. email netlock@mls.hac.com

I have no association with the above companies, nor have I tried the
products (yet).

---------------------------------------------------------------------------
Alastair Young                                     _               Ariel NH
Cadence Design Systems, Information Services     )/___     _     Red Hunter
555 River Oaks Parkway, 4B1                    __/(___)_*##/c 
San Jose CA 95134         Fax: (408)894-3487  / /\\|| \ /  \ Brakes'n'lites
alastair@cadence.com           (408)428-5278  \__/ ----'\__/  novel eh?
---------------------------------------------------------------------------
These statements and opinions are mine, not those of Cadence Design Systems

-----------[000421][next][prev][last][first]----------------------------------------------------
Date:      Thu, 21 Apr 94 23:53:16 GMT
From:      ftlofaro@unlv.edu (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   Re: AFS

In article <NADEL.145.000F8DB2@litc.lockheed.com> NADEL@litc.lockheed.com (Ron Nadel) writes:
>I'm trying to find out where I would get AFS software.  Like most sites, we 
>use NFS, but find it such a poor performer regardless of media, and I 
>understand that AFS (Andrews File System?) is supposed to be much better.  Can 
>someone tell me something about it, and where to go to get it?
>
>Thanks very much,
>
>Ron

If you think NFS is a poor performer, just wait until you try AFS!
It is DREADFUL! Don't even bother.
I was at an AFS site a little while back:

1. The servers had to be restarted each week, due to a memory leak.
Every week, thus, the servers were down for 15 minutes, during which time 
all the workstations were as good as dead. Here at UNLV (Kerberized NFS), 
the average uptime is measured in MONTHS. The main fileserver was up 
continuosly for over 240 days until a memory parity error caused an OS 
panic.
2. The AFS servers crashed quite a bit. With 20 servers, they were at 
one point having a couple crashes a *day*. Race conditions in the 
fileserver code.
3. It is incompatible with NFS, and no free implementations exist. No 
implementations exists at all, except from one company. They might 
scream patent/copyright violation if you make something try to interoperate 
with it. No such problem with NFS. There is no AFS for Linux, btw.
4. Really slow.
5. Evil way of being loaded into the UN*X kernel. Has to write /dev/kmem!
6. Trying to umount AFS server causes client to panic.
7. AFSD's (these are client side daemons, that are mandatory) can never 
be killed, and shutdown always thinks something is hung.
8. A bad packet on the AFS port (7001) caused thousands of workstations 
on the net (all the AFS clients) to fault in kernel mode and panic. The 
rebooting and recopying of files that occurs on reboot crippled the 
performance on the net.
9. Chmod permissions do not work correctly.
10. Non-standard. Need an AFS guru if things go wrong.
11. No support for sockets, named pipes or device special files. You get 
a kernel message complaing about it whenever a user even tries to make 
a named pipe, UN*X socket, etc.
12. More overheard than NFS.
13. Harder to setup and administrate. Much more complicated than NFS, with 
very little added benefit.
14. Just too plain un-UN*Xlike.
15. There is more, but I thik you've heard enough.



-----------[000422][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 00:53:11 GMT
From:      Chris Hopen
To:        comp.protocols.tcp-ip
Subject:   Any ttcp ports for Winsock 1.1 available



Hello All,

Does anyone know of any ports of the 'ttcp' program for Winsock 1.1?

I have checked around a bit and cannotfind anything.  Any help, ideas
or thoughts would be appreciated.

Thanks,
   Chris


-----------[000423][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 94 01:00:19 GMT
From:      tsai@teetot.acusd.edu (Allen Tsai)
To:        comp.sys.sun.admin,comp.protocols.ppp,comp.protocols.tcp-ip
Subject:   sliplogin daemon does not die after phone hangup

Has anyone had experiences on cslip-2-6 slip software, which make Sparc as
slip server.  Finally i am able to get cslip-2-6 to run on Sparc/SunOS4-1-3. 
However sliplogin daemon does not exit when phone hung up; 
Even ttya-rts-dtr-off is set to false and ttya-ignore-cd to true 
according to FAQ. It seems not matter what ttya-ignore-cd is true or not.
This cause next person not to be able to dial in since getty can not be
respawned.  i would like to hear from you if you know any behavior of
serial line (tty*) and modem control on Sparc. Thanks in advance.

	Allen Tsai
	University of San Diego
	tsai@teetot.acusd.edu
-- 
Allen Tsai 
University of San Diego
tsai@teetot.acusd.edu

-----------[000424][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 01:08:58 GMT
From:      csc3bem@cabell.vcu.edu (Bryan E. Miller)
To:        comp.protocols.tcp-ip
Subject:   Large number of subnets


Does anyone out there have any experience with TCP/IP with a
large number of remote sites (350+).  I'm looking for tips on
how to lay out the IP subnets.  Each remote site will have 
approximately 10 nodes.  We want to use SNMP for management.

Bryan

-----------[000425][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 01:39:56 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Is 'Congestion Avoidance and Control' online anywhere?

In article <1994Apr21.094612.16646@unipalm.co.uk> ian@unipalm.co.uk (Ian Phillipps) writes:
>
>This paper by Van Jacobson is cited in RFC 1122 [TCP:7] as a "MUST",
>but is itself not an RFC, nor can I find it on line anywhere.
>
>It apparently is the only description of the "slow start" algorithm.
> ...

There are descriptions of "slow start" in all of the many copies 4.3BSD
TCP source code around the world.  Two can be found by reading the FAQ's
for comp.os.386bsd.announce.


Vernon Schryver    vjs@rhyolite.com

-----------[000426][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 01:57:18 GMT
From:      harpua@rs6.tcs.tulane.edu (Alexander Oliver)
To:        comp.protocols.tcp-ip,alt.security
Subject:   Re: Encyption of data between nodes across the net

padgett peterson (padgett@tccslr.dnet.orl.mmc.com) wrote:

: For one-time communications, I suspect that off-line encryption using
: PGP/ViaCrypt, SecureExchange, or some other system would be easier
: to use.

I'd like to get the PGP encrypting/decrypting files.

Does anyone know an anonymous ftps site from where they can be had?

Thanks.

Catch y'all on the rebound, 
			  --Alex
                           (harpua@mailhost.tcs.tulane.edu)


-----------[000427][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 04:43:06 GMT
From:      cjross@bbn.com (Jonathan Ross)
To:        comp.protocols.tcp-ip,comp.unix.programmer
Subject:   When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?

Recently I accepted responsibility for a batch queue system homegrown at
BBN, and I've traced a nasty bug in it to the IRIX 4.0.x implementation
of SO_SNDBUF.  In an attempt to optimize large transfers from the queue
server to its clients, the original author chose to set the server's
SO_SNDBUF (and clients' SO_RCVBUF) to 16K bytes.  This seems to have had
the opposite of the desired effect; the server occasionally gets stuck
in write() calls, even when a client is simultaneously calling read().
The delay is most obvious when server and client are on the same host.
If I remove the setsockopt() buffer size hacks, the problem disappears.

This makes me wonder -- why aren't numerous other network-related IRIX
applications stuck in the mud?  Is it really that unusual to try and
tune socket buffer sizes?  Is there much point to it at all?  My various
UNIX manuals only document what SO_SNDBUF does; none explain its effects
(or lack thereof) on performance.  Please enlighten me, and I'll
summarize.

(If you have an IRIX 4.0.x system and would like to see this bug in
action, my original version of this post (which garnered no responses)
is on comp.sys.sgi.bugs; do a subject search on "SNDBUF".  It includes a
demo program with buffer parameters settable from the command line.)

Followups to comp.protocols.tcp-ip.

=====  Jonathan Ross  =====  <cjross@bbn.com>  =====  (617) 873-3272  =====
  WARNING: article may contain flammable material.  Do not expose to open
    flames.  In case of accidental ignition, douse keyboard with water.

-----------[000428][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 17:03:14 -0700
From:      lstowell@pyrnova.mis.pyramid.com (Lon Stowell)
To:        comp.protocols.tcp-ip
Subject:   Re: Detecting lost TCP connection?

In article <1994Apr22.133701.12195@isc-br.isc-br.com> bruceo@loki.isc-br.com (Bruce Oscarson) writes:
>
>
>I am having problems detecting the loss of a TCP connection on my 
>server when a client workstation goes down or the network fails.
>(Not a problem when client process crashes or exits).
>
>It is critical for me to detect this loss.  In some cases, the 
>clients have locked critical tables in my database, and then the 
>client machine goes down, or the network fails, and I am left with 
>locked tables for a non-existent client.
>
>I am aware of SO_KEEPALIVE, but have been told that there is no 
>standard duration for this 'heartbeat'.  One of our TCP 
>implementations has about a 5 minute timeout.  I have also heard of 
>timeouts of over a day!  Additionally, I need a solution that is not 
>dependent on a specific TCP/IP implementation, since our server 
>currently runs on three different architectures, with more to 
>come.

  In that case keepalives should be considered absolutely verboten, as
  some implementations don't support them at all, some don't allow
  you to adjust them, and some do even worse things. 

>
>I have considered implementing a simple heartbeat between my 
>server and clients, but I am not certain what type of interval I 
>would need for Windows clients (non-preemptive os), plus we have 
>many other client/servers that would require the same changes.
>
   If you absolutely need dead detection, there is no other way to do
   it.  You may want to use a tuneable config file for the timer
   intervals.  I have seen some implementations that take down all the
   circuits for intervals caused by flush daemons hogging CPU's, still
   from experience I will categorically state that this
   dead-connection detection is best done at the layer that is
   guaranteed to exist from the same vendor on all systems.  



-----------[000429][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 10:28:10
From:      fks@ftp.com  (Frances K. Selkirk)
To:        comp.protocols.tcp-ip
Subject:   Re: Winsock under Win-OS/2

Sorry for the late continuition of this thread...

In article <60.16811.4471.0N19AAD5@canrem.com> john.childs@canrem.com (John Childs) writes:

> Has anybody gotten Win-OS/2 to recognize winsock.  If so who's TCP/IP
> product were you using?

Version 1.3 of our PC/TCP for OS/2 includes a module for support of
our DOS and Windows applications, including our winsock.dll, which is
included. 

Cheers,

--
Frances K. Selkirk                                        fks@ftp.com
FTP Software, Inc.        Technical Support            (800) 382-4FTP
---------------------------------------------------------------------
Get our support newsletter from       | FTP support - support@ftp.com
ftp.ftp.com or our BBS (508-659-6240) | FTP sales   -    info@ftp.com


-----------[000430][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 11:08:55
From:      timc@sni.com.au (Tim Cullen)
To:        comp.protocols.tcp-ip
Subject:   MS-TCP/IP-32 Beta 2 and wsockets.dll

I am using WfW 3.11 and  MS TCP/IP-32 Beta 2 driver.
I have a client windows application that uses \windows\wsockets.dll (19/12/93) 
When I start the application (which attempts to connect to a host) I get the 
following errors (inside dialog boxes):      
         Socket 123
         ATTACH Error at TCP socket call

Note: I can successfully use other Winsock programs such as Trumpet, XVision.
         The application works with WfW TCP/IP 1.0a.

I am unsure how wsockets.dll communicate with MS TCP/IP-32 winsock driver.

Are these compatible ? Any suggestions ?
Regards,

Tim Cullen (timc@sni.com.au)
Unix Product Manager
Siemens Nixdorf Information Systems Pty Ltd
655 Pacific Highway, St Leonards, NSW, 2065, AUSTRALIA
Voice: +61-2-430-2154, Fax: +61-2-439-5734

-----------[000431][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 16:37:54 -0500
From:      mikea@MCS.COM (Mike Andrews)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Variable IP subnets with RIP on Wellfleet and Cisco routers

Help!  I'm  in multi-vendor, variable subnetting, finger pointing, IP hell!

The problem:

I have one reserved Class B address space onto which I'll eventually have have
to fit over 600 networks.  To do this, I set different subnet masks in
various places:
24 bits, 255.255.255.0   27 bits, 255.255.255.224 and 30 bits, 255.255.255.248.

I'm using Wellfleet and Cisco routers in a topology like this (simplified):

_____________                                                      ___________
Wellfleet BLN|  1.111----  Ethernet 128.1.1.96 --------------1.110| Cisco 7000
             |               255.255.255.224                      |
    HQ       |                                                    |    HQ
             |  1.158----  Ethernet 128.2.1.127                   |
             |               255.255.255.224                      |
_____________                                                     ____________
     | 128.1.101.1                                                     |
     |                                       128.1.100.0  Frame Relay  |
     | Frame Relay 128.1.101.0               255.255.255.0             |
     |       255.255.255.0                                       ____________
     |                                                            Cisco 2000
     | 128.1.101.51                                    Ethernet   128.1.21.0
_____________                                                    255.255.255.0
Wellfleet  LN|                                                       City C
             | 128.1.51.61 ----- Ethernet 128.1.51.32 ---
             |                     255.255.255.244
   City B    |
             | 128.1.52.5 -------- T1 128.1.52.4 ----------- | 128.1.52.6
_____________|                       255.255.255.248           City B remote


I'm running only IP and RIP over the links.  Also IPX, but only on the ethernet
segments.  I'm using IP tunneling on the WAN.
The main problem is that when I set the netmask on the directly connectly 
ethernet interface on the Cisco 7000 properly to 255.255.255.224 and do
a show ip route it says, correctly, "Network 128.1 is variably subnetted...
255.255.255.0 on Ethernet.... 255.255.255.0 on Serial..."  But the Cisco will
not route any traffic from the 128.1.1.96 ethernet to the Frame Relay network.
I incorrectly set the netmask on the ethernet interface to 255.255.255.0
and now it routes. But then it broadcasts an incorrect route to a 128.1.1.0 LAN
which screws up any chance of using other 128.1.1 subnets.

The Wellfleets have no trouble routing over the entire WAN.  I can ping a host
on a LAN behind the 128.1.52.6 host from a host on the the 128.1.1.96 LAN.

My Wellfleet SE says that there's nothing wrong with my topology.  

I had a long talk with two Cisco techs who said variously:

- It won't work.             (IT DOES WORK! - mostly)
- Don't variably subnet.     (I have to.)
- You can't do this with RIP.  (Then why can't the cisco route it's LOCALLY
				ATTACHED interfaces?)
- Wellfleets use a non-standard RIP that includes the subnet mask in the RIP
  updates.                   (Is this true?)
- You must run OSPF.         (I'll consider this. But the wellfleets don't
				 require OSPF)
- Use the same subnet mask everywhere.
				( I can put more hosts on ethernet than on
				a transisional router-router subnet.  I want
	                        to use an appropriate subnet mask where needed.)


If you have any experience with variable subnetting I'd appreciate your
suggestions or comments.   I'd also like info and experiences in installing
OSPF on a multi-vendor WAN like this.

Please reply via Email to mikea@mcs.net.  I'll summarize if there's interest.
Thanks in advance.

-- 
Mike Andrews                             mikea@Genesis.MCS.Com
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=-=-=-=
"The chances of Sun today are...uhhhnnnhh...iffy, if that."
       -Brant Miller, Chicago TV weatherman/radio DJ, 03/20/93

-----------[000432][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 10:46:25 GMT
From:      635376@rhds12.rz.fht-esslingen.de (GUANCIALINO MARIO)
To:        comp.protocols.tcp-ip
Subject:   beginner problems

Hello,

I have a Problem regarding TCP IP 

I have to write an application to sent a file via TCP IP (System V) to another 
Host.

If you can help me with some easy examples or by providing some hints,

please E-Mail me.


                                                     vvv
                                                    (. .)
                                                ooo0 (_) 0ooo
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=
                 Mario Guancialino  (The "Energizer")
                email:635376@rhds01.rz.fht-esslingen.de
              Und es gibt DOCH ein Leben nach dem Studium !!!!
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=


-----------[000433][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 12:35:04 GMT
From:      choang@tdkt.mn.org (Carl Hoang)
To:        comp.protocols.tcp-ip
Subject:   Multiple Winsock Connection

Is it possible to have more than 1 connection to the server socket program
from the same PC?  I'm running Windows and Winsock.  I could get 1
connection going OK but not 2.  Thanks.
   Carl Hoang   choang@tdkt.mn.org


-----------[000434][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 12:55:02 GMT
From:      jglazer@lansoft.usa.com
To:        comp.protocols.tcp-ip
Subject:   FTP Mirror script needed

I would like to setup a mirror script for my site to mirror another. 
I have never done this before and am not sure how to go about it.  
Is there some "standard" program out there that can check out a 
site's directory structure then run FTP to download only new files 
(deleting and old files too)?  It's a simple matter to have FTP 
connect and download a whole directory but that's kinda pointless if 
you already have 99% of the files.

Any help would be appreciated.

Thanks!
Jon

-------------------------------------------------------------------
Jon Glazer                   | jglazer@lansoft.usa.com
Lansoft USA, Inc.            |  
6665 Busch Blvd.             |   - other lansoft departments -
Columbus, Ohio  43229        | sales@lansoft.usa.com   
(614)-786-1713               | support@lansoft.usa.com

-----------[000435][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 13:17:35 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,alt.security
Subject:   Re: Encyption of data between nodes across the net

Alexander Oliver (harpua@rs6.tcs.tulane.edu) wrote:

> padgett peterson (padgett@tccslr.dnet.orl.mmc.com) wrote:
>
> > For one-time communications, I suspect that off-line encryption using
> > PGP/ViaCrypt, SecureExchange, or some other system would be easier
> > to use.
>
> I'd like to get the PGP encrypting/decrypting files.
>
> Does anyone know an anonymous ftps site from where they can be had?


Try these sites:


archie pgp:


Host archive.egr.msu.edu

    Location: /pub
      DIRECTORY drwxr-xr-x        512  Mar  9 18:58  pgp

Host arthur.cs.purdue.edu

    Location: /pub/pcert/tools/unix
      DIRECTORY drwxr-xr-x        512  Jul 31 1993  pgp

Host bloom-picayune.mit.edu

    Location: /pub/usenet-by-hierarchy/alt/security
      DIRECTORY drwxrwxr-x        512  Mar 14 00:17  pgp

Host cecelia.media.mit.edu

    Location: /pub
           FILE -rw-r--r--     321424  Nov 30 20:27  pgp

Host cs.huji.ac.il

    Location: /pub/security
      DIRECTORY drwxrwxr-x        512  Oct 26 19:26  pgp

Host csn.org

    Location: /mpj/public
      DIRECTORY drwxr-xr-x        512  Mar 14 20:30  pgp

Host dime.cs.umass.edu

    Location: /pub/rcf/exp/build/pgp-2.3
           FILE -rwxr-xr-x     241916  Mar 15 15:42  pgp
    Location: /pub/rcf/exp/build/pgp-2.3/src
           FILE -rwxr-xr-x     241916  Mar 15 15:41  pgp

Host f.ms.uky.edu

    Location: /pub2/NetBSD/NetBSD-current/src/usr.bin/file/magdir
           FILE -rw-r--r--        478  Dec 17 02:10  pgp

Host files1zrz.zrz.tu-berlin.de

    Location: /pub/mail
      DIRECTORY drwxr-xr-x       1024  Jan 11 1993  pgp

Host ftp.germany.eu.net

    Location: /pub/comp/msdos/local/utils
      DIRECTORY drwxr-xr-x        512  Jul 12 1993  pgp

Host ftp.uni-kl.de

    Location: /pub1/unix/security
      DIRECTORY drwxrwxr-x        512  Feb 24 1993  pgp

Host gatekeeper.dec.com

    Location: /.0/BSD/NetBSD/NetBSD-current/src/usr.bin/file/magdir
           FILE -r--r--r--        478  Dec 16 23:10  pgp

Host granuaile.ieunet.ie

    Location: /ftpmail-cache/ie/tcd/maths/ftp/src/misc
      DIRECTORY drwxr-xr-x        512  Dec  2 11:43  pgp

Host hpcsos.col.hp.com

    Location: /mirrors/.hpib1/NetBSD/NetBSD-current/src/usr.bin/file/magdir
           FILE -r--r--r--        478  Dec 17 00:10  pgp

Host info1.rus.uni-stuttgart.de

    Location: /afs/.rus.uni-stuttgart.de/sw/rs_aix32/pgp-2.3/bin
           FILE -rwxr-xr-x     211318  Aug 23 1993  pgp

Host info2.rus.uni-stuttgart.de

    Location: /afs/rus.uni-stuttgart.de/sw/rs_aix32/pgp-2.3/bin
           FILE -rwxr-xr-x     211318  Aug 23 1993  pgp

Host isy.liu.se

    Location: /pub/misc
      DIRECTORY drwxr-xr-x        512  Sep 19 00:00  pgp

Host jhunix.hcf.jhu.edu

    Location: /pub/public_domain_software/NetBSD/usr/src/usr.bin/file/magdir
           FILE -rw-r--r--        478  Jun  9 1993  pgp

Host josquin.media.mit.edu

    Location: /pub
           FILE -rw-r--r--     321424  Nov 30 20:27  pgp

Host minnie.zdv.uni-mainz.de

    Location: /pub/atari/misc
      DIRECTORY drwxr-xr-x        512  Dec 30 17:56  pgp

Host mintaka.lcs.mit.edu

    Location: /pub
      DIRECTORY drwxr-xr-x        512  Jun 18 1993  pgp

Host netcom.com

    Location: /pub/halliday
           FILE -rwx------     212992  Nov 27 20:21  pgp
    Location: /pub/kevitech
           FILE -rwxr-xr-x      89643  Dec  3 05:46  pgp
    Location: /pub/torin
      DIRECTORY drwx--x--x       4096  Jan 11 18:59  pgp

Host quepasa.cs.tu-berlin.de

    Location: /.4/pub/bsd/386bsd-0.1/unofficial/doc/software
           FILE -rw-rw-r--      12121  Feb  2 1993  pgp

Host sun.rz.tu-clausthal.de

    Location: /pub/atari/misc
      DIRECTORY drwxr-xr-x        512  Dec 30 18:56  pgp

Host walton.maths.tcd.ie

    Location: /src/misc
      DIRECTORY drwxr-xr-x        512  May 30 1993  pgp
    Location: /src/misc/pgp-2.0/src

Host xanth.cs.odu.edu

    Location: /pub
      DIRECTORY drwxrwxr-x        512  Oct 18 00:00  pgp




The latest release (to my knowledge) is version 2.3.

Cheers.


_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000436][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 13:37:01 GMT
From:      bruceo@loki.isc-br.com (Bruce Oscarson)
To:        comp.protocols.tcp-ip
Subject:   Detecting lost TCP connection?


I am having problems detecting the loss of a TCP connection on my 
server when a client workstation goes down or the network fails.
(Not a problem when client process crashes or exits).

It is critical for me to detect this loss.  In some cases, the 
clients have locked critical tables in my database, and then the 
client machine goes down, or the network fails, and I am left with 
locked tables for a non-existent client.

I am aware of SO_KEEPALIVE, but have been told that there is no 
standard duration for this 'heartbeat'.  One of our TCP 
implementations has about a 5 minute timeout.  I have also heard of 
timeouts of over a day!  Additionally, I need a solution that is not 
dependent on a specific TCP/IP implementation, since our server 
currently runs on three different architectures, with more to 
come.

I have considered implementing a simple heartbeat between my 
server and clients, but I am not certain what type of interval I 
would need for Windows clients (non-preemptive os), plus we have 
many other client/servers that would require the same changes.

I would appreciate any thoughts or experiences that you have in 
the area of detecting loss of a partner under TCP/IP.

Thanks in advance!
bruceo@mail.isc-br.com
--

Bruce Oscarson          | bruceo@mail.spk.olivetti.com
Olivetti North America  |
N7RWO                   | ma-bell  (509) 927-5437

-----------[000437][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 94 13:05:48 CET
From:      s22562@seb.se
To:        comp.protocols.tcp-ip
Subject:   RFC:s

Hello!

Can anyone tell me how I can get RFC:s for eg OSPF and CIDR from Internet
news. Is there any newsgroup where you can find this wonderful stuff?
Many thanks in advance!!!
Regards Bo

-----------[000438][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 14:30:25 GMT
From:      nrd1rls@nrd.ups.com (Richard Siddall Contractor)
To:        comp.protocols.tcp-ip,comp.unix.programmer,comp.unix.bsd
Subject:   HELP! Setting subnet mask under 4.3BSD

I need to set the subnet mask under an embedded OS that uses the
4.3 BSD sockets API.  The OS vendor cannot supply us with an
example of how to do this.  If anyone has a code sample, it would
be greatly appreciated.  What we need to know is where to insert
the new subnet mask in the ifreq structure passed to the ioctl()
call.

	Thanks.


-----------[000439][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 22:12 PDT
From:      adelman@tgv.com (Kenneth Adelman)
To:        comp.protocols.tcp-ip
Subject:   Re: When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?

In article <CooJME.7B9@calcite.rhyolite.com>, vjs@calcite.rhyolite.com (Vernon Schryver) writes...
 >In article <2p7kkq$f00@info-server.bbn.com> cjross@bbn.com (Jonathan Ross) writes:
 >>Recently I accepted responsibility for a batch queue system homegrown at
 >>BBN, and I've traced a nasty bug in it to the IRIX 4.0.x implementation
 >>of SO_SNDBUF.  In an attempt to optimize large transfers from the queue
 >>server to its clients, the original author chose to set the server's
 >>SO_SNDBUF (and clients' SO_RCVBUF) to 16K bytes.  This seems to have had
 >>the opposite of the desired effect; the server occasionally gets stuck
 >>in write() calls, even when a client is simultaneously calling read().
 >>The delay is most obvious when server and client are on the same host.
 >>If I remove the setsockopt() buffer size hacks, the problem disappears.
 >>
 >>This makes me wonder -- why aren't numerous other network-related IRIX
 >>applications stuck in the mud?  Is it really that unusual to try and
 >>tune socket buffer sizes?  Is there much point to it at all?  My various
 >>UNIX manuals only document what SO_SNDBUF does; none explain its effects
 >>(or lack thereof) on performance.  Please enlighten me, and I'll
 >>summarize.
 >>
 >>(If you have an IRIX 4.0.x system and would like to see this bug in
 >>action, my original version of this post (which garnered no responses)
 >>is on comp.sys.sgi.bugs; do a subject search on "SNDBUF".  It includes a
 >>demo program with buffer parameters settable from the command line.)
 >
 >IRIX has been shipping with the values that SO_SNDBUF and SO_RCVBUF control
 >set to 60*1024 for years.   Many common utilities, including gated, routed,
 >ping, ttcp, and mrouted, use those setsockopt()'s.
 >
 >The most likely reason you received no response to your original posting
 >is that those setsockopt()'s work for many people in many vendors' systems,
 >everyone figured it was a bug in your program, and no one felt like
 >debugging your program for you.  There are so many people who reflexively
 >reason "I can't make [arbitrary commonly used facility] work so it must
 >be a system bug", who are unable to conceive of the possibility that the
 >trouble might be a personal problem, that even legitimate bug reports in
 >commonly used facilities are sometimes ignored.
 >
 >Have you tried your program on other BSD TCP stacks?	Have you seen if
 >you can reproduce the same behavior with the "-b" arg to `ttcp`?
 >
 >Is there any chance you're seeing the Nagle algorithm in action?
 >
 >If you're using UDP and zillions of sockets and clients, have you ensured
 >that you are not running out of mbufs?  (see `netstat -m`)
 >
 >(No, please do not send your source to me.  I have more than enough of my
 >own bugs to find.)


    Actually, I believe there is a bug in the 4.3bsd-tahoe (and perhaps
later) TCPs that is tripped by using setsockopt() to LOWER the SO_SNDBUF
value.

    I seem to recall at the comment "Compare available window to amount..."
in tcp_output.c that the calculation of the variable "adv" could go negative.

							    Ken

-----------[000440][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 15:56:43 GMT
From:      jimc@jts.jts.com (Jim Carroll )
To:        comp.protocols.tcp-ip
Subject:   looking for CIDR clarification

Looking for clarification on what CIDR is all about, including
one or more examples.

Email response preferred, as I tend to be swamped by work.  :-(
-- 
* Jim Carroll * jimc@jts.com * JTS Computer Systems Ltd., Toronto, Ont.  *

-----------[000441][next][prev][last][first]----------------------------------------------------
Date:      22 Apr 1994 17:16:45 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: looking for CIDR clarification

In article <Coo4AK.D5F@jts.com> jimc@jts.jts.com (Jim Carroll ) writes:
    Looking for clarification on what CIDR is all about, including
    one or more examples.

Please see RFC 1519.

Tony



-----------[000442][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 94 17:51:48 GMT
From:      calvin@wshb.csms.com (Calvin Giles - WSHB Operator)
To:        comp.protocols.tcp-ip
Subject:   Help



  Hello Everyone,

	I needed to know if there was a way to connect (in real time)
  to a ftp site via our modem and TCP/IP.  We have been using ftpmail
  as we are not directly on internet.  However, using ftpmail we have
  a lot of no response from the ftpmailer.  We were wondering if we
  could access an ftpsite through our modem in real time.  If we can
  could someone tell us how?  We are running SCO Unix Open System
  Xwindows and have also purchased the SCO TCP/IP.  Any help would
  be greatly appreciated.

	Please reply to me at calvin@wshb.csms.com

-- 
...calvin@wshb.csms.com.        If it wern't for bad luck,
kamikazi"calvin"                I'd have no luck at all! 
>From a warped mind comes a warped product.
These are not my employers opinions.  In fact I don't know if there mine!

-----------[000443][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 18:08:22 GMT
From:      peterd@jamie.dev.cdx.mot.com (Peter Desnoyers)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   Re: 56K leased line VS V.FAST or has anyone measured latency lately?

kwia4000@bronto.zrz.TU-Berlin.DE (Manfred Kwiatkowski) writes:

>In article <CoKFIG.AEq@cichlid.com>, aab@cichlid.com (Andy Burgess) writes:
>> 
>> [..]
>> Does the latency of different modem brands really vary this much 
>> (100-300 msec)?
 
>Yes. Analog modems need enough horsepower to handle the LAPM protocol
>and sophisticated algorithms to decide on the packetizing strategy.

Outside of the LAPM protocol, the filters, echo cancellers, and other
analog goodies in a high-end modem are going to have significant
delays as well, so you can't get rid of modem delay by eliminating
LAPM and running synchronous. Going to 28.8, as Andy Burgess
suggested, would probably only increase the delay by adding more
stuff. I believe the only way to get modem latency as low as that for
a DDS line is to go down to 300 baud :-(

[I think the dial-up Naval Observatory clock supports 300bps for
exactly this reason]

				Peter Desnoyers
-- 

-----------[000444][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 21:27:50 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?

In article <2p7kkq$f00@info-server.bbn.com> cjross@bbn.com (Jonathan Ross) writes:
>Recently I accepted responsibility for a batch queue system homegrown at
>BBN, and I've traced a nasty bug in it to the IRIX 4.0.x implementation
>of SO_SNDBUF.  In an attempt to optimize large transfers from the queue
>server to its clients, the original author chose to set the server's
>SO_SNDBUF (and clients' SO_RCVBUF) to 16K bytes.  This seems to have had
>the opposite of the desired effect; the server occasionally gets stuck
>in write() calls, even when a client is simultaneously calling read().
>The delay is most obvious when server and client are on the same host.
>If I remove the setsockopt() buffer size hacks, the problem disappears.
>
>This makes me wonder -- why aren't numerous other network-related IRIX
>applications stuck in the mud?  Is it really that unusual to try and
>tune socket buffer sizes?  Is there much point to it at all?  My various
>UNIX manuals only document what SO_SNDBUF does; none explain its effects
>(or lack thereof) on performance.  Please enlighten me, and I'll
>summarize.
>
>(If you have an IRIX 4.0.x system and would like to see this bug in
>action, my original version of this post (which garnered no responses)
>is on comp.sys.sgi.bugs; do a subject search on "SNDBUF".  It includes a
>demo program with buffer parameters settable from the command line.)

IRIX has been shipping with the values that SO_SNDBUF and SO_RCVBUF control
set to 60*1024 for years.   Many common utilities, including gated, routed,
ping, ttcp, and mrouted, use those setsockopt()'s.

The most likely reason you received no response to your original posting
is that those setsockopt()'s work for many people in many vendors' systems,
everyone figured it was a bug in your program, and no one felt like
debugging your program for you.  There are so many people who reflexively
reason "I can't make [arbitrary commonly used facility] work so it must
be a system bug", who are unable to conceive of the possibility that the
trouble might be a personal problem, that even legitimate bug reports in
commonly used facilities are sometimes ignored.

Have you tried your program on other BSD TCP stacks?  Have you seen if
you can reproduce the same behavior with the "-b" arg to `ttcp`?

Is there any chance you're seeing the Nagle algorithm in action?

If you're using UDP and zillions of sockets and clients, have you ensured
that you are not running out of mbufs?  (see `netstat -m`)

(No, please do not send your source to me.  I have more than enough of my
own bugs to find.)


Vernon Schryver    vjs@rhyolite.com

-----------[000445][next][prev][last][first]----------------------------------------------------
Date:      Fri, 22 Apr 1994 09:22:15 +0800
From:      peter.lewis@info.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   Re: Proposed FTP LIST Specification

>It makes sense to have a pool of machine-readable LIST format
>recognition and decoding specifications, available to all, that many can
>contribute to (ideally FTP server authors), until servers can be made to
>universally support a more machine-readable variant of the LIST command.

An interesting idea, but not what I'd intended.  My intent with the LIST
spec proposal was to get the small number of non-conformant FTP servers
(probably less than 1% of current servers) to conform to an easily
parseable format which is sufficiently general to allow machine-specific
details.  At least the intent is to offer this as a valid choice for
server authors.

>       1.  A publicly available file specifies them all.  Clients
>           run with the latest version of the file at hand.

Viable.  I currently use a similar system in my archie client to keep the
list of archie servers up to date.
>
>       2.  Individual server sites offer their format specs in
>           the "ftplist" or "/etc/ftplist" files (clients look in
>           both places).

Hmmm, that's more than a little machine specific, although I suppose
servers could look for "RETR /etc/ftplist" and take it out as a special
case without the file even existing.  Sounds like a bit of a hack to me.

>These all assume a custom or robot client, but no changes to server
>programming.  Is this the purpose you had in mind with your "FTP LIST
>Specification", or have I missed the point?

No, the point was to document the LIST format in a way that can parse 99+%
of all current servers, and in a way that server implementers can see how
to conform, but with enough freedom to show any information they currently
show.

I'm against the current situation where we have N clients and M servers,
and thus require N*M parsing algorithms.  Your concept would also reduce
this to N parsing algorithms, although they would need to be quite
complicated in retrieving and interpretting some machine-readable parse
specification.

Enjoy,
   Peter.
_______________________________________________________________________
Peter N Lewis <peter.lewis@info.curtin.edu.au>       Ph: +61 9 368 2055

-----------[000446][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 1994 05:09:25 GMT
From:      mcstout@netcom.com (Mark C. Stout)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for RFC distrib. site

jmr@ibm1.nynexst.com wrote:
: I am looking for a few RFC documents:
 
: 	1. Is there a common distribution site privided by the ITF ?
: 	2. If not, can these RFC be purchased and from where ?
 
: Any hint would be helpful. Thanks in advance.

I get what's on-line at anonymous ftp at ds.internic.net /rfc.  It has what
looks to most, if not all the RFC's that are available on-line.  Latest one
is RFC1609.  They are in ASCII text format, though a few of them are also in
postscript.

If anyone knows where the off-line RFC's can be had, I'd like to complete my
collection, even if they are obsoleted by newer RFC's.

Mark


-- 
==========================================================================
E-Mail Address |  mcstout@netcom.com/mark_stout@ccm.fm.intel.com
Programming    |  C/C++, Visual Basic
Networking     |  Novell, Banyan, Pathworks 
Certification  |  Certified Netware Engineer
---------------+---------------------------------------------------------
                I multitask... I read in the bathroom!
=========================================================================
~
~
~
~

-----------[000447][next][prev][last][first]----------------------------------------------------
Date:      23 Apr 1994 05:58:53 GMT
From:      s900387@minyos.xx.rmit.EDU.AU (Craig Macbride)
To:        biz.sco.general,comp.protocols.tcp-ip,comp.sys.amiga,comp.sys.amiga.programmer
Subject:   Re: rwhod warped with an Amiga

jraja@snakemail.hut.fi (Jarno Tapio Rajahalme) writes:

>Just if the TCP/IP package happens to be AmiTCP/IP.

Yes. I should have mentioned it originally, but, yes, it's AmiTCP/IP 2.3
running on an Amiga 3000 with a Quicknet card.

>AmiTCP (the
>protocol stack) uses the BSD net/2 protocol code, and thus doesn't
>send ICMP replies to broadcasted udp packets.

Hmmm. This one sends a "udp port who unreachable" out as a broadcast!

>So, if the Amiga sends
>something back, there must be some daemon installed (by whoever did
>install the AmiTCP/IP on your Amiga in the first place). On the Amiga,
>you might check the file AmiTCP:db/inetd.conf to see if there is
>something strange there.

Seems odd. The person who installed it hasn't done anything weird, as he
doesn't know how to.

>They are included in the AmiTCP binary distribution because there was
>nothing better available at that time. The current AmiTCP version
>(2.3) has the NcFTP, which you might want to use instead of the "ftp".

I more recently noticed that was there, but it seems to try to do an anon
connection by itself when you start it. Note that I know almost nothing
about Amigas - I just want the silly thing to stop confusing SCO's poor,
fragile rwhod.

>Maybe you should update the rwhod of your Unix box?

The Unix box is far newer than the Xenix box, so I don't think updating
its rwhod is really a big option. 

>(Hmm. why don't you mention the "bad programming" here?)

I thought it was implied. The SCO rwhod certainly looks fragile.

-- 
 _--_|\		Craig Macbride	<craig@rmit.edu.au>
/      \
\_.--.*/	How many Anglican priests does it take to change a lightbulb?
      v		... "Change"? What's change?

-----------[000448][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 1994 09:32:46 +0000
From:      proyse@peeras.demon.co.uk (Phil Royse)
To:        comp.protocols.tcp-ip
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco route

In article <2p9g3i$bpj@genesis.MCS.COM> mikea@MCS.COM writes:

>Help!  I'm  in multi-vendor, variable subnetting, finger pointing, IP hell!
[stuff deleted]

>If you have any experience with variable subnetting I'd appreciate your
>suggestions or comments.  
>
>Please reply via Email to mikea@mcs.net.  

No, No, No.....  answers/responses via this newsgroup,  please!
Judging from the questions over the last few months, this issue of
subnetting is one of very widespread concern.

> I'll summarize if there's interest.
 
>Mike Andrews                             mikea@Genesis.MCS.Com

OK, Mike, it would be good if you could do that for any useful material
which misses this news group.

I'm looking to get some useful guidance on subnetting etc.....

FWIW, my understanding is that you cannot pass subnet masks in RIP
route adverts, therefore its probably a fluke that your packets are 
getting through OK on some routes but not on others.

Only OSPF allows subnet masks to be passed along with the routing info
and you need this to do variable length masks.  But you know this already.

-- 
Phil Royse     Comms Consultant  |  PRA Consulting Ltd.
TUDOR HOUSE                      |
12 Woodside Road, Purley
Surrey  CR8 4LN   (UK)        Tel: (+44) 81 645-986

-----------[000449][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 94 16:00:37 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Subnet Masks?????


To: rockbttm@hebron.connected.com, regarding subnet masking.

I believe the RFC you're looking for is RFC950, "Internet standard
subnetting procedure".  RFC1219 seems to have something to say on the
matter also, but it's not one of the "standard" RFC's.  RFC1122 has a
few official things to add, as the central standard on host protocol
requirements.

Subnetting is still a thorny mystery to me, but perhaps a thorough
reading of these RFC's would part the clouds.  There's a brief overview of
the concepts in TCP/IP Illustrated, W.R.Stevens, pp 43-45.

You might want to download rfc-index.txt (or some name like that) from
the /rfc directory at the ds.internic.net ftp site.  I've text-searched
that file on many occasion trying to find an RFC on a particular
subject, or to find what obsoletes what.  That's also a good spot to
find the RFC's themselves.

-- Bob Stein                              Internet mail: stein@gcomm.com
 ________________________________________________________________________
|                                                                        |
|  Galacticomm, Inc.                              (305) 583-5990 (voice) |
|  4101 SW 47 Avenue, Suite 101                   (305) 583-7846 (FAX)   |
|  Ft Lauderdale, Florida, USA, 33314             (305) 583-7808 (BBS)   |
|________________________________________________________________________|

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000450][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 1994 12:06:04 +0000
From:      fred@genesis.demon.co.uk (Lawrence Kirby)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   Re: 56K leased line VS V.FAST or has anyone measured latency lately?

In article <CoKFIG.AEq@cichlid.com> aab@cichlid.com "Andy Burgess" writes:

>Does the latency of different modem brands really vary this much 
>(100-300 msec)?

The older, large footprint USR Couriers were exceptional in this area with
a ping latency of about 130ms. The 16.8K HST Dual Standards were a serious
disappointment in this area with latencies more like 250ms. I haven't
reied the later models but I really hope that they have improved (V.34
ought to get sub 100ms latencies).

-- 
-----------------------------------------
Lawrence Kirby | fred@genesis.demon.co.uk
Wilts, England | 70734.126@compuserve.com
-----------------------------------------

-----------[000451][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 1994 18:13:54 GMT
From:      ritz@ritz.mordor.com (Chris Mauritz)
To:        comp.dcom.modems,comp.protocols.tcp-ip
Subject:   Re: 56K leased line VS V.FAST or has anyone measured latency lately?

Lawrence Kirby (fred@genesis.demon.co.uk) wrote:
: In article <CoKFIG.AEq@cichlid.com> aab@cichlid.com "Andy Burgess" writes:
 
: >Does the latency of different modem brands really vary this much 
: >(100-300 msec)?
 
: The older, large footprint USR Couriers were exceptional in this area with
: a ping latency of about 130ms. The 16.8K HST Dual Standards were a serious
: disappointment in this area with latencies more like 250ms. I haven't
: reied the later models but I really hope that they have improved (V.34
: ought to get sub 100ms latencies).

I have a pair of Microcom Deskporte FAST 28.8k modems that show
latencies of about 100ms over a 28.8k SLIP link between a pair
of Telebit Netblazer PN2's.

Regards,

Chris

-- 
Christopher Mauritz       |  Ask me about public access unix
ritz@mordor.com           |  and interactive internet services.
Mordor International BBS  |  BBS: (201)432-0060  8-N-1
Jersey City, NJ           |  FAX: (201)433-4222

-----------[000452][next][prev][last][first]----------------------------------------------------
Date:      23 Apr 94 03:35:16 +0900
From:      zcookbruc@cc.curtin.edu.au
To:        comp.protocols.tcp-ip
Subject:   Re: What is NetBIOS?

In article <2p44q9$b9a@cronkite.seas.gwu.edu>, timur@seas.gwu.edu (Timur Tabi) writes:
> Subject says it all.  I know it has something to do with peer-to-peer,
> since IBM's TCP/IP for OS/2 says that it's NetBIOS provides p2p service.

netBIOS gets applied in several contexts, the most common one (From my
exposure) is the device routing layer the task-to-task comms that is
barnacled into DOS when it gets a networking system.

For a program to talk to a network there must be a network interface, just
like there is a video/keyboard/disk etc interface. This interface forms
a (supposably) network independant layer that programs can use to communicate
over regardless of the transport.

This interface is called netbios on DOS machines, and their derivitives and
in true dos fashion no two network vendors betBIOSes look quite the same.

I think (Could be wrong) that most of the redirector runs through netbios,
so that dish, and printer services have a netbios packet shoved around them
and then the transport packet globbed around that.

> ------------------------------------------------------------------ Timur Tabi
> Contributing Editor for "OS/2 Monthly"        Internet:    timur@seas.gwu.edu
>                                               Fidonet: Timur Tabi @ 1:109/347

...BRU

-----------[000453][next][prev][last][first]----------------------------------------------------
Date:      Sat, 23 Apr 1994 18:57:27 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   This is the FAQ for comp.protocols.tcp-ip

Hi Folks,

	Following this note is a first cut at a FAQ for this newsgroup.  
Please email me any corrections, comments or additions.  

Later,
George


Internet Protocol Frequently Asked Questions

Maintained by: George V. Neville-Neil (gnn@netcom.com)

Version 1.0

Last Update:  April 21, 1994

************************************************************************

	The following is a list of Frequently Asked Questions, and
their answers, for people interested in the Internet Protocols,
including TCP, UDP, ICMP and others.  Please send all additions,
corrections, complaints and kudos to the above address.  This FAQ will
be posted on or about the first of every month.

************************************************************************

Glossary:

I felt this should be first given the plethora of acronyms used in the
rest of this FAQ.

IP: Internet Protocol.  The lowest layer protocol defined in TCP/IP.
This is the base layer on which all other protocols mentioned herein
are built.  IP is often referred to as TCP/IP as well.

UDP: User Datagram Protocol.  This is a connectionless protocol built
on top of IP.  It does not provide any guarantees on the ordering or
delivery of messages.  This protocol is layered on top of IP.

TCP: Transmission Control Protocol.  TCP is a connection oriented
protocol that guarantees that messages are delivered in the order in
which they were sent and that all messages are delivered.  If a TCP
connection cannot deliver a message it closes the connection and
informs the entity that created it.  This protocol is layered on top
of IP.

ICMP:  Internet Control Message Protocol.  ICMP is used for
diagnostics in the network.  The Unix program, ping, uses ICMP
messages to detect the status of other hosts in the net.

RFC: Request For Comment.  RFCs are documents that define the
protocols used in the IP Internet.  Some are only suggestions, some
are even jokes, and others are published standards.  Several sites in
the Internet store RFCs and make them available for anonymous ftp.

SLIP:  Serial Line IP.  An implementation of IP for use over a serial
link (modem).  CSLIP is an optimized (compressed) version of SLIP that
gives better throughput.

Bandwidth:  The amount of data that can be pushed through a link in
unit time.  Usually measured in bits or bytes per second.

Latency:  The amount of time that a message spends in a network going
from point A to point B.

Jitter:  The effect seen when latency is not a constant.  That is, if
messages experience a different latencies between two points in a
network.

RPC:  Remote Procedure Call.  RPC is a method of making network access
to resource transparent to the application programmer by supplying a
"stub" routine that is called in the same way as a regular procedure
call.  The stub actually performs the call across the network to
another computer.

Marshalling:  The process of taking arbitrary data (characters,
integers, structures) and packing them up for transmission across a
network.


Q) Are there any good books on IP?

A) Yes.  Please see the following:

Internetworking with TCP/IP Volume I
(Principles, Protocols, and Architecture)
Douglas E. Comer
Prentice Hall 1991

This volume covers all of the protocols, including IP, UDP, TCP, and
the gateway protocols.  It also includes discussions of higher level
protocols such as FTP, TELNET, and NFS.

Internetworking with TCP/IP Volume II
(Design, Implementation, and Internals)
Douglas E. Comer / David L. Stevens
Prentice Hall 1991

Discusses the implementation of the protocols and gives numerous code
examples.

Internetworking with TCP/IP Volume III (BSD Socket Version)
(Client - Server Programming and Applications)
Douglas E. Comer / David L. Stevens
Prentice Hall 1993

This book discusses programming applications that use the internet
protocols.  It includes examples of telnet, ftp clients and servers.
Discusses RPC and XDR at length.

The Design and Implementation of the 4.3 BSD Operating System
Samuel J. Leffler, Marshall Kirk McKusick, Michael J. Karels, John S.
Quarterman 
Addison-Wesley 1989

Though this book is a reference for the entire operating system, the
eleventh and twelfth chapters completely explain how the networking
protocols are implemented in the kernel.


Q)  Where can I find example source code for TCP/UDP/IP?

A)  Code from the Internetworking with TCP/IP Volume III is available
for anonymous ftp from:

arthur.cs.purdue.edu:/pub/dls


Q)  Are there any public domain programs to check the performance of
an IP link?

A)  

TTCP:  Available for anonymous ftp from....

Host gatekeeper.dec.com

    Location: /.0/BSD/NetBSD/NetBSD-current/othersrc
      DIRECTORY dr-xr-xr-x        512  Apr  8 09:57  ttcp
    Location: /.0/BSD/NetBSD/NetBSD-current/othersrc/ttcp
           FILE -r--r--r--       3885  Nov  7 03:35  ttcp.1
           FILE -r--r--r--      19225  Nov  7 03:35  ttcp.c

Host world.std.com

    Location: /src/wuarchive/graphics/graphics/mirrors/sgi.com/sgi/src/ttcp
           FILE -r--r--r--       3885  Oct  4 1991  ttcp.1
           FILE -r--r--r--      19170  May 17 1993  ttcp.c
           FILE -r--r--r--      13033  Sep  5 1989  ttcp.c-brl

There is suite of Bandwidth Measuring programs from gnn@netcom.com.
Available for anonymous ftp from ftp.netcom.com in
~ftp/gnn/bwmeas-0.3.tar.Z These are several programs that meausre
bandwidth and jitter over several kinds of IPC links, including TCP
and UDP.


Q) Where do I find RFCs?

A)  RFCs are available for anonymous ftp from the following servers.
You should pick the one geographically closest to you. 

North America

FTP.NISC.SRI.COM , NIC.DDN.MIL and nic.cerf.net

Austrailia and Pacific Rim

munnari.oz.au

Denmark

ftp.diku.dk

Germany

walhalla.informatik.uni-dortmund.de

Finland

funet.fi

Netherlands

mcsun.eu.net

Norway

ugle.unit.no

Sweden

sunic.sunet.se and chalmers.se

-- 
gnn@netcom.com

Gentelmen, I will not have you fighting in the War Room.
					--- The President in Dr. Stragelove

-----------[000454][next][prev][last][first]----------------------------------------------------
Date:      24 Apr 1994 02:31:02 -0400
From:      kchen@science.aaas.org (Ken Chen)
To:        comp.protocols.tcp-ip
Subject:   Problem: Unreadable packets from router.....

Hi folks:

We recently installed a NetWare Connect comm server onto our LAN.  After the
installation, we've been getting numerous errors on our SUN syslog stating that 
the it is getting packets from the NWC server/router that are unreadable.
The NWC is probably talking IPX and not IP since the rest of our LAN is 
IPX.  We've added IPX routers to our LAN before and never had this problem.

Anyone have any suggestions?  Anything would be appreciated!  Maybe the 
NWC is encapsulating the IPX packets with improper IP headers?  

Thanks in advance!
Ken

  -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
  Ken Chen                                |      Internet:     kchen@aaas.org
  Computer Specialist                     |      Phone:        (202) 326-7042
  American Association for the            |      Fax:          (202) 842-1711 
     Advancement of Science               | **Opinions stated may not 
  1333 H Street, NW, Washington, DC 20005 |   reflect that of the employer**
  -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

-----------[000455][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 01:19:31 GMT
From:      jcmurphy@acsu.buffalo.edu (Jeff Murphy)
To:        biz.sco.general,comp.protocols.tcp-ip,comp.sys.amiga,comp.sys.amiga.programmer
Subject:   Re: rwhod warped with an Amiga

>>They are included in the AmiTCP binary distribution because there was
>>nothing better available at that time. The current AmiTCP version
>>(2.3) has the NcFTP, which you might want to use instead of the "ftp".
 
>I more recently noticed that was there, but it seems to try to do an anon
>connection by itself when you start it. Note that I know almost nothing


NcFTP is an excellent ftp alternative. by default it assumes you are
attempting an anonymous connection. use the "-u" option to have it 
prompt you for a username/password combination instead.

(btw: it is available for unix, which is where it was originally developed
 i believe.. i use it all the time on the unix boxes i use...)

-- 
jcmurphy@acsu.buffalo.edu                            ece/cs             cit oss
opnsmurf@ubvms.bitnet				     standard disclaimers apply
''She walks in beauty, like the night - Of cloudless climes and starry skies.''

-----------[000456][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 04:53:30 GMT
From:      ibottema@alkaid.sce.carleton.ca (Ike Bottema)
To:        comp.protocols.tcp-ip
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco route

proyse@peeras.demon.co.uk (Phil Royse) writes:

>Only OSPF allows subnet masks to be passed along with the routing info
>and you need this to do variable length masks.  But you know this already.

Apparently so does IGRP if I'm not mistaken.  But then again it's a
proprietary protocol which won't help with Wellfleet units.

Ike Bottema


-----------[000457][next][prev][last][first]----------------------------------------------------
Date:      24 Apr 1994 06:04:53 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco route

In article <ibottema.767163210@alkaid.sce.carleton.ca>
ibottema@alkaid.sce.carleton.ca (Ike Bottema) writes: 
    
    >Only OSPF allows subnet masks to be passed along with the routing info
    >and you need this to do variable length masks.  But you know this already.
    
    Apparently so does IGRP if I'm not mistaken.  But then again it's a
    proprietary protocol which won't help with Wellfleet units.
    
Actually, IGRP does not pass subnet masks.  You need EIGRP to do this.

Tony

-----------[000458][next][prev][last][first]----------------------------------------------------
Date:      24 Apr 94 14:23:56
From:      drw@nevanlinna.mit.edu (Dale R. Worley)
To:        comp.protocols.tcp-ip
Subject:   Re: Encyption of data between nodes across the net

In article <766598295.222snx@ozinkl.pc.my> Geoffc@ozinkl.pc.my (Geoff Collins) writes:
   Without extensive changes to all TCP stacks on all workstations, and
   machines connected on each side, is there an easy way to implement some
   form of encryption, to ensure that proprietary data and other
   intelectual property passed between the two sites is protected, as the
   data passes across the net, where is might be intercepted by others?

In a perfect world, the routers that gateway from Office A to the
outside world and from Office B to the outside world would
encrypt/decrypt all outgoing packets that were to/from the other
office.  But I doubt that such a router exists.

Dale

Dale Worley		Dept. of Math., MIT		drw@math.mit.edu
--
The Garden of Eden was composed of green grass, thick forests, innocent
sex, and a sin worth leaving for.

-----------[000459][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 17:13:30
From:      hharamis@cohesive.com (Harry Haramis)
To:        comp.protocols.tcp-ip
Subject:   Looking for IP Address Librarian Software

Hello All,

I am looking for any software, commercial or shareware, that will act as a 
librarian for IP Addresses.  Advanced features would be great like,
GUI based, feeds DNS,  randomly selects hostnames, etc...

Does anybody know of such a beast?

I prefer email replies to: hharamis@cohesive.com

Thanks,

Harry Haramis
hharamis@cohesive.com

-----------[000460][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 16:19:58 GMT
From:      atkinson@itd.itd.nrl.navy.mil (Ran Atkinson)
To:        comp.protocols.tcp-ip,comp.protocols.snmp
Subject:   Re: MIL-STD for SNMP?

In article <CoMnq4.Csr@news.ess.harris.com> mwallace@sur3ax.ess.harris.com (Mark Wallace) writes:
>Question for a colleague:
>
>Does anyone know if there is a MIL-STD (military standard)
>number associated with SNMP?

Not that I know about.  In any event, the TCP/IP MIL-STDs are known
to be incorrect.  It is _always_ best to use the appropriate RFC
citations directly.

Ran
atkinson@itd.nrl.navy.mil



-----------[000461][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 18:58:28 GMT
From:      ian@unipalm.co.uk (Ian Phillipps)
To:        comp.protocols.tcp-ip
Subject:   Re: What is NetBIOS?

In article <1994Apr23.033516.1@cc.curtin.edu.au>,
 <zcookbruc@cc.curtin.edu.au> wrote:
>In article <2p44q9$b9a@cronkite.seas.gwu.edu>, timur@seas.gwu.edu (Timur Tabi) writes:
>> Subject says it all.  I know it has something to do with peer-to-peer,
>> since IBM's TCP/IP for OS/2 says that it's NetBIOS provides p2p service.
 
>netBIOS gets applied in several contexts, the most common one (From my
>exposure) is the device routing layer the task-to-task comms that is
>barnacled into DOS when it gets a networking system.
 
>For a program to talk to a network there must be a network interface, just
>like there is a video/keyboard/disk etc interface. This interface forms
>a (supposably) network independant layer that programs can use to communicate
>over regardless of the transport.
 
>This interface is called netbios on DOS machines, and their derivitives and
>in true dos fashion no two network vendors betBIOSes look quite the same.

Well, that's not really true.

netBIOS (or is it NETBios :-) defines a programmers' interface (See
Ralph Brown's Interrupt List for the gory details).  It doesn't really
define the packets that go out on the network.  There are two standards
that matter defacto (maybe three) - NETbeui and TCP/IP. The latter is
defined in RFC 1001 and 1002. The former is used by Microsoft. I think
you can run it over IPX (Novell) as well.

There are NETBios/TCP file servers which run on various Unixen, and Lan
Manager provides a server on (e.g.) OS/2. Windows for Workgroups will
run over NETbeui or TCP/IP - doesn't matter as long as the stack is
there and working.

We've tried this at Unipalm and have it working either *over* FTP's
PC/TCP stack, using RFC 100[12], and *alongside* using NETbeui.
Most TCP/IP vendors for DOS provide a NETBIOS support program.

Ian
-- 
Ian Phillipps. Tech support manager, Unipalm. News admin, pipex. Internic: IP4
Many system managers claim that holes in an NNTP stream are more valuable than
the data. Van Jacobson, RFC 1144


-----------[000462][next][prev][last][first]----------------------------------------------------
Date:      Sun, 24 Apr 1994 19:23:01 GMT
From:      kardel@immd4.informatik.uni-erlangen.de (Frank Kardel)
To:        comp.protocols.tcp-ip
Subject:   Re: NTP algorithm specification?

jwmanly@unix.amherst.edu (John W. Manly) writes:

>Hi there.  I was wondering if anyone can point me to an article or an RFC
>that discusses the algorithm that underlies the NTP protocol.
RFC1305

>The RFC on
>NTP itself (I forget the number) seems to only talk about the protocol being
>used for passing time around, not for actually doing the synchronization by
>taking things like network delay into account, particularly when using 
>multiple timesources.
No, NTP does take network delays and multiple sources into account. (its in the
RFC).

>If anyone has any suggested reading, please let me know.
Well, take the RFC and serveral files found on louie.udel.edu:/pub/ntp/doc/*.

The news group is comp.protocols.time.ntp.

Frank Kardel (time@informatik.uni-erlangen.de)

-----------[000463][next][prev][last][first]----------------------------------------------------
Date:      24 Apr 1994 21:58:53 -0000
From:      sunborn@news.delphi.com (SUNBORN@DELPHI.COM)
To:        comp.protocols.tcp-ip
Subject:   Subnetting Class C into Eight nets ?

I would like to subnet our class C network into six to eight subnets, can 
this be done ?


-----------[000464][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 11:41:14 -0700
From:      schmidt@tango.ics.uci.edu (Douglas C. Schmidt)
To:        comp.object,comp.client-server,comp.lang.c++,comp.unix.solaris,comp.protocols.tcp-ip
Subject:   OO network programming toolkit ACE 2.14 now available for ftp

	       ========================================

        The latest version of the ADAPTIVE Communication Environment
(ACE) object-oriented network programming toolkit is now available for
anonymous ftp from the ics.uci.edu (128.195.1.1) host in the
gnu/C++_wrappers.tar.Z file (approximately .5 meg compressed).  This
release contains contains the source code, documentation, and example
test drivers for C++ wrapper libraries and higher-level network
programming frameworks developed as part of the ADAPTIVE project at
the University of Calfornia, Irvine.

        o The contents of this release encapsulate the following 
          user-level BSD and System V Release 4 (SVR4) IPC facilities
          via type-secure, object-oriented C++ interfaces:

                o Internet and UNIX-domain sockets (including broadcasting) and TLI
                o Event multiplexing via select and poll
                o named pipes (FIFOs) and STREAM pipes (plus connld)
                o the mmap family of memory-mapping APIs
                o System V IPC (i.e., shared memory, semaphores, message queues)
                o SVR4 explicit dynamic linking facilities (i.e., dlopen/dlsym/dlclose)

        o In addition, there are also a set of higher-level network
          programming frameworks that integrate and enhance the
          lower-level C++ wrappers to support the dynamic
          configuration of concurrent network daemons composed of
          complex distributed application services 

        Many of the C++ wrappers and higher-level components have been
described in issues of the C++ Report, as well as in the proceedings
of (1) the 2nd Annual C++ World conference, October 1993, (2) the 11th
and 12th Annual Sun Users Group Conference in 1993 and 1994, (3) the
2nd International Workshop on Configurable Distributed Systems, March
1994, the 6th USENIX C++ Conference in April 1994, and the 9th OOPSLA
Conference to be held in October 1994.

        A relatively complete set of documentation and extensive
examples are included in the release.  The current release has been
tested fairly extensively on Sun workstations running Sun OS 4.1.2 and
Solaris 2.2.  I expect that most of the release will port easily to
other platforms.  If anyone is willing to help coordinate ports to
other platforms please let me know.

        A mailing list is available for discussing bug fixes,
enhancements, and porting issues regarding ACE.  Please send mail to
ace-users-request@ics.uci.edu if you'd like to become part of the
mailing list.

	       ========================================

CONTENTS OF THE RELEASE

        The following subdirectories are included in this release:

        . apps    -- complete applications written using the ACE wrappers
        . bin     -- utility programs for building this release such as g++dep
        . build   -- a separate subdirectory that keeps links into the main
                     source tree in order to facilitate multi-platform
                     build-schemes
        . include -- symbolic links to the include files for the release
        . lib     -- object archive libraries for each C++ wrapper library
        . libsrc  -- the source code for the following C++ wrappers:
                        ASX -- higher-level C++ network programming framework
                        Get_Opt -- a C++ version of the UNIX getopt utility
                        SOCK_SAP -- wrapper for BSD sockets
                        TLI_SAP -- wrapper for SVR4 TLI 
                        FIFO_SAP -- wrapper for FIFOS (named pipes)
                        SPIPE_SAP -- wrapper for SVR4 STREAM pipes and connld 
                        Log_Msg -- library API for a local/remote logging facility
                        Mem_Map -- wrapper for BSD mmap() memory mapped files 
                        Message_Queues -- wrapper for SysV message queues
                        Reactor -- a framework for event demultiplexing and dispatching
                        Semaphores -- wrapper for SysV semaphores
                        Service Configurator -- a framework for dynamically linking
                        Shared_Memory -- wrapper for SysV shared memory
                        Shared_Malloc -- wrapper for SysV/BSD shared mallocs
        . tests -- programs that illustrate how to use the various wrappers

        Please refer to the INSTALL file for information on how to
build and test the ACE wrappers.  Due to the large size of the release
(~2 Meg) I'm sorry that I will be unable to distribute the ACE
wrappers via email.  The BIBLIOGRAPHY file contains information on
where to obtain articles that describe the ACE wrappers and the
ADAPTIVE system in more detail.

        Also, please note that there is a companion tar file called
C++_wrappers_doc.tar.Z, which is approximately 2 Meg compressed.  This
file is in the same ftp/gnu directory as the source code distribution.
In this file is the following:

        . doc     -- LaTeX documentation (in both latex and .ps format)
        . papers  -- postscript versions of various papers describing ACE

	       ========================================

COPYRIGHT INFORMATION

        You are free to do anything you like with this code.  However,
you may not do anything to this code that will prevent it from being
distributed freely in its original form (such as copyrighting it,
etc.).  Moreover, if you have any improvements, suggestions, and or
comments, I'd like to hear about it!  It would be great to see this
distributed evolve into a comprehensive, robust, and well-documented
C++ class library that would be freely available to everyone.
Natually, I am not responsible for any problems caused by using these
C++ wrappers.

        Thanks,
        
                Douglas C. Schmidt
                (schmidt@ics.uci.edu)
                Department of Information and Computer Science
                University of California, Irvine
                Irvine, CA 92717
                Work #: (714) 856-4105
                FAX #: (714) 856-4056

	       ========================================

ACKNOWLEDGEMENTS
        
        Special thanks to Paul Stephenson for devising the recursive
Makefile scheme that underlies this distribution.  Also thanks to Olaf
Kruger for explaining how to instantiate templates for shared
libraries on SunOS 4.

-- 
His life was gentle, and the elements so            | Douglas C. Schmidt
Mixed in him that nature might stand up             | schmidt@ics.uci.edu
And say to all the world: "This was a man."         | ucivax!schmidt
   -- In loving memory of Terry Williams (1971-1991)| (714) 856-4105

-----------[000465][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 94 01:51:57 +0100
From:      alberto@cnuce.cnr.it (Alberto Mura)
To:        comp.protocols.tcp-ip
Subject:   IMAP server & client

I am looking for IMAP (Interactive Mail Access Protocol) software, both
server and client. Can anybody tell me the names of good programs?
For IBM PC client preference is for Windows 3.1 software.

Thanks in advance,

Alberto Mura
Dept. of Philosophy
Univ. of Pisa
Italy


-----------[000466][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 04:31:04 GMT
From:      dmag@engin.umich.edu (Daniel Demaggio )
To:        comp.protocols.tcp-ip
Subject:   Re: AFS

In article <1994Apr21.235316.6327@unlv.edu> ftlofaro@unlv.edu (Frank Lofaro) writes:
>In article <NADEL.145.000F8DB2@litc.lockheed.com> NADEL@litc.lockheed.com (Ron Nadel) writes:
>>I'm trying to find out where I would get AFS software.  Like most sites, we 
>>use NFS, but find it such a poor performer regardless of media, and I 
>>understand that AFS (Andrews File System?) is supposed to be much better.  Can 
>>someone tell me something about it, and where to go to get it?
>>
>>Thanks very much,
>>
>>Ron
>
>If you think NFS is a poor performer, just wait until you try AFS!
>It is DREADFUL! Don't even bother.
>I was at an AFS site a little while back:
 [ Extreme compression of previous post here ==> . ]

You must have been running an older version or something.. Here at UM, we run
tons of AFS servers, and I think most of our binaries are run from the net.
(except Mentor Graphics CAD, for obvious reasons).  You best bet is to check
out the FAQ for AFS. (see alt.filesystems.afs)
 My reasons for liking AFS are:
- Caching: When you read a file, it's cached on the local machiene.  Repeated
reads don't go over the network at all.  If the file is changed, *THE SERVER*
notifies your station to dump that file from it's cache.  For the typical user
(compile, edit, run) it's a big win..
- Replication: You can have *MULTIPLE* servers for the same (read-only)
information.  The load will be distributed accross them.  Big win for speed.
- Transparency: Your files can be moved around from server to server (by the
sysadmins to make space) and you will never know it.
- Backups. If you have a large volume, your chances of getting a consistant
backup are slim, except in AFS. A "backup" takes almost no time at all, and you
can copy the files off at your own pace.

ITD here at UM has even developed AFS-AppleTalk and AFS-NetWare translators
that work quite well (I use my same home directory wherever I am on campus,
even on the Macs!)
				-=Dan=-
-- 
dmag@umich.edu | When laws are outlawed,      | Ono-Sendai: the best
Dangerous  Dan | only outlaws will have laws. | Sim Stim decks

-----------[000467][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 12:57:28 -0500
From:      johns@oxygen.house.gov (John Schnizlein)
To:        comp.protocols.tcp-ip
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2p9g3i$bpj@genesis.MCS.COM>, mikea@MCS.COM (Mike Andrews)
wrote:

> Help!  I'm  in multi-vendor, variable subnetting, finger pointing, IP hell!
> I'm running only IP and RIP over the links..
> The main problem is that when I set the netmask on the directly connectly 
> ethernet interface on the Cisco 7000 properly to 255.255.255.224 and do
> a show ip route it says, correctly, "Network 128.1 is variably subnetted...
> 255.255.255.0 on Ethernet.... 255.255.255.0 on Serial..."  But the Cisco will
> not route any traffic from the 128.1.1.96 ethernet to the Frame Relay network.
> I incorrectly set the netmask on the ethernet interface to 255.255.255.0
> and now it routes. But then it broadcasts an incorrect route to a 128.1.1.0 LAN
> which screws up any chance of using other 128.1.1 subnets.
> 
> The Wellfleets have no trouble routing over the entire WAN.  I can ping a host
> on a LAN behind the 128.1.52.6 host from a host on the the 128.1.1.96 LAN.
> 
> My Wellfleet SE says that there's nothing wrong with my topology.  

When we had a Wellfleet SE he was also unrealistically optimistic.
Wellfleet running a non-standard RIP?
Are they supporting RIPv2?

Using variable subnets with RIP requires great skill because you have to
essentially make routers lie to each other. Since RIP has no way to tell
which subnet masks are associated with which subnets, you would have to
arrange that all the smaller subnets which would appear to be contained in
any larger subnet are reachable through the same path. Then you would need
to put the lies about subnet ranges into the router with static route
statements. Static routes do support variable subnets (assuming cisco
version 9.1 or better).

I would not do any of this. The preplanning of address range management is
too critical to long term success to have any chance of avoiding Murphey's
Law. OSPF was designed to support variable subnets and explicitly carries
mask info along with subnet value.

Re: Why your RIP almost works with your current configuration:
My guess is that the cisco7000 assumes a mask of 255.255.255.224 for any
RIP subnets advertised to it on ethernet 128.1.1.96. You should look in its
route table for the mask it associated with network 128.1.101.0 for
example.
My best theory as to why the Wellfleet appeared to work is that all your
128.1.1.XX subnets were reached through the ethernet on the BLN, or you had
proxy ARP turned on (down some other branch of the menu tree) and didn't
know it.  Can you tell us what mask the Wellfleet associated with subnet
128.1.1.96?

Running OSPF seems like your best way out of this pickle
-- 
Badges! we don't need no stinking badges!         |
disclaimer! we don't need no stinking disclaimer! | John M. Schnizlein
everybody knows nobody can represent the views of | johns@oxygen.house.gov
435 elected policy makers.                        | router jockey

-----------[000468][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 15:02:52 -0400
From:      Craig_Everhart@transarc.com
To:        comp.protocols.tcp-ip
Subject:   Re: AFS

Excerpts from netnews.comp.protocols.tcp-ip: 21-Apr-94 AFS Ron
Nadel@litc.lockheed. (315) 

> I'm trying to find out where I would get AFS software.  Like most sites, we 
> use NFS, but find it such a poor performer regardless of media, and I 
> understand that AFS (Andrews File System?) is supposed to be much better.  Can 
> someone tell me something about it, and where to go to get it? 

AFS is a commercial product of Transarc Corporation; you can find out
more via afs-sales@transarc.com. 

(Hope this isn't too commercial to violate ethics rulings, but it is the
answer to a direct question.) 

		Craig 
 

-----------[000469][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 13:45:40
From:      clarkb@netstar.com (Clark Bremer)
To:        comp.protocols.tcp-ip
Subject:   Re: DOS TCP/IP Stacks?

In article <43458@mindlink.bc.ca> KCARPENT@mindlink.bc.ca (Ken Carpenter) writes:
>From: KCARPENT@mindlink.bc.ca (Ken Carpenter)
>Subject: DOS TCP/IP Stacks?
>Date: 25 Apr 94 16:38:44 GMT
 
>I am looking for information on companies who offer TCP/IP stacks for DOS.
 
>Could anyone who knows of such products please email me with the company's
>name, and if possible, the phone number or address.

Probably the best [and appropriately priced :-( ] is PC/TCP from FTP Software:

FTP Software Inc.
26 Princess Street
Wakefield, MA 01880
(617) 256-0900

It has everything.  Dos and windows versions of telnet, ftp, ping; WinSock 
compliant under windows; NFS, pop-mail, nntp support, lpr printing, etc.  Very 
tunable, and works fine side-by-side with Novell Netware, sharing the ODI 
driver.  On the down side, it sucks down a big chunk of your precious 640K.  
There may be better windows-only stacks, but this is the best Dos product I've 
tried.  CB.

===========================================================================
          _  _               Clark Bremer     clarkb@netstar.com
         /  /_)              Software Engineer, NetStar Inc.
         \_/__)              10250 Valley View Road  MPLS, MN 55344

-----------[000470][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 09:05:14 GMT
From:      fontaine@sri.ucl.ac.be (Alain Fontaine)
To:        comp.protocols.tcp-ip
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2p9g3i$bpj@genesis.MCS.COM>, mikea@MCS.COM (Mike Andrews)
wrote:
> If you have any experience with variable subnetting I'd appreciate your
> suggestions or comments.   I'd also like info and experiences in installing
> OSPF on a multi-vendor WAN like this.
I do have Wellfleet's.
I do have Cisco's.
I do use several subnet masks on the same networks.
I do run OSPF.
I am quite happy (at least for this...).                       /AF

P.S. forget RIP if you use variable subnetting.

-----------[000471][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 09:09:07 +0000
From:      proyse@peeras.demon.co.uk (Phil Royse)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting Class C into Eight nets ?

In article <2peq2t$arm@news.delphi.com> sunborn@news.delphi.com writes:

>I would like to subnet our class C network into six to eight subnets, can 
>this be done ?

Yes.  (Stricly speaking not "six to eight" subnets).  If you set
a subnet mask of:

	255.255.255.192  you get  2 subnets of 62 hosts
	255.255.255.224           6            30
	255.255.255.240           14           14
	255.255.255.248           30            6
	255.255.255.252           62            2

The lost subnets (two, when you might think you should get four,
(00, 01, 10, 11) are becasue 00 and 11 cannot be used.

Read some of the other articles in this newsgroup.  This subject
is coming up more and more.

-- 

Phil Royse     Comms Consultant  |  PRA Consulting Ltd.
TUDOR HOUSE                      |
12 Woodside Road, Purley
Surrey  CR8 4LN   (UK)        Tel: (+44) 81 645-986

-----------[000472][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 09:18:33 GMT
From:      hunen@brc.medtronic.com (Roger Hunen)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting Class C into Eight nets ?

In article <vdh.41.00094741@info.fundp.ac.be> vdh@info.fundp.ac.be writes:
#In article <2peq2t$arm@news.delphi.com> sunborn@news.delphi.com (SUNBORN@DELPHI.COM) writes:
#>I would like to subnet our class C network into six to eight subnets, can
#>this be done ?
#
#From what I understand of subnets, you will have to devote 3 bits for
#the 'subnet address' and it leaves 5 bits for the host adresses, that's to
#say up to 32 hosts.  Perhaps less hosts as some ip adresses should be
#reserved for broadcasting (am i right on this point ?)
#
#Subnet mask would be set to 255.255.255.224

Suppose you have S bits for the subnet and H bits for the host number, then:

1. S+H = 24 for a subnetted class A network
   S+H = 16 for a subnetted class B network
   S+H =  8 for a subnetted class C network

2. Since the subnet number must not be all 0s or all 1s:
   A. Nsubnets = 2^S - 2
   B. S >= 2

3. Since the host number must not be all 0s or all 1s:
   A. Nhosts_per_subnet = 2^K - 2
   B. K >= 2

Therefore the subnet mask 255.255.255.224 (S=3,K=5) gives you 6 subnets with
30 hosts each.

Regards,
-Roger

-----------[000473][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 94 15:09:51
From:      hafner@informatik.tu-muenchen.de (Walter 'madhouse' Hafner)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,de.comm.internet,comp.protocols.ppp
Subject:   SUMMARY [was: help with ka9q setup needed]

Hello!

A couple of days I asked:

   I need a little bit help with my ka9q installation. It seems, I can't get
   ka9q to dial out!

   My configuration: 386DX/16 (Yes!), MD-DOS 5.0 with 4DOS 5.0D, QEMM 5.01
   and a couple of TSRs. The ka9q version is that of the e920603.zip file.
   My modem is a ZyXEL 1496E with the 5.01 firmware.

   The ka9q setup is as follows:

   [del. Read the original posting in case you are interested :-)]

   The (quite old) docu says about the 'dialer' command:

   dialer <iface> [<dialer-file> [<seconds> [<tests> [<hostid>]]]]

   The online-help says:

   dialer <iface> <timeout> [<raise script> <lower script>]

   From the docu I think I understand the parameters. But what do the
   parameters of the online-help mean? And what format is the right one? In
   doubt I tend to believe the online-help.

   [del.]

Quite a lot of people wrote back to me, most of them saying 'I have the
same trouble aswell'. :-)

Here's a short summary of hints I got (I can't tell you which of them
are working because I'm just coming back from a short-trip to the
mountains: I really needed it! :-) So I just did a little testing by
now. I'll do a third posting as soon as I find the fault.)

All text in [..] is by me.


====================
Colin Spensley <colin@alecto.demon.co.uk> writes:

This may or may not be a helpful suggestion. :-)

The service provider which I use (Demon Internet Systems) and several of
their subscribers have put together a version of KA9Q which has been
pretty heavily developed. I believe it is now widely regarded as one of
the most stable version of KA9Q and it is also quite well documented.

There is a new users package which includes an automated installation.
It is available on ftp.demon.co.uk in directory (from memory)
/pub/ibmpc/DIS. I think the file is DIS.EXE but you can look around. You
might consider trying this version and see whether it works better for
you. There are over 3000 DIS subscribers and a large portion of them use
this package so it definitely works. If nothing else it would produce a
'working' configuration for you to adapt.

The only drawback is that Demon use PPP by default rather than SLIP
(which you seem to be trying). I can't remember whether the SLIP support
is still compiled into the DOS version (I use an NT port of this
software) but in any case maybe you have the option of using PPP and,
regardless of the protocol, this software should most certainly dial!

[ I tested it: SLIP support is NOT compiled in. Unfortunately for me I
only have a SLIP feed. A friend of mine says, the DIS version would dial
ok when using PPP but wouldn't dial with SLIP. Strange ... However, I
didn't try it out.]


====================
ian@isis.demon.co.uk (Ian Smith) writes:

I can't help you with what you have, but I can tell you where to get
a KA9Q optimised for phone access. I'm using it now, and have done,
daily, for the last 18 months.
ftp.demon.co.uk  /pub/DIS/NET215.ZIP
They have several versions of their NET as well as PCElm and SNEWS.
Better, they have a thing called DIS which is a front end to it all
and will configure the three progs. There is also a complete package.

[ see above: The install package is excellent. Non-DIS sites just have
to do a little adjusting manually ... Unfortunately you can't use the
DIS-version config files for the vanilla KA9Q, because of some additions
by DIS to the program]


====================
jbledsoe@eosc.osshe.edu (Jeff W. Bledsoe) writes:

below is the batch file I use to start ka9q (Original version ?)
__________________________
net -d \spool autoexec.net
--------------------------

In your autoexec.net file:


"beginning info"

attach asy... etc
dialer sl0 slipdial       <<if you use a file called "slipdial">>
			  <<put it below the attach line even thought
			    it shouldn't matter.>>
"rest of info"


--------------------------------------------------
In your "slipdial" file use something like:

control down
wait 1000
control up
wait 1000
wait 2000
send "at\r"
wait 3000 "OK"
send "atdt#######\r"		# = phone number to dial
send "\r"
wait 60000 "login: "		The "login:" part must match what you will see
send "slip\r"			whatever login you use
wait 5000 "password:"		must match what you see again if PW used
wait 1000
send "yourpassword\r"


This should work or a slight variation of it.  HOPEFULLY  :-)

[unfortunately not. I already knew about the \r. Even if they are never
mentioned in the manuals ...]


====================
gh@wurstl.muc.de (gottfried huengsberg) wrote me a mail in german,
saying that he too can't get KA9Q to dial. Noe he's using the
wattcp/cutcp tools from dorm.rutgers.edu.


====================
wwalker@inhuf.ang.af.mil (William D. Walker) writes:

use the dialer line in the docs.
here is my setup

attach asy 0x2f8 3 slip comm2 1024 512 19200
ifconfig comm2 ipaddress 131.57.99.1
ifconfig comm2 netmask 0xffffff00
dialer comm2 modem.net 60 3 131.57.100.1

of course your using comm1 and calling it sl0.
might let me see your dialer file
start out your dialer with an ATZ

here is my modem.net file it is setup to receive

configure:
init "atz\r"
dial_cmd "atdt"
retries 1
#
#
execute:
control down
wait 2000
control up
wait 3000
init
wait 3000
send "ats0=1\r"

[ That are files for the DIS version. At least the wont's work with the
version I got from SimTel.]


====================
Jaroslaw Lis <jjlis@ict.pwr.wroc.pl> writes:

First, is dialing started ? You should get a message "...Dialing..."
I had problem with modem which echoed characters in command mode (standard
situation). KA9Q tried to send PING, it "received" this packet (echoed)
and won't try to dial, because it se IP packet from connection !
I had to disable command echo (ate0) BEFORE starting KA9Q.

try to dial manually (TIP SL0 starts a simple terminal). Do exactly the same
as in your DIALER file and check if modem will dial?


Hope this help

[ I hadn't the time to try this one]


====================
jkee@motown.ge.com stated that there is better software:

	I gave up on this also... Go get yourself The Windows Internet
	Tour Guide by M. Fraase. It comes with Chameleon software, a
	windows oriented package PPP/SLIP. A lot easier to install
	and a much nicer program.

[well ... I don't use MS-Windows]
	


By now I think probably it's really better to use the cutcp package with
the extern umslp dialer. The only but major drawback is the missing
multitasking environment!

On the other side: Maybe I just ftp the sources from DIS and compile ma
a very own version ... :-)

Thanks to all who answered to me!

-Walter Hafner
--
Walter Hafner____________________ hafner@forwiss.tu-muenchen.de
FORWISS Muenchen,            Forschungsgruppe Kognitive Systeme
Raum O-134,  Tel: 089/48095-220                   IRC: Majestix
http://www.forwiss.tu-muenchen.de/~hafner/index.html  (German!)

-----------[000474][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 11:06:43 GMT
From:      advisor@dorsai.org (Hector Llorens)
To:        comp.protocols.tcp-ip
Subject:   Software & Systems MTS - NOT New York City.

Distribution: world

Advisor, Sociedad Anonima is a high technology MBE with international 
multi disciplinary Members of Technical Staff. Please include the 
pertinent requested information: salary history, # of years of continuous 
experience from present, position, forte, references (optional). All 
replies including the above information will be process first.

Positions: Senior (7+ years), Mid (5+), Junior (3+), Entry (1+) years.
Function: Application, Systems/Networks, Mainframe, Tools
email/phone(work home):
years:
current compensation without benefits per year: $(USA)
Forte: see areas of technical skill put how current in years & number of 
years of continuous experience for most current.
references: email address and phone #.
us visa, if applicable:

Highly qualified foreign candidates will be considered for sponsorship.

technical skills areas: C,C++,UNIX,client server, object oriented design 
& development,distributed systems, multiprocessor architecture, CASE 
Tools, communication protocols & ISO/OSI standards, ISO 9000 Software 
Quality Assurance, X-Windows, M/S Windows, NT, OS/2, natural language 
processing, DSP, AI/Expert Systems, Business Process Re-engineerding,
CIM, SGML, & other  new technologies in the communications & information 
technology industries.

Please include the above request information & then your CV or resume.
All replies with this format will be acknowledge. Minimum requirement is 
One Years of continuous experience in any of the above skills in a 
non-academic enviroment is preffered.


-----------[000475][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 12:08:10 GMT
From:      khweis1@mvmhp.ciw.uni-karlsruhe.de (Karl-Heinz Weiss)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   announcement: GETNEWS

Announcement of GETNEWS
-----------------------

What it is:
it's a DOS-application, using a packetdriver

What it does: 
GETNEWS collects new news from all newsgroups and all 
newsservers you define in a simple textfile named 'groups.id' and
writes all articels to a single file, for later use by an offline-
reader. GETNEWS uses 'nntpget.exe', a programm I found on 
biochemistry.cwru.edu to query the newsservers.

GETNEWS is totally free. I add the pascal source, so you may
modify the articel-delimters ('rnews-header') if this format 
does'nt work with your offline reader. I also add 'COMTOOL',
a small dialer/terminalprogramm. It may be useful if you connect
to your newserver with slip.

You may ftp it from:
mvmpc9.ciw.uni-karlsruhe.de (129.13.118.9) /public/nos/getnews.zip
(ws_ftp-users should select 'ka9q' as servertype!)

Enjoy!
-------------------------------------------------------------------
K.H. Weiss            E-mail  : <khweis1@mvmhp.ciw.uni-karlsruhe.de>
Eulenweg 2            Phone   : (49)-721-608-2418
76536 Weingarten      Germany

-----------[000476][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 09:17:18 +0200
From:      vdh@info.fundp.ac.be
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting Class C into Eight nets ?

In article <2peq2t$arm@news.delphi.com> sunborn@news.delphi.com (SUNBORN@DELPHI.COM) writes:
>From: sunborn@news.delphi.com (SUNBORN@DELPHI.COM)
>Subject: Subnetting Class C into Eight nets ?
>Date: 24 Apr 1994 21:58:53 -0000
 
>I would like to subnet our class C network into six to eight subnets, can
>this be done ?

From what I understand of subnets, you will have to devote 3 bits for
the 'subnet address' and it leaves 5 bits for the host adresses, that's to
say up to 32 hosts.  Perhaps less hosts as some ip adresses should be
reserved for broadcasting (am i right on this point ?)

Subnet mask would be set to 255.255.255.224

Can anyone confirm this ?

Hope it helps

Vincent D'Haeyere

-----------[000477][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 14:38:40 GMT
From:      skov@ilx.com (John Skovron)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell and IPX

In article <2p4jj0$9nh@news.bu.edu>, John Adams <jna@bu.edu> wrote:
>If this is the wrong group, my apologies.
>
>Has anyone had experience in tunnelling IPX packets, coming from a novell
>server into IP for transmission via PPP ? 
>
>I'm trying to find a way to connect a remote novell client (some pc in
>the middle of nowhere) via PPP to our local (PC) novell network. 
>
>The network has both IP machines (unix boxes) and IPX machines (novell
>servers and clients) on the same wire. We accomplish this via 
>ODIPKT (Clarkson Packet Ethernet Drivers)... 
>
>I Just don't know how to move IPX over a PPP link! 
>
>thanks!
>		-john

Novell has a driver module to do just this, called IPTUNNEL.  Look up IPTUNNEL
in the Novell administrator's manuals.  Also, there's a RFC written by
someone at Novell for the transmission of IPX over IP.  I don't have the
RFC number handy, but if you search for IPX in the rfc-index you'll find it.
We have tested doing something similar using SLIP.  It worked OK in a test,
but we have not yet gone as far as "production" use.


-- 
-----------------------------------------------------------------------------
John Skovron
ILX Systems Inc.
skov@ilx.com

-----------[000478][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 94 15:07:48 GMT
From:      conbvs@melupl (Bill VerSteeg (Contract))
To:        comp.protocols.tcp-ip
Subject:   to SNAP or not to SNAP on token ring (that is the question)

I have been working on a token ring driver for an embedded system
tcp/ip stack, and have noticed a couple of different encapsulations
for IP over token ring. Is there a document describing the various
encapsulations? While I am at it, is there a good reference on token ring in 
general? I need to get my client company up to speed on token ring, and
if I can hand them a pointer to a well-written token ring primer, it
will save everyone invloved some time.

Thanks in advance

Please reply to bvs@ver.com

Bill VerSteeg
bvs@ver.com

-----------[000479][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 94 16:38:44 GMT
From:      KCARPENT@mindlink.bc.ca (Ken Carpenter)
To:        comp.protocols.tcp-ip
Subject:   DOS TCP/IP Stacks?

I am looking for information on companies who offer TCP/IP stacks for DOS.

Could anyone who knows of such products please email me with the company's
name, and if possible, the phone number or address.

Thanks,


Ken Carpenter
Research & Development
Delta Controls

--
Ken Carpenter
Research & Development
Delta Controls

-----------[000480][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 94 16:49:24 GMT
From:      cjross@bbn.com (jonathan ross)
To:        comp.protocols.tcp-ip
Subject:   Re: When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?

vjs@calcite.rhyolite.com (Vernon Schryver) writes:

> cjross@bbn.com (Jonathan Ross) writes:
>>In an attempt to optimize large transfers from the queue
>>server to its clients, the original author chose to set the server's
>>SO_SNDBUF (and clients' SO_RCVBUF) to 16K bytes.  This seems to have had
>>the opposite of the desired effect; the server occasionally gets stuck
>>in write() calls, even when a client is simultaneously calling read().
 
>There are so many people who reflexively
>reason "I can't make [arbitrary commonly used facility] work so it must
>be a system bug", who are unable to conceive of the possibility that the
>trouble might be a personal problem

Uh-huh.  About as many as reflexively assume that anyone who says "I
can't make [facility] work" has made no effort to verify that there
really is a problem.

>Have you tried your program on other BSD TCP stacks?

I have run test cases under SunOS 4.1.3, Solaris 2.3, and HP-UX A9.01.
None show write() delays.

>Have you seen if
>you can reproduce the same behavior with the "-b" arg to `ttcp`?

I have now.  It worked fine.  Thoroughly mystified, I downloaded
ttcp.c from ftp.sgi.com and compared it to the queue server's code.
The key difference is that ttcp.c sets SO_SNDBUF on the server socket
before calling accept(), while I call it for the fd returned from
accept().  Changing either program to use the other's call sequence
causes it to manifest the other's behavior.

Now, why does the order make such a difference under IRIX?  (I agree,
it's pointless to setsockopt() each fd returned by accept() when the
fd automatically inherits the properties of the original socket, but I
didn't write the code.  I will certainly fix it, though.)

>Is there any chance you're seeing the Nagle algorithm in action?

I have briefly reviewed his RFC, and don't see how it could account
for delays of up to four seconds on a connection between two processes
on the same host.

>If you're using UDP and zillions of sockets and clients

TCP, one socket, one client.


=====  Jonathan Ross  =====  <cjross@bbn.com>  =====  (617) 873-3272  =====
   We should not worry about what happens when computers pass the Turing
 test, but rather what happens when humans fail. -- <vs0r+@andrew.cmu.edu>

-----------[000481][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 17:08:07 GMT
From:      Vic Kamhi <kamhiv@pt.cyanamid.com>
To:        comp.protocols.tcp-ip
Subject:   ...but with no DNS, how.....?




I'm not exactly a "newbie" to TCP/IP, but I really haven't had to "look 
under the hood" before. I have successfully used ChameleonNFS for my TCP/IP
connectivity under WfWG3.11 for a while. Recently, I downloaded Microsoft's
TCPIP-32 beta (March beta, have downloaded, but not installed yet, beta 2).

One thing that it is missing is any DNS (noted in the README files as work
under construction) in even the latest beta. I have successfully gotten the
TCP/IP to work with both the included (admittedly Spartan) Microsoft apps as
well as some freeware on the bulletin boards (ie.; FTP, telnet, ping, etc).
These will work because I can specify (when I know it!) the actual IP address,
rather than the "alias" name. But things like Mosaic, Cello, Gopher, etc. 
DEPEND on having a DNS capability, or so I THOUGHT (this comes back to my
not having looked into the nuts and bolts). Yet, Microsoft talks about using
these programs. What am I missing? CAN I still do name resolving? If so, how?
(We DO have a server which does our name resolving; it's just not obvious to
me how to use it under Microsoft's TCPIP-32 beta. Under Chameleon, I WAS able
to understand how to get to do DNS with that server.)

Please help. Thanks, VIC

kamhiv@pt.cyanamid.com

P.S. Also, I've been looking for a freeware/shareware NFS to try with the
Microsoft TCPIP-32 product. Anybody have any recommendations? Any results?

-----------[000482][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 17:27:14 GMT
From:      tomkwong@netcom.com (Thomas Kwong)
To:        comp.unix.questions,comp.unix.sys5.r3,comp.protocols.tcp-ip
Subject:   [Q] Help: how to compile NSLOOKUP in ISC R3.2?

Hi!

I got the nslookup tool to test my DNS server.  However, I
am not able to compile it because it uses some BSD-specific
function:

> tomkwong@discover% make
> cc -o nslookup main.o getinfo.o debug.o send.o skip.o list.o subr.o commands.o -linet
> undefined                       first referenced
>  symbol                             in file
> _getshort                           getinfo.up
> _getlong                            debug.o
> sigmask                             subr.o
> sigblock                            subr.o
> sigsetmask                          subr.o
> yywrap                              commands.o
> ld fatal: Symbol referencing errors. No output written to nslookup
 
Is there anyone who got it compiled successfully (and working successfully)
in the Interactive UNIX System V Release 3.2?

Thanks a lot in advance!!!	Email preferred.

-Tom
-- 
Thomas Kwong (tomkwong@netcom.com)     __&__  |-    
Wilshire Associates, Incorporated.    ___\/__ __]   
Office: (310)451-3051 Ext. 227          [--]  [     
Home  : (310)641-8818 Ext. 911          )==]  [=    
                                       )  .|  [__\  

-----------[000483][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 17:34:22 GMT
From:      mycroft@duality.gnu.ai.mit.edu (Charles Hannum)
To:        comp.protocols.tcp-ip
Subject:   Delayed ack causing stalls


I've found a problem with the TCP delayed ack algorithm.  If the writer's
buffer becomes full before sending an entire window, the writer will stop
and the ack will be delayed and the transmission will be stalled pending
a timeout on (and transmission of) the delayed ack.

As an experiment, I've applied the following patch to my (NetBSD) kernel,
and it alleviates the problem.

The worst case for this change is that the writer sets the PSH bit on
every outgoing packet, in which case delayed ack is effectively disabled.
This is not an issue of correctness, however, and since most vendors use
the PSH bit a bit more intelligently, it doesn't seem like a serious
problem.

I'd appreciate comments on the appropriateness and correctness of this
change.

===================================================================
RCS file: /b/source/CVS/src/sys/netinet/tcp_input.c,v
retrieving revision 1.7
diff -c -2 -r1.7 tcp_input.c
*** 1.7	1994/04/12 18:07:46
--- tcp_input.c	1994/04/25 16:56:32
***************
*** 82,86 ****
  	    (tp)->seg_next == (struct tcpiphdr *)(tp) && \
  	    (tp)->t_state == TCPS_ESTABLISHED) { \
! 		tp->t_flags |= TF_DELACK; \
  		(tp)->rcv_nxt += (ti)->ti_len; \
  		flags = (ti)->ti_flags & TH_FIN; \
--- 82,89 ----
  	    (tp)->seg_next == (struct tcpiphdr *)(tp) && \
  	    (tp)->t_state == TCPS_ESTABLISHED) { \
! 		if ((ti)->ti_flags & TH_PUSH) \
! 			tp->t_flags |= TF_ACKNOW; \
! 		else \
! 			tp->t_flags |= TF_DELACK; \
  		(tp)->rcv_nxt += (ti)->ti_len; \
  		flags = (ti)->ti_flags & TH_FIN; \
***************
*** 445,449 ****
  			sbappend(&so->so_rcv, m);
  			sorwakeup(so);
! 			tp->t_flags |= TF_DELACK;
  			return;
  		}
--- 448,455 ----
  			sbappend(&so->so_rcv, m);
  			sorwakeup(so);
! 			if (ti->ti_flags & TH_PUSH)
! 				tp->t_flags |= TF_ACKNOW;
! 			else
! 				tp->t_flags |= TF_DELACK;
  			return;
  		}
--
- Charles Hannum
  NetBSD group
  Working ports: i386, hp300, amiga, sparc, mac68k, pc532.
  In progress: pmax, sun3.

-----------[000484][next][prev][last][first]----------------------------------------------------
Date:      25 Apr 1994 19:14:56 GMT
From:      smithb@esdev02 (Beck Smith)
To:        comp.client-server,comp.protocols.tcp-ip,comp.protocols.nfs
Subject:   Re: Stateless Servers and transport protocols

Andrew Francis (andrew@gmvt4.concordia.ca) wrote:
: In article <1994Apr12.220351.9774@msus1.msus.edu>
: manohar@msus1.msus.edu writes:
 
: >	I'm working on issues dealing with Statelessness in Servers, and from
: >what I've read, it seems that the main motivation for statelessness in
: >Client-server systems is protocol reliability.
 
:    Actually the main reason one wants a stateless server is that it simplifies
: the construction of transaction processors that has ACID properties (atomicity,
: consistency, isolation, durablity). In the case of NFS, the designers felt
: that it is easy to implement crash recovery, if the server does not need to 
: have memory of prior transactions in order to implement the current transact-
: ion (in short no state). 
In addition to crash recovery, stateless also allows the server usage to 
scale up.  The server does not have to allocate resources(memory, processes)
to maintain state for each client.  Therefore increasing the number of clients 
serviced would not always require an increase in the hardware or processing 
power requirements of the server (up to a certain point). 

-----------[000485][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 20:32:00 GMT
From:      wcs@anchor.ho.att.com (Bill Stewart +1-510-484-6204)
To:        comp.unix.sys5.r4,comp.sys.att,comp.protocols.tcp-ip
Subject:   rdate replacement for SVR4 time-adjustment?

SVR4 has system calls to adjust the date/time gradually.
Does anybody have a version of rdate (the TCP/IP remote date-setter)
which knows how to use this?  Alternatively, is there any free rdate
code around - I'm currently using a Wollongong-based TCP/IP which
doesn't have rdate, though it does answer rdate requests properly.

	Thanks;  Bill
--
# Bill Stewart       AT&T Global Information Solutions (new name for NCR!)
# 6870 Koll Center Pkwy, Pleasanton CA 94566  1-510-484-6204 fax-6399
# Email: bill.stewart@pleasantonca.ncr.com billstewart@attmail.com
# ViaCrypt PGP Key IDs 384/C2AFCD 1024/9D6465

Disclaimer: My cats are walking on my keyboard again.

-----------[000486][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 20:56:23 GMT
From:      jcmurphy@acsu.buffalo.edu (Jeff Murphy)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

Mike Andrews (mikea@MCS.COM) wrote:
>Help!  I'm  in multi-vendor, variable subnetting, finger pointing, IP hell!


	we are trying to do a similar thing.

	we have: a class B network using a universal mask of 255.255.255.0
	we have: several institutions (hospitals, etc) that ''hang'' off of
	us. some of these are class C networks. 
	the problem: if we connect a class C to us we either have to 
		A) waste one of our subnets connecting them
		B) use two of there scarce addresses.
	we would rather use our address space, so we decided to try to 
	subnet one of our networks with 255.255.255.252 so that we get 
	(i forget off hand) some number of 2-host networks. 
	we dont use OSPF, we use RIP.

	so i tried this...

	setup a test wellfleet router with class C subnetting, and 
	configured an interface on our BCN to 128.205.11.254 / 255.255.255.252

	this didnt work, the packets were routed (according to the 
	BCN's routing table) to 128.205.7.253. however, i was able to 
	telnet in from a cisco route (odd..) but nothing else.. (hosts or
	otherwise)...

	the test router could see the BCN, but not the other way...

	so i installed a static route, 128.205.11.0 -> 128.205.11.254
	which also didnt work. (this also presents a problem because i 
	dont want to use static routes for all of these mini-two-host-nets

	does anyone have suggestions as to how to set up such a network
	and have it route properly..

	heheh.. what a mess! thanks for any help tho!


	jeff m.

-- 
jcmurphy@acsu.buffalo.edu                            ece/cs             cit oss
opnsmurf@ubvms.bitnet				     standard disclaimers apply
''She walks in beauty, like the night - Of cloudless climes and starry skies.''

-----------[000487][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 09:34:55 -0700
From:      lstowell@pyrnova.mis.pyramid.com (Lon Stowell)
To:        comp.protocols.tcp-ip
Subject:   Re: to SNAP or not to SNAP on token ring (that is the question)

In article <Apr25.150748.45476@melupl> conbvs@melupl (Bill VerSteeg (Contract)) writes:
>
>I have been working on a token ring driver for an embedded system
>tcp/ip stack, and have noticed a couple of different encapsulations
>for IP over token ring. Is there a document describing the various
>encapsulations? While I am at it, is there a good reference on token ring in 
>general? I need to get my client company up to speed on token ring, and
>if I can hand them a pointer to a well-written token ring primer, it
>will save everyone invloved some time.

   RFC 1122 notes that you MUST NOT use the IEEE snap header but MUST
   rather use the SNAP header. 

   This is either AA or 55 depending on the bit order of your system
   and how it feeds the T/R chip. 

   There are a couple other things you should do to maintain
   compatibility.  You will want to support the NULL SAP '00' and
   TEST and XID llc frames on that SAP as well as the SNAP SAP
   or any other protocol-specific sap you'll be supporting.  The NULL
   sap is used by GOOD implementations for source route discovery--it
   is also a required SAP in 802.2 even though some fairly large
   vendors have obviously not read 802.2 or don't understand the 
   meaning of the phrase "Conformance Issue".  >:-)  

   I haven't really seen any GOOD references on Source Routing other
   than some of the flows in the IBM manual.  Most T/R's are source
   routed and it is best to consider that you may run into ones
   that ignore RIF, some bridges (usually from traditional enet
   vendors) don't properly forward Single Route broadcasts, and such.

   I'd get the 802.5 and 802.2 IEEE specs, you may also want to
   get IBM's Token Ring Architecture Reference--it is slanted a bit
   towards SNA but contains a lot of misc. info as well.   TI's
   TMS380C16 manual is a help (if you are using that chip).

   For cabling, etc. the best manuals I've found are IBM's.  They are
   a bit conservative on wire lengths. 
>
>Thanks in advance
>
>Please reply to bvs@ver.com
>
>Bill VerSteeg
>bvs@ver.com



-----------[000488][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 22:11:37 GMT
From:      bsmith@rahul.net (Bob Smith)
To:        comp.protocols.tcp-ip
Subject:   Multiple IP stacks on one host: the problem

The problem is to build a network load box which has multiple IP
addresses.  The load box should look like a router with 4 hosts behind it.
Sometimes the load box has a SLIP interface and sometimes it lets UDP
or TCP do its packet assembly.
Arranging the stack vertically, one configuration might be:

   MyApp1          MyApp2           MyApp3            MyApp4   \
  sockmod         sockmod          sockmod              |      |
   tcp1            tcp2             udp3                |      | pseudo-
    ip1             ip2              ip3                |      |  hosts
   slip1           slip2            slip3             slip4    |
   pts1            pts2             pts3              pts4     /
     |               |                |                 |
     |               |                |                 |
   ptm1            ptm2             ptm3              ptm4     \
   slip5           slip6            slip7             slip8    |
     |               |                |                 |      | router-
     ----------------------------------------------------      |  host
                             |                                 |
                            ip                                 |
                            le0                                /
                             |
                         (Ethernet)


-- 
Bob Smith <bsmith@rahul.net>

-----------[000489][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 22:13:36 GMT
From:      bsmith@rahul.net (Bob Smith)
To:        comp.protocols.tcp-ip
Subject:   Multiple IP stacks on one host: a solution


Please comment on the following solution.  What is good about it?
What is bad about it?  What pitfalls should I expect?

1.  Build and link into the kernel clones of ip, icmp, udp, and tcp.
	- Create directories in pack.d for the new drivers.  Create
	  entires as needed in sdevice.d and node.d.  My new drivers
	  were Bp, Bcmp, Bdp, and Bcp.
	- Write a shell/awk script which uses 'nm' to get the global
	  names from the Driver.o files in /etc/conf/pack.d/Bp, ..
	  As a start, get the names that are global and have nonzero
	  size. You may need to add names manually to this list.
	- Use the binary editor 'fm' to change the global names.
	  There are about 250 names to change so I wrote a simple
	  C program to do the binary edit for me.  Inputs to the C
	  program were the file name of the driver and the list of
	  global names to change found with the shell/awk script.
	- Edit the variable names in the space.c files to avoid
	  multiple definitions when the kernel is linked.
	- Relink the kernel.  The kernel size increases by about
	  130KB for each pseudo stack.  Do not be discouraged if
	  the link is not successful on the first try!

2.  Use STREAMS to assemble the pseudo stack.
	- open both sides of a pseudo-tty
		o open() /dev/ptmx, the master
		o grantpt() to the slave side
		o unlockpt() to clear the slave's lock flag
		o ptsname() get the slave name
		o open() the slave side
	- I_PUSH "slip" onto the master side
	- open() /dev/ip
	- I_LINK /dev/ip to the master tty
	- SIOCSIFNAME to set the interface name
	- ifconfig to set the local and 'pseudo' host IP addresses
	- I_PUSH "slip" onto the slave side
	- open() /dev/Bp
	- I_LINK /dev/Bp to the slave tty
	- SIOCSIFNAME to set the interface name
	- Bfconfig to set the local and 'pseudo' host IP addresses
	  (or use SIOCSIFADDR and SIOCSIFDSTADDR ioctls)
	- open() /dev/Bcmp
	- I_LINK /dev/Bcmp to /dev/Bp
	  (at this point you should be able to ping the pseudo host)
	- open() /dev/Bdp
	- I_LINK /dev/Bdp to /dev/Bp
	- open() /dev/Bcp
	- I_LINK /dev/Bcp to /dev/Bp

3.  Run an application on the pseudo host
	- open() /dev/Bcp
	- I_PUSH "sockmod"
	  (the above two steps are equivalent to a socket() call)
	- bind()
	- listen()
	- (your code goes here)


This is just an outline, not step-by-step directions.  I have tested
down to the point where I can ping the pseudo host.  Once I knew *what*
to do, the above steps took only two or three days to implement.

USEFUL REFERENCES:
Unix Network Programming by Stevens
Unix System V Network Programming by Rago
STREAMS Modules and Drivers published by Unix Press

Thanks
-- 
Bob Smith <bsmith@rahul.net>

-----------[000490][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 94 23:12:04 GMT
From:      tsai@teetot.acusd.edu (Allen Tsai)
To:        comp.protocols.tcp-ip
Subject:   gated config for slip lines

	I would like to hear from those who have configured gated for 
slip lines on SunOS4.1.3/IPC and thank you for your early response in advance.
	Allen Tsai
	University of San Diego
	tsai@teetot.acusd.edu
-- 
Allen Tsai 
University of San Diego
tsai@teetot.acusd.edu

-----------[000491][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 00:22:25 GMT
From:      we41901@vub.ac.be (SCHOOVAERTS KURT)
To:        comp.protocols.tcp-ip
Subject:   Mixing Ethernet & Arcnet

Hello,

I'm attemting to set up a network consisting of 4 IBM-compatible PC's
running OS/2 and 4 Amiga's. All computers are using TCP/IP as a
protocol.
The problem : the PC's are on ethernet, the Amiga's on arcnet. One of
the Amiga's also has an ethernetcard.
We don't the smallest thing about setting up these two networks, so that
every computer can communicate with another.
Has anybody have an idea about setting up a gateway; what about routing;
subnets, masks.

The project we are working on, is of an urgent kind. We need all the
help we can get!

Thanks in advance...

--
||
||  Kurt Schoovaerts
||  we41901@is1.vub.ac.be
||  kurts@p6.f500.n292.z2.k12.be
||  kurts@p6.f500.n292.z2.fidonet.org
|+---------------------------------------
+--------------------------------------------

-----------[000492][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 12:12:54 -0700
From:      moliver@pyramid.com (Mike Oliver)
To:        comp.protocols.tcp-ip
Subject:   Re: to SNAP or not to SNAP on token ring (that is the question)

In article <2pjfrf$igs@pyrnova.mis.pyramid.com>,
Lon Stowell <lstowell@pyrnova.mis.pyramid.com> wrote:
>In article <Apr25.150748.45476@melupl> conbvs@melupl (Bill VerSteeg (Contract)) writes:
>>I have been working on a token ring driver for an embedded system
>>tcp/ip stack, and have noticed a couple of different encapsulations
>>for IP over token ring. Is there a document describing the various
>>encapsulations?
>
>   RFC 1122 notes that you MUST NOT use the IEEE snap header but MUST
>   rather use the SNAP header. 

That wording is less than crystal clear.  It should say "MUST use SNAP
and MUST NOT use the IP LSAP reserved by the IEEE".  The exact wording
in RFC1122 is:

	Note that the standard IP encapsulation in RFC-1042 does not
	use the protocol id value (K1=6) that IEEE reserved for IP;
	instead, it uses a value (K1=170) that implies an extension
	(the "SNAP") which can be used to hold the Ether-Type field.
	An Internet system MUST NOT send 802 packets using K1=6.

("K1" is the value of the 802.2 Destination and Source LSAP fields.)

However, if you want the details then skip RFC1122 and go straight to
RFC1042, "A Standard for the Transmission of IP Datagrams over IEEE 802
Networks", which discusses Token Ring encapsulation and other issues
like Source Route usage.  RFC1042 pokes its nose a little too far into
the implementation for my taste but it does have the information Bill
was looking for.

>>         While I am at it, is there a good reference on token ring in 
>>general? I need to get my client company up to speed on token ring, and
>>if I can hand them a pointer to a well-written token ring primer, it
>>will save everyone invloved some time.
>
>   I haven't really seen any GOOD references on Source Routing other
>   than some of the flows in the IBM manual.  Most T/R's are source
>   routed and it is best to consider that you may run into ones
>   that ignore RIF, some bridges (usually from traditional enet
>   vendors) don't properly forward Single Route broadcasts, and such.
>
>   I'd get the 802.5 and 802.2 IEEE specs, you may also want to
>   get IBM's Token Ring Architecture Reference--it is slanted a bit
>   towards SNA but contains a lot of misc. info as well.   TI's
>   TMS380C16 manual is a help (if you are using that chip).

IBM's T-R Arch Ref is SC30-3374, for the manual-number spotters out
there, but it's not exactly light reading and it's not cheap either.
The "IBM T-R Introduction and Planning Guide", GA27-3677, is a much
gentler introduction to the subject.

The "TMS380 Second-Generation Token Ring User's Guide", 1605836-9702,
has some decent T-R background and is worth having around even if
you're not using that chipset.  TI's Semiconductor Products Info number
is (214) 644 5580.

>   For cabling, etc. the best manuals I've found are IBM's.  They are
>   a bit conservative on wire lengths. 

Cabling is covered in reasonable detail in the Intro and Planning
Guide, but if you're heavily into that sort of thing you should get
hold of the "IBM Cabling System Technical Interface Specification",
GA27-3773.  There are additional manuals specifically for unshielded
cable and fiber plant, but those references are listed in the Intro and
Planning manual so I'm not going to type the names and numbers in here.

Cheers, Mike.

moliver@pyramid.com
{allegra,decwrl,hplabs,munnari,sun,utai,uunet}!pyramid!moliver

-----------[000493][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 94 00:59:46 GMT
From:      ddl@harvard.edu (Dan Lanciani)
To:        comp.protocols.tcp-ip
Subject:   Re: When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?

In article <cjross.767292564@raitt.bbn.com>, cjross@bbn.com (jonathan ross) writes:

| I have now.  It worked fine.  Thoroughly mystified, I downloaded
| ttcp.c from ftp.sgi.com and compared it to the queue server's code.
| The key difference is that ttcp.c sets SO_SNDBUF on the server socket
| before calling accept(), while I call it for the fd returned from
| accept().  Changing either program to use the other's call sequence
| causes it to manifest the other's behavior.
| 
| Now, why does the order make such a difference under IRIX?  (I agree,
| it's pointless to setsockopt() each fd returned by accept() when the
| fd automatically inherits the properties of the original socket, but I
| didn't write the code.  I will certainly fix it, though.)

I suspect there are versions of sockets where the accept()ed socket
does not inherit that particular property (buffer size) from the
original.  As I recall, tcp_input calls sonewconn which sets the
buffer sizes from the parent and then calls the PRU_ATTACH routine
of the protocol.  At one time, tcp_attach set the buffer sizes to
the defaults without checking whether they were already set.  This
was fixed (although it isn't absolutely 100% clear that it was a bug :)
a while ago, but perhaps IRIX has the old code and/or deliberately
preserves the old behavior.

				Dan Lanciani
				ddl@harvard.*

-----------[000494][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 01:06:47 GMT
From:      kchao@boi.hp.com (Kent_Chao)
To:        comp.protocols.tcp-ip
Subject:   Re: FTP PCTCP ---> HP1200C Jet Direct?

Lee Thomas (temp@sneezy.cc.utexas.edu) wrote:
: Hi,
 
: I just bought an HP 1200C color printer with an ethernet card (Jet Direct).
: I need to print to it from an IBM RS/6000 using TCP/IP and PC's with FTP's 
: PCTCP software.  Does  anyone have any ideas or specific instructions on how
: to get either system to print to the HP 1200C using TCP/IP?  Please E-mail
: me if you have a solution or ideas.
 
: Thanks!
 
: -Lee

   I assume that you are running AIX on RS/6000.  AIX just adds support of
   HP JetDirect printing.  Talk to them.



-----------[000495][next][prev][last][first]----------------------------------------------------
Date:      Mon, 25 Apr 1994 11:58:50 +0800
From:      peter.lewis@info.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   Re: NTP algorithm specification?

In article <2pegulErcp@uni-erlangen.de>,
kardel@immd4.informatik.uni-erlangen.de (Frank Kardel) wrote:

>jwmanly@unix.amherst.edu (John W. Manly) writes:
>
>>Hi there.  I was wondering if anyone can point me to an article or an RFC
>>that discusses the algorithm that underlies the NTP protocol.
>RFC1305

There is also RFC1361 Simple NTP.  It describes how to implement an NTP
server that is only accurate to a second or so, although when I
implemented on as described, it didn't work with NetworkTime, so I'm a bit
confused as to how interoperable NTP and SNTP are (and I'm somewhat loath
to start wading thru the NTP RFC).

Enjoy,
   Peter.
_______________________________________________________________________
Peter N Lewis <peter.lewis@info.curtin.edu.au>       Ph: +61 9 368 2055

-----------[000496][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 02:19:18 GMT
From:      fzshenau@dale.ucdavis.edu (Greg Shenaut)
To:        comp.protocols.tcp-ip
Subject:   Portable Computers

We have several portable computers with ethernet cards in them and
would like to use them in two modes: (1) dialup PPP and (2) direct
connection at one of our three ethernet sites.  The question
is, how can these portable machines maintain constant IP addresses
with all this moving around?  Is there any way to do this?

The network is mixed DOS (PC/TCP) and BSD/386.

--
Greg Shenaut -- gkshenaut@ucdavis.edu

-----------[000497][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 02:29:55 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2phv9t$stl@news.sprintlink.net> paul@hawksbill.sprintmrn.com
writes: 

    There are only two ways to get a variably subnetted IP network to
    route:
    
          OSPF or static-routes
    
or Integrated IS-IS, or EIGRP.  You really need an IGP which can deal with
variable length subnet masks.

Tony

-----------[000498][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 10:54:53 -0500
From:      lgrande@hr.house.gov (Len Grande)
To:        comp.protocols.tcp-ip
Subject:   Telnet Session Timeout Problem

Hi all.

Hopefully someone can help shed some light on a problem we've encountered.

We have half a dozen users that remotely connect to an IBM RS6000 via a
frame relay circuit.  These remote users have DEC PCs (386s and 486s) with
3COM ethernet cards and run the Tun Telnet client. Users on the DEC PCs
Telnet over the frame relay circuit (which has a 32 Kbps CIR) to the RS6000
to run various applications.  The average round-trip delay for a 64 byte
ping packet between the PCs and the RS6000 is 145 ms, with high values of
300 ms sometimes seen.

The problem being experienced is that the Telnet sessions on these PCs
periodically hang, requiring a new session to be established to the RS6000.
I  have been unable to find any problems on the frame relay circuit that
might account for this behavior such high bit error rates, unusually high
transit delays, or frames being dropped by the network.  Unfortunately, I
have no control over the DEC PCs or the RS6000 as they are provided by a
third party vendor.  I have asked the vendor to look at tuning parameters
in the Telnet client such as window and packet sizes, which they are now
doing. 

Does anybody have any recommendations regarding tuning parameters in the
Telnet daemon running on the RS6000 that might be related to this problem?

Thanks in advance for your help.


Len       

-- 
Len Grande
House Information Systems
lgrande@hr.house.gov
"The views expressed are my own and are not those of my employer."

-----------[000499][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 02:46:21 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

Jeff Murphy (jcmurphy@acsu.buffalo.edu) wrote:

> 	so i installed a static route, 128.205.11.0 -> 128.205.11.254
> 	which also didnt work. (this also presents a problem because i 
> 	dont want to use static routes for all of these mini-two-host-nets
 
> 	does anyone have suggestions as to how to set up such a network
> 	and have it route properly..


I don't mean to sound facetious, but "it ain't gonna work."

There are only two ways to get a variably subnetted IP network to
route:

      OSPF or static-routes

Neither of which, in my opinion, are desirable.

Hindsight is 20/20; prior planning and IP engineering evaluation prior
to implemntation are absolutely critical. 

Best of luck.

_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000500][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 09:49:22 -0400
From:      reh@cs.umd.edu (Richard Huddleston)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP References (was: FAQ)

gnn@netcom.com (George Neville-Neil) writes:

First, I want to complement you on undertaking this task.  Second, I
want to suggest an addition to your reference list (along with discussing
some the reasons why I like it).

>Q) Are there any good books on IP?

The hands-down best text I have seen on TCP/IP is 

	TCP/IP Illustrated, Volume 1 (1993 Stevens Addison-Wesley)

First, Richard Stevens writes very clearly.  His coverage is deep, yet 
well-phrased enough to convey understanding without inducing sleep ;).

Every time I've got to pick up a Comer text, I've looked it over in the
bookstore and invariably concluded that I didn't need it.  Stevens' text,
on the other hand, struck me as "I've *got* to have this...".

Stevens is also fairly up to date.  It doesn't go into great depth on such
topics as CIDR and DHCP, which I suspect we're all going to be seeing a
lot more of, but it's good to see the topics in a text.  My recollection,
subject to error, is that Comer pretty much stays in the mainstream (as 
you would expect with an edition that's a couple of years old as I recall).

In my working life, I currently spend a good deal of time designing and 
setting up Internet firewalls.  It's a grunt job, no doubt, but it's hardly
cookie-cutter; every client has unique needs.  The Stevens text and the
Leffler, et al, text on BSD 4.3 are the only references I need to have with
me.  I am looking forward to supplementing my keep-with-me references with
the Bellovin/Cheswick text once it shows up in my local technical bookstore.

<end of rant>

-- 
Richard Huddleston 

In case it isn't obvious: personal opinions/correspondence 		

-----------[000501][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 03:28:46 GMT
From:      wier@merlin.etsu.edu (Bob Wier)
To:        comp.protocols.tcp-ip
Subject:   E-Mail <-> ftp Gateways for MVS?

I'm not sure if this is the right place to ask this... if
there is a better newsgroup, please suggest that via e-mail.

I've been asked to investigate a problem which was come
up recently. Suppose you have an ftp site on an IBM MVS
machine? This is difficult because they don't seem to 
respond to "normal" unix type commands such as CD and LS - - 
you have to know the directory you want in advance.  Suppose
further that you have some people who would like to retrieve
files via sending e-mail requests (ie, don't have ftp access).
Does anyone know something along the line of BITFTP that will
work with an MVS machine as a target, either thru a gateway or
which will actually run on the MVS machine?

I'm not real familiar with these kinds of problems since 
I just go ahead and ftp when I need stuff. However, my friend
has some people who are rather remotely located and can't
ftp direct, hence the need for some kind of e-mail <->
ftp server which can access the non-hierarchical structure
of MVS.

THANKS


    ======== insert usual disclaimers here ============
      Bob Wier, East Texas State U., Commerce, Texas
keeper of the Adobe Photoshop, MC68HC11, ICOM mailing lists
      wier@merlin.etsu.edu (watch for address change) 

-----------[000502][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 03:38:38 GMT
From:      kline@tampico.cso.uiuc.edu (Charley Kline)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

paul@sprintlink.net (Paul Ferguson) writes:

>      OSPF or static-routes
       ^^^^

>Neither of which, in my opinion, are desirable.


Oh for heaven's sake. OSPF works fine.

/cvk

-----------[000503][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 04:42:45 GMT
From:      starshum@netcom.com (Jolly Roger)
To:        comp.protocols.tcp-ip
Subject:   TCP source code anywhere?

Hello:

	Can anyone please inform me of any ftp site that I can find some
source codes writing for TCP applications?


Steve

-- 
                                             starshum@netcom.com

-----------[000504][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 04:56:25 GMT
From:      peril@extro.ucc.su.OZ.AU (Peter Lisle)
To:        comp.protocols.tcp-ip,comp.client-server
Subject:   IP ports

I am preparing guidelines for the use of Netwise RPCs in our
organisation. Netwise RPCs do not use a port-mapper to
assign IP ports dynamically; instead, services need to
be added to the services file for each RPC interface.

I want to make sure that the port numbers that I choose
do not conflict with anything else out there. I have
checked the Assigned Numbers RFC - so I am aware of
the 0-1023 'Well Known Ports' + the miscelaneous
'registered' ports in the 1024+ range. However, I
would like information of which port numbers the
SUN ONC/XDR port mapper and the OSF DCE port mapper
use (they obviously have a 'pool' out of which the port
numbers are dynamically allocated). I would like to
avoid these 'pools' so that conflicts are avoided! (eg: if
I do not avoid the SUN ONC/XDR mapper pool then
serious problems may be experienced when trying to use
NFS and Netwise RPCs concurrently).

I would appreciate it if someone who knows the answer to
this can mail me the response.

My email address is: auampdrv@ibmmail.com

Thank-you.

-----------[000505][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 11:59:58 -0400
From:      tenser@zenos.physci.psu.edu (Dan Cross)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP source code anywhere?

In article <starshumCounrA.I12@netcom.com>,
Jolly Roger <starshum@netcom.com> wrote:
>	Can anyone please inform me of any ftp site that I can find some
>source codes writing for TCP applications?

Not sure what platform you're working on, and that's where it really gets
tricky, but if you are working from a UNIX base, you can get tcp stuff from
any of the BSD sources out on the net...  (Mostly Net/2, don't think that
4.4lite has really gotten out yet?)  Other than that, UNIX Network Programming,
but Richard Stevens is a great book for learning how to program using UNIX
network calls...  (Covers sockets, TLI, other cool stuff...  :-)  Hope this
helps some!

	- Dan C.


-----------[000506][next][prev][last][first]----------------------------------------------------
Date:      26 APR 94 15:25:33 EST
From:      ciarfella@took.enet.dec.com (Paul Ciarfella)
To:        comp.protocols.tcp-ip
Subject:   Re: DOS TCP/IP Stacks?


In article <43458@mindlink.bc.ca>, KCARPENT@mindlink.bc.ca (Ken Carpenter) writes...
>I am looking for information on companies who offer TCP/IP stacks for DOS.
> 
>Could anyone who knows of such products please email me with the company's
>name, and if possible, the phone number or address.
> 

Digital Equipment Corporation's Pathworks product for DOS and Windows provides
a complete TCP/IP solution, including TCP/IP stack and utilities, such as
telnet, ftp, etc.

In addition, Digital's RoamAbout product line for mobile and wireless users
provides mobility and portability to PC-based TCP/IP users.  

With RoamAbout Mobile IP, a portable PC user keeps its permanent IP address 
wherever it goes.  Thus, a user can transparently move from subnet to subnet, 
using the services provided in the subnet that it is visiting while also 
being able to access the services provided in the user's home subnet.

In addition to RoamAbout Mobile IP, the RoamAbout product family includes the 
following:
    	RoamAbout Transport - provides client/server connectivity for
      	  PC-based TCP/IP users  over cellular, dialin or packet radio
    	RoamAbout Access Point - a wired to wireless LAN bridge that
	  operates stand-alone or in a DEChub 900 
    	RoamAbout PCMCIA Network Adapter and WaveLAN - provide wireless
    	  connectivity for PCs

For product information, call 1-800-DIGITAL.  

    ----------------------------------------------------------------------  
    Paul Ciarfella                             ciarfella@took.enet.dec.com
    Mobile & Wireless Networks                 508 486-7712
    Digital Equipment Corporation              Littleton, MA   USA
    ----------------------------------------------------------------------  


-----------[000507][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 18:08:18 -0400
From:      aem@symbiosis.ahp.com (a.e.mossberg)
To:        comp.protocols.tcp-ip
Subject:   Re: External Clock for Network Time Protocol

perry@brooks.af.mil (David Perry) writes:

>We're looking for an external clock (hardware) to interface to a server
>to provide time synchronization using something like network time protocol.
 
>Does anyone know of a good source for this type of product?  We want to
>interface to a Sun server.

If you're looking for types of suitable clocks try
ftp://louie.udel.edu/pub/ntp/doc/clock.txt for a list of suitable
clocks.  


aem

-- 
Andrew Mossberg      Network Administrator   Symbiosis Corporation,  Miami FL
(305) 597-4110 fax 597-4002, MD5OfPublicKey: 15784D117CC103912AEC427A3A99BA83

-----------[000508][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 18:30:03 -0400
From:      alatas@access.digex.net (Robert W. Mattox)
To:        comp.protocols.tcp-ip
Subject:   Subnet Masking Explanations

I hope that this is the correct forum for my query. I'm relatively new at the 
TCP/IP game. I have a good idea of network addressing, but I'm having a 
problem with the idea of subnetting. Could anyone point me to the appropriate 
RFC or other literature that will give me a comprehensive but well written 
explanation of the vagaries of the concept? I don't want something that is 
too technical since I'm a newbie, but I do want something that will explain, 
for example, using a host address of 99.3.80.1 with a subnet mask of 
255.255.255.0 on the same network as a host with an address of 99.3.80.1 with 
a subnet mask of 255.255.254.0. Do they conflict? or are they considered to 
be different network addresses?

Any help would be greatly appreciated. Please send replies to 
alatas@access.digex.net.

Thanks,



Robert Mattox
U.S. Department of State
Office of Foreign Missions
E-mail: alatas@access.digex.net
Phone: 1-202-895-3579
Fax:   1-202-895-3669


-----------[000509][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 12:27:54 GMT
From:      brianw@zeus.datasrv.co.il (Brian Zvi Weiss)
To:        comp.protocols.tcp-ip
Subject:   Leased-Lines Overseas

I realize this is not the perfect place for this posting but I can't seem
to find a topic to match...

Where can I get info about purchasing leased lines.  We have a US-based
subsidiary that would be handling this with the local providers.  At a
minimum, I need some initial cost guidelines.

Please respond by direct mail to brianw@datasrv.co.il

Thanks

--
======================================================================
Brian Zvi Weiss      | BI-DI Systems, Ltd.  | Editor, Hebrew Windows
brianw@datasrv.co.il | Yeda Am 14           | Magainze; Developer of
v: +972-3-613-1027   | Ramat Gan, ISRAEL    | Hebrew Windows support
f: +972-3-613-1029   | 52526                | in Lotus 1-2-3, Da Vinci
                                            | eMAIL, MSI CaLANdar,
"Opening Windows in Israel"                 | OPTUS FACSys, and others.
=======================================================================

-----------[000510][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 12:45:01 GMT
From:      w-rolph@ds.mc.ti.com (Don Rolph)
To:        comp.protocols.tcp-ip
Subject:   Re: IMAP server & client

In article <1994Apr25.015157.1@cnuce.cnr.it> alberto@cnuce.cnr.it (Alberto Mura) writes:
>From: alberto@cnuce.cnr.it (Alberto Mura)
>Subject: IMAP server & client
>Date: 25 Apr 94 01:51:57 +0100
>Summary: Looking for IMAP sofware
>Keywords: IMAP
 
>I am looking for IMAP (Interactive Mail Access Protocol) software, both
>server and client. Can anybody tell me the names of good programs?
>For IBM PC client preference is for Windows 3.1 software.
 
>Thanks in advance,
 
>Alberto Mura
>Dept. of Philosophy
>Univ. of Pisa
>Italy

ECS Mail is an imap2 based mail client with mime support.  I find it to be 
very nice.

Regards.
 
Don Rolph w-rolph@ds.mc.ti.com WD3 MS10-13 (508)-699-1263

-----------[000511][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 00:32:24 -0700
From:      beepy@nova.netapp.com (Brian Pawlowski)
To:        comp.client-server,comp.protocols.tcp-ip,comp.protocols.nfs
Subject:   Re: Stateless Servers and transport protocols

> In addition to crash recovery, stateless also allows the server usage
> to scale up.  The server does not have to allocate resources(memory,
> processes) to maintain state for each client.  Therefore increasing 
> the number of clients serviced would not always require an increase
> in the hardware or processing power requirements of the server (up
> to a certain point).

I don't believe I'm about to say this, but...

There is a lot of serious disagreement with the above statement
regarding scalability and stateless vs. stateful in the literature.

For a good time, someone interested in this subject might want
to read:

    [Howard88] Howard, J.H., M.L. Kazar, S.G. Menees, D.A.
    Nichols, M. Satyanarayanan, R.N. Sidebotham, and M.J.
    West, "Scale and Performance in a Distributed File
    System," ACM Transactions on Computer Systems 6(1).
    February, 1988. Primary reference on the Andrew File
    System-contrasts the performance and scalability of AFS
    and NFS-cites the lack of consistency guarantees as the
    cause of poor scalability of NFS file servers.
    
    [Satyanarayanan89] Satyanarayanan, M., "A Survey of
    Distributed File Systems," Annual Review of Computer
    Science, Annual Reviews, Inc. Volume 4, 1989. Also
    available as Technical Report CMU-CS-89-116, Dept. of
    Comp. Sci., Carnegie Mellon University. A survey of NFS,
    AFS and other distributed file systems with a comprehensive
    bibliography.
    
    [Nelson88a] Nelson, Michael N., Brent B. Welch and John
    K. Ousterhout, "Caching in the Sprite Network File
    System," ACM Transactions on Computer Systems 6(1).
    February, 1988. Also Computing Reviews, Vol. 30, No. 3,
    March 1989. Caching strategies, consistency protocol and
    performance results.

    [Mogul92] Mogul, Jeffrey C., "A Recovery Protocol for
    Spritely NFS," USENIX File System Workshop
    Proceedings, Ann Arbor, MI, USENIX Association,
    Berkeley, CA, May 1992. Second paper on Spritely NFS
    proposes a scheme for recovering state in a consistency
    protocol.
    
    [Mogul93] Mogul, Jeffrey C., "Recovery in Spritely NFS,"
    Research Report 93/2, Digital Equipment Corporation
    Western Research Laboratory, June 1993. Third paper on
    Spritely NFS describes the implementation of recovery.
    
    [Ousterhout90] Ousterhout, John K., "Why aren't Operating
    Systems Getting Faster as Fast as Hardware," Proceedings
    of the 1990 Summer USENIX Conference, Anaheim, June
    11-15, 1990. A description, in part, of the synchronous write
    bottleneck in NFS Version 2.

Some references will discuss directly the "scalability" aspect
of stateful vs. stateless. Follow the references in the papers
for more information. I'm sure other papers (more recent)
cover this ground further.

Brian Pawlowski


-----------[000512][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 12:56:48 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

Charley Kline (kline@tampico.cso.uiuc.edu) wrote:

> Oh for heaven's sake. OSPF works fine.

Whatever you say, pal.

_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000513][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 13:17:17 GMT
From:      landin@cherokee.nsuok.edu (Mark Landin)
To:        comp.protocols.tcp-ip
Subject:   Re: HP-UX, Am I there yet ?

lchutson@netcom.com (Lawrence Hutson) writes:

>Is this the "Best" place for disscussions on HP-UX ? I noticed the comments 
>on this group were for HP products. Sounded kinda proprietary ;-).

No, try comp.sys.hp.hpux. There are also other groups under comp.sys.hp
that may also be interesting to you.

-- 
*-----------------------------------------------------------------------------*
*  Mark C. Landin					Northeastern St. Univ *
*  landin@cherokee.nsuok.edu					Tahlequah, OK *
*   "Living in the pools, they soon forget about the sea" - Neil Peart, RUSH  * 

-----------[000514][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 13:52:28 GMT
From:      ncherry@cbnewsg.cb.att.com (neil j.cherry)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2pi2bu$71u@vixen.cso.uiuc.edu> kline@tampico.cso.uiuc.edu (Charley Kline) writes:
>paul@sprintlink.net (Paul Ferguson) writes:
>
>>      OSPF or static-routes
>       ^^^^
>
>>Neither of which, in my opinion, are desirable.
>
>
>Oh for heaven's sake. OSPF works fine.
>

Ahhhhhhhhhhhhhhhhhhh, but will OSPF play nice on a network between
Cisco and Wellfleet? (If you're using both routers with OSPF do let us
know though!)

NJC (Neil Cherry)


-----------[000515][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 14:24:08 GMT
From:      jeff@astph (Jeff Martin)
To:        comp.protocols.tcp-ip
Subject:   Hanging accept()!  Too many sockets?

The accept() call hangs after having opened 61 socket connections, thus
preventing our server from having 62 or more socket connections open.
The maximum open descriptors per process is set to 100 in the kernel
parameters and the server process in questions has fewer than 70 open.
What is the deal?  Is this an a TCP/IP implementation problem, a
kernel parameter configuration problem, or my mis-use and abuse of the
system calls?  It almost seems as if the implementation of the accept()
call was not prepared for the expanded number of open descriptors.
That is our system suggests NOFILES=60 as a default, but also says
that NOFILES=100 is the maximum.  Perhaps the accept() call will not
lets us handle the maximum?

We are running Interactive UNIX System V, Release 3.2.  The kernel
parameters of NFILE, total open descriptors per system, equals 800, and
NOFILES, total open descriptors per process, equals 100.  These values
are maxed.
-- 
Jeff Martin, dbms programmer,		Philadelphia Phillies
INET:	astph!jeff@attmail.com		Voice:	(814)234-8592x32
UUCP:	attmail.com!astph!jeff		FAX:	(814)234-1269
SLOW:	141 West Beaver, Suite A, State College, PA 16801

-----------[000516][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 94 22:58:31 -0500
From:      ajkim@miavx1.acs.muohio.edu (Michael Jordan-I WISH!)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   IP address assignment via MacTCP/ARA?

Hi Netters:
I have a problem...I have been trying to access the Internet via MacTCP and 
Apple Remote Access without success.  I have heard that there is a way to 
DYNAMICALLY assign the Mac an IP address to mirror that of the ARA server (use 
the same IP number)  

Is this possible?  If so, How do I do it from within the MacTCP Control panel?
I'd also like to get an IP address for my Mac (permanent...Do I just talk to my 
school's network administrator?)

How difficult is it to get a deidcated internet line?  I mean an IP address?  
Any help is most appreciated!  thanks in advance for any help!

-Albert Kim


-- 

ajkim@miavx1.acs.muohio.edu
akim@opus.mco.edu

"Get a T.O. Baby!"
      -D.V.


-----------[000517][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 94 14:46:51 GMT
From:      jonesjg@dg-rtp.dg.com (Greg Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: When is it (not) appropriate to use SO_SNDBUF and SO_RCVBUF?


In article <cjross.767292564@raitt.bbn.com> cjross@bbn.com (jonathan ross) writes:
> vjs@calcite.rhyolite.com (Vernon Schryver) writes:
> 
> > cjross@bbn.com (Jonathan Ross) writes:
> >>In an attempt to optimize large transfers from the queue
> >>server to its clients, the original author chose to set the server's
> >>SO_SNDBUF (and clients' SO_RCVBUF) to 16K bytes.  This seems to have had
> >>the opposite of the desired effect; the server occasionally gets stuck
> >>in write() calls, even when a client is simultaneously calling read().
> 

I have seen situations where setting a large recieve buffer size, can 
cause the transmitting machine to choke itself.  Changing the buffer 
size may affect the size of the TCP Window.  A large window may cause
the transmitter to send more packets than the ethernet driver can
handle - thus dropping the packets before hitting the wire.  This
causes TCP Retransmissions - slowing the overall data transfer.
In our TCP/IP we had to limit the size of the TCP Window to avoid 
choking some well known systems.  This is becoming more of a problem 
with CPUs that can build packets faster that ethernets can move them.


> I have now.  It worked fine.  Thoroughly mystified, I downloaded
> ttcp.c from ftp.sgi.com and compared it to the queue server's code.
> The key difference is that ttcp.c sets SO_SNDBUF on the server socket
> before calling accept(), while I call it for the fd returned from
> accept().  Changing either program to use the other's call sequence
> causes it to manifest the other's behavior.
> 
> Now, why does the order make such a difference under IRIX?  (I agree,
> it's pointless to setsockopt() each fd returned by accept() when the
> fd automatically inherits the properties of the original socket, but I
> didn't write the code.  I will certainly fix it, though.)

Setting an option on the parent socket should have the same affect as
setting it for each child, however, I have found that some TCP/IPs
do a poor job of copying all the state into the child.  I was looking
into problems in this area myself and found that BSD doesn't even
copy all the options over (TCP_NODELAY).

Has anyone else run into this problem?  My BSD code reading isn't the
best.  In this case the TCP_NODELAY is stored in the tcp_flags field
rather than the so_options field, when accept creates a new socket
it only copies the so_options?

> 
> =====  Jonathan Ross  =====  <cjross@bbn.com>  =====  (617) 873-3272  =====
 

-----------[000518][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 15:42:41 GMT
From:      mahmood@lambda.msfc.nasa.gov (Rafique Mahmood)
To:        comp.protocols.tcp-ip,comp.sys.sun.hardware,comp.sys.sun.admin
Subject:   Need multicast info. in Solaris-2 with FDDI interface.

Currently we are running Solaris 1 (sunOS 4.1.3). We use IP multicasting on 
ethernet and on FDDI. 

We are planning to upgrade the OS to Solaris-2. I know it has multicast in
the OS (in 4.1.3 I installed the multicast patch). But how to configure 
Solaris 2 with FDDI? I remember reading somewhere the problem with the driver
which does not work with this OS. I am not sure what is the problem I read.

I called 'sun' but haven't got the respond yet. I need to find out this really
soon. I will appreciate any response or suggesion.

Please email me the response.

Thanks,
Rafique

mahmood@lambda.msfc.nasa.gov


-----------[000519][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 16:03:52 GMT
From:      perry@brooks.af.mil (David Perry)
To:        comp.protocols.tcp-ip
Subject:   External Clock for Network Time Protocol

We're looking for an external clock (hardware) to interface to a server
to provide time synchronization using something like network time protocol.

Does anyone know of a good source for this type of product?  We want to
interface to a Sun server.

Thanks,


---

David W. Perry
Computer Sciences Corp.  
Brooks AFB, TX



-----------[000520][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 16:03:54 GMT
From:      rmadhok@anl.liv.usw.com (Raghu Madhok)
To:        comp.protocols.tcp-ip
Subject:   Re: Network Management Platform Comparisons

In article <CoGnCM.5K9@tigadmin.ml.com> root@devjam1.ml.com writes:

 >I am working on an evaluation of SNMP Network Management Tools.  I was
 >wondering
 >if anyone else out there has prepared an analysis of these tools. The
 >particular SNMP Network tools include HP/OV/Netmetrics/EASE, Concord Trakker, 
 >Cabletron Spectrum.  I have looked at others such as W&G, but for our needs
 >...
 >I'll share my feature/list/results with anyone else who wishes thanks. 
 
I would be interested in any information you would be willing to share.
Thanks.

...Raghu

netcomsv!anl!rmadhok
Kenetech Windpower

-----------[000521][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 17:59:50 GMT
From:      kline@tampico.cso.uiuc.edu (Charley Kline)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers


>Ahhhhhhhhhhhhhhhhhhh, but will OSPF play nice on a network between
>Cisco and Wellfleet? (If you're using both routers with OSPF do let us
>know though!)

I haven't done extensive testing, but they seem to, yes. Our network is
nearly all Proteon with a small smattering of Cisco (which seem to
interoperate reasonably) and I'm currently testing a Wellfleet BLN.


> Whatever you say, pal.

Well, the proof is in the pudding. I run a 70-router network with 300
subnets and 7 different subnet masks, and the only routing protocol I
use is OSPF, and I haven't had any problems whatsoever with it. I'm
sorry if people can't get it to work, but that's hardly the fault of
the protocol.

/cvk

-----------[000522][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 18:51:02 GMT
From:      mojumdar@fooba.ml.com (Mainak Mojumdar)
To:        comp.protocols.tcp-ip
Subject:   Help required for tcp-ip vs udp comparison

Can anybody comment on the advantages and disadvantages between the usage
of tcp/ip and udp for a client/server implementation on a Lan environment.

What concerns me the most is the connection time required in a tcp/ip 
approach, in a 300 user client/server approach, if for every transaction
a connection has to be opened and dropped subsequent to the completion of
the transaction.

Any experiences in using udp for a serious application.

Any comments are welcom.

Mainak Mojumdar
mojumdar@ml.com



-----------[000523][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 19:24:54 GMT
From:      sklower@oboe.CS.Berkeley.EDU (Keith Sklower)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <fontaine-250494110325@130.104.58.13>,
Alain Fontaine <fontaine@sri.ucl.ac.be> wrote:
}I do run OSPF.
}P.S. forget RIP if you use variable subnetting.

And, in article <767093566snz@peeras.demon.co.uk> Phil Royse writes:
}Only OSPF allows subnet masks to be passed along with the routing info
}and you need this to do variable length masks.  But you know this already.
 
}Phil Royse     Comms Consultant  |  PRA Consulting Ltd.

Evidently Phil is unaware of RFC-1388, or its contents, which is an
extension to RIP (called RIPv2) which passes subnetmasks.

However, the only open-available implementation I'm aware of is gated,
which also does OSPF.  It would be good if somebody from wellfleet or
cisco could authoritatively say whether their products support RIP version 2.

-----------[000524][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 20:07:23 GMT
From:      tli@cisco.com (Tony Li)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2pjpq6$3tk@agate.berkeley.edu> sklower@oboe.CS.Berkeley.EDU
(Keith Sklower) writes: 

    However, the only open-available implementation I'm aware of is gated,
    which also does OSPF.  It would be good if somebody from wellfleet or
    cisco could authoritatively say whether their products support RIP
    version 2. 

Keith,

Currently we do not support RIPv2.

Tony


-----------[000525][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 20:43:22 GMT
From:      scotta@kije.gsfc.nasa.gov (Scott Austin)
To:        comp.protocols.tcp-ip,misc.forsale.computers.other
Subject:   Bargains on Comer's 3 Volume TCP/IP set

Hi,

The Library of Computer and Information Sciences (LCIS), is currently
having a promo with Comer's 3 books.  You can save some bucks, since the
Publisher's prices on these is $164!  This offer is open to any
non-members (wish *I* could).  The scoop is this:

* Join LCIS for $4.95 + $7.95(S&H) and receive:
      
  Internetworking with TCP/IP, Volume I          Publisher's Price: $60 
     Principles, Protocols, and Architecture     2nd Edition
     By Douglas E. Comer

  Internetworking with TCP/IP, Volume II         Publisher's Price: $54
     Design, Implementation, and Internals       1st Edition
     By Douglas E. Comer and David L. Stevens

  Internetworking with TCP/IP, Volume III        Publisher's Price: $50
     Client/Server Programming and Applications  1st Edition
     By Douglas E. Comer and David L. Stevens

* In doing so, you "... agree to purchase at least one additional
  Selection or Alternate over the next 12 months.  As a member I can save
  up to 50% off the publisher's price"  (quote from the offer card).  As
  an example, these 3 books individually are offered to members at
  savings of 20-30% off the pub.  price.  But I have often seen good
  books at 1/2 price (periodic sales, bonus book plan, etc).  Shipping
  and handling costs are additional but reasonable.

* Books cover topics such as Mainframes, Unix, and PC's.  Email me
  for more info.

* "My membership is cancelable any time after I buy this additional book."
 
* "No-Risk Guarantee: If I am not satisfied -- for any reason -- I may
   return the INTERNETWORKING SET within 10 days.  My membership will be
   cancelled, and I will owe nothing."

* 15 times a year you'll receive the LCIS News (a catalog).  If you want
  the Main Selection, do nothing.  If you don't want it, send in the card.
  Sound familiar?  If you want details, email me.
  

I'll tell you this is a "classic" book series on TCP/IP; ask around.  Some
of the material gets deep, but for the price....   Again, the cost is $13
for now, plus the cost of your 1 obligatory book.   Roughly less than the
cost of one of Comer's books.

=============
Motive check:
=============

When I first decided to post this offer, it was strictly for your
information; take it or leave it.  When I called to make sure this offer
was good for anyone, the operator told me, with each referral I would get
referral credit (read that: free books).  I love books, but this *not* my
motive [put down that flame].  If you doubt me on that, you can directly
send in your request (no money; they'll bill you) to:

		The Library of Computer and Information Sciences
                3000 Cindel Drive
                Delran, NJ 08075-9889
  You should refer to the INTERNETWORKING SET (#00274), cost of $4.95 +S&H,
  and the obligation of 1 book over the next year.  Sign it and you're on
  your way.

The other choice is to give me your name and address, I'll tack on my
address and account info and send it off to them for you.  And for *all*
this effort [blood, sweat, and root beers], I'll get some free books.

Thanks for your time,
Scott No-Affiliation-With-LCIS Austin
scotta@cnt.com


-----------[000526][next][prev][last][first]----------------------------------------------------
Date:      26 Apr 1994 22:40:50 GMT
From:      David_Chatfield@ccm.sc.intel.com (Dave Chatfield)
To:        comp.protocols.tcp-ip
Subject:   Looking for UNIX/AIX real-time shared session over TCP/IP

We are actively seeking UNIX/AIX-based packages which can allow
a real-time, shared data session between two TCP/IP participants.
The ideal would be to allow dual-access to CAD-drawings, but
something as modest as a shared space for text or simple drawing
would be a good start. Anon-FTP availability would be very helpful.

Please reply to the list or to my account above. Thanks..

  -Dave



-----------[000527][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 1994 22:55:00 GMT
From:      kate@wti.com (Kathy Samec)
To:        comp.protocols.tcp-ip
Subject:   Novell/TCP

Hello, I have a question to the wizards out there. 
I have been asked by a friend to help them connect a unix system up
to a small novell network. I think they are using netware lite? I don't 
know exactly. What I need to know is, What software can I get on the PC that
will allow concurrent use of the network card(accton) or what PC TCP software
is the best these days. In the past, I used NCSA and WD cards. Does NCSA stuff
still exist? What drivers are needed? Does it work with the Accton card? I 
haven't dealt with PC networks in a few years and am out of touch a bit. They
also mentioned using novell dos 7 or something like that.

Thanks for any info.

-- 
*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*
Kathi Samec			- Everything is beautiful and
Santa Barbara Studios		  Nothing hurt.
kate@wti.com					-Stoney Stevenson

-----------[000528][next][prev][last][first]----------------------------------------------------
Date:      Tue, 26 Apr 94 23:09:34 GMT
From:      martiner@int-evry.fr (Olivier MARTINERIE)
To:        comp.protocols.tcp-ip
Subject:   RS6000 & TOKEN RING

Hi! 

Here's the problem of a friend:

"
* We have 2 rings linked by bridges.

* A machine connected on ring 2 try to link a risc 6000 on ring 1.

* This machine send a frame with correct TCP/IP addresses.

* But the RS6000 does not answer that frame.

* The same frame, sent by another machine on ring 2, is recognised by the 
rs6000.

(machine working on ring 1: IBM Controller
machine not working: Mac Data controller)

ANY IDEAS ???

"

Thanks for answering, here or in my email !

Olivier.

----------------------------------------------------------------------------
Olivier MARTINERIE                                      Amiga 4000-30
Phone: +33 1 60 76 64 69		          	Amiga 1000	
Fax: +33 1 46 32 33 68
IRC Nickname: KAZIM				National Institute of
E-Mail: martiner@galaxie.int-evry.fr		Telecoms (France)
----------------------------------------------------------------------------


-----------[000529][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 94 11:52:11 -0700
From:      "Quinn Erickson" <p01048@psilink.com>
To:        comp.protocols.tcp-ip,comp.sys.ibm.pc.rt
Subject:   IBM RT to PC ftp lock up's

Does anyone have any experience with ftp from an IBM RT to a PC.  We 
are having problems getting reliable communications between these two
machines.  The RT is running AIX ver 2.2.1 and the PC is running DOS 5.0
with LAN Workplace for DOS ver 4.01.  Occasionally a file transfer will
lock up the PC causing a reboot and the ftp process hangs on the RT.

If anyone has any experience and/or information concerning this 
problem, please let us know.  Thanks for your help!


-----------[000530][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 00:51:36 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

Charley Kline (kline@tampico.cso.uiuc.edu) wrote:

> Well, the proof is in the pudding. I run a 70-router network with 300
> subnets and 7 different subnet masks, and the only routing protocol I
> use is OSPF, and I haven't had any problems whatsoever with it. I'm
> sorry if people can't get it to work, but that's hardly the fault of
> the protocol.

Okay -- let me rephrase my earlier post.

Wellfleet's implementation of OSPF seems to have matured, with
various patch-releases of 7.56 and 7.57 software. We are preparing
to deploy an extension to a pilot network which will consist of
upwards to 50 routers (two BCNs & the remainder AFNs) which will
run solely 7.70. This is our only Wellfleet network employing
OSPF out of the 80 or so that we designed/deployed/support/manage.

(On cisco, we use IGRP or EIGRP.)

Once this pilot has proven successful, the entire buildout of this
single network will be well over 1500 routers, both domestically and
internationally.

Yes, OSPF seems stable IN SOME environments, as it has appeared to
have stabilized in our current pilot scenarion. There are SERIOUS
problems which can occur, as when more than one router tries to 
act as the designated router in a single area (more a circuit
failure issue, than a design issue, although we have "designed
around" this occurance in the future buildout).

cisco's implementation of OSPF doesn't appear to be quite as
mature. (OBopinion.)

Final thought: OSPF is not for everyone. In fact, I think that
it is a "tailored" IGP and should be used as such.

Cheers,

_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000531][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 01:13:31 GMT
From:      wisecomm@rahul.net (Wise Communications)
To:        comp.protocols.tcp-ip
Subject:   HELP! query unusTCP send buffer size using socket

How do I find out how many more bytes a TCP pipe can accept before
either it would block or it would not accept any more data?  I am
using socket interface.  I would like to know beforehand that all
n bytes would be accepted when I do a send of n bytes.

Thanks

Ralph Lao
lao@wise.com 
-- 
Wise Communications <wisecomm@rahul.net>

-----------[000532][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 02:25:36 GMT
From:      jgs@hadron.merit.edu (John G. Scudder)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2pkcuo$94r@news.sprintlink.net>,
Paul Ferguson <paul@hawksbill.sprintmrn.com> wrote:
[...]
>Final thought: OSPF is not for everyone. In fact, I think that
>it is a "tailored" IGP and should be used as such.

So what's a "tailored" IGP?  

--John Scudder

-----------[000533][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 15:04:07 -0700
From:      lstowell@pyrnova.mis.pyramid.com (Lon Stowell)
To:        comp.protocols.tcp-ip
Subject:   Re: RS6000 & TOKEN RING

In article <1994Apr26.230934.9159@int-evry.fr> martiner@int-evry.fr (Olivier MARTINERIE) writes:
>Hi! 
>Here's the problem of a friend:
>* We have 2 rings linked by bridges.
>
    You may want to post the make of the bridge, as some don't quite
    emulate the IBM bridge as well as others...and it could be
    a "personality mismatch" between how the station and the bridge
    handles source routing discovery broadcasts.  Some bridges don't
    forward Single Route Broadcasts by default. 

>* A machine connected on ring 2 try to link a risc 6000 on ring 1.
>* This machine send a frame with correct TCP/IP addresses.
   
    To what MAC address, and does that frame have a RIF field
    included?  If not, you may just have to turn Source Routing
    on in the machine.  

    Or it may be an issue where the bridges don't forward SRB's and
    that is what the machine uses.  Some use ARB's some SRB's.



-----------[000534][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 04:01:04 GMT
From:      gnb@bby.com.au (Gregory Bond)
To:        comp.protocols.tcp-ip
Subject:   Can someone point me to the RFC on Variable IP subnets?

Seeing the thread on RIPv2 and variable subnets came as something of a
shock to me because I was firmly of the opinion that a subnet mask was
constant across a given IP net.  Obviously, I'm wrong, so can someone
point out an RFC that explains just how the hell you decide where a
packet belongs with variable subnet masks?

A quick scan of (a possibly out of date) rfc-index.txt didn't show
anything obvious...

Greg.

--
Gregory Bond <gnb@bby.com.au> Burdett Buckeridge & Young Ltd Melbourne Australia

Atilla The Hun's Maxim: If you're going to rape, pillage and burn, be sure to 
do things in that order.  -- P. J. Plauger, Programming On Purpose, p147


-----------[000535][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 94 11:56:00 -0400
From:      grant.porter@canrem.com (Grant Porter)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP dial-up - async vs sync

Can anyone tell me the pros and cons of using synchronous communications versus
async when using TCP/IP? Does async allow multiple users using TN3270 to
connect over one modem?

-----------[000536][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 05:21:09 GMT
From:      medin@dscs.arc.nasa.gov (Milo S. Medin)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2pkcuo$94r@news.sprintlink.net>, paul@sprintlink.net (Paul Ferguson) writes:
...
|> Wellfleet's implementation of OSPF seems to have matured, with
|> various patch-releases of 7.56 and 7.57 software. We are preparing
|> to deploy an extension to a pilot network which will consist of
|> upwards to 50 routers (two BCNs & the remainder AFNs) which will
|> run solely 7.70. This is our only Wellfleet network employing
|> OSPF out of the 80 or so that we designed/deployed/support/manage.
|> 

At Ames, the institutional network is currently buily mostly of 
Wellfleets, with the exception of 2 cisco's, which my group manages
that interface the center to the outside world.  It all runs OSPF,
with variable length subnets, and works just fine.  You do have
to have a reasonably recent rev of Cisco code, as it's had it's 
problems in the past.  Cisco is tuning it's OSPF implementation,
and I expect it to improve, although it works pretty well with
our Proteon's on the WAN.

|> (On cisco, we use IGRP or EIGRP.)
|> 

That's fine, if your needs are limited enough to be met by them,
and that you don't mind running a protocol which locks you in 
to one routing vendor.  Only you can answer that questions of course.

|> Once this pilot has proven successful, the entire buildout of this
|> single network will be well over 1500 routers, both domestically and
|> internationally.
|> 
|> Yes, OSPF seems stable IN SOME environments, as it has appeared to
|> have stabilized in our current pilot scenarion. There are SERIOUS
|> problems which can occur, as when more than one router tries to 
|> act as the designated router in a single area (more a circuit
|> failure issue, than a design issue, although we have "designed
|> around" this occurance in the future buildout).
|> 

In complex nets (NSI's routing is probably considered complex), 
you need to be running pretty new cisco code.  In simpler 
situations, the older code worked fine.  It just wasn't very
robust.

I don't understand what you are talking about regarding designated
routers in an area.  A DR is elected on a given network (like an
ethernet, or X.25 network), and if you have multiple such networks in
an area, you'll usually have multiple DR's.  This is just fine!
What problems do you see because of this??


|> cisco's implementation of OSPF doesn't appear to be quite as
|> mature. (OBopinion.)
|>

I think their current code is pretty close to the quality of the
WF code we run (and their code is getting better too).  I do
think it's fair to say that Proteon's OSPF code probably sets the
standard for implementation quality.  It does more, with less
memory, and less CPU, in less time than either the Cisco or
WF colde.  John Moy is truly an amazing programmer.  But that's
not saying that the existing WF and Cisco implementations don't work.
They in general work just fine.

 
|> Final thought: OSPF is not for everyone. In fact, I think that
|> it is a "tailored" IGP and should be used as such.
|> 

I don't exactly know what you mean by tailored.  Certainly, the IETF doesn't
think so, since it's speced out in the draft updated router requirements RFC.

I quote: "A router which implements any routing protocol (other
      than static routes) MUST IMPLEMENT OSPF (see Section
      [7.2.2]) and MUST IMPLEMENT RIP (see Section [7.2.4]).
      A router MAY implement additional IGPs."

  

						Thanks,
						   Milo

-----------[000537][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 07:00:00 GMT
From:      ehammeg@hvstsg3.ns-nl.att.com (0396103-hammega)
To:        comp.protocols.tcp-ip
Subject:   OSI-CLNP versus IP

I am interested in finding out more about the differences
between the OSI-CLNP and IP.

-CLNP speaks of "segmenting" while IP
 speaks of "fragmenting". Are these the same? How do these mechanisms work?

-Both protocols speak of "lifetime" "source routing" and "adressing"
 Do they mean the same thing ?

-CLNP knows a "total length of originating PDU". IP doesn't.
 How about this?

Any answers would be appreciated.


Erik Hammega





-----------[000538][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 08:32:58 GMT
From:      shlam@ie.cuhk.hk (Alan S H Lam)
To:        comp.protocols.ppp,comp.protocols.tcp-ip
Subject:   Help: bootp in cslip and network routing via slip inteface


We had installed cslip2.7 on SunOS 4.1.3. and supposed this
workstation is connected by some remote pc by the etherslip 
driver there.

I have two questions

1) can cslip support bootp as terminal server does? If 
the bootp.2.1 package (or other packages) can do the job?

2) we discover that the remote pc can only ping the dail in
workstaion. We cannot ping other workstaions on the network
even the route table have been set as follows:

# netstat -rn
Routing tables
Destination          Gateway              Flags    Refcnt Use
Interface
137.189.97.241       137.189.96.19        UH       1      223
sl0
127.0.0.1            127.0.0.1            UH       1      3377
lo0
137.189.0.0          137.189.96.19        U        8      118865
le0


where 137.189.97.241 is the remote pc ip.

Is there any way I can access other workstaions via the slip
connection?


I would very appreciate it if you could send me a e-mail and give
me some hint to solve the above problem. Thanks.


--
Alan S. H. Lam
Department of Information Engineering, CUHK, Hong Kong
E-mail: shlam@ie.cuhk.hk 
Tel: (852) 609 8364 Fax: (852) 603 5032

-----------[000539][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 17:27:02 -0400
From:      rayrizzuto@aol.com (RayRizzuto)
To:        comp.protocols.tcp-ip
Subject:   TLI t_optmgmt and no output buffering

Help!  I need to find out how to specify (using TLI on a Pyramid system) that
TCP not buffer outgoing data on a connection, but rather send it immediately,
even if the amount to be sent.  I've used setsockopt for sockets, but can't
find equivilent information on the TLI version of the interface.

Thanks in advance for any help you may lend.

Ray Rizzuto

-----------[000540][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 94 17:19:08 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Problem: Unreadable packets from router.....


IK>Anyone have any suggestions?  Anything would be appreciated!  Maybe the
IK>NWC is encapsulating the IPX packets with improper IP headers?

I had understood that IP and IPX were independently encapsulated into
Ethernet packets.  IP datagrams are put into Ethernet type 0x0800
packets.  (I'm sorry, I can't find the Ethernet type for IPX
encapsuation.)  On our Ethernet, TCP/IP and IPX/SPX never seem to
collide, above physical layer, anyway.

-- Bob Stein

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000541][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 13:24:13 GMT
From:      zeeff@zip.eecs.umich.edu (Jon Zeeff)
To:        comp.protocols.tcp-ip
Subject:   Adding compression to a 56k leased line?

I am using a 56k leased line for a tcp/ip over ethernet connection.  I'd like
to increase the speed by adding compression.   Ideally this would all
be in one box - ethernet router, compression, and csu/dsu.

Does anyone know of reasonably priced solutions?  Black Box wants
$2500 at each end just for compression and the csu/dsu.


-----------[000542][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 14:04:52 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

John G. Scudder (jgs@hadron.merit.edu) wrote:

> So what's a "tailored" IGP?  

The "tailored" reference is my own term, for lack of a more
approriate phrase. IGP is Internal Gateway Protocol...


_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000543][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 14:12:18 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

Milo S. Medin (medin@dscs.arc.nasa.gov) wrote:

> |> (On cisco, we use IGRP or EIGRP.)
> |> 
 
> That's fine, if your needs are limited enough to be met by them,
> and that you don't mind running a protocol which locks you in 
> to one routing vendor.  Only you can answer that questions of course.

Using IGRP doesn't necessarily limit you to a single vendor. You can 
certainly redistribute to RIP on locally attached networks/wide area
segments that don't speak IGRP. With "class C" addresses, this is a 
no-brainer.

With the implementation of EIGRP, you can choose not to summarize
learned routes, which gives you the ability to invoke (E)IGRP on
a subnet of a "class B" address, allowing you to effectively
redistribute as well.

The argument that IGRP limits you to a single vender only holds
true if you are referring to the wide area implementation, and
only in particular configurations (not all). 


_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000544][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 14:13:29 GMT
From:      anw@allen.ess.harris.com (Allen Williams)
To:        comp.protocols.tcp-ip
Subject:   ?? Multicast Backbone (MBONE) Routers ??



Does anyone out there know which of the major router vendors support MBONE
(RFC 1075)?
If yes, do any of the routers have a "switching hub" type architecture, where
only the packets destined for a particular node appear on that node's segment
(the unique "spoke")?

-- 

Thanx,

Allen                                                    Voice: (407)727-5280
anw@allen.ess.harris.com                                 Fax:   (407)729-1985


-----------[000545][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 14:42:52 GMT
From:      jgs@hadron.merit.edu (John G. Scudder)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2plre4$gtt@news.sprintlink.net>,
Paul Ferguson <paul@hawksbill.sprintmrn.com> wrote:
>John G. Scudder (jgs@hadron.merit.edu) wrote:
>
>> So what's a "tailored" IGP?  
>
>The "tailored" reference is my own term, for lack of a more
>approriate phrase. IGP is Internal Gateway Protocol...

That much was clear.  Let's try this again:  What do you MEAN by
"tailored" IGP?  (It was suggested to me in private mail that this is
an IGP with pinstripes, or my guess was that in the case of RIP it's an
IGP which laces up the back and has really, really long sleeves.)

--John Scudder

-----------[000546][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 15:22:22 GMT
From:      ncherry@cbnewsg.cb.att.com (neil j.cherry)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

In article <2pjkqm$9jl@vixen.cso.uiuc.edu> kline@tampico.cso.uiuc.edu (Charley Kline) writes:
>
>>Ahhhhhhhhhhhhhhhhhhh, but will OSPF play nice on a network between
>>Cisco and Wellfleet? (If you're using both routers with OSPF do let us
>>know though!)
 (I wrote that)
>
>I haven't done extensive testing, but they seem to, yes. Our network is
>nearly all Proteon with a small smattering of Cisco (which seem to
>interoperate reasonably) and I'm currently testing a Wellfleet BLN.
>
>
>> Whatever you say, pal.
 (I don't know who wrote that, but Charles didn't say I wrote it either).
>
>Well, the proof is in the pudding. I run a 70-router network with 300
>subnets and 7 different subnet masks, and the only routing protocol I
>use is OSPF, and I haven't had any problems whatsoever with it. I'm
>sorry if people can't get it to work, but that's hardly the fault of
>the protocol.
>

Charles, are the Cisco's and Wellfleets connected via ethernet, serial,
or a mix of both? If serial what line encap are you using?

Thanks NJC (Neil Cherry)

>/cvk



-----------[000547][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 16:52:52 GMT
From:      d00n@crash.cts.com (Kevin Spousta)
To:        comp.protocols.tcp-ip
Subject:   Re: Winsock under Win-OS/2

In <940422102810@whip.ftp.com> fks@ftp.com  (Frances K. Selkirk) writes:

>Version 1.3 of our PC/TCP for OS/2 includes a module for support of
>our DOS and Windows applications, including our winsock.dll, which is
>included. 

Frances,

I've been given a project that necessitates running TCP/IP on an OS/2 box and 
am having no luck with IBM's TCP/IP 2.0 products due to the fact the TCP/IP
kernel does not 'advertise' itself in a DOS box (Even with their DOS/WIN access
kit) so I was wondering, does FTP's IP product support ping/telnet/finger/ftp
from DOS-os/2 sessions?  AND, does it come with daemons for sendmail, ftp
etc?

Thanks!


-----------[000548][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 17:02:56 GMT
From:      d00n@crash.cts.com (Kevin Spousta)
To:        comp.protocols.tcp-ip
Subject:   NCSA Telnet:  Token Ring works, E-net doesn't...

Well, after surfing the net for the past two days looking for an answer to my
current problem, I've come to the point where it's time to start asking
questions.

I've got an old Epson 386sx that I've set up as a remote workstation (PC 
Anywhere) and the 'owner' of this machine would like to be able to dial in
and telnet out.  Piece of cake, right?  Not so.  Using NCSA Telnet v2.3.07
adnd an IBM Token Ring card, I'm able to make it work with no problems.
Since the IBMTOKEN packet driver will NOT work concurrently with IPX, this
configuration was tossed out the door in favor of an Intel EtherExpress 16 
Flash card using the EXP16.COM (Dated 1-28-92) packet driver.

The packet driver is loaded and the card initializes just fine, however, when
I try to telnet or FTP somewhere, I get "Network Jammed - Probable Broken Wire"
errors.  I know it's not a broken wire because I can log into the netware
boxes just fine.  I can even run the ODI stuff and make the IP and IPX live
happily together and still log into the NW box, but still no IP.

I spent the better part of yesterday checking router configs to see if IP was
being filtered out, which isn't the case.  Now I'm begining to think the
packet driver is not compatible with this card or just out of date.

In an attempt to locate a newer packet driver, I've bounced all over the planet
to every FTP site on my list (ok, only about 30 of 'em) and have not had any 
luck.  I've even tried an archie search.  the readmes indicate drivers are
available at simtel20.army.mil, but trying to FTP (or even ping) to this site
nets me a "unknown host" error.  

Does anyone know the IP address (or REAL host name) of the simtel site?
AND, what kinds of things should I be looking for with this E-net problem?


"Sometimes I hate my job"  - me, yesterday



-----------[000549][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 94 03:11:03 -0600
From:      pearceh@isg001.rec.bp.com (Huw Pearce)
To:        comp.protocols.tcp-ip
Subject:   Re: External Clock for Network Time Protocol


In article <2pje18$ivf@xenon.brooks.af.mil>, perry@brooks.af.mil
(David Perry) writes:
|>We're looking for an external clock (hardware) to interface to a
|>server
|>to provide time synchronization using something like network time
|>protocol.
|>
|>Does anyone know of a good source for this type of product?  We want
|>to
|>interface to a Sun server.
|>
|>Thanks,
|>
|>
|>---
|>
|>David W. Perry
|>Computer Sciences Corp.  
|>Brooks AFB, TX
|>
|>
|>


We have bought a GPS receiver from Trimble Navigation. The cost here
in the UK
was approx 500 pounds sterling. It is currently attached to a Sun 4
system for 
test purposes. But is scheduled to be moved to a Sun SPARC station, or
if this 
not available a Digital DECstation 3100.
 
My colleague who has set it up is getting better than 2 milli-second
accuracy 
in the test environment. The software being used is 'xntp', I am not
sure what 
revision level but it is greater than "m".

We are shortly to roll out to an environment where there will be the
machine 
with the GPS receiver providing a time source to 10 next level down
'secondary' 
servers which will be supporting in excess of 200 systems. If we find
the right 
clients for Mac's, PC's and other systems this could rise to 500+ (at
this site).

Something we have yet to organise is access to additional reference
clocks within 
the company (or externally) to try and alleviate any possible clock
"runaways"!!

-- 
Huw Pearce                              E-mail:
pearceh@rcsv3.dnet.bp.com
BP Research & Engineering                      
pearcedh@rcwusr.rcw.bp.com
B200 R148
Chertsey Road
Sunbury-on-Thames, Middlesex TW16 7LN
UK

-----------[000550][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 94 23:44:50 MDT
From:      u-mwong%peruvian.cs.utah.edu@cs.utah.edu (Michael Wong)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Problem with FTPing to RS6000 with Novell LWP for DOS...HELP!!!

   Hi...I've got a major problem that needs to be resolved ASAP and am
hoping someone here could help me...

   I am using a PC with Novell's LAN Workplace for DOS V4.1 (with Patch N)
and am trying to FTP to an IBM RS6000 (AIX) as well as a PC (I think it's
DOS-based).  I am using Novell's TCP/IP stack and a 3Com 3C509 ethernet card.
I am writing a simple FTP client program, but I am using LWP's FTP client
for testing and verification.  The problem I'm about to describe occurs on
both my FTP program AND Novell's FTP program.  So, I will describe my 
problem with Novell's FTP program.

   Now the problem...  When I connect to the RS6000 and try to send a file
or perform a directory listing, the server seems to hang and NOT respond
back to the client.  Here's an example of some FTP debug info I got:

	ftp> dir
	---> PORT 190,1,53,20,4,3
	200 PORT command successful.
	---> LIST
	(up to a minute or so delay)
	425 Can't create data socket (190.1.65.20,20): Address already in use.

Sometimes it will work and sometimes it won't.  However, I don't care about
performing a "dir" as much as transferring a file to the FTP server.
I mentioned the above only because I think it's related to my file
transfer problem.  Note that both a file transfer and a directory listing
require usage of the "control connection" AND a "data connection", if you're
familiar with the FTP protocol.  The server also doesn't seem to respond 
back when attempting a file transfer or after a file transfer.  Here's an 
example of some more FTP debug info:

	ftp> put testfile.ext
	---> PWD
	257 "/adv/ftp" is current directory.
	---> CWD testfile.ext
	550 testfile.ext: No such file or directory.
	---> PORT 190,1,53,20,4,7
	200 PORT command successful.
	---> STOR testfile.ext

At this point, sometimes I get the following return message after a moment:

	425 No data connection.
	put: I/O error

At other times, I get this:

	150 Opening data connection for testfile.ext.
	1983 bytes transmitted in 1 seconds (1983 bytes/s)

Then it hangs here, with the server not sending back "226 Transfer completed".
The question is, WHY!?!?!?!?!?!  In this second situation, after a minute,
I'll get lots of "approximate" characters (a tilde on top of another tilde).
and return back to the "ftp>" prompt.  When I quit out of ftp and ping the
RS6000, it says "No response from ...", as if my PC got taken down from the
network.  After a reboot, I can ping again.

  I did notice that when a data connection WAS opened and the file was sent
and the client's data connection closed, the FTP server seemed to have a hard
time closing the received file (I saw it had a "0" length file during the
wait), and because it wasn't closed, was unable to send back "226 Transfer 
completed".  I am also sure that there isn't a machine with the same IP 
address as mine.

   Does anybody have any ideas???  Is it a conflict with Novell sockets and 
IBM sockets?  Is the RS6000 low on socket resources?  After a lot of testing 
last Sunday and last night for the afternoon and evening, I noticed that 
my problems persisted until the later evening (i.e. after 7p); the 
problems cleared up and the file transfers worked!!!!!!  Is this 
coincidence or what?!?!  I'm afraid it won't work the next afternoon...

   Any help or ideas are GREATLY appreciated!!!!!  By the way, I have talked
to Novell Tech. Support, but the guy didn't know of any problems like this...

-Mike
-- 
Michael Wong 
u-mwong@peruvian.cs.utah.edu
University of Utah

-----------[000551][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 18:11:27 GMT
From:      howlw@niccolo.gsfc.nasa.gov (Wes Howl 301-794-2043 S825)
To:        comp.protocols.tcp-ip
Subject:   Tapping the loopback network interface on SunOs 4.1.3

How can I passively access the data flowing on the
loopback network interface know as "lo0" (SunOs 4.1.3)?
  
I have a working program which uses the STREAMS module "nit_if" to tap 
the ethernet network interface "le0".  When I try same program on the network 
interface "lo0", I get neither errors nor input.  I get the same results with 
the Sun program "etherfind" (i.e. data on le0 and nothing on lo0).  Etherfind 
accepts lo0 as a valid interface yet behaves as if there is no packet activity,
despite the fact that I've verified its presence via "netstat -I".

    The loopback network interface, while similar in many ways to real 
(i.e. underlying hardware) network interfaces,  seems somehow different
with respect to using "nit_if" on it.  If there is a mechanism other 
than "nit_if" which must be used, then I'll use it; I just don't know what 
it is.  The ftp address of an example program would also suffice as an answer.

							Thanks 


							Wes

-----------[000552][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 18:59:21 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting Class C into Eight nets ?

In article <2peq2t$arm@news.delphi.com> sunborn@news.delphi.com (SUNBORN@DELPHI.COM) writes:
>I would like to subnet our class C network into six to eight subnets, can 
>this be done ?

Depends on how many hosts you want to put on each subnet.  With a 3-bit
subnet field you can have up to 6 subnets with up to 30 hosts on each.
With a 4-bit subnet field you can have up to 14 subnets with 14 hosts on
each.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000553][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 20:18:35 GMT
From:      harp@martha.utcc.utk.edu (Sean Harp)
To:        comp.protocols.tcp-ip,comp.unix.solaris,comp.unix.admin,comp.unix.misc
Subject:   Re: Help


	I have been wondering about the pros and cons of a certain
hypothetical system configuration.  The system configuration is as
follows:

                                 __________
ETHERNET                        /          \
                   ________    /            \
-------------------|BRIDGE|---/              \
               \_  --------   |   FDDI       |
                 \            |     RING     |
                  \___        \              /
                      \__      \            /
                         \     /\__________/
                          UNIX/     
                          SERVER     

The UNIX server has both an ethernet interface and an FDDI.  Because
the network is bridged, the host has the same IP address on both
interfaces.  Both interfaces in the situation are live.  What I would
like to have your opinion on is:

1) What are the benefits derived by this configuration? (i.e. redundancy?)

2) What problems would this configuration cause? (i.e. broadcast
   packets being received by the same machine on two different interfaces)

Any comments would be greatly appreciated.

Sean Harp

-----------[000554][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 21:02:58 GMT
From:      optel@phoque.info.uqam.ca ()
To:        comp.protocols.tcp-ip
Subject:   netmask to reduce class C Network (?)

:
Subject: netmask to reduce class C Network (?)
Newsgroups: comp.protocols.tcp-ip
Summary: 
Keywords: 

Can I use a netmask to reduce the network portion of a class C address?
I have 2 class C addresses 193.220.250 and 193.220.251 each currently supporting
about 200 users and would like to rehome them into an Ethernet Switch and 
therefore stop using a router between them.

I have:
    <--------- net ----------> <-host->
    NNNNNNNN NNNNNNNN NNNNNNNN HHHHHHHH

I want:
    <-------- net ---------><---host-->
    NNNNNNNN NNNNNNNN NNNNNNHH HHHHHHHH 

Thank you for helping me out 

Marc Louarduzzi
opTEL@phoque.info.uqam.ca 

-----------[000555][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 22:10:32 GMT
From:      dsiebert@icaen.uiowa.edu (Doug Siebert)
To:        comp.protocols.tcp-ip
Subject:   Example of raw socket code to send TCP SYN packet?


Does anyone have any example of some code using raw sockets on a Unix system
to send a single TCP SYN packet?  What I want to do is create a raw socket,
send a SYN packet at a host (to get it to think I want to open a connection)
but give a bogus IP address in the packet so the ACK won't get back to me.
This will leave a half-open/half-closed socket stuck in the listen queue, so
that I can fill up the listen queue, tickling this particular BSD TCP bug, so
that the engineer at HP I am talking to about fixing the bug will be able to
reproduce it.  Any help out there?  Even if you don't have code that will do
all this, even something that will do part of this (maybe create the raw
socket and build a TCP packet of some sort, with checksum -- I could probably
hack the rest)  I don't want to reinvent the wheel, or have to read too much
to figure out what goes where in the packet, how to calculate the checksum,
etc.  Thanks!


-- 
Douglas Siebert
dsiebert@isca.uiowa.edu

-----------[000556][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 22:29:27 GMT
From:      jliang@mystery.Stanford.EDU (June Liang)
To:        comp.dcom.lans,comp.dcom.lans.fddi,comp.dcom.lans.etherent,comp.dcom.lans.token-ring,comp.protocols.tcp-ip
Subject:   Traffic pattern on new type of application (e.g. Lotus Notes)

Hi,

Since there is a lot of talk about companies running "groupware" in
their enviornment, I would like to know if anyone actually look into
what type of bandwidth requirement and traffic pattern such type of
applications is currently using and what type of traffic requirement
is expected in the near future.  For example, someone told me that
Lotus Notes generates a lot of server to server traffic.  But he has
no idea of how much traffic is actually consumed.  Anyway, any
information or speculation is useful.  Thanks a lot.

JL
jliang@mystery.stanford.edu



-----------[000557][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 22:46:59 GMT
From:      jcmurphy@acsu.buffalo.edu (Jeff Murphy)
To:        comp.protocols.tcp-ip
Subject:   Re: Tapping the loopback network interface on SunOs 4.1.3

Wes Howl 301-794-2043 S825 (howlw@niccolo.gsfc.nasa.gov) wrote:
>How can I passively access the data flowing on the
>loopback network interface know as "lo0" (SunOs 4.1.3)?
>  

i believe that you can't. because of the way sun's NIT device is written,
the machine can not ''see'' traffic that *it* generates. 

since the loopback device will always contain only traffic generated by
the machine itself.. nit will never see anything (i.e. no traffic on the
device). 

try using tcpdump to capture outgoing traffic from your machine..

unix% tcpdump src host [your machine] &
unix% ping -s [some other machine]

or try:

unix% tcpdump -i lo0 


you shouldn't see anything...


hope this clears things up!

jeff
-- 
jcmurphy@acsu.buffalo.edu                            ece/cs             cit oss
opnsmurf@ubvms.bitnet				     standard disclaimers apply
''She walks in beauty, like the night - Of cloudless climes and starry skies.''

-----------[000558][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 09:44:28 +0800
From:      carl@oversteer.library.uwa.edu.au (Carl Brewer)
To:        comp.protocols.tcp-ip,comp.sys.sun.hardware,comp.sys.sun.admin
Subject:   Re: Need multicast info. in Solaris-2 with FDDI interface.

In article <1994Apr26.154241.26538@lambda.msfc.nasa.gov>,
Rafique Mahmood <mahmood@lambda.msfc.nasa.gov> wrote:
>Currently we are running Solaris 1 (sunOS 4.1.3). We use IP multicasting on 
>ethernet and on FDDI. 

Ditto, but no FDDI :)

>We are planning to upgrade the OS to Solaris-2. I know it has multicast in
>the OS (in 4.1.3 I installed the multicast patch). But how to configure 
>Solaris 2 with FDDI? I remember reading somewhere the problem with the driver
>which does not work with this OS. I am not sure what is the problem I read.

The main problem is that the Sod 2.x Multicast stuff is pretty out of date
(as I understand it) and you'll have to fix it up with 3.1 stuff (the Sod 2 
stuff is IP multicast 2.x from memory, someone knows more than me on this 
though :) ).  

Someone's ported mrouted 2.2 I think,  but I'm not sure about 3.1 (or 3.2b)
Maybe the IP multicast list should move to a USENET group?

>
>I called 'sun' but haven't got the respond yet. I need to find out this really
>soon. I will appreciate any response or suggesion.

Don't downgrade to solaris!



-- 
Carl Brewer				Ph :61-9-380-1893 | #include \
Systems/Network Officer, Reid Library   Fax:61-9-380-1012 | <std_disclaimer.h>
University of Western Australia		carl@oversteer.library.uwa.edu.au
Merlin, where are you?  Call your Dragon, to weave a mist ....        "BEABLE"

-----------[000559][next][prev][last][first]----------------------------------------------------
Date:      Wed, 27 Apr 1994 23:49:16 GMT
From:      vjs@calcite.rhyolite.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Adding compression to a 56k leased line?

In article <2plp1t$eos@zip.eecs.umich.edu> zeeff@zip.eecs.umich.edu (Jon Zeeff) writes:
>I am using a 56k leased line for a tcp/ip over ethernet connection.  I'd like
>to increase the speed by adding compression.   Ideally this would all
>be in one box - ethernet router, compression, and csu/dsu.
>
>Does anyone know of reasonably priced solutions?  Black Box wants
>$2500 at each end just for compression and the csu/dsu.

You could always run PPP with CCP over the line, compressing in the host.
The drafts for PPP compression are pretty stable.


Vernon Schryver    vjs@rhyolite.com

-----------[000560][next][prev][last][first]----------------------------------------------------
Date:      27 Apr 1994 23:59:14 GMT
From:      paul@sprintlink.net (Paul Ferguson)
To:        comp.protocols.tcp-ip,comp.dcom.sys.wellfleet,comp.dcom.sys.cisco
Subject:   Re: Variable IP subnets with RIP on Wellfleet and Cisco routers

John G. Scudder (jgs@hadron.merit.edu) wrote:

> That much was clear.  Let's try this again:  What do you MEAN by
> "tailored" IGP?  (It was suggested to me in private mail that this is
> an IGP with pinstripes, or my guess was that in the case of RIP it's an
> IGP which laces up the back and has really, really long sleeves.)

Okay -- I'll buy that. Or perhaps an IGP cross-dresser?  

Actually, what I MEAN, is that OSPF is geared towards networks in which
transferrence of the subnet mask is desirable (variably subnetted 
networks eg.). Most of the driving factors in implementing OSPF are
specialized networks, where a specific routing requirements exist.

_______________________________________________________________________________
Paul Ferguson                         
US Sprint 
Enterprise Internet Engineering                    tel: 703.904.2437 
Herndon, Virginia  USA                        internet: paul@hawk.sprintmrn.com

-----------[000561][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 01:02:27 GMT
From:      schaumann@oms2.usmc.mil (David K Schaumann)
To:        comp.dcom.lans.token-ring,comp.protocols.tcp-ip
Subject:   LANWatch on NDIS/IBM LSP

I have seen some people out here talk about their config.sys having the file
ftpsoft.dos in it.  From what I could tell, that was LANWatch running on top
of NDIS.  I would like to know how, and if I need the file ftpsoft.dos, where 
is it.  I found the file in the zipped file kslan.exe on ftp.ftp.com, but it was
encrypted.  Does anyone know where else to look?

Please help!!!!!!

Dave

Dave Schaumann
Okinawa   Japan
InterNet: schaumann@oms2.usmc.mil
or: schaumann@okr-smtp1.usmc.mil

-----------[000562][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 03:21:01 GMT
From:      albert@dn.itg.telecom.com.au (Albert Fu)
To:        comp.protocols.tcp-ip
Subject:   IP RIP Question in multi-subnet/network environment

This is a question regarding the correct behaviour of a router
connected to a segment that has 2 IP subnets defined, but only 
one of which is defined on the router.

Here is a diagram that exemplified the configuration:
(fictitious network used, netmask 255.255.255.0)

  +-----------+                            +-----------+
  |  Router1  |                            |  Router2  |
  +-----------+                            +-----------+
        |150.150.1.1                             |150.150.1.2
        |150.150.2.1                             |
        |                                        |
    |------------------------------------------------|

Both routers run the RIP protocols, and have connections to other
networks/subnets. During the periodic RIP update (every 30s), Router1 
sends out two sets of RIP updates to the directed subnet broadcast
addresses of its interface, namely 150.150.1.255, and 150.150.2.255. 
Both the RIP packets are sent to the broadcast mac address 
(0xffffffffffff).

Router2 will process the RIP packets destined to the 150.150.1.255 
subnet broadcast. The question is what should it do with the
RIP packets sent by Router1 to the 150.150.2.255 subnet broadcast?

a) drop the rip packets as these packets are mac layer broadcasts
   and therefore should not be routed.

b) route the packets back to 150.150.1.1 since this router knows how
   to get to the 150.150.2.0 subnet.

(a) and (b) appears to be the behaviours of two different router 
vendors. Does anyone know which is the correct one?

Albert Fu.


-----------[000563][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 94 03:23:32 GMT
From:      tom@wcc.oz.au (Tom Evans)
To:        comp.protocols.tcp-ip
Subject:   Signed port bug in which TCP-IP?

I remember an article from a few weeks ago where it was reported that
"Brand-X" TCP/IP didn't work if the TCP port number from "Brand-Y" was
over 32768.

If someone can remember what "Brand-X" was, and/or direct me to a site
that archives comp.protocols.tcp-ip where I can FTP the last few
months' worth, I'd appreciate it. We have a report of the box we make
not performing FTP to Solaris 2.x, and I know we use ports over 32768.
The Solaris patch-README files are both extremely unhelpful and
numerous :-)

Thanks

========================
Tom Evans  tom@wcc.oz.au
Webster Computer Corp P/L, 11 Glenvale Crescent Mulgrave, Melbourne 3170
Victoria, Australia 61-3-560-1100  FAX ...560-0067  A.C.N. 004 818 455


-----------[000564][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 94 10:00:40 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Re: DOS TCP/IP Stacks?


>>I am looking for information on companies who offer TCP/IP stacks for DOS.
 
>Probably the best [and appropriately priced :-( ] is PC/TCP from FTP
>Software.

I have found the following TCP/IP stacks for DOS and am currently
evaluating Ipswitch's Piper/IP.  Here are their and other relevant
numbers:

       Ipswitch            (617) 246-1150     claims 6K footprint in
                                              conventional memory if you
                                              use their protected-mode
                                              version

       FTP Software        (508) 685-3300     been around for a long
                                              time, written in assembly

       Beame & Whiteside   (919) 831-8989     benchmarked very fast

       Wollongong Group    (415) 962-7100

There are many other TCP/IP vendors who are aiming for Windows
developers only.  This seems to be where all the heat is in this market.
We are only interested in DOS for performance reasons.  There's a good
performance review in:

       Data Communications, 2/94, pp 71-84

Here's a good paper with more companies, and lots of interesting info:

       anonymous ftp ftp.cac.psu.edu
       /pub/dos/info/tcpip.packages

-- Bob Stein                            Internet mail: stein@gcomm.com
 ______________________________________________________________________
|                                                                      |
|  Galacticomm, Inc.                            (305) 583-5990 (voice) |
|  4101 SW 47 Avenue, Suite 101                 (305) 583-7846 (FAX)   |
|  Ft Lauderdale, Florida, USA, 33314           (305) 583-7808 (BBS)   |
|______________________________________________________________________|

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000565][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 05:59:04 GMT
From:      edward@e0sun29.ccl.itri.org.tw (Chao-Chi Yang 37h3100)
To:        comp.protocols.tcp-ip
Subject:   [Q] Software Bridge over FDDI and ATM.


Hi All,


    We want to construct an ATM testbed which two FDDI rings are
linked by ATM LAN.  Before this, we have no knowledge about how
to build such environment except IP router.  We don't like
to build onto TCP/IP stack because it will hurt performance.
In the first step, we only want to implement a workable software bridge. 
Would somebodies like to give us any suggestions about:


1) What API will we need ? 
	
   For the ATM end, we want to use Fore ATM API, but for FDDI end we don't
   know how to get the FDDI MAC frame.
   I have heard that 'netfind' is based on NIT to get raw ethernet data.
   Can we make use of NIT (Network Interface Tap) ? and where is the document
   of NIT ?

2) Where can I find some sample programs for similiar task ?

3) Could we get much better performance than IP router ?


----------------------------------------------------------------
ps. Software Bridge

	SunSparc 10, equipped with
	one NP-FDDI card, and
	one Fore SBA-200 ATM card, running on
	SunOs 4.1.3



--
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#
| __ ___                                                |
|  |||           ´­´Â¦N                                 |
|__|||_          Chao Chi Yang (Edward)                 |
|                                                       |
|   at Advance Tech. Center, CCL, ITRI, Taiwan          |
|       						|
|   TEL 	886-35-917178                           |
|   FAX 	886-35-820098                           |
|   Email       edward@e0sun3.ccl.itri.org.tw           |
|                                                       |
 #~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#

-----------[000566][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 06:22:21 GMT
From:      d00n@crash.cts.com (Kevin Spousta)
To:        comp.protocols.tcp-ip
Subject:   Re: NCSA Telnet: Token Ring works, E-net doesn't...

Never mind.  I found a newer packet driver..  Works like a charm.


-----------[000567][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 14:45:00
From:      cearley_k@wizard.colorado.edu (Kent Cearley)
To:        comp.protocols.tcp-ip
Subject:   What constitutes a duplicate TCP packet?

We have a problem with a host sending two tcp acknowledgements... one with 
WIND=0 and no data, followed immediately with a WIND=256, also no data. Both 
the sequence numbers and ack numbers are the same on both packets. The problem 
is, a client (FTP Inc.'s WTN3270) sees the first packet, assumes it cannot 
transmit, and appears to be discarding the second packet without the WIND=256 
registering.

FTP Inc. claims the RFC specifies if the SEQ# are the same its a duplicate 
packet. If so, how is this situation handled? All of our other clients 
recognize the second packet as advertising an available window. 



-----------[000568][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 09:54:20 GMT
From:      cmclal@lizzie.arh.cdc.com (A. Sudhakarlal)
To:        comp.protocols.tcp-ip
Subject:   Can a workstation have broadcast address as it's IP address?

Can a workstation on a TCP/IP based ethernet have IP broadcast address as 
it's host address? We have some routers which accept X.X.X.255 ( A class C
address) as a host address. Because of this RIP updates do not reach all the
routers on the LAN, and thus routing tables are inaccurate.

Your ideas would help me in fixing code for the router. Thanks.

Lal

-----------[000569][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 16:41:50 -0400
From:      gt0567a@prism.gatech.edu (Mathieu Claude Hans)
To:        comp.protocols.tcp-ip
Subject:   packet losses, WHY


-Transmission using the UDP protocol:
 
 
I beleive all the packet losses which occur on Internet are
because of:
        - congestions at routers and at the host
        - error transmission, which in the case of UDP protocol
implies dropping the packet.
        - expiry of time to live because of bad routing.

Are there any other reasons why packets are lost on the network?
 
What are the percentages of each - is it 80% loss because of 
congestion, 10% because of error transmission ...? -
 
-- 
Mat Hans (-:
Georgia Tech, school of Electrical and Computer Engineering 
Internet: 	gt0567a@acme.gatech.edu

-----------[000570][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 20:28:39 -0500
From:      lin@cs.purdue.EDU (John Chueng-Hsien Lin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and UDP examples

In article <Coz5L1.I1u.3@cs.cmu.edu> gz+@CS.CMU.EDU (Gregory Zelesnik) writes:

   [...]

> Does anyone know where I can get more examples of simple clients and servers
> to look at?

   Try Comer and Stevens "Internetworking with TCP/IP Vol.III" book.

   The example source code of the book can be found in

   ftp.uu.net:/published/books/comer.internetworking3.src.tar.Z.

   [...]

>    I am having trouble connecting to gwen.cs.purdue.edu and was wondering if anyone
>    else had another pointer to this or similar source.

   The IP address of gwen.cs.purdue.edu is 128.10.3.8.

John Lin (lin@cs.purdue.edu)




-----------[000571][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 12:41:36 GMT
From:      fenner@cmf.nrl.navy.mil (William C. Fenner)
To:        comp.protocols.tcp-ip
Subject:   Supernetting on hosts

A local Internet provider just announced that they ran out of addresses on
their class C address, and so got assigned a block of 16, x.x.32 to x.x.47.
They said "This means that our new network number is x.x.32.0, and our new
subnet mask is 255.255.240.0, and the broadcast address is x.x.47.255 ."

I was under the impression that few, if any, hosts were able to deal with
supernetting.  I tried ifconfig'ing a spare ethernet interface on a Sun
running 4.1.3, and there didn't appear to be the proper netmask associated
with the route in the routing table, as tracerouting to anything outside of
the first class C would go via default.

Are these people asking for trouble, or is SunOS in the minority?  Most of
the hosts in this space will probably be dynamic SLIP/PPP connections.

  Bill
-- 
Bill Fenner                  fenner@cmf.nrl.navy.mil

-----------[000572][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 14:21:03 GMT
From:      yjapts@info.vub.ac.be (Yoeri J. Apts)
To:        comp.lang.c,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Sources from book 'INTERNETWORKING WITH TCP/IP vol II'

I am currently reading INTERNETWORKING WITH TCP/IP vol II (by Douglas
E. comer and David L. Stevens). This book contains a lot of sample code
and so, being lazy like every programmer, I'm wondering if these sources
are electronically available 'anywhere'?

I am especially interested in de code concerning SNMP (Client-Server)

 Yoeri Apts                         Tel: +32(0)2-6412977
 yjapts@info.vub.ac.be              Fax: +32(0)2-6412870
                                   ISDN: +32(0)2-9239936
 Dept. INFO/TW - Digital Telecommunications Group.
 Brussels Free University
 Pleinlaan 2, 1050 Brussel
 BELGIUM

-----------[000573][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 14:58:12 GMT
From:      gz+@CS.CMU.EDU (Gregory Zelesnik)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP and UDP examples

I've recently been experimenting with writing simple clients and servers between
Sun hosts running BSD 4.3 using sockets and the TCP/IP and UDP/IP protocols. 
I've been coding up examples from a book by W. Richard Stevens entitled 
"UNIX Network Programming."  The TCP/IP examples work fine, but I'm having 
trouble getting the UDP examples to work.

Does anyone know where I can get more examples of simple clients and servers
to look at?  In a previous post to this newsgroup (32042) Chris Chlap of 
the University of Canberra, Australia writes:

> An excellent book on this is Comer and Stevens "Internetworking with TCP/IP
> Vol. 2" published by Prentice-Hall. The code in the book is available
> from:
>
> SUN Version: gwen.cs.purdue.edu in /pub/Xinu
> PC Version: csc.canberra.edu.au in /pub/ise/xinu
>
> The above site also contains recent manual pages.
> 
> Chris Chlap
> University of Canberra, Australia

I am having trouble connecting to gwen.cs.purdue.edu and was wondering if anyone
else had another pointer to this or similar source.

Thanks for any help you can give,
Greg

-- 

Gregory Zelesnik
School of Computer Science
Carnegie Mellon University
Pittsburgh, PA 15213-3980

(412) 268-3164
gz@cs.cmu.edu

-----------[000574][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 15:57:10 GMT
From:      rsmullen@ida.org (Robert Smullen)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: IP address assignment via MacTCP/ARA?

The bottom line, without getting into "deencapsulating" IP packets and other
technical garbage:  You can't get there without some hardware.  If you have
a connection to the network via either a GatorBox or a FastPath, you're OK
but if your "server" is just a Mac with an ethernet card.  NO GO.

I ended up buying a LanRover /E 2.0 which supports TCP/IP and IP FOrwarding.
It works like a champ.

Sorry I didn't have better news.

Been There- Done That,
J.R. Smullen


Michael Jordan-I WISH! (ajkim@miavx1.acs.muohio.edu) wrote:
: Hi Netters:
: I have a problem...I have been trying to access the Internet via MacTCP and 
: Apple Remote Access without success.  I have heard that there is a way to 
: DYNAMICALLY assign the Mac an IP address to mirror that of the ARA server (use 
: the same IP number)  
 
: Is this possible?  If so, How do I do it from within the MacTCP Control panel?
: I'd also like to get an IP address for my Mac (permanent...Do I just talk to my 
: school's network administrator?)
 
: How difficult is it to get a deidcated internet line?  I mean an IP address?  
: Any help is most appreciated!  thanks in advance for any help!
 
: -Albert Kim


: -- 
 
: ajkim@miavx1.acs.muohio.edu
: akim@opus.mco.edu
 
: "Get a T.O. Baby!"
:       -D.V.


-----------[000575][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 16:33:40 GMT
From:      bgoldstein@jplsp2.jpl.nasa.gov (Bruce Goldstein)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: IP address assignment via MacTCP/ARA?

In article <2pomcm$a3p@dgis.dtic.dla.mil>, rsmullen@ida.org (Robert
Smullen) wrote:

> The bottom line, without getting into "deencapsulating" IP packets and other
> technical garbage:  You can't get there without some hardware.  If you have
> a connection to the network via either a GatorBox or a FastPath, you're OK
> but if your "server" is just a Mac with an ethernet card.  NO GO.
> 
> I ended up buying a LanRover /E 2.0 which supports TCP/IP and IP FOrwarding.
> It works like a champ.
> 
> Sorry I didn't have better news.
> 
> Been There- Done That,
> J.R. Smullen
> 
> 
> Michael Jordan-I WISH! (ajkim@miavx1.acs.muohio.edu) wrote:
> : Hi Netters:
> : I have a problem...I have been trying to access the Internet via MacTCP and 
> : Apple Remote Access without success.  I have heard that there is a way to 
> : DYNAMICALLY assign the Mac an IP address to mirror that of the ARA server (use 
> : the same IP number)  
 
> : Is this possible?  If so, How do I do it from within the MacTCP Control panel?
> : I'd also like to get an IP address for my Mac (permanent...Do I just talk to my 
> : school's network administrator?)
 
> : How difficult is it to get a deidcated internet line?  I mean an IP address?  
> : Any help is most appreciated!  thanks in advance for any help!
 
> : -Albert Kim
> 
> 
> : -- 
 
> : ajkim@miavx1.acs.muohio.edu
> : akim@opus.mco.edu
 
> : "Get a T.O. Baby!"
> :       -D.V.
There is a software solution put out by Sonic Systems, PowerBridge/TCP
and SuperBridge/TCP that encapsulates/deencapsulates TCPIP
into/from AppleTalk.  It runs on a Mac.  Have no personal experience
with this product, though.

-----------[000576][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 00:54:05 -0500
From:      Samuel.Leo@f1.n700.z6.ftn.air.org (Samuel Leo)
To:        comp.protocols.tcp-ip
Subject:   Access Novell from Unix with IPX protocol

Is there anyone known how to access a Novell file server from 
a SCO Unix with SCO IPX/SPX?
How can I get the detail of Novell Core Protocol in HK?

... Maximus 2.00
 # Origin: TAIC Maximus-CBCS - Home of HK PC User Group (6:700/1)

-----------[000577][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 17:06:27 GMT
From:      ollie@viper.dev.cdx.mot.com (Paul Oliveira)
To:        comp.protocols.tcp-ip
Subject:   Re: Adding compression to a 56k leased line?

zeeff@zip.eecs.umich.edu (Jon Zeeff) writes:

>I am using a 56k leased line for a tcp/ip over ethernet connection.  I'd like
>to increase the speed by adding compression.   Ideally this would all
>be in one box - ethernet router, compression, and csu/dsu.
 
>Does anyone know of reasonably priced solutions?  Black Box wants
>$2500 at each end just for compression and the csu/dsu.

We here at Motorola / Codex sell a dsu/csu w/compression also.  Our
price is around $2400, but ours has mux, remote front panel, lpda and our
own network management.

If you want more info on it send me an e-mail and I will gladly
get you all the info you need.

Paul Oliveira

-----------[000578][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 17:17:19 GMT
From:      peter@nmti.com (Peter da Silva)
To:        comp.protocols.tcp-ip,comp.unix.solaris,comp.unix.admin,comp.unix.misc
Subject:   Re: Help

In article <1994Apr27.201835.10629@martha.utcc.utk.edu>,
Sean Harp <harp@martha.utcc.utk.edu> wrote:
> The UNIX server has both an ethernet interface and an FDDI.  Because
> the network is bridged, the host has the same IP address on both
> interfaces.

It needs two IP addresses. An IP address belongs to a network interface,
not a system.
-- 
Peter da Silva                                            `-_-'
Network Management Technology Incorporated                 'U` 
1601 Industrial Blvd.     Sugar Land, TX  77478  USA
+1 713 274 5180                       "Hast Du heute schon Deinen Wolf umarmt?"

-----------[000579][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 94 22:23:07
From:      drw@runge.mit.edu (Dale R. Worley)
To:        comp.protocols.tcp-ip
Subject:   Data compression on leased lines

If I remember correctly, someone here wanted to know if there was
decent data compression available for leased lines.  I've recently
discovered an article in Network World (28 Mar 94) that a group of
vendors are developing a standard for data compression for 56kb/s
lines.  The intention is to build it into CSU/DSUs.  I assume that at
least some of the vendors in the group already offer data compression
for 56kb/s lines:
	Adtran
	Advanced Hardware Architectures
	AT&T Paradyne
	Conklin Instrument
	Cray Communications
	CrossComm
	DCP Research
	FastComm
	General DataComm
	Hayes Microcomputer Products
	IBM
	Integrated Network Corp.
	Magnalink Communications
	Motorola Codex
	Racal-Datacomm
	STAC, Inc.
	Telemax
	Transcend, Inc.
	TxPort
	Tylink
	UDS Motorola
	U.S. Robotics

Dale

Dale Worley		Dept. of Math., MIT		drw@math.mit.edu
--
... and thereof do I repent:  I only plucked an occasional flower when
I might have gathered an ample harvest of fruit -- such are the just
grounds for the regrets I have ...
-- D. A. F. Sade, "Dialogue between a Priest and a Dying Man"

-----------[000580][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 17:45:15 +0000
From:      pyers@pyers.demon.co.uk (Pyers Symon)
To:        comp.protocols.tcp-ip
Subject:   IP headers in UDP,TCP

When TCP or UDP are constructing the pseudo-headers that they use as a 
checksum, they incorporate the IP address in those headers. Since TCP & UDP
are at a higher level in the protocol suite that IP, how is this done ?

Sorry if this is a FAQ, I've been asked it by a friend!!
-- 
Pyers Symon

-----------[000581][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 00:39 -0400
From:      holdrege@eisner.decus.org (Matt Holdrege)
To:        comp.protocols.tcp-ip
Subject:   Need TCPdump for 5.4R2.10 of DG-UX

Does anyone know where the source of tcpdump is. Preferably something that
will compile on DG-UX.

Tanx!

-Matt@phs.com

-----------[000582][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 18:37:03 GMT
From:      Heidi_Epp@utpe.pe.utexas.edu (Heidi Epp)
To:        comp.protocols.tcp-ip
Subject:   need multicast info



I'm interested in experimenting with multicast, but am unsure how
to get started.   I'm quite familiar with socket programming, but
have yet to track down all the necessary steps to set up a multicast.
Can someone point me to a good reference (or exisiting code) for
writing multicast code to send between Unix boxes (Suns and HPs).

Thanks in advance.

-----------[000583][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 20:12:48 GMT
From:      phuang@stimpy.eecis.udel.edu (Po-I Huang)
To:        comp.protocols.tcp-ip
Subject:   What's the difference between kludge line mode and linemode of Telnet ?

 Hello :
  
     Have anybody know what's the difference bwtween kludge line mode(line-at-a-time) and linemode(real mode). Any information about these two modes of Telnet
would be appreciated.

   Thanks for your time.

-----------[000584][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 21:18:46 GMT
From:      cliffb@skeeter.bdb.com (Cliff Bedore)
To:        comp.protocols.tcp-ip,misc.forsale.computers.other
Subject:   Re: Bargains on Comer's 3 Volume TCP/IP set

Scott Austin (scotta@kije.gsfc.nasa.gov) wrote:
: Hi,
 
: The Library of Computer and Information Sciences (LCIS), is currently
: having a promo with Comer's 3 books.  You can save some bucks, since the
: Publisher's prices on these is $164!  This offer is open to any
: non-members (wish *I* could).  The scoop is this:
 
: * Join LCIS for $4.95 + $7.95(S&H) and receive:
:       
:   Internetworking with TCP/IP, Volume I          Publisher's Price: $60 
:      Principles, Protocols, and Architecture     2nd Edition
:      By Douglas E. Comer
 
:   Internetworking with TCP/IP, Volume II         Publisher's Price: $54
:      Design, Implementation, and Internals       1st Edition
:      By Douglas E. Comer and David L. Stevens
 
:   Internetworking with TCP/IP, Volume III        Publisher's Price: $50
:      Client/Server Programming and Applications  1st Edition
:      By Douglas E. Comer and David L. Stevens

...
...
: Scott No-Affiliation-With-LCIS Austin
: scotta@cnt.com


The only problem is that I joined in March (maybe Feb) and I'm still waiting
for two of the Volumes


Cliff



-----------[000585][next][prev][last][first]----------------------------------------------------
Date:      Thu, 28 Apr 1994 21:31:30 GMT
From:      mebane@ralvm12.vnet.ibm.com
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Problem with FTPing to RS6000 with Novell LWP for DOS...HELP!!!

In article <1994Apr27.234450.8704@hellgate.utah.edu> u-mwong%peruvian.cs.utah.edu@cs.utah.edu (Michael Wong) writes:
>   Hi...I've got a major problem that needs to be resolved ASAP and am
>hoping someone here could help me...
>
>
>-Mike
>--
>Michael Wong
>u-mwong@peruvian.cs.utah.edu
>University of Utah

I'll take a stab since I have worked on our ftp and ftpd server quite a
bit. The 425 imples that the server cannot open the data connection to
the client. Lots of possible reasons, but often because of an improper
socket setup.

I test often with a RISC6000 running AIX. The ftpd server on that
platform seems to be very compliant. It sounds like your problem is
related to the tcp or ip layers on the client side since both your
clients see the problem. If you have a sniffer, this would help a lot to
see what the packets look like.

Cummins A. Mebane III
DOS/Windows TCP/IP Development

-----------[000586][next][prev][last][first]----------------------------------------------------
Date:      28 Apr 1994 21:35:13 GMT
From:      nelson@crynwr.crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnet Masking Explanations

In article <9404261830.AA03475@mattox.digex.net> alatas@access.digex.net (Robert W. Mattox) writes:

   I hope that this is the correct forum for my query. I'm relatively
   new at the TCP/IP game. I have a good idea of network addressing,
   but I'm having a problem with the idea of subnetting. Could anyone
   point me to the appropriate RFC or other literature that will give
   me a comprehensive but well written explanation of the vagaries of
   the concept? I don't want something that is too technical since I'm
   a newbie, but I do want something that will explain, for example,
   using a host address of 99.3.80.1 with a subnet mask of
   255.255.255.0 on the same network as a host with an address of
   99.3.80.1 with a subnet mask of 255.255.254.0. Do they conflict? or
   are they considered to be different network addresses?

Here are some simple rules to follow:

Different hosts must always have different IP addresses.  No if, ands,
or buts.  Subnetting doesn't affect that.

Hosts on the same wire (possibly separated by repeaters or bridges
but on the same side of a router) must all be using the same subnet
mask, otherwise some of them won't be able to talk to others of them.

Logically and the value for each IP address on a subnet with the
subnet mask.  The result must be the same for each machine on the
subnet (otherwise, they're on different subnets).

You must use more subnet mask bits than your IP address class [1].
For 0.0.0.0 through 127.255.255.255, you must use at least 255.0.0.0.
For 128.0.0.0 through 191.0.0.0, you must use at least 255.255.0.0.
For 192.0.0.0 on up, you must use at least 255.255.255.0.

    [1] Unless your router can talk CIDR, when you can use fewer
    subnet mask bits than the class of the network address would dictate.

--
-russ <nelson@crynwr.com>      ftp.msen.com:pub/vendor/crynwr/crynwr.wav
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)    | Quakers do it in the light
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000587][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 01:34:51 GMT
From:      mike@cs.curtin.edu.au (Mike Schulze)
To:        comp.protocols.tcp-ip
Subject:   Re: Tapping the loopback network interface on SunOs 4.1.3

jcmurphy@acsu.buffalo.edu (Jeff Murphy) writes:

>Wes Howl 301-794-2043 S825 (howlw@niccolo.gsfc.nasa.gov) wrote:
>>How can I passively access the data flowing on the
>>loopback network interface know as "lo0" (SunOs 4.1.3)?
>>  
 
>i believe that you can't. because of the way sun's NIT device is written,
>the machine can not ''see'' traffic that *it* generates. 
 
>since the loopback device will always contain only traffic generated by
>the machine itself.. nit will never see anything (i.e. no traffic on the
>device). 
 
>try using tcpdump to capture outgoing traffic from your machine..
 
>unix% tcpdump src host [your machine] &
>unix% ping -s [some other machine]
 
>or try:
 
>unix% tcpdump -i lo0 
 
>you shouldn't see anything...
 
>hope this clears things up!

	This problem disappears if you are game enough to hack your kernel
and install bpf.  You don't need if_le.c -- just copy the kernel modules
from the MBONE stuff (they just happen to have the bpf hooks compiled
in), and follow the instructions for installing bpf (comes with the
tcpdump source).  I had to install:

	if_le.o,
	if_lo.o (if you want loopback tapping :-),
	if_ether.o,
	if_subr.o. (The last two are to satisfy the multicast support in if_le)

Here is a demo:

????# uname -a
SunOS ???? 4.1.3 2 sun4c
????# dmesg|grep bpf
bpf: le0 attached
bpf: lo0 attached
????# tcpdump-bpf -i lo0
tcpdump-bpf: listening on lo0
09:39:05.018837 ????.cs.curtin.edu.au.1025 > ????.cs.curtin.edu.au.659: udp 76
09:39:05.026117 ????.cs.curtin.edu.au.659 > ????.cs.curtin.edu.au.1025: udp 28
09:39:06.268808 ????.cs.curtin.edu.au.1025 > ????.cs.curtin.edu.au.659: udp 76
09:39:06.274810 ????.cs.curtin.edu.au.659 > ????.cs.curtin.edu.au.1025: udp 28
09:39:11.059497 ????.cs.curtin.edu.au.1025 > ????.cs.curtin.edu.au.659: udp 76
09:39:11.069184 ????.cs.curtin.edu.au.659 > ????.cs.curtin.edu.au.1025: udp 28
09:39:12.319486 ????.cs.curtin.edu.au.1025 > ????.cs.curtin.edu.au.659: udp 76
09:39:12.330377 ????.cs.curtin.edu.au.659 > ????.cs.curtin.edu.au.1025: udp 28
^C
8 packets received by filter
0 packets dropped by kernel


	From memory I found what I needed at UC Irvine (sorry, can't remember
the address).  Good luck.

								...mike
--
Mike Schulze
Department of Computer Science
Curtin University of Technology, Perth, W.A
mike@cs.curtin.edu.au

-----------[000588][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 01:47:16 GMT
From:      edward@e0sun29.ccl.itri.org.tw (Chao-Chi Yang 37h3100)
To:        comp.lang.c,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: Sources from book 'INTERNETWORKING WITH TCP/IP vol II'

Yoeri J. Apts (yjapts@info.vub.ac.be) wrote:
: I am currently reading INTERNETWORKING WITH TCP/IP vol II (by Douglas
: E. comer and David L. Stevens). This book contains a lot of sample code
: and so, being lazy like every programmer, I'm wondering if these sources
: are electronically available 'anywhere'?
 
: I am especially interested in de code concerning SNMP (Client-Server)
 
:  Yoeri Apts                         Tel: +32(0)2-6412977
:  yjapts@info.vub.ac.be              Fax: +32(0)2-6412870
:                                    ISDN: +32(0)2-9239936
:  Dept. INFO/TW - Digital Telecommunications Group.
:  Brussels Free University
:  Pleinlaan 2, 1050 Brussel
:  BELGIUM

You can look at 'arthur.cs.purdue.edu:/pub/dls'.

--
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#
| __ ___                                                |
|  |||           ´­´Â¦N                                 |
|__|||_          Chao Chi Yang (Edward)                 |
|                                                       |
|CCL-ITRI  						|
|                                                       |
|   TEL 	886-35-917178                           |
|   FAX 	886-35-820098                           |
|   Email       edward@e0sun3.ccl.itri.org.tw           |
|   Address     E000, Bldg. 11, 195 Sec. 4,             |
|               Chung Hsing Rd., Chutung, Hsinchu,      |
|               Taiwan 310, R.O.C.                      |
|                                                       |
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#

-----------[000589][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 02:32:58 GMT
From:      gwright@world.std.com (Gary R Wright)
To:        comp.unix.internals,comp.unix.misc,comp.unix.bsd,alt.books.technical,comp.protocols.tcp-ip,comp.unix.programmer
Subject:   Re: Book on BSD's network implementation

In article <2pk67g$1k8@irz401.inf.tu-dresden.de>,
Ulf Kieber <uk1@irz.inf.tu-dresden.de> wrote:
>I'm looking for a book, paper, commented code, etc. on the BSD
>kernel's implementation of networking.  I gotta port a new
>protocol family from SunOS to OSF/1 whitout most of the sources
>of OSF/1, and with no SunOS sources at all.

Although it won't help you in the short term, Rich Stevens and I have
teamed up to write TCP/IP Illustrated Volume 2: The Implementation.
The text will include the code, detailed commentary, figures, tables,
etc. describing the popular BSD implementation of the TCP/IP protocols.
We anticipate that Volume 2 will be available late this year.

We'll post a message when the table of contents and other information
becomes available via ftp from aw.com.

-----------[000590][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 09:33:33
From:      jasonh@jsb.co.uk (Jason Holloway)
To:        comp.protocols.tcp-ip
Subject:   POP3 server and IMAP capability..

Hello..

Anyone out there know the whereabouts of a POP3 server and an IMAP server for 
use on a Unix machine? Ideally sources and binaries. (Any Unix)

Please email me at: jasonh@jsb.co.uk

Many Thanks..

Regards, Jason.


-----------[000591][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 06:22:27 GMT
From:      terence@hitech.po.my (Terence Tan)
To:        comp.protocols.tcp-ip
Subject:   Tandem & TCP/IP

Can anybody suggest any companies which provide software/hardware
to get a Tandem to speak TCP/IP??
-- 
------------------------------------------------------------------------
A4000/040/Internet: terence@hitech.po.my(Malaysia)     Terence Tan
120megs /               /*//_\                         "Amigo Man" 
6 megs /              \*//     \miga                  Mostly Harmless .. 

-----------[000592][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 06:55:04 GMT
From:      etxmesa@eos.ericsson.se (Michael Salmon)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP dial-up - async vs sync

In article <60.16961.4471.0N19C2E3@canrem.com>
grant.porter@canrem.com (Grant Porter) writes:
|> Can anyone tell me the pros and cons of using synchronous communications versus
|> async when using TCP/IP? Does async allow multiple users using TN3270 to
|> connect over one modem?

I don't think that there are any advantages in using async
communications apart from simplicity, if you can use sync then do so.

Multiple users aren't determined by the link level implementation, it
is TCP/IP or more specifically ppp or slip that enables multiple
connections on one modem connection.

-- 

Michael Salmon

#include	<standard.disclaimer>
#include	<witty.saying>
#include	<fancy.pseudo.graphics>

Ericsson Telecom AB
Stockholm

-----------[000593][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 94 12:00:14 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   Telnet option negotiation


I'm starting to get the feeling that there's more to Telnet option
negotiation that meets the eye; that it's not as straightforward as
pure command-reply pairs, particularly with the ECHO option.  (Are
there two ECHO options?  Client-echo and server-echo?  Are they
automatically mutually exclusive?)

Does anyone know of a good description of the minimum option negotiation
features that Telnet clients and daemons must support?

-- Bob Stein                            Internet mail: stein@gcomm.com
 ______________________________________________________________________
|                                                                      |
|  Galacticomm, Inc.                            (305) 583-5990 (voice) |
|  4101 SW 47 Avenue, Suite 101                 (305) 583-7846 (FAX)   |
|  Ft Lauderdale, Florida, USA, 33314           (305) 583-7808 (BBS)   |
|______________________________________________________________________|

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000594][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 94 12:00:15 EST
From:      stein@gcomm.com
To:        comp.protocols.tcp-ip
Subject:   External Clock for Network Time Protocol


IP>We're looking for an external clock (hardware) to interface to a server
IP>to provide time synchronization using something like network time protocol.

You might be interested in a twist to the approach you describe.  The
November 93 issue of Boardwatch magazine describes how to get the time
via the Internet from the National Institute if Standards and
Technology, formerly the National Bureau of Standards.  These are the
guys with the cesium clocks that keep much better time than most mortals
would ever need.  (By the way, the thing called a second is now DEFINED
in terms of the peculiar little rhythms of cesium atoms.)  Apparently
they've been doing more than just changing their name.

According to the article, if you "telnet 132.163.135.130 13" you get a
string with the current time and then get kicked off immediately.  The
same site has an FTP server with programs and source code to help you
with your timekeeping, including PCTIME in the /pub/acts directory and
NISTIMEN in the /pub/daytime directory.

-- Bob Stein                            Internet mail: stein@gcomm.com
 ______________________________________________________________________
|                                                                      |
|  Galacticomm, Inc.                            (305) 583-5990 (voice) |
|  4101 SW 47 Avenue, Suite 101                 (305) 583-7846 (FAX)   |
|  Ft Lauderdale, Florida, USA, 33314           (305) 583-7808 (BBS)   |
|______________________________________________________________________|

===============================================================================
| ... The Galacticomm Demo System - 305.583.7808 - Home of The Major BBS! ... |
===============================================================================


-----------[000595][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 09:26:07 GMT
From:      pgw@datcon.co.uk (Peter Waters)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for RFC distrib. site

Mark C. Stout (mcstout@netcom.com) wrote:
: jmr@ibm1.nynexst.com wrote:
: : I am looking for a few RFC documents:
 
: : 	1. Is there a common distribution site privided by the ITF ?
: : 	2. If not, can these RFC be purchased and from where ?
 
: : Any hint would be helpful. Thanks in advance.
 
: I get what's on-line at anonymous ftp at ds.internic.net /rfc.  It has what
: looks to most, if not all the RFC's that are available on-line.  Latest one
: is RFC1609.  They are in ASCII text format, though a few of them are also in
: postscript.
 
: If anyone knows where the off-line RFC's can be had, I'd like to complete my
: collection, even if they are obsoleted by newer RFC's.
 
: ~
: ~
You can obtain RFC's from nic.ddn.mil using ftp in the directory /rfc

RFC 1600 contains a list of all rfc's.
: ~
: Peter

-----------[000596][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 17:00:03 PDT
From:      burnhosp@vmsmail.gov.bc.ca
To:        comp.protocols.tcp-ip
Subject:   Re: Novell and IPX - Tunneling

The RFC you are looking for is 1234.

Good luck.

Chris Hansen
Burnaby Hospital Society
-- 
Disclaimer: The opinions and statements contained in this posting are the sole
responsibility of the author and have not in any way been reviewed or approved
by my employer or any network service.

-----------[000597][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 08:59:01 +0100
From:      mark@plato.ucsalf.ac.uk (Mark Powell)
To:        comp.unix.sys5.r4,comp.sys.att,comp.protocols.tcp-ip
Subject:   Re: rdate replacement for SVR4 time-adjustment?

In article <WCS.94Apr25153200@anchor.att.com>,
Bill Stewart +1-510-484-6204 <wcs@anchor.ho.att.com> wrote:
>SVR4 has system calls to adjust the date/time gradually.
>Does anybody have a version of rdate (the TCP/IP remote date-setter)
>which knows how to use this?  Alternatively, is there any free rdate
>code around - I'm currently using a Wollongong-based TCP/IP which
>doesn't have rdate, though it does answer rdate requests properly.

xntp which uses the network time protocol, is probably the best way
to go. It will sync. your time with any sites you choose over the
Internet. We've only recently got an IP connection to the Internet.
Even before that I installed xntp v3.3 on this Dell SVR4 box and 
a Sparc server here, because the SVR4 box was losing minutes every
day. This enabled the Dell box to sync. it's time with the Sparcserver
which had a very good RTC. Now we have an IP connection the sparcserver
syncs. it's time with an upstream time server and all machines on the
campus sync with it. It's a very neat solution. All UNIX boxes on the
campus have the same time.

>	Thanks;  Bill
># Bill Stewart       AT&T Global Information Solutions (new name for NCR!)
-- 
System Administrator

root@plato.ucsalf.ac.uk		root%plato.ucsalf.ac.uk@nsfnet-relay.ac.uk
Work:	+44 61 745 3376

-----------[000598][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 17:41:21 PDT
From:      burnhosp@vmsmail.gov.bc.ca
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer/monior" FREEWARE

We use "The Beholder" and "The Gobbler", both written by the
Data Network Performance Analysis Group (DNPAP) of the Delft Univer
University of Technology. Both versions are FREE and work 
VERY well! They are software-only products and require the
following hardware configurations:
8086, 80286, 80386, 80486 (SX or DX)
recommend at least 286/20 MHz for busy networks
Hard disk drive (10Mb is fine)
Video display (recommend VGA colour, if possible)
Network Interface Card
recommend WD8003, NE2000 or 3COM)
Keyboard required only for installation and startup
Software: MS-DOS 3.x or higher

The data can be obtained three different ways:
- by looking at the screen display
- by requesting the data as SNMP variables
- by using TFTP to collect files containing the data

A standard LAN can contain several of these monitoring stations,
each containing several network interfaces.

The main design goals of these two products were:
- minimum loss of packets
- continuous operation
- compliance to standards wherever possible
- ease of use
capable of monitoring all traffic on one or more ethernet segments. 
They can be easily integrated into SNMP-based management
environments like Sun Net Manager(tm)

Info on how to get these products can be obtained by E-mailing
JPMvOorschot@et.tudelft.nl (Project leader)
His name is Jan van Oorschot

Hope this helps...
Chris Hansen
Burnaby Hospital Society

These are my views... you can share them with me if you like!!!
-- 
Disclaimer: The opinions and statements contained in this posting are the sole
responsibility of the author and have not in any way been reviewed or approved
by my employer or any network service.

-----------[000599][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 17:51:41 PDT
From:      burnhosp@vmsmail.gov.bc.ca
To:        comp.protocols.tcp-ip
Subject:   Re: Help for beginner

This is handled by the IANA (Internet Assigned Numbers Authority)
and is farmed out by the IR (Internet Registry) to the DDN NIC
Defense Data Network Network Information Centre. You can FTP the
info from NIC.DDN.MIL in the netinfo directory. The files you may
want to get are:
internet-number-template.txt
domain.template.txt
   and
in-addr-template.txt

Chris Hansen
Burnaby Hospital Society

These are my views... you can share them with me if you like!!!
-- 
Disclaimer: The opinions and statements contained in this posting are the sole
responsibility of the author and have not in any way been reviewed or approved
by my employer or any network service.
P.S. That should read    domain-template.txt   ...

-----------[000600][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 11:10:18 GMT
From:      scoggin@delmarva.com (John K Scoggin Jr)
To:        comp.protocols.tcp-ip
Subject:   Re: Tandem & TCP/IP

In article 3542@hitech.po.my, terence@hitech.po.my (Terence Tan) writes:
> Can anybody suggest any companies which provide software/hardware
> to get a Tandem to speak TCP/IP??

Why not ask Tandem?  We have TCP/IP software from Tandem on our VLX system.
Warning, though - it is really a strange (and in some ways, broken) implementation
compared to any Unix system...

	- John


---
+---------------------------------------------------------------------+    
|  John K. Scoggin, Jr.			Email: scoggin@delmarva.com   |
|  Supervisor, Network Operations              scoggin@ee.udel.edu    |
|  Delmarva Power & Light Company       Phone: (302) 451-5200         |
|  500 N. Wakefield Drive               NOC:   (800) 388-7076         |
|  Newark, DE 19714-6066		Fax:   (302) 451-5321         |
|  					Ham:   N3??? (real soon now!) |
|  The opinions expressed are not those of Delmarva Power, simply the |
|  product of an over-active imagination...                           |
|  Just a pothole-patcher on the Information Superhighway.            |
+---------------------------------------------------------------------+



-----------[000601][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 11:16:20 GMT
From:      perretc@eiga.unige.ch (Perret Cédric)
To:        comp.protocols.tcp-ip
Subject:   HELP: PPP AND WINSOCK please





Hello, i've some questions about PPP

1) It is possible to use Mosaic over PPP ?

2) IF yes, how (is chameleon a good idee) ?

3) I'm trying with Chameleon, but i'm wonder if all my applications that worked with Trumpett
   winsock will continue to Work (Mosaic, Winftp, Trumpett NewsReader, Xwin...)

4) Will it work on both PPP and ethernet (it is the version taken from netmanage, but i'm 
   waiting for the version that comes with "The Windows Internet Tour Guide) ?

5) I've also heard about a driver called etherPPP.zip. Will it allow me to run Mosaic over 
   PPP? 

6) If yes, will it work by just replacing my normal packet driver by etherppp and let 
   the trumpett winsock, winpkt ?

7) My only purpose is to use PPP with applications that works with trumpett winsock. Coud
   you tell wich product is better to use and how to install it. (what should i use with this
   product: packet driver , NDIS ...)

Please, answer me. It becomes more and more important each days.
If possible E-mail me, but i'll have a look on the groups.
Thanks.
				PERRETC@EIGA.UNIGE.CH
				PERRET CEDRIC

    

-----------[000602][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 13:10:34 GMT
From:      russ@madhaus.uucp (Russell Sutherland)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   DNS and variable length subnets

Lets assume we have a class C network:  a.b.c.0 that is subdivided into 4
equal sized subnets (mask 255.255.255.192):

        cat     a.b.c.0
        dog     a.b.c.64
        bird    a.b.c.128
        insect  a.b.c.192

Where the "animal names" will correspond to the domain name space
for that particular network. So for example ip address: a.b.c.3
might have a fully qualified domain name: calico.cat.animal.zoo

Lets also assume that one machine on each of the networks
will provide authoritative name service for that network.

Question: What do the named.boot files look like on each of these four name
servers? In particular, the tricky part seems to be the in-addr.arpa stuff
which appears to me only to work if divided on the natural byte boundaries of
the ip address.

Do these four bits of contiguous ip address space have to served by one (and
not four) name server(s)?

---

-- 

Russell Sutherland		Bell:	   (416)-978-0470
UTCC, Network Development 	Fax:	   (416)-978-6620
University of Toronto		E-mail:    russell.sutherland@utoronto.ca

-----------[000603][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 13:57:28 GMT
From:      brucej@tau-ceti.isc-br.com (Bruce T. Jorgens)
To:        comp.protocols.tcp-ip
Subject:   Re: Adding compression to a 56k leased line?

In article <2plp1t$eos@zip.eecs.umich.edu> zeeff@zip.eecs.umich.edu (Jon Zeeff) writes:
>From: zeeff@zip.eecs.umich.edu (Jon Zeeff)
>Subject: Adding compression to a 56k leased line?
>Date: 27 Apr 94 13:24:13 GMT
 
>I am using a 56k leased line for a tcp/ip over ethernet connection.  I'd like
>to increase the speed by adding compression.   Ideally this would all
>be in one box - ethernet router, compression, and csu/dsu.
 
>Does anyone know of reasonably priced solutions?  Black Box wants
>$2500 at each end just for compression and the csu/dsu.

I recently attended an Internetworking seminar put on by Cisco.  They are 
working on a 3 differentlevels of compression for use with their routers.  The 
three levels were: header compression, link compression and data compression.  
Header and Link compression are complete, they expect to ship product with 
data compression in the late summer '94 timeframe.

This wouldn't be a single box solution, but a Cisco 2501 w/ IP router s/w 
lists for $2500 and a CSU/DSU should be about $500.

Maybe someone from Cisco would like to verify this information and give an 
actual date?


#####  ###   ####   Bruce T. Jorgens            brucej@tau-ceti.isc-br.com
  #   #   #  #   #  Internet On-Ramp Inc.       (509) 927-RAMP
  #   #   #  ####   E. 3724 11th
#####  ###   #   #  Spokane, WA 99202                         




-----------[000604][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 94 22:31:01 -0500
From:      ajkim@miavx1.acs.muohio.edu (Michael Jordan-I WISH!)
To:        comp.sys.mac.system,comp.sys.mac.comm,comp.protocols.tcp-ip
Subject:   ARA->MacTCP->InterNet?? HOW??


Hello: 
This is a FOLLOWUP to my recent questions about ARA and MacTCP...Thanks to all 
who helped me out these past few days...I've been able to get a few more 
answers from our network administrator here (He seems reluctant to reveal info 
to me...he think a lot of this stuff is "classified" or something!  can you 
believe that!)

Anyways, here's what I've got at MY SYSTEM NETWORK.
(1) Operating system is DEC Pathworks
(2) There are numerous AppleTalk ZONES, but I'm uncertain where the GATEWAY box
    is (which zone) and if it even exists!
(3) The ARA server is running Shiva LanRover/L OS.  I have heard that this may
    be able to distribute IP addresses as needed.  

MY QUESTION is:
How quickly does the IP address appear in the MacTCP control panel when it's 
set to the proper ZONE?  Will the numbers just appear (IF I set it to SERVER 
mode in the Control Panel) like magic!

Also, Are Domain Name Servers only located at my local node or can I use any on 
the Internet?  

Any answers or help is greatly appreciated!  Thanks in advance!

-Albert

ajkim@miavx1.acs.muohio.edu
akim@opus.mco.edu



-----------[000605][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 14:47:36 GMT
From:      jeff@zis.ziff.com (Jeff Macdonald)
To:        comp.protocols.tcp-ip,misc.forsale.computers.other
Subject:   Re: Bargains on Comer's 3 Volume TCP/IP set

However the three books could not total over $100 or $105.




In article <Cozn7C.30I@cjbsys.bdb.com> cliffb@skeeter.bdb.com (Cliff Bedore) 
writes:>From: cliffb@skeeter.bdb.com (Cliff Bedore)
>Subject: Re: Bargains on Comer's 3 Volume TCP/IP set
>Date: Thu, 28 Apr 1994 21:18:46 GMT
 
>Scott Austin (scotta@kije.gsfc.nasa.gov) wrote:
>: Hi,
 
>: The Library of Computer and Information Sciences (LCIS), is currently
>: having a promo with Comer's 3 books.  You can save some bucks, since the
>: Publisher's prices on these is $164!  This offer is open to any
>: non-members (wish *I* could).  The scoop is this:
 
>: * Join LCIS for $4.95 + $7.95(S&H) and receive:
>:       
>:   Internetworking with TCP/IP, Volume I          Publisher's Price: $60 
>:      Principles, Protocols, and Architecture     2nd Edition
>:      By Douglas E. Comer
 
>:   Internetworking with TCP/IP, Volume II         Publisher's Price: $54
>:      Design, Implementation, and Internals       1st Edition
>:      By Douglas E. Comer and David L. Stevens
 
>:   Internetworking with TCP/IP, Volume III        Publisher's Price: $50
>:      Client/Server Programming and Applications  1st Edition
>:      By Douglas E. Comer and David L. Stevens
 
>...
>...
>: Scott No-Affiliation-With-LCIS Austin
>: scotta@cnt.com


>The only problem is that I joined in March (maybe Feb) and I'm still waiting
>for two of the Volumes


>Cliff






Jeff Macdonald
Ziff Information Services
Ziff Communications
10 President's Landing
Medford, MA  
jeff@zis.ziff.com

-----------[000606][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 16:38:49 GMT
From:      art@acc.com (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: IP headers in UDP,TCP

In article <767555115snz@pyers.demon.co.uk> pyers@pyers.demon.co.uk writes:
>When TCP or UDP are constructing the pseudo-headers that they use as a 
>checksum, they incorporate the IP address in those headers. Since TCP & UDP
>are at a higher level in the protocol suite that IP, how is this done ?

Well, TCP or UDP better know who the packet is being sent to, so it already
knows the dest address.  If TCP or UDP doesn't provide the source IP address
to IP, there is usually a routine to call to ask IP what source address it
going to use for a given dest IP address.

Art


-----------[000607][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 16:47:53 GMT
From:      gek@ncn.uucp (Geir Kristensen)
To:        comp.protocols.tcp-ip,sni.svr4
Subject:   Loosing connection telnet/SVR4

[ Article crossposted from comp.protocols.nfs ]
[ Author was gek@ncn.uucp ]
[ Posted on Fri, 29 Apr 1994 16:45:57 GMT ]


When using PCNFS 4.0a Telnet or Emulators on top of PCNFS4.0s like JSB
or GLINK we experience that we loose connection after about 2Hour10minutes of 
no activity like just standing in the shell or a application.

This happens when communicating with some SVR4 hosts.
When testing from other  hosts to this SVR4 hosts or from other telnet 
clients the connection is not droped.

Any ideas about this folks ?

Sorry if this is a FAQ.

Geir
-- 
Geir Kristensen        ! siemens !
+47 22749558           ! nixdorf !     (These are my opinion's not SNI's)
gek@siemens-nixdorf.no              nerv: kristensen.fbu[@sni.de|@sni-usa.com]
-- 
Geir Kristensen        ! siemens !
+47 22749558           ! nixdorf !     (These are my opinion's not SNI's)
gek@siemens-nixdorf.no              nerv: kristensen.fbu[@sni.de|@sni-usa.com]

-----------[000608][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 16:48:12 GMT
From:      leonard@telcom.arizona.edu (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP dial-up - async vs sync


In article <2pqb08$ob0@erinews.ericsson.se>, etxmesa@eos.ericsson.se 
(Michael Salmon) writes:

||> Can anyone tell me the pros and cons of using synchronous communications versus
||> async when using TCP/IP? Does async allow multiple users using TN3270 to
||> connect over one modem?
|
|I don't think that there are any advantages in using async
|communications apart from simplicity, if you can use sync then do so.

No advantages to asynch?  How about price, speed, and availability?

Most all computers come with asynch ports built in.  Adding in
a synch port may be impossible, or else will cost hundreds of 
dollars.  Asynch comm servers cost as little as $100 per port.
A synch comm server [router] typically will cost $1000 a port
or more.

Asynch links can easily take advantage of V.42bis compression
to deliver 57.6Kbps - 115.2Kbps of throughput, for only
$200-500/modem.  But compressing synch modems are MUCH more 
expensive, and I doubt they can put out the bandwidth.

Once upon a time, synch was preferable to asynch, because it
avoided the 20% overhead wasted by asynch's framing bits.  But
in the modern era of V.42, "asynch" modem connections actually
use synch framing over the phone line, so none of this 
bandwidth is wasted.

(followups to comp.dcom.modems.)

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
  \ Don't lock yourself into open systems. /


-----------[000609][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 17:00:34 GMT
From:      leonard@telcom.arizona.edu (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: packet losses, WHY


In article <2pp72e$jem@acme.gatech.edu>, gt0567a@prism.gatech.edu (Mathieu Claude Hans) writes:
|
|-Transmission using the UDP protocol:
| 
| 
|I beleive all the packet losses which occur on Internet are
|because of:
|        - congestions at routers and at the host
|        - error transmission, which in the case of UDP protocol
|implies dropping the packet.

Only if UDP checksumming is turned on, or if the error occurs
in the header.

|        - expiry of time to live because of bad routing.
|
|Are there any other reasons why packets are lost on the network?

Lessee ...

- Filters installed on routers (I put a filter on my router which
  silently discards packets you send to my TFTP port)

- Vanished bits (I powercycled my bridge, and the frames that 
  were in it evaporated.)


|What are the percentages of each - is it 80% loss because of 
|congestion, 10% because of error transmission ...? -

Congestion: 			72.14%
Errored packets delivered: 	 2.37%
TTL expired due to bad routing:	 1.14%
Routing filters:		 4.44%
Vanished bits:			14.72%
Poltergeists:			 0.03%

(total may not add up to 100% because I just made up these numbers
anyway)

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
  \ Don't lock yourself into open systems. /

-----------[000610][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 17:19:15 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   Re: Help required for tcp-ip vs udp comparison

mojumdar@fooba.ml.com (Mainak Mojumdar) writes:

>Can anybody comment on the advantages and disadvantages between the usage
>of tcp/ip and udp for a client/server implementation on a Lan environment.
 
>What concerns me the most is the connection time required in a tcp/ip 
>approach, in a 300 user client/server approach, if for every transaction
>a connection has to be opened and dropped subsequent to the completion of
>the transaction.
 
>Any experiences in using udp for a serious application.


Well, first of all you don't have to open and close the TCP
connection for every transaction.  You could have the connection
kept open.

You might also consider using RPC instead of TCP or UDP.  RPC is 
implemented on top of both, but gives you an easier interface for
client/server applications.

One issue for deciding between TCP and UDP is whether you want the 
IPC channel you are using between the client and server to be 
reliable.  UDP is unreliable, TCP is supposed to be reliable.
If all of your clients and servers are on a LAN (such as an Ethernet)
they the reliablity under light to moderate loads should be about the
same for UDP and TCP.  I.e. not many packets are lost on a single
Ethernet string that is not under high load.

It probably makes sense to look at _Internetworking with TCP/IP Volume III_
which is all about client/server computing with the IP protocols.

Hope this helps.

Later,
George


-- 
gnn@netcom.com

Gentelmen, I will not have you fighting in the War Room.
					--- The President in Dr. Stragelove

-----------[000611][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 17:23:00 GMT
From:      gnn@netcom.com (George Neville-Neil)
To:        comp.unix.internals,comp.unix.misc,comp.unix.bsd,alt.books.technical,comp.protocols.tcp-ip,comp.unix.programmer
Subject:   Re: Book on BSD's network implementation

gwright@world.std.com (Gary R Wright) writes:

>In article <2pk67g$1k8@irz401.inf.tu-dresden.de>,
>Ulf Kieber <uk1@irz.inf.tu-dresden.de> wrote:
>>I'm looking for a book, paper, commented code, etc. on the BSD
>>kernel's implementation of networking.  I gotta port a new
>>protocol family from SunOS to OSF/1 whitout most of the sources
>>of OSF/1, and with no SunOS sources at all.

SOrry, I missed the initial posting.  YOu should probably look
at the Net BSD sources, available on many sites, including
gatekeeper.dec.com.

Also pick up _The Design and Implementation of the 4.3 BSD Unix
Operating System_ by Leffler, Karels, McKusick and Quarterman.
This is THE book on BSD Unix.

Later,
George

--
gnn@netcom.com

Gentelmen, I will not have you fighting in the War Room.
					--- The President in Dr. Stragelove

-----------[000612][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 94 18:57:53 GMT
From:      ftlofaro@unlv.edu (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   Re: What constitutes a duplicate TCP packet?

In article <cearley_k.33.000EC06D@wizard.colorado.edu> cearley_k@wizard.colorado.edu (Kent Cearley) writes:
>We have a problem with a host sending two tcp acknowledgements... one with 
>WIND=0 and no data, followed immediately with a WIND=256, also no data. Both 
>the sequence numbers and ack numbers are the same on both packets. The problem 
>is, a client (FTP Inc.'s WTN3270) sees the first packet, assumes it cannot 
>transmit, and appears to be discarding the second packet without the WIND=256 
>registering.
>
>FTP Inc. claims the RFC specifies if the SEQ# are the same its a duplicate 
>packet. If so, how is this situation handled? All of our other clients 
>recognize the second packet as advertising an available window. 
>
>

How else is it supposed to update the window if it does not have any data 
to send? The SEQ satys the same. The ACK can stay the same too. Imagine this

TCP A sends packet
TCP B ACKS it
TCP B increases window (perhaps since attached process finally read some 
data from the sockets, but way after the acceptable time for an ACK, so 
the ACK had to be sent before)

2nd packet will have same SEQ/ACK and a new window. FTP software's client 
is brain damaged to ignore it. The data would be a dup, but the window 
info shouldn't be thrown out, it is more recent (thus more accurate). 
That isn't very interoperable (I believe the RFC's also say "be conservative 
in what you do, be libreal in what you accept". Otherwise you won't be 
interoperatable).



-----------[000613][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 19:36:12 GMT
From:      khm@powell.cs.jt.dk (Kurt Hangaard Marcussen)
To:        comp.protocols.tcp-ip
Subject:   Re: REPOST: What is NetBIOS?

timur@seas.gwu.edu (Timur Tabi) writes:

>Sorry to post this twice, but I really need an answer quick, and no one's
>responded to me yet.
 
>>Subject says it all.  I know it has something to do with peer-to-peer,
>>since IBM's TCP/IP for OS/2 says that it's NetBIOS provides p2p service.
>-- 
 
>------------------------------------------------------------------ Timur Tabi
>Contributing Editor for "OS/2 Monthly"        Internet:    timur@seas.gwu.edu
>                                              Fidonet: Timur Tabi @ 1:109/347

Hi Timur

NetBIOS is an old dog! It's the IBM PC's standardized API (Application
Programming Interface) to networks. This means, that if you have a PC
with a network adaptor of some kind and possibly some kind of driver
software on top of it, then you could have a NetBIOS interface software
on top of this driver giving you the NetBIOS API (the NetBIOS API may
also be provided directly in the driver software). Given this NetBIOS
API, you can now use all NetBIOS compliant networking software with
your PC/adaptor configuration (try it before you believe it!). The
network interface provided by the NetBIOS API is quite simple, so you
may also write your own applications for it.

Good luck

Kurt

--
    _/    _/  _/    _/  _/     _/  Name:   Kurt H. Marcussen, P-117
   _/_/_/    _/_/_/_/  _/_/ _/_/           Jydsk Telefon a/s, Sletvej 30
  _/  _/    _/    _/  _/  _/ _/            8310 Tranbjerg J., DK-Denmark
 _/    _/  _/    _/  _/     _/     e-mail: khm@cs.jt.dk

-----------[000614][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 20:05:11 GMT
From:      roy@mchip00.med.nyu.edu (Roy Smith)
To:        comp.protocols.tcp-ip
Subject:   Destination Unreachable

	Can somebody explain the following to me?  The following three
machines are all on the same IP subnet (possibly with an ethernet bridge
between some of them, but that shouldn't matter): 128.122.205.115 is a
DECstation running Ultrix 4.2a, 128.122.205.96 is an SGI Iris running some
flavor of IRIX, and 128.122.205.14 is a Macintosh which isn't running
anything because it's turned off at the moment.  128.122.135.9 is another
SGI on another subnet (connected to another port on the same Cisco router,
if that matters).

	When I ping from 205.115 to 205.14, I get Desination Unreachables
and I don't understand why (see ping output below).  One came from 127.0.0.1
(i.e.  the loopback interface), and others from the two SGI's (205.96 and
135.9).  What does this mean?  What I expected was to just not get any
response from the powered-off host, and can't figure out what the Dest
Unreachables mean and why I'm getting them.

mchip00> ping -v 128.122.205.14
PING 128.122.205.14: 56 data bytes
36 bytes from 127.0.0.1: icmp_type=3 (Dest Unreachable)
x00: x00240045
x04: x00003fd8
x08: x000001ff
x0c: x0100007f
x10: x0100007f
x14: xc1b20303
x18: x00000000
x1c: x007b0045
x20: x00003fd7
x24: x0000111e
x28: x0100007f
x2c: x0100007f
icmp_code=3
36 bytes from 128.122.205.96: icmp_type=3 (Dest Unreachable)
x00: x00240045
x04: x00003d2b
x08: x000001ff
x0c: x60cd7a80
x10: x73cd7a80
x14: x41120303
x18: x00000000
x1c: x00800045
x20: x0000df3f
x24: x0000111e
x28: x73cd7a80
x2c: x60cd7a80
icmp_code=3
36 bytes from 128.122.135.9: icmp_type=3 (Dest Unreachable)
x00: x00240045
x04: x0000ba9b
x08: x000001fe
x0c: x09877a80
x10: x73cd7a80
x14: xf4620303
x18: x00000000
x1c: x007b0045
x20: x0000fc3f
x24: x0000111d
x28: x73cd7a80
x2c: x09877a80
icmp_code=3
36 bytes from 128.122.205.96: icmp_type=3 (Dest Unreachable)
x00: x00240045
x04: x00003d43
x08: x000001ff
x0c: x60cd7a80
x10: x73cd7a80
x14: xaf610303
x18: x00000000
x1c: x007c0045
x20: x00001940
x24: x0000111e
x28: x73cd7a80
x2c: x60cd7a80
icmp_code=3
36 bytes from 128.122.205.96: icmp_type=3 (Dest Unreachable)
x00: x00240045
x04: x00003d49
x08: x000001ff
x0c: x60cd7a80
x10: x73cd7a80
x14: xb5610303
x18: x00000000
x1c: x007c0045
x20: x00007840
x24: x0000111e
x28: x73cd7a80
x2c: x60cd7a80
icmp_code=3
^C
----128.122.205.14 PING Statistics----
15 packets transmitted, 0 packets received, 100% packet loss
-- 
Roy Smith <roy@nyu.edu>
Hippocrates Project, Department of Microbiology, Coles 202
NYU School of Medicine, 550 First Avenue, New York, NY 10016
"This never happened to Bart Simpson."

-----------[000615][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 20:38:57 GMT
From:      dennis@cauchy.math.nwu.edu (Dennis Director)
To:        comp.protocols.tcp-ip
Subject:   Can I test ethernet card from off net.


I have a NE2000 compatible card installed in a
Linux box, running Linux 1.0.
I don't have a net connection in my office yet.

Is there some way to configure and test the
installation up to the card (loopback at card?)
so that when I get the connection I'll be that
much closer to having things work.
More importantly, I like to test systems in
modules.  Thanks for any ideas,  dennis@math.nwu.edu


-----------[000616][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 21:40:24 GMT
From:      Ian_Gilchrest@walnut.prs.k12.nj.us (Ian Gilchrest)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP "sniffer" software - freeware anywhere?

In article <xK2PKXs.aeayrs@delphi.com>, Allan Eayrs <aeayrs@delphi.com>
wrote:

> Aaron Leonard <leonard@telcom.arizona.edu> writes:
>  
> >|I have heard of "freeware" TCP/IP "netwatcher" software.  Does anyone know
> >|where I can find such a thing?  It would be EXTREMELY helpful to me right
> >|now.
>  
> I'm running into the same situation.  If the software is written for a DOS or
> DOS-able environment, I could sure use it.
>  
> Allan Eayrs
> MKRail Corporation
> Boise ID
 
-- 
For macs, there are programs such as TCPWatcher and Appletalk Manager,
which are both shareware and availiable at any info-mac ftp site.  For DOS,
I believe there is a program called netwatch that comes with shareware PCIP
software.

-----------[000617][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 21:50:11 GMT
From:      ganesh@TFS.COM (Ganesh Vaidee)
To:        comp.protocols.tcp-ip
Subject:   Help for beginner

Hello,
	I have a very basic question about IP addressing. I know that
	if the IP address of my PC is 170.145.99.4, then the
	subnet that i am assigned to is 99.4 and I am on class B network.
	Now the question, who determines the class and the overall 
	net id that i have got, that is who determines the number 170.145 in
	the above case.

	Thanx a lot, u can send replies to ganesh@tfs.com



-----------[000618][next][prev][last][first]----------------------------------------------------
Date:      29 Apr 1994 21:51:17 GMT
From:      cricket@nsr.hp.com (Cricket Liu)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: DNS and variable length subnets

Russell Sutherland (russ@madhaus.uucp) wrote:

: Do these four bits of contiguous ip address space have to served by one (and
: not four) name server(s)?

You could also delegate at the last octet, though that's messy.  In other
words, you could delegate d.c.b.a.in-addr.arpa from the c.b.a.in-addr.arpa
name server.  That implies, however, that you've got as many subdomains as
you do hosts whose IP addresses begin with a.b.c.  Yech.

--
cricket

cricket@hp.com / Hewlett-Packard Professional Services / Englewood, CO

-----------[000619][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 23:00:45 GMT
From:      ganesh@TFS.COM (Ganesh Vaidee)
To:        comp.protocols.tcp-ip
Subject:   Subnet mask

Is it possible to have a subnet mask of 255.255.0.0 
Does this mean that there are o bits for subnet id and 0 bits for host id.
How is that possible. help. thanx

-----------[000620][next][prev][last][first]----------------------------------------------------
Date:      Fri, 29 Apr 1994 23:01:54 GMT
From:      etolrb@etn.ericsson.se
To:        comp.protocols.tcp-ip
Subject:   API to FTP client


I need an API to a FTP client (ftp_get(..), ftp_put(..) etc.) for my 68K based 
embedded system that is to be run under the pSOS+ RTOS which has a TCP/IP stack 
with FTP but requires a console to run the FTP client. 

I am evaluating my different options.

Anyone out there with any good advice ?.

I will highly appreciate your opinions.	

-----------[000621][next][prev][last][first]----------------------------------------------------
Date:      30 Apr 1994 00:03:03 GMT
From:      mkail@mars.pl.uhc.com (Mike Kail)
To:        comp.lang.c,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: Sources from book 'INTERNETWORKING WITH TCP/IP vol II'

In article <2ppov4$107@news.csie.nctu.edu.tw>, edward@e0sun29.ccl.itri.org.tw (Chao-Chi Yang 37h3100) writes:
|> Yoeri J. Apts (yjapts@info.vub.ac.be) wrote:
|> : I am currently reading INTERNETWORKING WITH TCP/IP vol II (by Douglas
|> : E. comer and David L. Stevens). This book contains a lot of sample code
|> : and so, being lazy like every programmer, I'm wondering if these sources
|> : are electronically available 'anywhere'?
 
|> : I am especially interested in de code concerning SNMP (Client-Server)
|> 
 ...
|> 
|> You can look at 'arthur.cs.purdue.edu:/pub/dls'.
|> 
Or better yet,
	ftp.uu.net:/published/books

-- 
+---------------------------------------------------------------+
| Mike D. Kail                      AT&T:   (612) 945-8044      |
| ProviderLink Developer            FAX:    (612) 945-8430      |
| United HealthCare Corp.           E-Mail: mkail@uhc.com       |
+---------------------------------------------------------------+
| "Nothing great was ever achieved without enthusiasm."         |
|  - Ralph Waldo Emerson                                        |
+---------------------------------------------------------------+

-----------[000622][next][prev][last][first]----------------------------------------------------
Date:      30 Apr 1994 00:43:09 GMT
From:      sklower@oboe.CS.Berkeley.EDU (Keith Sklower)
To:        comp.protocols.tcp-ip
Subject:   Re: OSI-CLNP versus IP

In article <Cowos1.7C@nntpa.cb.att.com>,
0396103-hammega <ehammeg@hvstsg3.ns-nl.att.com> wrote:
}I am interested in finding out more about the differences
}between the OSI-CLNP and IP.

RFC1561 (Use of CLNP in TUBA Environments) draws some high-level
parallels.  The exact differences are best left to comparing
the actual specifications.  There is work going on to make an official
specification for CLNP available in ASCII via anonymous ftp as part
of a goodwill gesture on the part of ISO towards the IETF.

}-CLNP speaks of "segmenting" while IP
} speaks of "fragmenting". Are these the same? How do these mechanisms work?

The intent is the same.  There are some minor differences, such as you point
out below.

}-Both protocols speak of "lifetime" "source routing" and "adressing"
} Do they mean the same thing ?
}
}-CLNP knows a "total length of originating PDU". IP doesn't.
} How about this?


Yeah, how about that ;-(  It means that if you have multiple network
address translator boxes and a receiving end hyper-scrupulous CLNP
implementation that carefully enforced that the total length of originating
PDU fragment field were the same in all packets, you'ld be out of luck.

However, many implementations are forgiving, and only take that particular
field as a hint.  Dave Katz suggested also setting that field to zero
if you wanted to translate IP to CLNP, although that would be even
more likely to be tossed.

-----------[000623][next][prev][last][first]----------------------------------------------------
Date:      30 Apr 94 03:26:02 GMT
From:      markw@antimatr.hou.tx.us (Mark Whetzel)
To:        comp.protocols.tcp-ip,comp.sys.ibm.pc.rt
Subject:   Re: IBM RT to PC ftp lock up's

In article <2976547406.1.p01048@psilink.com>, p01048@psilink.com (Quinn Erickson) writes:
> Does anyone have any experience with ftp from an IBM RT to a PC.  We 
> are having problems getting reliable communications between these two
> machines.  The RT is running AIX ver 2.2.1 and the PC is running DOS 5.0
> with LAN Workplace for DOS ver 4.01.  Occasionally a file transfer will
> lock up the PC causing a reboot and the ftp process hangs on the RT.

The only time I have seen problem like this myself, was with PC/TCP software
on the PC.  Somehow the IP window size was being mis-negotiated.
PC/TCP version 2.01 worked, 2.03 broke it, and 2.05 and later fixed it.
However.. this hang was consistant.  The problem was solely on the PC side.

Window pacing size problems, and improper packet size maximum will result in
hangs due to lost packets, and similar problems.  Make sure if you
are running on ethernet that the RT /etc/net has the inetlen=1500.
Earlier versions of AIX/RT defaulted this size incorrectly to 1518.

You might check for network wiring problems that may be introducing 
errors into the network.  This can account for intermittent problems,
and might explain a lockup due to a lost packet.
-- 
AIX..... NOT just another UNIX. (tm)
Mark Whetzel                          | My own RT system.. My own thoughts..
DOMAIN: markw@antimatr.hou.tx.us      | IBM RT/135 running AIX 2.2.1 
UUCP ..!menudo!lobster!antimatr!markw | comp.sys.ibm.pc.rt FAQ maintainer.

-----------[000624][next][prev][last][first]----------------------------------------------------
Date:      Sat, 30 Apr 1994 03:34:42 GMT
From:      davidk@csn.org (David Kirkpatrick)
To:        comp.protocols.tcp-ip
Subject:   pcnfsd



Hi,

Does anyone know where to get source code for pcnfsd?  I need to connect
a PC running DOS/Windows to a Stratus running their version of UNIX
(called FTX).  Problem is, FTX doesn't come with pcnfsd, so I guess I
need to try to compile it myself.  I'm running PC/TCP version 2.3 on the
Windows machine, and already have tcp/ip running between the two 
machines....

Thanks,
davidk


--
------------------------
-- David Kirkpatrick  --
-- Southwest Software --
-- davidk@csn.org     --

-----------[000625][next][prev][last][first]----------------------------------------------------
Date:      Sat, 30 Apr 1994 13:27:45 -0500
From:      opie@panix.com (Matthew S. "Opie" Warren)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm,comp.sys.novell
Subject:   BootP on Novell's NLM and MacTCP

Hi.

Novell's BootP NLM does not work with our Macintoshes at one site; oddly
enough, it works fine at another site.  

At the working site, the Macs ask the server for an IP number, show up in
the list of assigned IP numbers, receive the IP number and happily continue
about their business.

At the non-working site, the Macs ask the server for an IP number, show
upin the list of assigned IP numbers, but never receive the IP number they
need.  In other words, the server receives the BootP broadcast, sends back
an answer, but the answer never gets there or is not properly received.

At both sites, our PCs can receive IP numbers via BootP just fine.

What's up with this?  The only difference I can think of is that in the
non-working site, the BootP server is in a different subnet from the
requesting Macs.  Our networking people, however, assure me that this would
not matter.

If you have any ideas, I would be eternally grateful.  Thanks in advance.

Matt


Configuration details:
System 7.1, MacTCP 2.0.4, Quadra 650, built-in ethernet
Netware 3.11, BootP NLM version 1.1
Compaq 486/33M, Novell DOSUP9, TCPIP 4.1.2
Class B internet, subnet mask 255.255.255.0
WellFleet routers

-- 
Matthew S. "Opie" Warren     opie@panix.com
"Life is so short and so complicated.  Can I have some of your pizza?"

-----------[000626][next][prev][last][first]----------------------------------------------------
Date:      Sat, 30 Apr 1994 14:13:29 GMT
From:      prime@herring.dr.att.com (Anthony R. Davis)
To:        comp.protocols.tcp-ip,alt.bbs.internet
Subject:   IP over ATM


Does anyone have any information or can point me to information regarding the
transmission of IP over ATM. Also of interest is any information discussing ATM
to the desktop. Please respond to prime@dr.ATT.COM. Thanks in advance for any
help provided!!

Tony
prime@dr.ATT.COM

-----------[000627][next][prev][last][first]----------------------------------------------------
Date:      Sat, 30 Apr 1994 15:40:13 GMT
From:      jcmurphy@acsu.buffalo.edu (Jeff Murphy)
To:        comp.protocols.tcp-ip
Subject:   Re: Can I test ethernet card from off net.

Dennis Director (dennis@cauchy.math.nwu.edu) wrote:
>Is there some way to configure and test the
>installation up to the card (loopback at card?)
>so that when I get the connection I'll be that
>much closer to having things work.
>More importantly, I like to test systems in
>modules.  Thanks for any ideas,  dennis@math.nwu.edu


you could terminate the card, assign your self an IP, and broadcast ping
the ''subnet'' you've put yourself on.. you should see and answer from 
your own machine.. i believe the kernel wont catch this and send it
to the loopback device.. so it should offer a good way of testing 
whether the actual hardware is configured correctly...

if you dont have a bnc connection on the card (i assume it does if it
is an ne2000), you will have to most likely use the aui port with a transciever
and some terminators.
-- 
jcmurphy@acsu.buffalo.edu                            ece/cs             cit oss
opnsmurf@ubvms.bitnet				     standard disclaimers apply
''She walks in beauty, like the night - Of cloudless climes and starry skies.''

-----------[000628][next][prev][last][first]----------------------------------------------------
Date:      30 Apr 1994 17:32:46 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: IP headers in UDP,TCP

In article <767555115snz@pyers.demon.co.uk> pyers@pyers.demon.co.uk writes:
>When TCP or UDP are constructing the pseudo-headers that they use as a 
>checksum, they incorporate the IP address in those headers. Since TCP & UDP
>are at a higher level in the protocol suite that IP, how is this done ?

The interface between IP and the next layer for a received packet is
required to include the IP address to which the packet was addressed.
Similarly, the interface for a packet to be sent has to include the source
address for the packet.

Computing the checksum isn't the only reason for this.  In a TCP
connection, the source address of packets being sent must match the
destination address of packets being received.

This isn't the only interface like this.  The interface between the device
driver and IP must indicate whether a received packet was a MAC-level
broadcast (so that no ICMP errors will be sent as a result).
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000629][next][prev][last][first]----------------------------------------------------
Date:      30 Apr 1994 17:40:49 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: DNS and variable length subnets

In article <Cp0v9n.K16@gpu.utcc.utoronto.ca> russ@madhaus.uucp (Russell Sutherland) writes:
>Lets assume we have a class C network:  a.b.c.0 that is subdivided into 4
>equal sized subnets (mask 255.255.255.192):

Hmm, your Subject line said you were asking about variable-length subnets,
not equal-sized subnets.  However, the answer is the same....

>Do these four bits of contiguous ip address space have to served by one (and
>not four) name server(s)?

You can use four name servers for name->info lookups, but all the
IN-ADDR.ARPA entries will have to be on a single master.  This domain is
tied to the byte boundaries, and knows nothing of subnet masks.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

barmar@think.com          {uunet,harvard}!think!barmar

-----------[000630][next][prev][last][first]----------------------------------------------------
Date:      Sat, 30 Apr 1994 19:40:38 GMT
From:      mintz@cup.hp.com (Ken Mintz)
To:        comp.protocols.tcp-ip
Subject:   Re: packet losses, WHY

Mathieu Claude Hans (gt0567a@prism.gatech.edu) wrote:

> -Transmission using the UDP protocol:
> I beleive all the packet losses which occur on Internet are
> because of:
>         - congestions at routers and at the host
>         - error transmission, which in the case of UDP protocol
> implies dropping the packet.
>         - expiry of time to live because of bad routing.
> Are there any other reasons why packets are lost on the network?

  Besides other responses:

  - Lack of space in the receive (socket) buffer on the host.  (Not exactly
    congestion.  The message might be too large to ever fit.)

  - "Congestion" on the outbound host.  Specifically, too many packets queued
    up for the LAN driver to send.  (In most BSD implementations, this is 
    detectable as ENOBUFS returned by send().)

  - Deferred transmission Ethernet time-out.  (Not exactly "error
    transmission", if that means collision.) 

  - PMTU ICMP error.  (Not likely for UDP, but some implementation might
    permit enabling PMTU for UDP, useful only for certain application
    protocols, e.g. request-response.)

-- Ken Mintz

-----------[000631][next][prev][last][first]----------------------------------------------------
Date:      1 May 1994 07:09:08 -0500
From:      pall@cs.ait.ac.th (Mano Pallewatta)
To:        comp.protocols.tcp-ip
Subject:   A question on EGP

While being connected to the Internet core gateways is it possible for two
autonomous systems to have additional connection between them ?. (not via
core systems). Commer's Internetworking with TCP/IP Vol.1 says that EGP
restricts the Internet to a tree structure. 

Thanks in advance, 

+-----------------------------------+---------------------------------------+
|  Mano Pallewatta                  |    Internet: pall@cs.ait.ac.th        |
|  Computer Science Program         |                                       |
|  Asian Institute of Technology    |    Phone: +66 2 524 5720              |
|  GPO Box 2754, Bangkok 10501      |    Fax  : +66 2 524 5721              |
|  Thailand.                        |                                       |
|                                   |                                       |
+-----------------------------------+---------------------------------------+



END OF DOCUMENT