The 'Security Digest' Archives (TM)

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

ARCHIVE: TCP-IP Distribution List - Archives (1992)
DOCUMENT: TCP-IP Distribution List for September 1992 (331 messages, 166491 bytes)
SOURCE: http://securitydigest.org/exec/display?f=tcp-ip/archive/1992/09.txt&t=text/plain
NOTICE: securitydigest.org recognises the rights of all third-party works.

START OF DOCUMENT

-----------[000000][next][prev][last][first]----------------------------------------------------
Date:      31 Aug 92 23:20:45 +0100
From:      lc1bocv1@cine88.cineca.it
To:        comp.protocols.tcp-ip
Subject:   kermit for decstation

Is possible to connect a decstation 3100 to an MSDOS computer
with a rs232 line using kermit ?
I've triede whit decnetdos  ,but my ethernet card (3comm351) seems
to be incompatible with my olivetti computer.
Any help is welcome !!!
thanks in advances.

fabrizio ravegnani
lc1bocv1@cine88.cineca.it


-----------[000001][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 92 01:10:25 GMT
From:      dyer@spdcc.com (Steve Dyer)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,arpa.bind,arpa.nets
Subject:   Re: Maintaining DNS resource records for sites with many hosts

In article <1992Aug31.163615.16509@tc.cornell.edu> kakazu@cornell.edu writes:
>I would like to see if anyone has developed a front end for maintaining
>resource records that named uses, particularly for sites that have many
>hosts.
>
>What we do is maintain a file in hosts.txt format (RFC 952), which is
>sucked in nightly by a bunch of scripts which produce files with the
>records.  This has been working fine for several years, but it is less
>than forgiving for typos, error checking illegal addresses, etc.
>I think this method is fine if you don't have many hosts, but we are
>now approaching several thousand hosts.
>
>I would like to find some type of user friendly front end where one
>could input the DNS info, do some error checking, and have the resource
>record files automagically generated.

DECathena uses its service management system, Moira, as a front end
for generating input to BIND.  This feature will be available in the
next (1.1) release.

Of course, DECathena/Moira is a bit more than you might need, since
it also deals with maintaining users, groups, file systems, printcap
files, and an armslength of other info which an installation needs
to keep track of.


-- 
Steve Dyer
dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer

-----------[000002][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 92 11:28:19 GMT
From:      starrenb@ptt-iat (John Starrenburg)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   HELP with hanging NetCI in UB's TCP BNS/NDIS with UCX 2.0

Whenever I do a DISCONNECT from The Connection Assistant (NetCI), the session
manager that is integrated with Ungermann-Bass' TCP/IP stack for the PC,
TCP BNS/NDIS, and I am connected to a VAX with UCX 2.0, NetCI hangs. The
problem is probably because UCX 2.0 sends a logoff message after the TCP/IP 
FINish, which is conform the RFC. When tested against MultiNet or a UNIX box,
everything works fine. These stacks don't send any further data after the FIN 
from the PC. The PC itself works fine after the hanging NetCI, but when I
try to start my Windows Comms package DynaComm (or Reflection), no NetCI
prompt appears. NetCI uses software INT 6Bh for communication with these 
comms pacakges. The reason why I think it is NetCI that is failing is because
the DOS versions of FTP and Telnet work fine. I assume these progs interface
with the TCP/IP stack directly, instead of the comms packages which use 
NetCI as session manager.
Hope this is enough information for those who might know the answer. Any 
help and/or clues are GREATLY APPRECIATED, because we are approaching a
deadline.
Thanx in advance.
____
John
-- 
---------------------------------------
John Starrenburg    I&AT    PTT Telecom
Internet: starrenb%ptt-iat@nluug.nl

-----------[000003][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 92 14:41:21 GMT
From:      jdarnold@HQ.Ileaf.COM (Jonathan Arnold)
To:        comp.os.os2.networking,comp.protocols.tcp-ip
Subject:   Re: I need 3COM 3c505 Ethernet Plus driver for IBM's OS/2 TCP/IP

In article <BtnEGE.Lw3@fmsrl7.srl.ford.com> probert@fires1.srl.ford.com writes:
>I need the 3COM 3c505 EtherLink Plus driver for IBM's OS/2 TCP/IP
>software.  I am running OS/2 2.0 and the driver must conform to the
>NDIS MAC specification.

The OS/2 TCP/IP package comes with the NDIS drivers for the 3COM card.
I just installed TCP/IP and have the 3COM EtherLink Plus.  The driver
is called ELNKII.OS2.

+=================================+================================+
| Jonathan Arnold                 |  Interleaf, Inc.               |
| jdarnold@ileaf.com              |  Prospect Place                |
| Home: (617)335-5457             |  9 Hilside Avenue              |
| Bus: (617)290-4990 x5489        |  Waltham, MA 02154             |
+=================================+================================+


-- 
+=================================+================================+
| Jonathan Arnold                 |  The Complete Baseball BBS     |
| jdarnold@ileaf.com              |  All Your Baseball Info!       |
| Home: (617)335-5457             |  Stats - Box Scores - News     |

-----------[000004][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 92 15:51:33 GMT
From:      antonis@helios.ntua.gr (Antonis Kyriazis)
To:        comp.protocols.tcp-ip
Subject:   OSPF in 3COMs NetBuilderII


Did someone install successfully OSPF routing in 3COMs NB2s?
I posted an article in bit.listserv.3com-l but I think it
is inactive and I don't want to use bandwidth here...

thanks
+-------------------------------------------------------------------+
|     Antonis Kyriazis                                              |
| Networks & Communications       e-mail: antonis@intranet.gr       |
| INTRACOM sa                                                       |
| 19.5 km Marcopoulo Ave.          fax:   +30-1- 66 44 379          |
| Peania 190 02                           +30-1- 66 43 718          |
| GREECE                          phone:  +30-1- 66 44 961-5        |
|                                         +30-1- 88 43 715          |
+-------------------------------------------------------------------+




-----------[000005][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 1992 17:06:24 GMT
From:      eajjs@cbnewsf.cb.att.com (john.sottile)
To:        comp.protocols.tcp-ip
Subject:   Re: Why does my slip line die overnight?

In article <z5h1H206b9@atlantis.psu.edu> barr@pop.psu.edu (David Barr) writes:
>In article <1992Aug30.205245.29936@psg.com> randy@psg.com (Randy Bush) writes:
>>ghawkins@unix1.tcd.ie (George C. Hawkins) writes:
>>> When I leave my machine overnight and come back in the morning I
>>> find my slip connection has died on me.
>>
>>We see the same thing here in Orygun's RAINet.  We believe (urban legend)
>>that the telco does an 02:00 blast to clear the lines.
>
>Yep, i've seen similar experiences here, only here it's around 1:30.
>It won't necessarily be all the lines, but a significant portion of them.
>


Between the hours of 2am and 5am, the telecos telephone switches go into
diagnostics.  On older switch types (analog) can get noise on the
line when this happens, but it doesn't last long (the noise).

The dropped line is purely UNintentional.  The switches must go through
diags and 2am is a good time to start.  

Also, when switches are upgraded (software or hardware), the telco will
"Phase" a machine to get it to start with the new load.  98% of the time
no calls are dropped, however, if something went wrong with a load,
the machine will recover and a few calls could be dropped.  Usually,
the recovery will drop all calls, but not always.

I would suspect a PBX first (reacting to stuff on the line) before
the telco.  :-)

=============================
John Sottile                             "That Guy has to Oil his Brakes!"
AT&T Bell Labs                           #include <std/disclaimer/who-me?.h>
jjs@cblph.att.com

Let me add this:
I don't speak for AT&T and the ideas presented here are my own from my
own observations.

-----------[000006][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 92 18:25:05 GMT
From:      parker@mprgate.mpr.ca (Ross Parker)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP break - terminal server?


I am currently connecting a number of systems in our computer room via
serial cables to a Xyplex terminal server. Connections to individual ports
on this server are made using the 'xyp_ptyd' daemon supplied by Xyplex,
running on a central console 'host' system. This daemon essentially makes
a connection between each port on the serial side of the terminal server and
a pseudo-tty on the UNIX host system. I can then connect to these individual
pseudo-ttys using the 'console' package from Ohio State, but that's irrelevant
to this problem...

The problem is that I need to be able to generate a break signal on one of
the serial lines on the terminal server. I have no problem in modifying the
Xyplex xyp_ptyd code in order to do so, but I don't know what/how to send
to the terminal server (over the network connection) to tell it that I want
it to generate a break over the serial line.

Can anyone help? Is there a 'normal' way of passing this kind of thing through
TCP/IP? The Xyplex manual seems to indicate that it will do the same thing
in the opposite direction - i.e. a break on one of the serial lines can be
propogated through the terminal server and over TCP/IP to a host system... how?

Thanks...

Ross

-- 
Ross Parker           |                            | DoD# 1188   KotHFJ
MPR Teltech Ltd.      | You can't erase the dream, | '88 FJ1200
Burnaby, B.C., Canada | You can only wake me up... | parker@mprgate.mpr.ca

-----------[000007][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 1992 18:40:22 GMT
From:      jmfres11@ucs.usl.edu (Karthikeyan Gurnswamy)
To:        comp.protocols.tcp-ip
Subject:   source code for SMTP


Hi !,
I am looking for a public domain source code for SMTP. Can someone
help me ? 

Karthik
/----------------------------------------\--------------------------\
| Karthikeyan Guruswamy                  | Why does'nt my           |
| Graduate Student, Computer Science     |                          |
| Center for Advanced Computer Studies   |    NET always WORK ?     | 
| Univ. of SouthWestern Louisiana        |                          |
| Lafayette, LA 70501                    | -- Life beats any logic  |
| Email: jmfres11@ucs.usl.edu            |                          |
|                                        \--------------------------\
| The opinions stated are solely mine and does not reflect the views|
| of anyone in my organization ...                                  |
\-------------------------------------------------------------------/ 

-----------[000008][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 1992 19:15:23 GMT
From:      bourman@hpcc01.corp.hp.com (Bob Bourman)
To:        comp.protocols.tcp-ip
Subject:   Re: Digital Pathway ?




	ALSO:

	LeeMah DataCom Security Corp.
	3948 Trust Way
	Hayward, Ca. 94545

	Pat Sebers
	(510) 786-0790

	Both Digital Pathways and LeeMah are working with Enigma Logic using
	the token cards to access several different types of secure access
	devices. EL is working on a UNIX based solution as well.

	BOBB

-----------[000009][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 92 20:58:51 GMT
From:      mgic@mixcom.mixcom.com (mgic)
To:        comp.protocols.tcp-ip
Subject:   server that won't die (normally)


System: RS/6000 AIX 3.2.1

I've written a server that is started by inetd, reading
and writing stdin and stdout.

The server uses select() to determine when data is
ready to be read. If an (socket) error is detected, 
the server quits.

The problem is that the server never dies if the client
quits abnormally. When the client quits normally, it sends
the server an "end of session" message, and the server quits normally.

Why isn't the server being informed that the client no longer exists
so that select() will report an error? Or, what method should
be used to determine if the stream no longer exists?

Thank you.

Dean
-- 

-----------[000010][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 92 21:14:47 GMT
From:      ad0@cbnewsj.cb.att.com (amarjet.dhingra)
To:        comp.protocols.tcp-ip
Subject:   ksh history and .nfs* files question

Please email your help directly to att!mtgzfs3!ad0

Occassionly, ksh loses history and I noticed that sometimes the commands
that I am giving are going into a file that has name like .nfs34D5 (.nfs*).
Then at some point in time this file goes away and so does part of
history and I am back to a file upto 2 days old.  I think it happens
after our weekly reboots of the server SUNs when every workstation
is rebooted automatically.  I just use the default history file name.

Any ideas for fixing this problem?

Thanks in advance.

Amarjit

-----------[000011][next][prev][last][first]----------------------------------------------------
Date:      Tue, 1 Sep 1992 21:58:03 GMT
From:      dsiebert@icaen.uiowa.edu (Doug Siebert)
To:        comp.protocols.tcp-ip
Subject:   What is the minimum set of options a telnet server must negotiate?

I'm wondering what the minimum required set of telnet options the server side
of a telnet connection (the telnetd) is?  That is, what options is it required
to respond to, and how should it respond to requests for negotiation on the
options it declines to support?
 
I'm asking because I have an application into which I may want to try to
include some very primitive telnetd capabilities, so that I can negotiate to
turn off echo, turn on character at a time mode, and get the terminal type of
the connecting client.  Would it be sufficient to just properly negotiate
these options at the start and then refuse ANY negotiation after that time?

-- 
/-----------------------------------------------------------------------------\
| Doug Siebert                             | "I don't have to take this abuse |
| Internet:  dsiebert@isca.uiowa.edu       |  from you - I've got hundreds of |
| NeXTMail:  dsiebert@chop.isca.uiowa.edu  |  people waiting in line to abuse |
|     ICBM:  41d 39m 55s N, 91d 30m 43s W  |  me!"  Bill Murray, Ghostbusters |
\-----------------------------------------------------------------------------/

-----------[000012][next][prev][last][first]----------------------------------------------------
Date:      01 Sep 92 22:11:05 GMT
From:      jallard@microsoft.com (James 'J' Allard)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.dcom.lans
Subject:   Re: TCP/IP programming reference material wanted

In article <1992Aug30.011626.730@ccsvax.sfasu.edu> a_millsbs@ccsvax.sfasu.edu writes:
>I am intersted in trying to do some tcp/ip programming on some MS-DOS systems
>and I was wondering if anyone out there could recommend some good books or
>documents.  This is my FIRST attempt at tcp/ip pragramming so any references
>of general nature are valuable not just those that apply to ms-dos.

Probably the most widely referenced book for TCP/IP programming is
Stevens' UNIX Network Programming which is widely available (I think it's
a Prentice Hall title). It explains the basics of sockets, etc.

However, if you want to do MS-DOS TCP/IP programming, the interface will
vary from implementation to implementation. Your best bet is to get in
touch with the manufacturer of the stack that you'll be working on to get
the correct library and header files. Most MS-DOS based interfaces are 
based on BSD sockets of some flavor.

Finally, if you're thinking about doing some Windows programming over TCP/IP,
it's probably worth your time to check out the recently released Windows 
Sockets specification available on ftp.uu.net under the /vendors/microsoft
directory. The specification was jointly developed and endorsed by about
20 vendors in the community and will be the new interface by which to 
access TCP/IP via Windows. Windows NT will have a Windows Sockets compliant
interface as well.

Hope this helps,

J. Allard
Microsoft Corp.


-----------[000013][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 92 00:19:49 GMT
From:      ddl@burrhus.harvard.edu (Dan Lanciani)
To:        comp.os.os2.networking,comp.protocols.tcp-ip
Subject:   Re: I need 3COM 3c505 Ethernet Plus driver for IBM's OS/2 TCP/IP

In article <1992Sep1.144121.674@HQ.Ileaf.COM>, jdarnold@HQ.Ileaf.COM (Jonathan Arnold) writes:
| In article <BtnEGE.Lw3@fmsrl7.srl.ford.com> probert@fires1.srl.ford.com writes:
| >I need the 3COM 3c505 EtherLink Plus driver for IBM's OS/2 TCP/IP
| >software.  I am running OS/2 2.0 and the driver must conform to the
| >NDIS MAC specification.
| 
| The OS/2 TCP/IP package comes with the NDIS drivers for the 3COM card.
| I just installed TCP/IP and have the 3COM EtherLink Plus.  The driver
| is called ELNKII.OS2.

	Not quite.  ELNKII.OS2 is the driver for the 3Com EtherLink II,
i.e., the 3c503.  This card is completely different from the 3c505 =
EtherLink Plus.  The correct driver is ELNKPL.OS2.  Beware that there
are several versions of this driver floating around and some of them
have serious bugs.  You might want to get all you can find and try them...
(Based on the differing sizes of the ELNKPL.OS2 drivers, I am *guessing*
that some contain replacement code for the on-board cpu and some do not.
In general, I have had better luck with the larger drivers.)

				Dan Lanciani
				ddl@harvard.*

-----------[000014][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 1992 03:30:28 GMT
From:      peter@cujo.curtin.edu.au (Peter N Lewis)
To:        comp.protocols.tcp-ip
Subject:   FTP RFC and multi-line replies?

Hi All,

I've noticed that some ftp servers implement multi-line replies like this:

123-Message
123-Message line 2
123-Message line 3
123 Mesage last line

Now, my reading of this section of the RFC:

         Thus the format for multi-line replies is that the first line
         will begin with the exact required reply code, followed
         immediately by a Hyphen, "-" (also known as Minus), followed by
         text.  The last line will begin with the same code, followed
         immediately by Space <SP>, optionally some text, and the Telnet
         end-of-line code.

            For example:
                                123-First line
                                Second line
                                  234 A line beginning with numbers
                                123 The last line

         The user-process then simply needs to search for the second
         occurrence of the same reply code, followed by <SP> (Space), at
         the beginning of a line, and ignore all intermediary lines.  If
         an intermediary line begins with a 3-digit number, the Server
         must pad the front  to avoid confusion.

In particular the last sentence, would indicate that the above format is in
fact illegal (although any client should be able to cope with both, since
they should be searching for a line that starts with '123<space>').  As
well as that it looks much worse than padding each message line with
spaces:

123-Message
    Message line 2
    Message line 3
123 Mesage last line

But thats neither here nor there.  I'm wondering if there is any
disagreement on this point, is the first format actually valid?

Thanks for any info,
   Peter.

_______________________________________________________________________
Peter N Lewis, NCRPDA, Curtin University       peter@cujo.curtin.edu.au
GPO Box U1987, Perth WA 6001, AUSTRALIA             FAX: +61 9 367 8141

-----------[000015][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 1992 04:59:12 GMT
From:      jeremy@cowan.edu.au (J.LAIDMAN)
To:        comp.protocols.tcp-ip
Subject:   CUTCP throwing me out!

I hope someone can help.  I've been using Clarkson Uni's TELBIN for quite a
while now with the occasional problem of being thrown back to DOS for no
apparent reason, as if I logged out.  Some times I've had two sessions going
and one of them disappears.

The problem has arisen on PS/2s, clones, token ring and ethernet, WD cards,
NE cards, etc, etc...  The only common element is the network:

2 x Novell 3.11 hosts, 1 x Unix Host (AIX 2); I'm loading novell and packet
drivers to connect to both types of hosts.

I haven't seen this problem reported before so I've no clues.  Anyone else
seen this?

Jeremy

---------------------------------------------------------------------
Jeremy Laidman, Computer Support Officer
Department of Computer Science                 Phone: (61 9) 370 6648
Edith Cowan University                         Fax:   (61 9) 370 2910
Perth, Western Australia                       j.laidman@cowan.edu.au

-----------[000016][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 92 05:33:37 GMT
From:      paul@frcs.Alt.ZA (Paul Nash)
To:        comp.protocols.tcp-ip,comp.protocols.ibm,comp.sys.novell
Subject:   High capacity TCP/IP - SNA gateways

A while back I posted here asking about high capacity Novell - SNA
gateways, to replace or failing (and discontinued) Novell gateways.
We have pretty much decided on a route to follow, and I would like to
bounce the idea off the net.community, in case we have it all wrong.

We want to get access from an Ethernet LAN to a large-ish MVS system,
and _also_ get access to Unix hosts on the LAN from SNA 3270
terminals.  We currently have two Novell SNA gateways, converting
Ether to Token Ring, thence via a pair of Mohawk MC400 devices into
the MVS system.  This is massively unstable.  Users in the SNA cloud
access the MVS' VTAM, thence via IBM's TCP/IP an ELC, and so into the
LAN.  This works, but costs us 2 - 4 MIPS on the MVS box.

We are planning to rip all this out, and replace with a large Unix box
(something like an IBM RS/6000 or Sun 670/MP).  This machine will have
the usual Ethernet adapter, plus a channel-attatch into the MVS.  This
must handle between 600 and 700 simultaneous connections, doing vt100
to 3270 translation.  The Unix box will also have a 48 kbps SDLC link
into the SNA cloud, for incoming connections from SNA 3270 terminals.
We expect between 100 and 200 simultaneous connections this way, with
suitable 3270 - vt100 mapping providing access to Unix hosts on the
LAN.  Software would be either vendor-supplied, or something like Open
Connect II.

Has anyone out there had any experience of something like this?  If
so, was it good or bad?  Any ideas about hardware (type, size) and
software?  Any input at all would be most welcome.

I last worked on IBM comms about 10 years back, and want to avoid it
if at all possible, hence this scheme :-).

	paul
 
 ---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---
 Paul Nash                                         paul@frcs.Alt.ZA
 Box 12475, Onderstepoort, 0110 South Africa         +27-12-5611879

      "You don't want to get locked into open systems" -- IBM

-----------[000017][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1992 08:32:18 GMT
From:      leres@ace.ee.lbl.gov (Craig Leres)
To:        comp.protocols.tcp-ip
Subject:   Re: Monitor program for catching duplicate IP addrs?

Peter Richards writes:
> >In article <jqymwsq.ptrubey@netcom.com> ptrubey@netcom.com (Phil Trubey) writes:
> >>Does anyone know of a utility program that will scan an Ethernet
> >>and build a table of IP # to ethernet # mappings and check for
> >>duplicate IP #s?  Thanks for any info,
> I was wondering if there is anything like this for Suns running SunOS 4.1.1?

I have something called arpwatch which maintains a database of
ethernet/ip address pairings. It emails significant changes. (I pipe
these into a local newsgroup called lbl.arpwatch.) To use it you must
first install the Berkeley Packet Filter. SunOS 4 is one operating
systems supported by bpf.

Arpwatch is available via anonymous ftp from ftp.ee.lbl.gov
(128.3.112.20). Set binary mode and retrieve the compressed tarchive
arpwatch-1.0.tar.Z; bpf is packaged as part of tcpdump. The current
version of tcpdump is also found on ftp.ee.lbl.gov in the compressed
tarchive tcpdump-2.2.1.tar.Z.

		Craig

-----------[000018][next][prev][last][first]----------------------------------------------------
Date:      Wed, 02 Sep 1992 15:45:20
From:      jbvb@vax.ftp.com  (James B. VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: What is the minimum set of options a telnet server must negotiate?

In article <1992Sep1.215803.7063@news.uiowa.edu> dsiebert@icaen.uiowa.edu (Doug Siebert) writes:

    I'm asking because I have an application into which I may want to try to
    include some very primitive telnetd capabilities, so that I can negotiate
    to turn off echo, turn on character at a time mode, and get the terminal
    type of the connecting client.  Would it be sufficient to just properly
    negotiate these options at the start and then refuse ANY negotiation after
    that time?

Yes, as long as you could accept the options you want in an arbitrary
order and avoid infinite option loops.  You will need to continue to
parse IACs for the duration of the connection, though, though, and it
is a good idea to always handle IAC AYT...

James B. VanBokkelen		2 High St., North Andover, MA 01845
FTP Software Inc.		voice: (508) 685-4000 fax: (508) 794-4488


-----------[000019][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 1992 12:00:53 GMT
From:      martin@axis.se (Martin Gren)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP Print server that supports LPD

Newsgroups: comp.protocols.tcp-ip.ibmpc
Subject: Re: TCP/IP printer server with LPD
Summary: 
Followup-To: 
Distribution: world
Organization: Axis Communications AB, Lund Sweden
Keywords: 


Alan Luchuk> In article <16851C670.PA9690@utkvm1.utk.edu>, PA9690@utkvm1.utk.edu
Alan Luchuk>  (Alan Luchuk) wrote:
Alan Luchuk> 
Alan Luchuk> >Are special purpose devices available that do LPD?  Something
Alan Luchuk> >that plugs into an Ethernet on one side and a printer on the
Alan Luchuk> >other and doesn't require additional software/hardware.
Alan Luchuk> 

My company makes a TCP/IP printer server that supports lpd as well as many
other TCP/IP protocols. Further to this it also support Novell IPX
concurrently - with a lot of features. The box takes up to 4 printers.
If you have any further interest - please send me an email, or to
"info@axis.se" or "info@axisinc.com".

/ Martin

Martin Gren                email: martin@axis.se
Axis Communications AB     Tel: +46 46 19 18 38
S - 223 70 LUND,  SWEDEN   Fax:   +46 46 13 61 30
-- 
Martin Gren                email: martin@axis.se
Axis Communications AB     Tel: +46 46 19 18 38
S - 223 70 LUND,  SWEDEN   Fax:   +46 46 13 61 30

-----------[000020][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 92 15:22:31 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Implementation of Distributed Database in SUN RPC

In article <1992Aug31.185338.29597@stortek.com> pae@blackcat.stortek.com (Phil Earnhardt) writes:
>
>Netwise's RPC TOOL has certainly addressed the issue of different styles of
>dispatching...there are single-threaded dispatchers, multi-processing
>dispatchers (forking off a process for new requests) on systems providing
>multi-processing, and multi-threaded dispatchers (spawning a new thread for
>new requests) on systems providing multi-threading capabilities.

Can anyone direct me to anything *available* that really describes
the Netwise product, complete with examples?  I keep seeing claims
about how great it is, but I've never seen examples of why it's so
good.

	Rich Stevens  (rstevens@noao.edu)

-----------[000021][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 92 20:45:56 GMT
From:      oak@odda.er.sintef.no (Olav Kvittem)
To:        comp.protocols.tcp-ip
Subject:   How "reliable" is UDP

Hello,

I have run some ttcp tests sending UDP-datagrams to test a 34Mbps/FDDI/Ethernet
network.

If I run the test on a Sun :

packetsize	KB/sek		CPU usage	Mbps    Interface
8000		2437		96%		20.0  	Ether
4000		2051		98%		16.4  	Ether
1400		1166		98%		 9.3  	Ether
1000		 941		97%		 7.5	Ether

From this I deduce that many packets does not leave the machine before they are
dropped. Even with an FDDI interface very few of the packets make it out of the
machine at 8192 packetsize (size of data for socket write operation). 

If I do the same tests on a DECStation Ultrix system it seems to be more "reliable" :

Ethernet :
8192		1051		30%		 8.0	Ethernet
8192		3033		92%		24.2    FDDI 

We see that the CPU usage is low keeping the Ethernet UDP at near maximum
Ethernet speed.  That might indicate that the DECsystem have a blocked
write for UDP sockets while the Sun kernel accepts UDP packets at full speed and
drops them if buffers fill up ? Perhaps there are no buffers in a UDP-implemntation ?

Does anyone have any explanation or observations for this behaviour ?

Regards
  Olav
 
FDDI :
-- 
   Olav Kvittem : SINTEF

      RFC Address     Olav.Kvittem@delab.sintef.no
      OR Address      C=no;ADMD=" ";PRMD=uninett;
                      O=sintef;OU=delab;S=Kvittem;G=Olav
      Postal Address  SINTEF DELAB
                      N-7034 Trondheim
      Phone           +47-7-532586(FAX)
      Phone           +47-7-596981
      Description     Research Scientist  -  Communication Networks


-----------[000022][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 92 21:13:53 GMT
From:      eclarke@ibmpa.awdpa.ibm.com (Eric S. Clarke)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.os.os2.misc,comp.os.os2.networking
Subject:   Re: Need help with getting IBM's TCP/IP running under OS/2

In article <Btp7wu.2Bv@fmsrl7.srl.ford.com>, probert@fires1.uucp (Neal W. Probert) writes:
|> I just installed the 3COM 3c505 driver for OS/2 2.0 for use with IBM`s
|> TCP/IP package.  Here's what I get from STARTUP.CMD:
|> 
|> CALL C:\TCPIP\BIN\TCPSTART.CMD
|> CONFIGURING TCP/IP
|> add net default: router 128.5.192.3: (null): Network is unreachable
.
.
.
|> -- 
|>     FORD       | Neal W. Probert   E3154 SRL    | probert@fires1.srl.ford.com
|>  SCIENTIFIC    | Ford Scientific Research Labs  | 313-845-8178 FAX 313-337-5581
|> RESEARCH LABS  | Dearborn, MI 48121-2053        |

Have you installed the latest CSD for the TCP/IP package? 

The base TCP/IP should show version number 1.20.1 and CSD level of UB02231
when you do a syslevel.  The LAPs package should show version number 2.01
and a CSD level of WR06000

-- 

Eric S. Clarke                     INTERNET:  eclarke@ibmpa.awdpa.ibm.com 
1055 Joaquin Road                      UUCP:  uunet!ibmsupt!eclarke
Moutain View, CA 94043             IBM VNET:  ECLARKE at AUSTIN
415-694-3858  T/L: 466-3858            

-----------[000023][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 92 21:13:54 GMT
From:      oak@odda.er.sintef.no (Olav Kvittem)
To:        comp.protocols.tcp-ip
Subject:   Dynamically setting the TCP window is hard ?

Hello,

I want to set the TCP window selectively for high performance  sessions on a high
bandwith*delay network.  

I have tried to set the max windowsize for TCP on BSD-type systems with the
setsockopt( SO_RCVBUF ). That means setting the TCP internal receive buffer.  By
this TCP hopefully will announce its entire buffer to the peer TCP.  This works
well with SunOs 4.1.1/2, Cray/Unicos v6.013 and HP/UX systems, but  DECstation
5000 Ultrix v4.2 does not announce a higher window than 32K on the initial SYN
response on a TCP connection setup even if the buffersize is confirmed to be set
higher by getsockopt.  

What is the actual mechanism for TCP to determine max inital receive window
size and for me to influence it ?

Common to all systems above except Unicos is that they does not accept a higher 
buffersize than about 50K - why is that ?

I guess that you can generate a kernel with fixed TCP buffersize larger than
50K. Then all TCP session for that system would have let say 64K window. Would that be more harmful to slow speed, normal delay , 64Kbps, networks than a 4K window ?

Regards
  Olav

-- 
   Olav Kvittem : SINTEF

      RFC Address     Olav.Kvittem@delab.sintef.no
      OR Address      C=no;ADMD=" ";PRMD=uninett;
                      O=sintef;OU=delab;S=Kvittem;G=Olav
      Postal Address  SINTEF DELAB
                      N-7034 Trondheim
      Phone           +47-7-532586(FAX)
      Phone           +47-7-596981
      Description     Research Scientist  -  Communication Networks


-----------[000024][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1992 23:17:16 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP break - terminal server?

The TELNET protocol includes a Break command, which is IAC BRK (255 243);
see p.12&14 of RFC 854.  Whether the Xyplex terminal server actually
translates this to an RS-232 break I don't know.  Also, if xyp_ptyd is
trying to provide a transparent pathway, it might double the IAC.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000025][next][prev][last][first]----------------------------------------------------
Date:      Wed, 2 Sep 1992 23:33:12 GMT
From:      pascoe@rocky.gte.com (Dave Pascoe)
To:        comp.protocols.tcp-ip
Subject:   SLIP for SunOS 4.1.1 - where?

I've searched high and low for public domain SLIP server software for 
SunOS 4.1.1 and I've only had a little luck.  There are a bunch of files out 
in anonymous ftp land called slip-4.1.shar.Z.  This looks like it's probably 
what I want but I keep getting "Unexpected fi" errors when I try to run sh 
on it to extract the sources.  Also, does this version have the capability to 
use VJ compression?

There is also cslipbeta.tar.Z, but, according to the README file, shouldn't 
be attempted unless you are a real kernel hacker.  Has anyone got this to
work on SunOS 4.1.1 without hacking?

Is there another package out there that I just might be missing?  I used
archie but these two packages appear to be the only two available.

I just installed ppp but haven't had a chance to try it out.  Is there any
problem with running both SLIP and PPP on the same machine?  I think the
answer is no, but I'm still getting my feet wet.
--
Dave Pascoe		
Internet: pascoe@rocky.gte.com

-----------[000026][next][prev][last][first]----------------------------------------------------
Date:      Thu, 3 Sep 1992 00:05:33 GMT
From:      jjensen@convex.com (James Jensen)
To:        comp.protocols.tcp-ip
Subject:   Re: Dynamically setting the TCP window is hard ?

oak@odda.er.sintef.no (Olav Kvittem) writes:
- I want to set the TCP window selectively for high performance  sessions on a high
- bandwith*delay network.  
- 
- I have tried to set the max windowsize for TCP on BSD-type systems with the
- setsockopt( SO_RCVBUF ). That means setting the TCP internal receive buffer.

- well with SunOs 4.1.1/2, Cray/Unicos v6.013 and HP/UX systems, but  DECstation
- 5000 Ultrix v4.2 does not announce a higher window than 32K on the initial SYN
- 
- What is the actual mechanism for TCP to determine max inital receive window
- size and for me to influence it ?
- 
- Common to all systems above except Unicos is that they does not accept a higher 
- buffersize than about 50K - why is that ?
- 
- I guess that you can generate a kernel with fixed TCP buffersize larger than
- 50K. Then all TCP session for that system would have let say 64K window. Would that be more harmful to slow speed, normal delay , 64Kbps, networks than a 4K window ?
- 
It sounds like you are doing the right thing.  Unless you have the
sources to the kernel there probably isn't a lot more you can do about it.
The larger buffer may still be buying you some performance even though
it isn't advertised.  It should allow the window to stay all the way
open even if the application's read is delayed.

The problem with BSD based tcp is that it had hidden assumptions, and
some not so hidden, that they were going over ethernet or some other network 
that didn't need a big window.  This has slowly changed, but modifying tcp 
is pretty scary.  tcp_input() has over 1000 lines in one routine full of 
gotos and subroutines with magic side effects.

I suspect the DECstation still has the window stuff declared as signed
16 bit numbers.  I have no idea where the 50k limit is coming from.

One thing you might want to watch out for is that some implementations
of tcp don't work very well if the socket buffers aren't the same size
on both sides of the connection.

Jim Jensen - jjensen@convex.com

-----------[000027][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 00:36:00 GMT
From:      pstevens@Metaphor.COM (Paul Stevens)
To:        comp.protocols.tcp-ip
Subject:   Broadcasting and Subnets

After looking in all the usual places (Comer, Stevens, Tanenbaum) 
there's still some loose ends I don't quite understand.

When I want to send a broadcast packet to all stations on a net or
subnet I set the IP address so that it contains the network/subnet
numbers but set the host portion to all 1's.  This is, of course,
net/subnet OR'ed with the inverse of the subnet mask.  Presumably 
this style of addressing even works for a net/subnet that might be
several hops away (i.e. the routers along the route will use the 
net/subnet fields of the IP address to route the packet to the proper
net/subnet).  My basic question concerns how to get a the correct
subnet mask.

Comer discusses using an ICMP message that can be sent to a gateway 
to learn the subnet mask for a given network, but doesn't list any 
programatic interfaces to this type of query.

What I don't understand is how networks and subnets get *named*.
The documentation on getnetbyname seems to imply that it uses
/etc/networks to return information and also does not give the subnet
mask.

Is this all hooked into Domain Name Servers somehow?  I know that
locally on a Unix system one could use the combination of /etc/networks
and /etc/netmasks to get this information.  But that implies that
each host would have to have these files kept up-to-date for all nets
that might be communicated with.  There must be a missing piece here
that allows these things to be administered more centrally and for
inquiries to be made remotely.

+------------------------------------------------------------------------+

   Paul Stevens                        {apple|decwrl}!metaphor!pstevens
   Metaphor Computer Systems           pstevens@metaphor.com
   Mountain View, CA

-----------[000028][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 03:21:00 GMT
From:      sob@hsdndev.UUCP (Scott Bradner)
To:        comp.protocols.tcp-ip
Subject:   Re: OSPF in 3COMs NetBuilderII

In article <3272@ariadne.csi.forth.GR> antonis@helios.ntua.gr (Antonis Kyriazis) writes:
>
>Did someone install successfully OSPF routing in 3COMs NB2s?

This & 10 other OSPF vendors were tested at Harvard for the past 3 days,
all work quite well. A report will be in an upcoming issue of Network
World.  I'll post more results after the publication.

Scott

-----------[000029][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 04:12:38 GMT
From:      lars@spectrum.CMC.COM (Lars Poulsen)
To:        comp.protocols.tcp-ip
Subject:   Re: Secure TCP-IP

In article <1992Aug28.083228.16663@hplb.hpl.hp.com>
   paola@hplb.hpl.hp.com (paola fulchignoni) writes:
>Does anyone know whether a "secure TCP" or "secure IP" exist (providing
>security
>services such as access control, source authentication, integrity, etc.)?

To secure traffic at the IP level requires a trusted physical network.
The military people have made some efforts in this direction, and some
of these efforts have been picked up by the financial community.
Esentially, you put a transmogrifier (authenticator/encryptor) between
your machine and the real network. The machines on the "red" side of the
encryptor box now form a closed network of trusted hosts, and the boxes
may optionally talk to a central control center which may potetially
open windows to talk to specific hosts attached directly to the
underlying insecure "black" network in a controlled manner. Such boxes
have been seen between a host's X.25/V.35 connector and the CSU/DSU.
They have also been spotted between the host's ethernet AUI connector
and its transceiver.  This technology is expensive. And of course you
have to trust the people who sold you the box.

The transport level could easily be modified to perform an
authentication handshake with a control center at connection startup
time, but then (of course) it would not be TCP any more. The US Air
Force has done some studies of variations of this concept. (Mostly with
parallel protocol modules sitting ARP-like between TCP and IP). I don't
think this has led to any large scale implementation.

Most people put this in the application layer. Maybe the best known
example is Kerberos. There is also work being done on a secure RPC
mechanism, and thus a secure NFS. Several workstation vendors have
formed a Trusted Systems Interoperability Group to ensure that the
multiple implementations of this spec really will be interoperable.
TSIG is loosely affiliated with IETF. HP is a major participant.
-- 
/ Lars Poulsen, SMTS Software Engineer	Internet E-mail: lars@CMC.COM
  CMC Network Products / Rockwell Int'l	Telephone: +1-805-968-4262	
  Santa Barbara, CA 93117-3083		TeleFAX:   +1-805-968-8256

-----------[000030][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1992 04:33:13 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: FTP RFC and multi-line replies?

In article <peter-020992112136@134.7.50.3> peter@cujo.curtin.edu.au (Peter N Lewis) writes:
>         If
>         an intermediary line begins with a 3-digit number, the Server
>         must pad the front  to avoid confusion.
>
>In particular the last sentence, would indicate that the above format is in
>fact illegal

I think that sentence was misworded.  It should have said, "If an
intermediary line begins with a 3-digit number followed by space, ...."
There's clearly no reason why a client should be confused by intermediary
lines prefixed with "123-".

>But thats neither here nor there.  I'm wondering if there is any
>disagreement on this point, is the first format actually valid?

Certainly.  The rule is that the first line must begin with "123-", the
last line must begin with "123<space>", and the intermediary lines must not
begin with "123<space>".  The above sentence was intended to easy
interoperability with clients that look for any
<digit><digit><digit><space> prefix rather than trying to match the
original ("Be conservative in what you send").

I've seen some FTP servers whose greeting messages contain:

230-If your ftp client has problems with the extended messages (or if you
230-simply find such messages annoying), connect again and use a password 
230-that begins with "-" to disable these messages.

So I suppose there are some clients out there that can't handle multiline
replies.  However, they may have trouble with them in any format.

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000031][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1992 08:18:00 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

In article <2490@cronos.metaphor.com> pstevens@Metaphor.COM (Paul Stevens) writes:
>When I want to send a broadcast packet to all stations on a net or
>subnet I set the IP address so that it contains the network/subnet
>numbers but set the host portion to all 1's.  This is, of course,
>net/subnet OR'ed with the inverse of the subnet mask.  Presumably 
>this style of addressing even works for a net/subnet that might be
>several hops away (i.e. the routers along the route will use the 
>net/subnet fields of the IP address to route the packet to the proper
>net/subnet).  My basic question concerns how to get a the correct
>subnet mask.

There's no common method for this.  In general, hosts are only expected to
know the subnet mask for their local networks.  Hosts are only supposed to
use subnet masks to determine whether a destination can be reached directly
or must go via a router.  All non-local addresses are equivalent to hosts.

Recent routing protocols transmit subnet masks along with routes, but these
are intended only for use by routers, not individual hosts.

Sending broadcasts to remote networks is frowned upon, so there's no
special support to make it easy.  If you want to do it, you're on your own
for a mechanism to learn the appropriate broadcast address for that subnet.

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000032][next][prev][last][first]----------------------------------------------------
Date:      Thu, 3 Sep 1992 08:40:30 GMT
From:      wmchung@eng.ie.cuhk.hk (chung wai man)
To:        comp.protocols.tcp-ip
Subject:   UDP Send Problem

Hi! I am facing a problem using UDP sockets. I have set the socket to be
 non-blocking. When I use recvfrom() call to receive packets, when nothing
is on the socket, the call fails and return -1. After the failure of 
recvfrom(), I cannot send a packet of 4 bytes only. The errno returned is 35,
meaning that the operation should be blocking.

I am really headached over this problem. I welcome any insights, solution or
comments to me.

Pls email directly to me: wmchung@eng.ie.cuhk.hk

Thanks a lot!

-----------[000033][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 10:24:23 GMT
From:      wihuri@polaris.utu.fi (Pauli Wihuri)
To:        comp.protocols.tcp-ip
Subject:   XGATE gateway from MHS to SMTP ?


Good morning/noon/afternoon/evening/night/whatever (but not day)!

We are at University of Turku searching for the possibilities to use
EMail straight from MS-DOS machines without having to establish
connection into mainframes in order to send email.

We use LANtastic as our LAN. We are testing Futurus Team for a PC
EMail program. 

We need to test any MHS-server working with LANtastic and we need to
test also gateway from MHS to SMTP EMail in our SUN/Unix-mainframe. We
heard about PD/Freeware/Shareware-product called XGATE.

Could someone please give me pointer to that program? I do not have
access on Compuserve but anywhere where is anonymous ftp available on
Internet.

With thanks in advance,

Pauli Wihuri
office:		Computing Centre, University of Turku, SF-20500 Turku, Finland
internet:	wihuri@utu.fi
fax:		+358-(9)21-633-6560
phone:		+358-(9)21-633-6518

-----------[000034][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 14:08:59 GMT
From:      sweeney@Ingres.COM (Tony Sweeney)
To:        comp.protocols.tcp-ip
Subject:   Re: server that won't die (normally)

In article <1992Sep01.205851.29040@mixcom.com> mgic@mixcom.mixcom.com (mgic) writes:
>
>System: RS/6000 AIX 3.2.1
>
>I've written a server that is started by inetd, reading
>and writing stdin and stdout.
>
>The server uses select() to determine when data is
>ready to be read. If an (socket) error is detected, 
>the server quits.
>
>The problem is that the server never dies if the client
>quits abnormally. When the client quits normally, it sends
>the server an "end of session" message, and the server quits normally.
>
>Why isn't the server being informed that the client no longer exists
>so that select() will report an error? Or, what method should
>be used to determine if the stream no longer exists?
>
You want to setsockopt(..., ..., SO_KEEPALIVE, ..., ...); The select will
(eventually) return an error. Note that the keepalive time is a kernel
parameter and cannot be set by the user.
>Thank you.
>
>Dean
>-- 

Hope this isn't too wide of the beam.

Tony.

-----------[000035][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 14:37:35 GMT
From:      neeraj@metrix.UUCP (Neeraj Sangal)
To:        comp.protocols.tcp-ip
Subject:   Re: Monitor program for catching duplicate IP addrs?

> Does anyone know of a utility program that will scan an Ethernet
> and build a table of IP # to ethernet # mappings and check for
> duplicate IP #s?  Thanks for any info,
> I was wondering if there is anything like this for Suns running SunOS 4.1.1?

You can use the NetMetrix Load Monitor to detect duplicate IP addresses.
The NetMetrix Load Monitor maintains a pretty complete mapping of MAC
and Network Addresses seen on the network and can be used to detect
duplicate addresses for most major protocols. NetMetrix runs on SunOS 4.x.x.

Neeraj Sangal
Metrix Network Systems, Inc.            1 Tara Blvd, Nashua, NH 03062
neeraj@metrix.com                       (603) 888-7000

-----------[000036][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 14:39:04 GMT
From:      epc1@quads.uchicago.edu (Merlin)
To:        comp.protocols.tcp-ip
Subject:   21 step guide to setting up DNS

Thanks


-- 
En Pum Cho					
University of Chicago Class of 1994
math & economics major
epc1@midway.uchicago.edu

-----------[000037][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 16:07:33 GMT
From:      leonard@telcom.arizona.edu (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

In article <184hnoINN7tu@early-bird.think.com>, barmar@think.com (Barry Margolin) writes:
| 
| In article <2490@cronos.metaphor.com> pstevens@Metaphor.COM (Paul Stevens) writes:
| >When I want to send a broadcast packet to all stations on a net or
| >subnet I set the IP address so that it contains the network/subnet
| >numbers but set the host portion to all 1's.  This is, of course,
| >net/subnet OR'ed with the inverse of the subnet mask.  Presumably 
| >this style of addressing even works for a net/subnet that might be
| >several hops away (i.e. the routers along the route will use the 
| >net/subnet fields of the IP address to route the packet to the proper
| >net/subnet).  My basic question concerns how to get a the correct
| >subnet mask.
| 
| There's no common method for this.  In general, hosts are only expected to
| know the subnet mask for their local networks.  Hosts are only supposed to
| use subnet masks to determine whether a destination can be reached directly
| or must go via a router.  All non-local addresses are equivalent to hosts.
| 
| Recent routing protocols transmit subnet masks along with routes, but these
| are intended only for use by routers, not individual hosts.
| 
| Sending broadcasts to remote networks is frowned upon, so there's no
| special support to make it easy.  If you want to do it, you're on your own
| for a mechanism to learn the appropriate broadcast address for that subnet.

The best and easiest approach is just to use ALL 1's (255.255.255.255) as your
broadcast address.  That way you don't have to worry about your subnet mask,
or for that matter even what your own IP address is.  Recommended by 
RFC-1122, sec. 3.3.6 (discussion section.)

This won't let you send a broadcast to a remote network, of course, but that
won't work anyway.  Use multicast for this.

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
|>   If it ain't broke, break it.

-----------[000038][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 92 16:27:15 GMT
From:      chris@frcs.Alt.ZA (Chris Old)
To:        comp.protocols.tcp-ip
Subject:   ka9q router unstable under heavy load

We have a mini internet, with four ka9q routers on the main ethernet
backbone, connecting remote sites via slip.  The version of ka9q has a
version number of 911229 (7 months ago).  All we did to it was undef the
various hardware and software options that we didn't need in config.h.
Otherwise it is stock standard.

The slip lines run over 9600 async lines (well actually sync, but we have
async to sync converters) and the serial ports are fitted with 164550's.
We used to have 16550's, but we changed them to see if that fixed the
problem described below.

Now two of the routers connect sites with very little load, and are very
stable.  The other two routers hang on average 3 times a day.  By hang I
mean that they don't answer pings on the local ethernet, and on the rare
occasions that they have keyboard and screen plugged in, the keyboard is
dead.

The loaded routers have ~8 telnet sessions being routed through each
router, onto the ethernet.  The telnet sessions are running a menu driven
piece of software on a unix machine, and so have to carry a lot of data
(redrawn screens etc.).

At one stage I doubled the MTHRESH, NIBUFS, IBUFSIZE and DEFNSOCK in
config.h, but this made no difference.

There is no hardware or software difference between the routers that
crash and those that don't, except hostname and ip addr in autoexec.net.
We used to use 286s, but now they are 386s and this makes no difference.

Any clues anyone?

chris

-----------[000039][next][prev][last][first]----------------------------------------------------
Date:      Thu, 3 Sep 1992 19:23:19 GMT
From:      dcm@csn.org (David C. Menges)
To:        comp.protocols.tcp-ip
Subject:   RS/6000 SLIP to Annex?

We have a customer that wants to run SLIP on a RS/6000.  Apparently
the SLIP that came with AIX can't handle the three prompts our Annex
requires to login.  The other option is to port a public domain SLIP
to the RS/6000.  Any experience here greatly appreciated - please
mail me as I don't frequent this group.

	David Menges, Colorado SuperNet, dcm@csn.org, 303-273-3471.

-----------[000040][next][prev][last][first]----------------------------------------------------
Date:      Thu, 3 Sep 92 20:04:01 GMT
From:      cliffb@cjbsys.bdb.com (cliff bedore)
To:        comp.protocols.tcp-ip
Subject:   Re: Why does my slip line die overnight?

In article <1992Aug30.205245.29936@psg.com> randy@psg.com (Randy Bush) writes:
>ghawkins@unix1.tcd.ie (George C. Hawkins) writes:
>> When I leave my machine overnight and come back in the morning I
>> find my slip connection has died on me.
>
>We see the same thing here in Orygun's RAINet.  We believe (urban legend)
>that the telco does an 02:00 blast to clear the lines.
>-- 
>randy@psg.com   ...!uunet!m2xenix!randy

I've had this happen and it turned out in our case that we were going through a
local PBX that did a midnight reset.  

Cliff


-----------[000041][next][prev][last][first]----------------------------------------------------
Date:      Thu, 03 Sep 1992 21:13:53 GMT
From:      jeanpaul@duteca.et.tudelft.nl (J.P.M. van der Jagt)
To:        comp.sys.dec,comp.unix.ultrix,comp.protocols.tcp-ip,comp.protocols.ppp
Subject:   Source devicedriver SLIP/CSLIP for DEC/Ultrix wanted

Hi All,

I'm trying to install CSLIP (Compressed Serial Line Internet Protocol)
on Ultrix3.1 on a DEC2100.

I already have SLIP running. The kernel-object if_sl.o is distributed
with Ultrix and works fine. But the troughput on 9600 baud is a bit low.
Hence the CSLIP. I ftp'd the Van Jacobson stuff which can be found on
numerous ftp sites and started configuring and compiling. The code contains
a version of if_sl.c, which is derived from 4.3bsd and which is altered to
support the sl_compress routines, which are in the file slcompress.c.
I'm not succeeding in compiling the if_sl.c (the SLIP driver). It seems
that Ultrix differs too much from the 4.3bsd.

So, the question: Does anyone have the source of the if_sl.o that comes
with Ultrix3.1, so that I can add the compress calls to it.
Or: Does someone in any other way got CSLIP running on Ultrix?

It has taken me a whole week so far and I'm getting desperate...
Please email or post.
Regards, John-Paul
-- 
[] J.P.M. van der Jagt [] Systems manager [] Delft University of Technology []
[] Email: jeanpaul@duteca.et.tudelft.nl [] Phone: 31-15-781366 [] unix unix []
[] - He sells C-shells by the seashore - Lauren Bacall - Key Largo - 1948 - []

-----------[000042][next][prev][last][first]----------------------------------------------------
Date:      Thu, 3 Sep 1992 21:20:29 GMT
From:      tengi@princeton.edu (Christopher Tengi)
To:        comp.protocols.tcp-ip
Subject:   looking for tn3270d

I am looking for a tn3270 daemon so that folks around here can telnet
from our VM system to a unix system, and be able to do things in a
full screen mode.  Strangely enough, one of the things that might be
wanted is to tn3270 back out of the unix system to another VM system.
I know that it sounds ugly, but, unfortunately, there really is a need
for this behaviour.

So has anybody out there heard of/written such a beast?  I checked
archie for .*3270.*, but only found things that looked like the user
side, not the server side.

					/Chris

==========----------==========---------+---------==========----------==========

	UUCP:	  ...princeton!tengi		VOICEnet: 609-258-6799
	INTERNET: tengi@princeton.edu		FAX:      609-258-3943
	BITNET:	  TENGI@PUCC

-----------[000043][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1992 22:52:53 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

In article <1992Sep3.090735.3657@arizona.edu> Leonard@Arizona.EDU writes:
>The best and easiest approach is just to use ALL 1's (255.255.255.255) as your
>broadcast address.  That way you don't have to worry about your subnet mask,
>or for that matter even what your own IP address is.  Recommended by 
>RFC-1122, sec. 3.3.6 (discussion section.)

One issue that's not raised in that section is the situation when multiple
IP subnets are on the same link.  The choice between limited broadcasts and
subnet broadcasts affects which group of machines will receive the
broadcasts, and may affect the behavior of the protocols you use.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000044][next][prev][last][first]----------------------------------------------------
Date:      Fri, 4 Sep 1992 02:57:23 GMT
From:      weber@world.std.com (Bob Weber)
To:        comp.protocols.tcp-ip
Subject:   location of various address space proposals

is there someplace where the various proposals to the ietf
regarding solutions to the problem of exhausting the ip address
space are available via anon ftp?
thanks
Bob Weber
-- 
------------------------------------------------------------------------
Robert Weber                       Voice: 617/570-0791 
Northeast Consulting Resources     Fax:   617/523-0150 
85 Devonshire Street, 4th floor   

-----------[000045][next][prev][last][first]----------------------------------------------------
Date:      Fri, 04 Sep 92 09:39:38 GMT
From:      pauln@nuustak.csir.co.za (Paul Nash)
To:        comp.compression,comp.dcom.modems,comp.protocols.misc,comp.protocols.tcp-ip
Subject:   Any experiences with Compressing TDMs?

We are considering installing a pair of Symplex Datamizer compressing
TDMs, to optimise the use of an international leased line.  However, we
would like to find out others' experiences first.  For those who don't
know, the device is basically a TDM with a data compressor attatched to
the aggregate channel.

We would like to know of any good or bad experiences in the following
areas:

1) Reliability:  

2) Throughput:  the local marketing people claim 4:1.  From experiences
                with modems, I would guess that 2:1 is a better bet in
		the real world, but would like to hear what sorts of
		throughput people actually get.

3) Satelites:  we want to use these over a satelite link.  Does the
               increased latency have any nasty side-effects?

The types of traffic that we plan to shove down the line include X.25,
SL/IP or PPP and SNA (perhaps).

If there are any other comments, or other products, I would appreciate
the info.  Mail or posting is great, and I will summarise any mail I get,
if there seems to be any interest.
-- 
 ---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---
 Paul Nash                                         paul@frcs.Alt.ZA
 Box 12475, Onderstepoort, 0110 South Africa         +27-12-8413050

     "God isn't dead, He's just metabolically challenged" -- me

-----------[000046][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 13:34:56 GMT
From:      ccg@tcdsp1.mmm.com ("Charles Ganzhorn")
To:        comp.protocols.tcp-ip
Subject:   Re: looking for tn3270d

In article <1992Sep3.212029.28661@Princeton.EDU>, tengi@princeton.edu (Christopher Tengi) writes:
|> I am looking for a tn3270 daemon so that folks around here can telnet
|> from our VM system to a unix system, and be able to do things in a
|> full screen mode.  Strangely enough, one of the things that might be
|> wanted is to tn3270 back out of the unix system to another VM system.
|> I know that it sounds ugly, but, unfortunately, there really is a need
|> for this behaviour.
|> 
|> So has anybody out there heard of/written such a beast?  I checked
|> archie for .*3270.*, but only found things that looked like the user
|> side, not the server side.

McData, Open Connect Systems, IBM, and Interlink all make TCP/IP
products for IBM mainframes.  Although as of last year, Interlink only
had a MVS product.

These products implement a telnet server (you don't need anything
particularly special to receive tn3270 connections if the CLIENT is
tn3270.) and a telnet client.

Charles.
--
Charles Ganzhorn			E/Mail:  ccganzhorn@mmm.com
3M IS&DP Network Services		AT&T:	 +1 612 736 7122
St. Paul, MN				FAX:	 +1 612 736 7689

-----------[000047][next][prev][last][first]----------------------------------------------------
Date:      Fri, 4 Sep 1992 13:41:22 GMT
From:      bob@MorningStar.Com (Bob Sutterfield)
To:        comp.sys.dec,comp.unix.ultrix,comp.protocols.tcp-ip,comp.protocols.ppp
Subject:   Re: Source devicedriver SLIP/CSLIP for DEC/Ultrix wanted

In article <1992Sep03.211353.19114@donau.et.tudelft.nl> jeanpaul@duteca.et.tudelft.nl (J.P.M. van der Jagt) writes:
   I already have SLIP running. But the troughput on 9600 baud is a
   bit low.  Hence the CSLIP.

CSLIP won't improve your throughput much as compared with SLIP, but it
will help a lot with interactive responsiveness.

CSLIP uses the clever techniques described in RFC-1144 to reduce the
size of the TCP header, and its contribution to the protocol overhead
cost of each packet that must traverse the link.  The TCP header
comprises a very large fraction of a typical single-data-character
interactive telnet packet, but it's a relatively smaller part of a
typical large FTP packet.

So the effect on throughput won't be as dramatic as the effect on
perceived interactive latency.  If you're hoping for a big throughput
jump, you may be disappointed with the degree of improvement that
results from all the work you're expending.

-----------[000048][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1992 11:03:00 +0200
From:      urlichs@smurf.sub.org (Matthias Urlichs)
To:        comp.protocols.tcp-ip,comp.dcom.isdn
Subject:   Re: TCP/IP over ISDN

In comp.protocols.tcp-ip, article <1992Aug31.183707.18401@gandalf.ca>,
  djarrett@gandalf.ca (Dave Jarrett) writes:
> In article <1992Aug28.114002.25416@stc.nato.int> koelman@stc.nato.int writes:
> >> 
> >>  - When the idle timer expires, the ISDN server issue a B-channel 
> >>    connect request to the client ISDN station. The user-user data field
> >>    in the connect request contains a "poll" command. 
> >> 
... which would require both sides of the ISDN connection to keep state about
the TCP connections crossing the ISDN link and to send bogus TCP packets to
check if the system on the other side is still there. IMHO, this is not a
reasonable requirement.

> >> This approach has two advantages over polling using SYNs:
> >>  2. This approach is LAN protocol independant.
> >> 
Since TCP/IP is LAN protocol independent by definition, anything you do would
have to be.
> >
> >This would imply that you get bandwidth for free through the
> >user-to-user data  (unless that is charged separately).
> >I don't think PTTs are that stupid.
User-user data has some valid applications, such as the transmission of
login information, or to check if your voice mail system has any new messages
without actually calling in. ;-)

> >If they are then let's send all our IP traffic this way...
> >(the MTU may be limited though :-))

Q.931 does limit the number and length of such data packets.
The ISDN switch might also raise some alert if you're continuously trying
to open connections, which are all mysteriously rejected; the main reason
is that such processing causes an excessive load on the switch. 
> 
> It is true that user-user data is an optional facility in Q.931.

Unfortunately, the original request came from Germany.
We don't have Q.931 (yet). We also don't have user-user data; maximum
user information transmittable during call setup is 10 bits in forward
direction (device address (1 decimal digit) and 2nd byte of service type)
and 6 bits back (the cause value). So polling is possible, though I'd rather
use it to ask if there's anything in the UUCP queue for the calling system.

-- 
"Life would be much simpler and things would get done much faster if it
weren't for other people"
		-- Blore
-- 
Matthias Urlichs  --  urlichs@smurf.sub.org -- urlichs@smurf.ira.uka.de   /(o\
Humboldtstrasse 7 -- 7500 Karlsruhe 1 -- Germany  --  +49-721-9612521     \o)/

-----------[000049][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 15:49:47 GMT
From:      Jerrod@Lilly.Com (Jerrod T. Carter)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.dcom.lans
Subject:   Re: TCP/IP programming reference material wanted



James 'J' Allard, jallard@microsoft.com writes:
>Finally, if you're thinking about doing some Windows programming over
 TCP/IP,
>it's probably worth your time to check out the recently released Windows 
>Sockets specification available on ftp.uu.net under the
 /vendors/microsoft
>directory. The specification was jointly developed and endorsed by about

Is this spec available at some other FTP site? I am unable to get
connected to
either of the addresses that the name server shows.

-----------[000050][next][prev][last][first]----------------------------------------------------
Date:      Fri, 4 Sep 1992 16:04:38 GMT
From:      bob@MorningStar.Com (Bob Sutterfield)
To:        comp.unix.bsd,comp.protocols.tcp-ip
Subject:   getnetmaskent()?

BSD's getnetent() family of routines works with the contents of
/etc/networks.  Is there a similar getnetmaskent() family for the
information contained in /etc/netmasks?  I note that the Tahoe (or is
it Reno?) <netdb.h> has no mention of a subnet mask, nor of a file
where it might be found.

Must I actually open() the file and parse it myself?  How barbaric! :-)

-----------[000051][next][prev][last][first]----------------------------------------------------
Date:      Fri, 4 Sep 1992 16:48:24 GMT
From:      ronf@panther3.panther.mot.com (Ron Feigen)
To:        comp.sys.dec,comp.unix.ultrix,comp.protocols.tcp-ip,comp.protocols.ppp
Subject:   Re: Source devicedriver SLIP/CSLIP for DEC/Ultrix wanted

Hi Bob,

   I just replied to a post re:SLIP refering to MST but I forgot your e-mail
address.  Here is a copy of the post, perhaps you should contact him.

 
To: dcm@csn.org
Subject: Re: RS/6000 SLIP to Annex?
Newsgroups: comp.protocols.tcp-ip
In-Reply-To: <Bu0oIy.FtB@csn.org>
Organization: Motorola Panther Project, Chandler, AZ
Cc: 
Bcc: 

In article <Bu0oIy.FtB@csn.org> you write:
>We have a customer that wants to run SLIP on a RS/6000.  Apparently
>the SLIP that came with AIX can't handle the three prompts our Annex
>requires to login.  The other option is to port a public domain SLIP
>to the RS/6000.  Any experience here greatly appreciated - please
>mail me as I don't frequent this group.
>
>	David Menges, Colorado SuperNet, dcm@csn.org, 303-273-3471.

David,


  There is an outfit called MorningStar Technology in either Clevland or
Cincinnati OH.  They sell a commercail PPP (like SLIP but better).  I have
used it with Sun O/S and was VERY, VERY pleased.  They have great tech support
as well.  The price of the product is around $400.00. 


-- 

>
Ron Feigen
ronf@panther.mot.com
-- 

>
Ron Feigen
ronf@panther.mot.com

-----------[000052][next][prev][last][first]----------------------------------------------------
Date:      Fri, 4 Sep 1992 19:13:01 GMT
From:      sommerfeld@apollo.hp.com (Bill Sommerfeld)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP reliability

In article <1992Sep4.162458.1@hei.unige.ch> droux@hei.unige.ch writes:

   These two entities are communicating through AF_INET family
   sockets of SOCK_STREAM type. Our Problem is, that a receiver implemented on
   a NeXT receives much more packets than the sender implemented on 
   a Sun workstation sends (1.3 : 1).

   The data exchange is made by normal read() and write() calls followed by an 
   acknowledgement. We transmitt packets of 5000 bytes of unstructered data.
   If a sender send 300 packets (as a concrete example) the receiver receives
   over 400 packets and about 120 of them are invalid. We consider this as a
   serious problem since our system is based on realibility of TCP sockets.
   Is there a special option on sockets or another method to avoid these
   problems?

No, you have a fundamental misunderstanding about TCP.

TCP is byte oriented, not "record" or "packet" oriented.

With sockets, when you do a read() on a TCP socket, your process
either blocks until something is available, or returns immediately
having read the data immediately available (or a prefix of it, if you
passed a buffer shorter than the amount queued for read).

If you want to do record-oriented things over TCP, you have to define
your own record boundaries (e.g., with a byte count followed by the
bytes), and count carefully on the receive end.

					- Bill









-----------[000053][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 19:24:27 GMT
From:      ejcsci@engr.uark.edu (Ed Cooley (sci386))
To:        comp.protocols.tcp-ip,comp.unix.sysv386
Subject:   ISC V3.0 TCP Problems

We are having two problems with an installation of TCP/IP on ISC UNIX V3.0

1 - Telnet are getting a Connection closed by foreign host.  This only 
occurs on the vt (virtual terminals) or standard terminals running multiview.
It even occurs with a loop-back telnet connection.

2 - We are getting blocked at some gateway.  Here at the U of A, we have
access past our gateway to other machines on campus, but get timeouts 
when trying to ping, telnet, ftp off campus.

Any help or ideas would be greatly appreciated.

Thanks in advance,
Ed Cooley

Please respond to ejcsci@engr.uark.edu

-----------[000054][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 16:24:58 +0200
From:      droux@hei.unige.ch
To:        comp.protocols.tcp-ip
Subject:   TCP reliability


TCP/IP BERKELEY SOCKETS: Transmission problems with doublicated packets
-----------------------------------------------------------------------

We are currently implementing a distributed system providing the use of
different computers (different UNIX workstations) connected by Internet.
Our System is based on a server - client model using berkeley sockets.

These two entities are communicating through AF_INET family
sockets of SOCK_STREAM type. Our Problem is, that a receiver implemented on
a NeXT receives much more packets than the sender implemented on 
a Sun workstation sends (1.3 : 1).

As TCP is known as sequenced, reliable and two-way connection based, this
fact is surprising to us. Is it possible to avoid this ? 

We implemented the mechanism as follows:

OPERATING SYSTEMS:

NeXT Cube (System Release 2.1) and SUN Sparcstation 1+ (Sun OS 4.1)

CONNECTION:

a)Server side:
  
  if ((socket_fd = socket(AF_INET, SOCK_STREAM, 0)) < 0)  /* create sockets */
    err_sys("client: error during socket creation");
  our_bzero((char *) &sock_addr,sizeof(sock_addr));
  sock_addr.sin_family = AF_INET;           /* initializes socket addresses */ 
  sock_addr.sin_port = 0;
  sock_addr.sin_addr.s_addr = INADDR_ANY;
  servlen = sizeof(struct sockaddr_in);
  if (bind(socket_fd, (struct sockaddr *)&sock_addr, servlen))
    err_sys("inet_bind_fix: error binding socket");
  if (getsockname(socket_fd, (struct sockaddr *)&sock_addr, &servlen))
    err_sys("inet_bind_fix: error getting socket name");
    
The port number is transferred using inetd to the client.

b)Client side:

  if ((socket_fd = socket(AF_INET, SOCK_STREAM, 0)) < 0)  /* create sockets */
    err_sys("client: error during socket creation");
  if ((foreign_addr = gethostbyname(machine)) == 0)       /* get the server */
    err_sys("unknown host");
  our_bzero((char *) &sock_addr, servlen);
  copy_struct(foreign_addr->h_addr,&sock_addr.sin_addr,
	foreign_addr->h_length);
  sock_addr.sin_family = AF_INET;
  sock_addr.sin_port = htons(port);
  do
  {                                          /* connect to the power server */
    connect_status = connect(socket_fd,
      (struct sockaddr *)&sock_addr, servlen);  
    connect_retries--;
  } while((connect_status < 0) && (connect_retries));
    
SOCKET OPTIONS:

#define SOCKET_BUFFER_LEN 10000                 /* buffer size of a channel */

  int buf_len = SOCKET_BUFFER_LEN;         /* set new socket buffers length */
  
  if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, 
                 (char *)&buf_len, sizeof(buf_len)) == -1)
    err_sys("set_sock_buf: cannot set new socket buffer length");
  if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, 
                 (char *)&buf_len, sizeof(buf_len)) == -1)
    err_sys("set_sock_buf: cannot set new socket buffer length");

The data exchange is made by normal read() and write() calls followed by an 
acknowledgement. We transmitt packets of 5000 bytes of unstructered data.
If a sender send 300 packets (as a concrete example) the receiver receives
over 400 packets and about 120 of them are invalid. We consider this as a
serious problem since our system is based on realibility of TCP sockets.
Is there a special option on sockets or another method to avoid these
problems?

Special Comments:
1. If the server and the client are running on the same computer, the problem
   does not occur.
2. If the NeXT (receiver client) is swapping the number of faulty packets is
   increasing.
3. The number of faulty packets is somehow depending on their size.

We would be very happy about some information as our application is part
of our graduation thesis.

Nicolas Droux & Daniel Liebhart
Biel School of Engineering, Switzerland
Computer Science Dpt.
E-mail: droux@hei.unige.ch
   

-----------[000055][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1992 19:53:01 GMT
From:      trier@slc6.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip,comp.unix.sysv386
Subject:   Re: ISC V3.0 TCP Problems

General advice for any TCP installation with problems similar to those you
encounter:  Check your IP address to make sure it does not duplicate that of
another host.  Make sure your address is in the correct net and subnet for
the cable you are on, and make sure the configured netmask is correct.
Finally, check your configured broadcast address for consistency with the
other hosts on your subnet.

If you want to route off your subnet, make sure you have an appropriate
router configured in, or if you want to use wiretapping for routing info,
make sure that the router is broadcasting a routing protocol that your host
can understand.

From experience here, I'd say that this is the kind of problem that is best
solved by a local networking guru who can sit down at your console and _see_
what is going wrong.  :-)

    Stephen

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000056][next][prev][last][first]----------------------------------------------------
Date:      04 Sep 92 20:02:01 GMT
From:      philc@msmgate.mrg.uswest.com (Phil Corchary)
To:        comp.protocols.tcp-ip
Subject:   ip segment size negotiation  NEED HELP!

HELP!!

Not being a tcp/ip guru or even expert, I am running into a problem that
I 
can't resolve.

the situation: transfering data over a private wire EtherNet (no routing, 
two David 10bT hubs, fully certified) between Mac IIci (8/80 Asante 
Ethernet adapter using  Apple drivers in System 7.0.1 with TuneUp 1.1.1 
and MacTCP 1.1) and Tandem RISC based minimain.

problem: as reported by a sniffer on the line: the segment size coming
down from the Tandem is 1500 ... going up from the Mac to the Tandem
the segment size is being negotiated down to 590 ... WHY?? Is MacTCP
doing this?? I was reading (Comer's Internetworking with TCP/IP)
that it is a SUGGESTED spec, not a requiredpart of the spec for 
this to happen.

Has anyone seen this ... or can anyone help. We are attempting to
benchmark 
for max throughput, and this problem is severly impacting this test ...

ANY HELP APPRECIATED ... please MAIL replies ...

philc

=======================================================================
======
*  Opinions expressed are soley those      | Flames Cheerfully Ignored   
   *
*  of the author. No one else wants them.  |                             
   *
=======================================================================
======
*                    U S West Marketing Resources Group, Inc.            
   *
*          publishers of the U S West Direct Yellow Page Directories     
   *
*                and other Yellow Page and Marketing products            
   *
=======================================================================
======
* Internet:   philc@uswmrg.mrg.uswest.com (preferred)                    
   *
*             phil_corchary@msmgate.mrg.uswest.com (secondary)           
   *
* AppleLink:  PCORCHARY                                                  
   *
* Snail Mail: U S West Marketing Resources, Inc.                         
   *
*             9394 W Dodge Road                                          
   *
*             Omaha, NE 68114                                            
   *
* Office:     402.384.3584                                               
   *
* FAX:        402.384.3862                                               
   *
=======================================================================
======

-----------[000057][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 21:09:44 GMT
From:      paul@sci.ccny.cuny.edu (Paul Chen)
To:        comp.protocols.tcp-ip
Subject:   Test (ignor please)

Test!
-- 
Paul Chen                                CC   CC  N   N  Y   Y   
System Manager                          C    C    N N N    Y      
Science Computer Facility                CC   CC  N   N    Y    -SCIENCE
City College of CUNY, J302            	Internet: paul@sci.ccny.cuny.edu

-----------[000058][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 92 21:27:45 GMT
From:      ms@Germany.EU.net (Marc Sheldon)
To:        comp.protocols.tcp-ip,comp.dcom.isdn
Subject:   Re: TCP/IP over ISDN

urlichs@smurf.sub.org (Matthias Urlichs) writes:
> Unfortunately, the original request came from Germany.
> We don't have Q.931 (yet). We also don't have user-user data; maximum
> user information transmittable during call setup is 10 bits in forward
> direction (device address (1 decimal digit) and 2nd byte of service type)
> and 6 bits back (the cause value). So polling is possible, though I'd rather
> use it to ask if there's anything in the UUCP queue for the calling system.

Sorry to correct you there, but there are not even that many Bits
to transmit. The second byte of the service type (i believe it is
called "Additional information") ist not available because the applicable
German Standard (1TR6) disagrees with itself on the point of transmission
of AI if you use the 64 kBit clear channel data service. According
to the 1TR6 the switch can either transmit AI transparently (so quoted
somewhere at the beginning of the standard) or it can set it to zero
(so quoted somewhere at the end of the SAME standard). We had several
problems with this as we used AI as a protocol identifier just to find
that some german PTT switches (namely those from Alcatel) set AI to
zero and some (namely those from Siemens) pass AI through. After we
informed the DBP Telekom about this inconsistency we were told that
the software on the Siemens switches would be changed to set AI to
zero, thereby making our use of AI impossible. So much for passing
User-to-User-Information.

Cheers,
	Marc
--
Marc R. Sheldon                             e-mail : ms@Germany.EU.net
EUnet Germany
University of Dortmund, CS-Department
P.O.Box 500500                              voice  : +49 231 7552444

-----------[000059][next][prev][last][first]----------------------------------------------------
Date:      Sat, 5 Sep 92 00:58:48 GMT
From:      terry@cs.weber.edu (A Wizard of Earth C)
To:        comp.unix.bsd,comp.protocols.tcp-ip
Subject:   Re: getnetmaskent()?

In article <BOB.92Sep4120430@volitans.MorningStar.Com> bob@MorningStar.Com (Bob Sutterfield) writes:
>BSD's getnetent() family of routines works with the contents of
>/etc/networks.  Is there a similar getnetmaskent() family for the
>information contained in /etc/netmasks?  I note that the Tahoe (or is
>it Reno?) <netdb.h> has no mention of a subnet mask, nor of a file
>where it might be found.
>
>Must I actually open() the file and parse it myself?  How barbaric! :-)

Actually, you need to look at the sources for ifconfig, which is capable
of printing out the netmask.  The netmask is an artifact of the paramters
to ifconfig initially, not an artifact of some attribute of your host
address.  This is especially true if you are subnetting some class of
subnet... for instance, 137.190.16.x as a subnet of the class B network
137.190.x.x.

One thing I generally do, but upon which you wouldn't be able to rely, is
to make a /etc/hosts entry for the netmask based on the network name.  Then
you can gethostent() it.


					Terry Lambert
					terry_lambert@gateway.novell.com
					terry@icarus.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.
-- 
-------------------------------------------------------------------------------
                                                       terry@icarus.weber.edu
 "I have an 8 user poetic license" - me
-------------------------------------------------------------------------------

-----------[000060][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 92 01:28:29 GMT
From:      tomb@studsys.mscs.mu.edu (Tom Baas)
To:        comp.protocols.tcp-ip
Subject:   help setting up SLIP

I am working on setting up SLIP under ISC Unix SVr3.2
I am learing thru reading and experimenting.  Right now I don't have
any problems mainly because I have not been set up on the other end yet.

However, in the meantime.... Any suggestions, instructions, problems, 
recommendations?

I know there are no specifics, but I don't have any yet.  Just, talk to me 
about it.  What can I expect? How do I set things up? What should I look out 
for?  How does it work?  Can I mount the other systems filesystem on my machine?

Thanks to anyone in advance.  :)
-- 
I can accept e-mail and Voice-mail at:
tbaas!tom   or   tom@tbaas.mil.wi.us    or ...spool!tbaas!tom  
or Voice at: 1-414-377-4038

-----------[000061][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1992 03:45:45 GMT
From:      trier@odin.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: ip segment size negotiation  NEED HELP!

In article <1992Sep4.160201.22932@uswmrg.mrg.uswest.com> philc@msmgate.mrg.uswest.com (Phil Corchary) writes:
>problem: as reported by a sniffer on the line: the segment size coming
>down from the Tandem is 1500 ... going up from the Mac to the Tandem
>the segment size is being negotiated down to 590 ...

I just ran into this problem with a VAX running VMS and WIN/TCP.  A PC/IP-
derived FTP couldn't fetch from it, because the VAX wasn't defaulting to a
536-byte maximum segment size (MSS) as it should have.  When I modified the
PC TCP to send the TCP MSS option with an MSS of 536, the VAX *still* sent
1500-byte packets.

One solution is to reduce the MTU on the Tandem to something smaller, if
possible.  This will reduce throughput when talking to machines capable of
the maximum MTU, but will provide compatibility with the Mac.

The solution to our problem, though, was to upgrade WIN/TCP to its current
release.  :-)

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000062][next][prev][last][first]----------------------------------------------------
Date:      Sat, 5 Sep 1992 05:20:37 GMT
From:      graeme@ccu1.aukuni.ac.nz ( Graeme Moffat)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

barmar@think.com (Barry Margolin) writes:

>One issue that's not raised in that section is the situation when multiple
>IP subnets are on the same link.  The choice between limited broadcasts and
>subnet broadcasts affects which group of machines will receive the
>broadcasts, and may affect the behavior of the protocols you use.

What I have observed in this situation (multiple subnets on same physical
cable) is that *many* hosts don't recognise x.y.z.255 as a broadcast on
subnet x.y.z when they are in subnet x.y.w, and respond by resending the
packet to MAC address 00FFBAD1DEAD, and an ICMP redirect to the original
sender! Lovely.

-- 
Graeme Moffat                g.moffat@aukuni.ac.nz \ Time wastes us all, 
Computer Aided Design Centre,  Fax: +64-9-366-0702 /  our bodies & our wits
School of Engineering,    Ph: +64-9-737-999 x8384 /  But we waste time,
University of Auckland, Private Bag, Auckland, NZ \   so time & we are quits

-----------[000063][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 92 11:56:14
From:      Steinar.Haug@delab.sintef.no (Steinar Haug)
To:        comp.unix.programmer,comp.protocols.tcp-ip
Subject:   Re: TCP reliability

> We are currently implementing a distributed system providing the use of
> different computers (different UNIX workstations) connected by Internet.
> Our System is based on a server - client model using berkeley sockets.
> 
> These two entities are communicating through AF_INET family
> sockets of SOCK_STREAM type. Our Problem is, that a receiver implemented on
> a NeXT receives much more packets than the sender implemented on 
> a Sun workstation sends (1.3 : 1).
> 
> As TCP is known as sequenced, reliable and two-way connection based, this
> fact is surprising to us. Is it possible to avoid this ? 

Yes, TCP is sequenced and reliable. But it does *not* provide "record" (i.e.
packet) boundaries which are visible to the application. You must build these
on top of TCP yourself if you need them. (Or you could use UDP, which *will*
allow you to preserve packet boundaries - but then you must build the
necessary reliability yourself on top of UDP. Much easier to make a small
layer on top of TCP which does what you want.)

The operating system is free to turn one or more of your write() calls into
as many packets as it sees fit. I.e.,

- one write call may turn into many packets (write 50000 bytes and you will
 get approx. Ethernet 34 packets)

- many write calls may turn into one packet (call write many times with 1 byte
buffers, and the OS will probably try to coalesce your writes into bigger
packets)

Similarly, on the receiving side, each read() may return less data than you
asked for - this is also perfectly acceptable.

> The data exchange is made by normal read() and write() calls followed by an 
> acknowledgement. We transmitt packets of 5000 bytes of unstructered data.
> If a sender send 300 packets (as a concrete example) the receiver receives
> over 400 packets and about 120 of them are invalid. We consider this as a

I'm not sure I understand what you mean by saying "120 of them are invalid".
Could you be more specific? In the first place, what do you mean by a packet?
Is it one write() call, or is it for instance an Ethernet packet? From your
description it sounds more like one write() call.

TCP *is* reliable, if correctly implemented. I have used TCP on Suns for
years, and it works just fine. I have no reason to believe that it does not
work correctly on the NeXT...

Steinar Haug, system/networks administrator
SINTEF DELAB, University of Trondheim, NORWAY
Email: Steinar.Haug@delab.sintef.no, 
	sthaug@idt.unit.no, steinar@tosca.er.sintef.no

-----------[000064][next][prev][last][first]----------------------------------------------------
Date:      Sat, 5 Sep 1992 09:04:00 +0000
From:      vadim@cix.compulink.co.uk (Vadim Lebedev)
To:        comp.protocols.tcp-ip
Subject:   Re: ka9q router unstable under

In-Reply-To: <1992Sep3.162715.9190@frcs.Alt.ZA> chris@frcs.Alt.ZA (Chris Old)

 in article:  : <1992Sep3.162715.9190@frcs.Alt.ZA>
  chris@frcs.Alt.ZA (Chris Old) writes

> We have a mini internet, with four ka9q routers on the main ethernet
> backbone, connecting remote sites via slip.  The version of ka9q has a
> version number of 911229 (7 months ago).  All we did to it was undef the
> various hardware and software options that we didn't need in config.h.
> Otherwise it is stock standard.
 [text deleted]

Chris, there are a lot of things that coul go wrong...
1) You shoul use STACKS=0,0 in your config.sys
2) You should verify the presense of 16550 support in your version of ka9q,
    if there is no you can add it very easily
3) You say the he heavy loaded routers have ~8 telnet seesion passing
   though, are they on the same UART or on different ones?


  Vadim.

-----------[000065][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1992 01:59:29 GMT
From:      mfitz@devnull.West.Sun.COM (Mark Fitzpatrick - Systems Engineer)
To:        comp.protocols.tcp-ip
Subject:   ?trapping a failed...NEVER MIND

I figured it out.... signal (SIGPIPE, handler);

Adios,
Mark


-----------[000066][next][prev][last][first]----------------------------------------------------
Date:      Sun, 6 Sep 1992 22:28:36 GMT
From:      dank@blacks.jpl.nasa.gov (Daniel R. Kegel)
To:        comp.protocols.tcp-ip
Subject:   Is there a Internet distance measure?

Hi,
I'm looking for an automatic way to select the 'nearest' host from a list,
e.g. such that an Archie client automatically chooses a nearby
site rather than one on the other side of the globe.

Anyone know how to do this?
- Dan K. (dank@blacks.jpl.nasa.gov)

-----------[000067][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 92 14:07:24 PST
From:      lukec@terapin.com (Luke Chastain)
To:        comp.protocols.tcp-ip
Subject:   Kermit

Is Kermit anygood to use?

                                       - thanx


-----------[000068][next][prev][last][first]----------------------------------------------------
Date:      Mon, 07 Sep 92 07:58:23 GMT
From:      fitz@wang.com (Tom Fitzgerald)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

graeme@ccu1.aukuni.ac.nz ( Graeme Moffat) writes:

> What I have observed in this situation (multiple subnets on same physical
> cable) is that *many* hosts don't recognise x.y.z.255 as a broadcast on
> subnet x.y.z when they are in subnet x.y.w, and respond by resending the
> packet to MAC address 00FFBAD1DEAD, and an ICMP redirect to the original
> sender! Lovely.

Wow, I thought I'd seen all the bad side effects of mixing subnets on the
same cable.  What kind of systems are these?

Another common side effect is that all the systems on x.y.w ARP for the IP
address x.y.z.255, causing a real broadcast storm.

-- 
Tom Fitzgerald   Wang Labs       fitz@wang.com   "I went to the universe today;
1-508-967-5278   Lowell MA, USA                   It was closed...."


-----------[000069][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1992 00:06:45 -0700
From:      tli@skat.usc.edu (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing to disconnected subnets

In article <Bu8Gnx.1zz@world.std.com> weber@world.std.com (Bob Weber) writes:
    
    Does OSPF and/or IGRP support disconnected subnets of the 
    same Class B in this fashion?  Can this work under any 
    circumstances? 

Yes, OSPF will support this.  You will need at least version 9.1 to do
this with ciscos.  To do this with IGRP, you will have to assign a
secondary subnet across the entire backbone.  All routers on the
backbone should have a secondary address from this subnet.

-- 
Tony Li - Escapee from the USC Computer Science Department          tli@usc.edu
		       The net is not what it seems.

-----------[000070][next][prev][last][first]----------------------------------------------------
Date:      Mon, 7 Sep 1992 15:29:26 GMT
From:      stewart@tweety.cis.udel.edu (John Stewart)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

In article <dank.715818516@blacks> dank@blacks.jpl.nasa.gov (Daniel R. Kegel) writes:
>I'm looking for an automatic way to select the 'nearest' host from a list,
>
>Anyone know how to do this?

There is a utility called "traceroute" that (among other things) counts the
number of hops between your site and the destination.

Good luck.

/jws

-----------[000071][next][prev][last][first]----------------------------------------------------
Date:      Mon, 7 Sep 1992 19:08:06 GMT
From:      fff@microplex.com (Fred Fierling)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

In article <dank.715818516@blacks> dank@blacks.jpl.nasa.gov (Daniel R. Kegel) writes:

>I'm looking for an automatic way to select the 'nearest' host from a list,
>e.g. such that an Archie client automatically chooses a nearby
>site rather than one on the other side of the globe.

To achieve this the latitude and longitude for a given host would have to be
listed in the DNS records.  All I've seen so far is some discussion (I think
in this newsgroup) about storing such information in a new type of record or
in existing record types.  I think it's a good idea.

For your purposes wouldn't you be more interested in the host with the best
connectivity not proximity?  If you can connect to a host on the other side
of the country through a T1 link, that would be better than a host across
the street through a 9600 bps line.  Although I suppose if you had a choice
between two hosts, both connected by a T1 link, you'd prefer the closest one...
-- 
Fred Fierling    fff@microplex.com       Tel: 604 875-1461   Fax: 604 875-9029
Microplex Systems Ltd   265 East 1st Avenue   Vancouver, BC   V5T 1A7,  Canada

-----------[000072][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 92 19:39:03 GMT
From:      dank@blacks.jpl.nasa.gov (Daniel R. Kegel)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

Thanks to all who responded.
Perhaps I'll send out an ICMP ping packet to all the hosts on the list
that match some simple heuristic (e.g. same top-level domain preferred),
and take the first host to respond as the 'closest'.

Eventually, it might be nice to access the SoftPages database to
determine the answer from global routing info; I hope they make it available 
via some other method than x.500 until that is universally accepted.
- Dan Kegel

Original question:
>I'm looking for an automatic way to select the 'nearest'
>host from a list e.g. such that an Archie client
>automatically chooses a nearby site rather than one
>on the other side of the globe.

From: Rahul Dhesi <dhesi@rahul.net>
>[Use traceroute; it gives # of hops and delay in ms. Rather expensive to run.]

From: Glenn Mansfield <glenn@aic.co.jp>
> There is a project called the SoftPages project (SPP) which is 
> attempting to solve just this problem- it finds the the closest
>  repository (ftp-server)containing the file sought by a user.
> It makes use of a X500 based directory service that contains the
> above mentioned network configuration info.
> The project is currently in the experimental stage- for more info
> you can write to spp@aic.co.jp or to join the discussion group
> write to spp-request@aic.co.jp. Related documents are available 
> from ftp.tohoku.ac.jp in ~ftp/pub/spp .

From: Earle Ake <ake@DAYVC.Dayton.SAIC.COM>
>	I would use the ping utility and traceroute to find the shortest path
>(traceroute) to a site and the shortest propagation time (ping) to the site.

From: cmaeda@MC8.MACH.CS.CMU.EDU
>Only heuristic methods (ie you're on your own).  In theory, internet
>domain names have nothing to do with geography.  "Near" is also a
>function of connectivity which is also unrelated to geography.
>Basically you would have to manage a lot of the kind of information
>that the internet was designed to let you ignore.

-----------[000073][next][prev][last][first]----------------------------------------------------
Date:      Mon, 7 Sep 1992 21:04:12 GMT
From:      ckd@eff.org (Christopher Davis)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

Fred> == Fred Fierling <fff@microplex.com> 

 Fred> If you can connect to a host on the other side of the country
 Fred> through a T1 link, that would be better than a host across the
 Fred> street through a 9600 bps line.

I have personally seen a pathological example of this.  Two sites,
literally two blocks apart, that (at the time) had packets between them
route through Maryland, Michigan, and New Jersey to get back to
Cambridge, Massachusetts.  It really was closer to get to machines in
Virginia than it was to get to the machine down the street.

Now the packets just have to cross the Charles River twice :)
--
Christopher Davis   | ]CALL -151       
<ckd@eff.org>       | *300: 20 DD FB 60
System Administrator| *3D0G            
EFF +1 617 864 0665 | ]CALL 768        

-----------[000074][next][prev][last][first]----------------------------------------------------
Date:      Mon, 7 Sep 1992 21:55:40 GMT
From:      jhfreeman@iworks.ecn.uiowa.edu (James H Freeman)
To:        comp.protocols.tcp-ip
Subject:   10base-T Networking, How can I tell if existing wiring works?

We are planning on converting a lab from serial connections to 10baseT.
Instead of replacing the wire, we would like to try to use the existing wire.
It was installed 3 years ago, but I do not know how twisted it is.  The 
building is going to be remodeled in 1 year, so we want to limits costs.

It is a small project: 8 connections that run on 2 pair wire, put RJ45's
on each end & connect to hub.  Each run is ~50 meters, we just got a 
Microtest pair scanner to help with the testing.

What things will I have to look out for & what tips does the gurus on
the net have on installation?

Hints I have heard of & are planning to implement:
	Replace 66 blocks with 110 or Krone.
	Twist the ends of the wires all the way to the blocks.
	Use RJ45's


Lastly, Is this the correct pinout for 2 pair 10base-T on both ends?

Diagram of RJ45 socket
>   1      8
> +----------+          Pin  Function
> | |||||||| |           1   Transmit data +
> | |||||||| |           2   Transmit data -
> |          |           3   Receive data +
> +-+      +-+           6   Receive data -
>   +-+  +-+
>     +--+        (Taken from the AT&T AUI adapter installation manual)


Thanks in advance.
Bruce Cantrall

PS. My IEEE spec book on 10baseT is coming in the mail soon.  This may answer
some of the questions.

------\   Bruce Cantrall, Vax & Apollo System Manager
------/   Cornell College Academic Computing 
Mail      600 First St, Mount Vernon, IA 52314 U.S.A.
Telephone 319.895.4414  Internet:  Bruce@Cornell-Iowa.edu

-----------[000075][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 92 21:59:11 GMT
From:      trevor@trevan.uucp
To:        comp.protocols.tcp-ip
Subject:   Problems with Intel Slip Driver

I have installed the Intel slip driver on a Consensys Unix System V rel 4.0.3.
When using a dialup a slip session can be started and finished ok.
When someone tries subsequently to login the /etc/issue message is displayed
without carrage returns. Any character typed in at the login prompt clears the
call. The slip driver pops the ldterm and ttcompat modules before attaching.
It looks as though the modules are not being pushed by ttymon before restarting the service.

I would be grateful for any help.
Has anyone ported dialupip for SysV?

-- 
				regards trevor
------------------------------------------------------------------------------
Trevor Harris				UUCP: trevor@trevan.uucp
Trevan Designs Limited
5 Blackwater Close, Ash,
Aldershot, Hants, UK.
GU12 6JS

-- 
				regards trevor
------------------------------------------------------------------------------
Trevor Harris				UUCP: trevor@trevan.uucp
Trevan Designs Limited

-----------[000076][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 92 21:59:17 GMT
From:      todd@toolz.uucp (Todd Merriman)
To:        comp.protocols.tcp-ip,comp.unix.sysv386
Subject:   Re: ISC V3.0 TCP Problems

trier@slc6.ins.cwru.edu (Stephen C. Trier) writes:

>General advice for any TCP installation with problems similar to those you
>encounter:  Check your IP address to make sure it does not duplicate that of
>another host.  Make sure your address is in the correct net and subnet for
>the cable you are on, and make sure the configured netmask is correct.
>Finally, check your configured broadcast address for consistency with the
>other hosts on your subnet.

I would like to be pointed to a source of information that covers these
subjects a bit better than the ISC installation manual.

Any suggestions?

* * * * * * * * * * * * * * * * * * * * * * * * * * * * *********************
* Todd Merriman - Software Toolz, Inc. +1 706 889 8264  * Maintainer of the *
* 8030 Pooles Mill Dr., Ball Ground, GA 30107-9610      * Software          *
* UUCP: ...!emory!slammer!toolz!todd                    * Entrepreneur's    *
* Internet:  todd%toolz.uucp@mathcs.emory.edu           * Mailing List      *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * *********************

Never let your schooling interfere with your education.

-----------[000077][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 92 22:52:51 GMT
From:      shri@legato.cs.umass.edu (H.Shrikumar{shri@ncst.in})
To:        comp.protocols.tcp-ip
Subject:   Re: Is info on the Delta-t protocol available electronically?

In article <STEINAR.HAUG.92Aug21174944@delab.sintef.no> 
Steinar.Haug@delab.sintef.no (Steinar Haug) writes:

>I would like to understand the Delta-t protocol better, and therefore I need
>more info. My questions are:
>
>- Is any info on the Delta-t protocol available electronically? Anonymous
>FTP would be fine.
>
>- Is the Delta-t protocol still being used?


  This does not answer the questions above directly ... but just adds a
  comment to it.

  Back a year or more ago, I had implemented a protocol that ws inspired
by the delta-T protocol, and attempted to Unify UDP and TCP in some
sense, and work well over Thin-Wires. (below 300 bauds). The product of
that work was called ALAP-GUT (with ALAP providing an IP like service
with cut-thru routing to avoid packet framing delays) and GUT providing
a Generalised User Transport Service.

   The effort was to gateway both TCP and UDP traffic back and forth
and was achieved quite OK.

   In fact, this proved to be good enough to run telnet and talk over
lines as slow as 50 bauds (yup, I needed to set up large Wide Area
networks using Teleprinter lines !! Nightmarish ... but I'm not
talking in my sleep :-) , running multiple hops, with the instant
connection establishment features of GUT, adapted from Delta-T and the
cut-thru source-routing of ALAP contributing very well.

   I guess I must wrap it up and write about it sometime ... Guess
I need some pushing ! ;-)

   Is Delta-T being used someplace else ... I'd be very very keen to know!


-- shrikumar ( shri@legato.cs.umass.edu, shri@iucaa.ernet.in )

PS: Since this is the firt time I've had the oportunity to mention
this, let me mention my tnx to Jon Postel for digging out the 
elusive IEN 116 for me when I was looking desparately for it, and
also to the 60 students of PGDST90 at NCST, Bombay, who very eagerly
implemented the stuff with great enthusiasm. I'd dedicate this to
them one day, when I get to wrapping it up.


-----------[000078][next][prev][last][first]----------------------------------------------------
Date:      Mon, 7 Sep 1992 22:57:38 GMT
From:      gtoal@ibmpcug.co.uk (Graham Toal)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

In article <CKD.92Sep7170409@loiosh.eff.org> ckd@eff.org (Christopher Davis) writes:
:I have personally seen a pathological example of this.  Two sites,
:literally two blocks apart, that (at the time) had packets between them
:route through Maryland, Michigan, and New Jersey to get back to
:Cambridge, Massachusetts.  It really was closer to get to machines in
:Virginia than it was to get to the machine down the street.
:
:Now the packets just have to cross the Charles River twice :)

My log in to my mail machine goes from britain to the usa to the netherlands.
The packets are rather damp when they arrive and take about as long to get
there as by UPS. :-(

G
-- 

-----------[000079][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 00:14:20 GMT
From:      weber@world.std.com (Bob Weber)
To:        comp.protocols.tcp-ip
Subject:   Routing to disconnected subnets


Consider the following situation that might occur within a firm's 
enterprise wide internet.

subnets<->Router<->Bridged backbone network<->Router<->subnets

The bridged backbone network has IP addresses within a flat, 
unsubnetted Class B address space. The  other routers and their 
associated networks have IP addresses within a single subnetted 
Class B address space.

Does OSPF and/or IGRP support disconnected subnets of the 
same Class B in this fashion?  Can this work under any 
circumstances? If
not, what are the alternatives, going to lots of Class C address 
around the edges?
 
Thanks
Bob Weber
-- 
------------------------------------------------------------------------
Robert Weber                       Voice: 617/570-0791 
Northeast Consulting Resources     Fax:   617/523-0150 
85 Devonshire Street, 4th floor   

-----------[000080][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 92 00:57:24 GMT
From:      dsc3jfs@nmrdc1.nmrdc.nnmc.navy.mil (Jim Small)
To:        comp.protocols.tcp-ip
Subject:   IP using MAIL (or other) as transport medium


A while back, someone mentioned using MAIL as the transport medium
for IP.  I think the reason was to have a tcp connection through a 
network so firewalled that only SMTP connections would get through  . .

Anyway, I have a similar desire.  Except I'll be using either UDP or
TCP packets between two sites.  (I can call into one site but not with slip,
the second site is on a distant network that I cannot call locally, but
that I wish to access as though directly connected . . .)

Would ANYONE with suggestions with existing code to do this or something
similar, please contact me.  (RS6000 code prefered . ... [cuz I'm lazy])

-- 
I hate the 3B2
The 3B2 can bite me.

-----------[000081][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 92 01:45:36 GMT
From:      jbryans@beach.csulb.edu (Jack Bryans)
To:        comp.protocols.tcp-ip
Subject:   Re: Kermit

In article <lukec.1duo@terapin.com> lukec@terapin.com (Luke Chastain) writes:
> Is Kermit anygood to use?

Yes.

Jack

-----------[000082][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 07:16:40 EDT
From:      Peter M. Weiss <PMW1@psuvm.psu.edu>
To:        comp.protocols.tcp-ip
Subject:   Re: 10base-T Networking, How can I tell if existing wiring works?

For excellent discussions on Ethernets and 10Base-T, review the
notebook archives of big-lan on bitnet node suvm or ~ftp syr.edu.

/Pete - pmw1@psuvm.psu.edu

-----------[000083][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 04:11:54 GMT
From:      skl@wimsey.bc.ca (Samuel Lam)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing to disconnected subnets

In article <Bu8Gnx.1zz@world.std.com>, weber@world.std.com (Bob Weber) wrote:
>Can this work under any circumstances?

One way to do it is to allocate one of the subnets from the
subnetted class B network to run on the bridged backbone
parallel to the unsubnetted class B network.  You will then
be running two IP (sub)networks on the bridged backbone where
one of them "glues" all the disconnected class B subnets
together.  i.e. Your subnetted class B is now contiguous.

You do have to pay extra attention in configuring the
backbone routers, though.

...Sam
-- 
<skl@wimsey.bc.ca>

-----------[000084][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 09:06:37 GMT
From:      jrg@otter.hpl.hp.com (John Grinham)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

Is traceroute a public domain utility, or is it a vendor specific program ?
If it's PD does anyone know where I could pick up a copy ? 


regards,
John.

-----------------------------------------------------------------------------
John Grinham.  HP Labs ISC, Filton Road, Stoke Gifford, Bristol BS12 6QZ, UK.
Tel: (0)272 799910	Fax: (0)272 790554 	Telex: 449206
jrg@hplb.hpl.hp.com
-----------------------------------------------------------------------------


-----------[000085][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 92 09:19:43 GMT
From:      amoss@huji.ac.il (Amos Shapira)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

dank@blacks.jpl.nasa.gov (Daniel R. Kegel) writes:

>Thanks to all who responded.
>Perhaps I'll send out an ICMP ping packet to all the hosts on the list
>that match some simple heuristic (e.g. same top-level domain preferred),
>and take the first host to respond as the 'closest'.

If you are going on the "ping" approach then you might be interested in
the fping utility posted on the net a few weeks ago.  It will alow you to ping
several hosts simultanously.

In general, I think you just better find (with traceroute) what's the closest
server network-wise and wire it into the configuration of the programme, with
the other servers as a backup.  Though there are chances that this server might
not be the best one to use all the time, I estimate the chances of a really
drastic change due to network failures don't worth the time and network
resources required to check the closest server every time you run the utility.

Also note that hosts are able to answer ping without being fully functional
(e.g. during system boot or shutdown or while in single user), so the ping
aproach is not 100% foolproof.

Cheers,
--
--Amos Shapira (Jumper Extraordinaire)
C.S. System Group, Hebrew University, Jerusalem, Israel
amoss@cs.huji.ac.il

-----------[000086][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 14:06:36 GMT
From:      idmrmb@gsusgi1.gsu.edu (Mark Buffington)
To:        comp.protocols.tcp-ip
Subject:   What is NQS and/or DQS?

Hello. I am trying to find out info on NQS/DQS. I have heard various
sources allude to them as some form of remote queuing for background
processes. If you can give me a push in the right direction I'd
greatly appreciate it!

Thanks,
 mark..
-- 
  __      __
 |__)|\/||__)    R. Mark Buffington     |  DBA - Database Group
 |  \|  ||__)    idmrmb@gsusgi1.gsu.edu |  GSU Wells Computer Center
----------------------------------------|  University Plaza
"I'm not bad, I'm just drawn that way." |  Atlanta, Georgia 30303

-----------[000087][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 14:44:59 GMT
From:      tengi@princeton.edu (Christopher Tengi)
To:        comp.protocols.tcp-ip
Subject:   Re: looking for tn3270d

In article <1992Sep4.133456.21335@mmm.serc.3m.com>, ccg@tcdsp1.mmm.com ("Charles Ganzhorn") writes:
|> In article <1992Sep3.212029.28661@Princeton.EDU>, tengi@princeton.edu (Christopher Tengi) writes:
|> |> I am looking for a tn3270 daemon so that folks around here can telnet
|> |> from our VM system to a unix system, and be able to do things in a
|> |> full screen mode.  

[text deleted]

|> 
|> McData, Open Connect Systems, IBM, and Interlink all make TCP/IP
|> products for IBM mainframes.  Although as of last year, Interlink only
|> had a MVS product.
|> 
|> These products implement a telnet server (you don't need anything
|> particularly special to receive tn3270 connections if the CLIENT is
|> tn3270.) and a telnet client.
|> 

Thanks for the info, but I don't need code to run on the VM side.
What I need is a way to make the UNIX system behave as thought it were
a VM system (ie as though it had the FAL or equiv. telnetd running),
for "pass thru" telnet purposes, and also for full screen programs
running on the UNIX box.  I realize that alot of work would be
involved, and was hoping that somebody might have already walked down
this path.  If not, my management may or may not (probably not) decide
that it is a worthwhile way to allocate my time.


==========----------==========---------+---------==========----------==========

	UUCP:	  ...princeton!tengi		VOICEnet: 609-258-6799
	INTERNET: tengi@princeton.edu		FAX:      609-258-3943
	BITNET:	  TENGI@PUCC

-----------[000088][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 92 15:16:10 GMT
From:      elefante@flash.ATC.Olivetti.Com (GUEST Luigi Elefante)
To:        comp.protocols.tcp-ip
Subject:   SLIP driver for UNIX 4.0


Hi,

I download a version of slip driver for Unix 4.0 from wuarchive.wustl.edu.
I would like to know an explanation about it, if possible.
I read in slattach(1M) manual pages:
	The - option causes stdin to be used as the device to be attached
	to the SLIP driver. This option is used to convert a remote login
						   ^^^^^^^^^^^^^^^^^^^^^^
	line to a SLIP connection.
	^^^^^^^^^^^^^^^^^^^^^^^^^

and in README file:
	The - option.... This is only to be used by a dial slip utility
	so that a slip user can login and use the same line as the device
	to be attached to slip.

I don't understand what do you mean this and how the slip utility work.

slattach -d /dev/tty00 9600 sl0    and
slattach host sl0 		   work correctly.


Can you help me to understand this?

Thanks in advance for Your attention

Your sincerely
Massimo Barontini
Olivetti consultant
e-mail: maxb@king.ICO.Olivetti.Com


-----------[000089][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 15:18:26 GMT
From:      dcarr@gandalf.ca (Dave Carr)
To:        comp.compression,comp.dcom.modems,comp.protocols.misc,comp.protocols.tcp-ip
Subject:   Re: Any experiences with Compressing TDMs?

In <1992Sep04.093938.26595@nuustak.csir.co.za> pauln@nuustak.csir.co.za (Paul Nash) writes:

>We are considering installing a pair of Symplex Datamizer compressing
>TDMs, to optimise the use of an international leased line.  However, we
>would like to find out others' experiences first.  For those who don't
>know, the device is basically a TDM with a data compressor attatched to
>the aggregate channel.

First, the units are STATISTIC multiplexers, not Time Division Multiplexers.
Big difference.  A traditional TDM will just interleave the data from
multiple channels, usually at the bit level, into one data stream.
TDMs are transparent to the data.

A statmux typically allocates the composite link bandwidth on demand.
Thus, one channel can use up to all the bandwidth available on the link
if no other channels have nothing to send.

>We would like to know of any good or bad experiences in the following
>areas:
 
>1) Reliability:  

Better than a traditional TDM.  Statmuxes run error corrected links.
As far as hardware goes, the Symplex box (at least the last one I looked
at) used dual Z80's, DRAM, EPROM, etc.  Should have a low MTBF. 
The've been making the product for some 10 or more years.  I would
hope it's reliable by now :-)

>2) Throughput:  the local marketing people claim 4:1.  From experiences
>                with modems, I would guess that 2:1 is a better bet in
>		the real world, but would like to hear what sorts of
>		throughput people actually get.

Well, there are several factors here.  First, Symplex does spoofing.
They emulate the protocol to remove unnecessary overhead from polls and
protocol headers and trailers.  This is protocol dependent and will
vary between applications.

Second,  async data overhead (stop, start bits) are removed.

Third, the symplex unit uses several algorithms in parallel, and choses
the best, not just MNP5 or V.42 bis.  I believe it is superior in this
respect.

But be realistic.  It depends on the data.  I would be comfortable
with a claim of 3:1 for these units over a spead of applications.

>3) Satelites:  we want to use these over a satelite link.  Does the
>               increased latency have any nasty side-effects?

The statmux/compression delay is insignificant compared to the .25 second
delay from the satellite hop.  In fact, it should improve the line
delay.  Two reasons for this.  First the compressed packet is shorter
and will get tghere sooner.  It is also less likely to be corrupted.
Satellite hops don't have the best error rate :-)  The statmux will
restansmit sooner than the application can.

>The types of traffic that we plan to shove down the line include X.25,
>SL/IP or PPP and SNA (perhaps).
 
>If there are any other comments, or other products, I would appreciate

I would recommend our compressing stat mux only if cost or number
of channels are issues.

I don't think we compare to Symplex w.r.t. compression.  Their's is a
fine unit.  I believe we did or still do OEM it.
-- 
Dave Carr               | dcarr@gandalf.ca       | It's what you learn, 
Principal Designer      | TEL (613) 723-6500     | after you know it all,
Gandalf Data Limited    | FAX (613) 226-1717     | that counts. 

-----------[000090][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 15:22:59 GMT
From:      ralf@chpc.org (Ralph Valentino)
To:        comp.protocols.tcp-ip
Subject:   RPC memory cleanup

I've written an RPC service and it works, except that it has a memory
leak.  I've added a number of xdr_free()'s everywhere I could think to
put one and that cut the leak from 1k/request to 100bytes/request, but
I just can't figure out where to put the remainder.  (I'm assuming
it's a missing rpc related *_free() because the program itself never
malloc()'s.)

Is there a list of which calls need [xdr|svc]_free()'s.  The nutshell
book isn't very useful here.  Does anyone have a suggestion as to how
I can approach tracking this down?  The server uses tcp and udp, both
client and server calls, one way, followup, and broadcast datagrams,
and does it's own multitasking in a single thread of execution (no
threads or forks).  It handles 1000+ requests a day, so even 100 bytes
per request becomes significant.

If it matters, the machines are a number of Decstation 5100's and
3100's.  The _svc.c _clnt.c and _xdr.c files were created through
rpcgen (though I had to hand modify _xdr.c by hand).

-Ralph
===============
Ralph Valentino   (ralf@chpc.org)  (ralf@wpi.wpi.edu)
Hardware Engineer,  Worcester  Polytechnic  Institute
Center for High Performance Computing, Marlborough MA

-----------[000091][next][prev][last][first]----------------------------------------------------
Date:      Tue, 8 Sep 1992 21:47:23 GMT
From:      pascoe@rocky.gte.com (Dave Pascoe)
To:        comp.protocols.tcp-ip,comp.sys.sun.admin
Subject:   SunOS 4.1.1 panics with SLIP-4.1

System:  SPARC-2 running SunOS 4.1.1
Situation: Just installed slip-4.1 and was attempting to test it

There were two separate panic situations:

#1:

Sep  8 08:42:02 rocky slip-attach[491]: attaching csl0: local 100.14.40.6 
remote 100.14.40.50 mask 255.0.0.0
Sep  8 08:51:35 rocky vmunix: /home: file system full
Sep  8 08:51:36 rocky vmunix: /home: file system full
Sep  8 09:01:38 rocky su: 'su root' succeeded for pascoe on /dev/ttyp1
Sep  8 09:14:14 rocky vmunix: assertion failed: vp->v_stream == stp, file: ../
../os/str_io.c, line: 609
Sep  8 09:14:14 rocky vmunix: panic: assertion failed
Sep  8 09:14:14 rocky vmunix: syncing file systems... [20] [18] [14] [8] [3] 
done


#2:

Sep  8 16:59:35 rocky slip-attach[2389]: attaching csl0: local 100.14.40.6 
remote 100.14.40.50 mask 255.255.255.0
Sep  8 16:59:35 rocky slip-attach[2389]: csl17091: TCP header compression  and
enabled100.14.40.50
Sep  8 17:13:07 rocky vmunix: BAD TRAP
Sep  8 17:13:07 rocky vmunix: pid 70, `in.routed': Data fault
Sep  8 17:13:07 rocky vmunix: kernel read fault at addr=0xff151090, pme=0x0
Sep  8 17:13:07 rocky vmunix: Sync Error Reg 80<INVALID>
Sep  8 17:13:07 rocky vmunix: pc=0xf804a764, sp=0xf813ed58, psr=0x114006c2, 
context=0xc
Sep  8 17:13:07 rocky vmunix: g1-g7: f805781c, 200, 0, f3, f8262000, f816d400,
 f816a000
Sep  8 17:13:07 rocky vmunix: Begin traceback... sp = f813ed58
Sep  8 17:13:07 rocky vmunix: Called from f8057a18, fp=f813edb8, args=ff66a180
8d ff66a191 ff149158 3 1
Sep  8 17:13:07 rocky vmunix: Called from f8127034, fp=f813ee20, args=ff66a180
 ff035309 ff66a191 ff66a180 80 ff1490e0
Sep  8 17:13:07 rocky vmunix: Called from f8126d6c, fp=f813ee80, args=ff0016f8
 1 ff0359b0 ff001ae8 1 0
Sep  8 17:13:07 rocky vmunix: Called from f803f604, fp=f813eee0, args=0 0 
f8145c00 110001e6 0 ff0016f8
Sep  8 17:13:07 rocky vmunix: Called from f803f4a4, fp=f813ef40, args=114000c1
 f8126d14 0 0 114001e7 f8204b4c
Sep  8 17:13:07 rocky vmunix: Called from f8005f34, fp=f813efa0, args=f81055a8
 114000c1 1 1 ff001ed4 f8204b4c
Sep  8 17:13:07 rocky vmunix: Called from f805cd04, fp=f8261ca0, args=100 
114001a2 f8145c00 f81f6968 118006e1 0
Sep  8 17:13:07 rocky vmunix: End traceback...
Sep  8 17:13:07 rocky vmunix: panic: Data fault
Sep  8 17:13:07 rocky vmunix: zs1: silo overflow
Sep  8 17:13:07 rocky vmunix: syncing file systems... done
Sep  8 17:13:07 rocky vmunix: 00000 low-memory static kernel pages
Sep  8 17:13:07 rocky vmunix: 00888 additional static and sysmap kernel pages
Sep  8 17:13:07 rocky vmunix: 00000 dynamic kernel data pages
Sep  8 17:13:07 rocky vmunix: 00184 additional user structure pages
Sep  8 17:13:07 rocky vmunix: 00000 segmap kernel pages
Sep  8 17:13:07 rocky vmunix: 00000 segvn kernel pages
Sep  8 17:13:07 rocky vmunix: 00130 current user process pages
Sep  8 17:13:07 rocky vmunix: 00089 user stack pages
Sep  8 17:13:07 rocky vmunix: 01291 total pages (1291 chunks)
Sep  8 17:13:07 rocky vmunix: 
Sep  8 17:13:07 rocky vmunix: dumping to vp ff035f64, offset 55096
Sep  8 17:13:07 rocky vmunix: 1291 total pages, dump succeeded
Sep  8 17:13:07 rocky vmunix: rebooting...

Is there some patch from Sun for SunOS 4.1.1 that will prevent this?  Any 
hints will be appreciated.
-- 
Dave Pascoe			|  Amateur Radio:  KM3T
Internet: pascoe@rocky.gte.com	|  Packet Radio: km3t @ w2xo.pa
GTE/SCSD - Needham Heights, MA	|  (617) 455-5704

-----------[000092][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 92 23:48:59 GMT
From:      wcs@cbnewsh.cb.att.com (Bill Stewart 908-949-0705 erebus.att.com!wcs)
To:        comp.protocols.tcp-ip,comp.mail.misc
Subject:   POP2 vs POP3 mail server protocols?


What are the advantages and disadvantages of using POP2 vs. POP3?
I'm running the Sun souped-up pop2 server that comes with PC-NFS,
and the popper pop3 server from ftp.cc.berkeley.edu, so I can cover
clients that only support one protocol or the other, but which one
should I use for clients that can talk both protocols?
Is pop3 more reliable, or faster, or does it just support extra fancy
features that don't affect basic post-office applications?

Thanks;  Bill Stewart wcs@anchor.ho.att.com
-- 

				Pray for peace;      Bill
#Bill Stewart 908-949-0705 wcs@anchor.ho.att.com AT&T Bell Labs 4M312 Holmdel NJ
# Any technology distinguishable from magic is not sufficiently advanced ...

-----------[000093][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 01:50:37 GMT
From:      pae@teal.csn.org (Phil Earnhardt)
To:        comp.protocols.tcp-ip
Subject:   Re: Implementation of Distributed Database in SUN RPC

In article <1992Sep2.152231.2819@noao.edu> rstevens@noao.edu (W. Richard Stevens) writes:
>In article <1992Aug31.185338.29597@stortek.com> pae@blackcat.stortek.com (Phil Earnhardt) writes:
>>
>>Netwise's RPC TOOL has certainly addressed the issue of different styles of
>>dispatching...there are single-threaded dispatchers, multi-processing
>>dispatchers (forking off a process for new requests) on systems providing
>>multi-processing, and multi-threaded dispatchers (spawning a new thread for
>>new requests) on systems providing multi-threading capabilities.
>
>Can anyone direct me to anything *available* that really describes
>the Netwise product, complete with examples?  I keep seeing claims
>about how great it is, but I've never seen examples of why it's so
>good.
>
>	Rich Stevens  (rstevens@noao.edu)

Netwise's number is (303) 442-8280. They're connected to USENET; you may be
able to get some examples sent to you via email. Actually, I'm a bit suprised
that no one there saw this thread...

I believe that the book _Managing Sun Networks_ has a chapter on the ONC
version of the Netwise RPC TOOL. There should be examples there.

Netwise's RPC compiler is supposed to be available as a front end to ONC RPC.
I don't know the details of if/when/where this is available. Perhaps Chuck can
give us some information on that...this was announced in press releases a
while back.

>       Rich Stevens  (rstevens@noao.edu)

--phil

-----------[000094][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 04:00:37 GMT
From:      cdaly@access.digex.com (CHRIS DALY)
To:        biz.sco.general,comp.protocols.tcp-ip
Subject:   Terminal Servers

Can anyone give me any suggestions on what type of terminal servers to buy.

I have a LAT - Tcp/ip ethernet network that includes a Dec MicroVax II,
Sco Xenix, AUX, SUN and Sco Unix systems. I would like to buy some terminal
servers that would let me share printers, terminals, and modems with all the
hosts. 


The terminal servers MUST support both LAT and Tcp/ip and allow a port
on the terminal server to be shared with both LAT and TCP at the same time. 
This would mean that if both the VAX and a UNIX host were sending print
jobs to the terminal server printer, the Server would be able to have one 
of the hosts Queue it jobs until the other is finished.  I would like to
share modems with all the HOSTS and have both dial IN and OUT from ANY of
the HOSTS.  The unix hosts should be able to use programs like cu, uucp, kermit
to connect to the terminal server port to dial out with.

Please EMAIL any suggestions or comments.   Thanks.

					cdaly@access.digex.com

-----------[000095][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 07:30:37 GMT
From:      graeme@ccu1.aukuni.ac.nz ( Graeme Moffat)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets

fitz@wang.com (Tom Fitzgerald) writes:

>> What I have observed in this situation (multiple subnets on same physical
>> cable) is that *many* hosts don't recognise x.y.z.255 as a broadcast on
>> subnet x.y.z when they are in subnet x.y.w, and respond by resending the
>> packet to MAC address 00FFBAD1DEAD, and an ICMP redirect to the original
>> sender! Lovely.
>Wow, I thought I'd seen all the bad side effects of mixing subnets on the
>same cable.  What kind of systems are these?

- RS6000 (AIX 3.1 and 3.2)  It's my gateway machine, and is correctly
configured, as far as I am aware. I can stop this behaviour in the other
non-gateway ones by turning ip forwarding & redirects off, (which are on by
default, in a deliberate design decision by IBM, despite rfc 1122). AIX
claims to run BSD 4.3 & now 4.4 networking code.
- Acorn running unix
- From memory, I think one was a Sun clone
- Many were PC's (inferred from the MAC addresses), OS unknown
Unfortunately (well, fortunately, really :) I can no longer see most since
we have started to break our net up with cisco's.

-- 
Graeme Moffat                g.moffat@aukuni.ac.nz \ Time wastes us all, 
Computer Aided Design Centre,  Fax: +64-9-366-0702 /  our bodies & our wits
School of Engineering,    Ph: +64-9-737-999 x8384 /  But we waste time,
University of Auckland, Private Bag, Auckland, NZ \   so time & we are quits

-----------[000096][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 15:35:43 CDT
From:      hens@snipe.cray.com (Tom Stephens)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?


In article <1992Sep9.184943.26790@gordian.com>, johnk@gordian.com (John Kalucki) writes:

> 
> Traceroute can be difficult to compile for certain OS types.


Yes this is the case.  We have had trouble with it on Sequent's.  Anyone have source that (when compiled) run on a Sequent??? 
--
---------------------------------
Tom Stephens
Corporate Computing and Networks, Chippewa Falls, WI
Cray Research Inc.  hens@cray.com

-----------[000097][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 12:10:06 GMT
From:      pahountis@ncf.al.alcoa.com (Julia Pahountis)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   NFS Over Satellite Link

Has anyone, or do you know of anyone, who has done NFS over a satellite link?  (Can it be done?)
I would appreciate any information on how it was accomplished and what kind of problems were
encountered.

Please send e-mail to pahountis@ncf.alcoa.com.

Thanks,

Julia Pahountis
Alcoa Technical Center
Alcoa Center, PA  15069
pahountis@ncf.al.alcoa.com

-----------[000098][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 12:54:39 GMT
From:      ljuno@zeus.bram.cdx.mot.com (Lance Juno)
To:        comp.protocols.tcp-ip
Subject:   FAQ - where ?

Can someone email me the location of the FAQ for this group ?

Thanks.


--
[ Lance Juno                              |  Motorola Codex                  ]
[ Internet: ljuno@dnbf01.bram.cdx.mot.com |  400 Matheson Blvd. West,        ]
[ Phone:        +1 416 507 7200           |  Mississauga, Ontario   L5R 3M1  ]
[ Fax:          +1 416 507 7236           |  Canada                          ]

-----------[000099][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 14:59:55 GMT
From:      micjrs@mica.mic.ki.se (Richard Soderberg, MD)
To:        comp.protocols.tcp-ip
Subject:   Echo "problem" in Telnet

hmm...

We're running an IBM 4381 and have recently installed TCPIP for MVS V2R2.
The Telnet server in this package *refuses* to do any echoing regardless of how 
the telnet client begs for it during the initial session negotiation.
Goes like this:

Telnet client       MVS Telnet Server
_____________       _________________
  Do Echo
                      Won't Echo
  Do Echo
                      Won't Echo

 (Gives up)

According to the appropriate RFC:s (as i understand them), an orthodox Telnet 
server is not required to supply echo but other Telnets i know do so anyway, 
worse, a lot of our users expect us to supply echo.( hiding passwords also 
becomes a problem if the client has to supply a local echo). Is it possible to 
do something about this?

Thanx in advance
/RS


#include <standard.disclaimer.h>
Richard Soderberg, MD 
Facsimile:   Int + 46 8 33 04 81
Talk phone: Int + 46 8 728 80 00

-----------[000100][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 15:29:48 GMT
From:      jzwiebel@pgl-devsvr.den.mmc.com (John Zwiebel (303)977-1480)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

In article <1992Sep9.121006.14055@udel.edu>, pahountis@ncf.al.alcoa.com (Julia Pahountis) writes:
|> Has anyone, or do you know of anyone, who has done NFS over a satellite link?  (Can it be done?)

Haven't done it but I think the turn around time to a geostationary satellite
would be a major problem since its up there at 24,000 miles.  Multiply that 
times 4 to include all your links and you start approaching the RPC time out.
You'd have to increase the number of biods on the clients and nfsd on the 
servers to increase "window" size (the number of outstanding nfs requests)
and have to increase the default timeout for retransmission to avoid too many
repeat requests on your links.  Both of these cause a decrease in NFS
performance on a "normal" NFS path.  (Too many nfsd waste CPU cycles on the
server, too many biods does the same thing on the client)

John Zwiebel

-----------[000101][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1992 15:48:22 GMT
From:      don@bodovipa.cis.brown.edu (Donald Wright)
To:        comp.protocols.tcp-ip
Subject:   Recommended seminars,conferences?

Can anyone recommend a seminar series or conference that teaches
intermediate      	
to advanced TCP-IP networking, routing, SNMP, etc.  I have received a
number of flyers from various training companies, but nothing that makes
anyone stand
out from the others.  Also, if there are any UNIX systems administration 
classes that contain the above topics, I would be interseted in those as
well.	

																																																Don Wright
																																																Network Support
																																																Brown University
																																																Voice: 401-863-7405
																																																Fax:   401-863-7329

-----------[000102][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 16:09:21 GMT
From:      add@philabs.philips.com (Aninda V. Dasgupta)
To:        comp.protocols.tcp-ip
Subject:   Setting socket send and receive buffer sizes

I am writing a TCP server and a client under SunOS.
I have four questions :

1) Most of my transfers are about 40-60 bytes in size.
   However, one in about 10 transfers is a quarter of
   a MByte in size (image files).  I am setting the
   send and receive buffer sizes in both the client and
   server to be the maximum allowed, viz. 52,000 bytes.
   Will this affect the throughput for the smaller (and
   more frequent) 60 byte packets?

2) In the server, is it absolutely necessary to call 
   the setsockopt() function right after the socket() call?
   Or can I wait till after the bind() call to set the send 
   and receive buffer sizes? 

3) What about new connections accepted by the server?  Do
   I have to set the buffer sizes for all the fd's that
   are returned by the accept() call?
   (I suspect that the buffer sizes are exchanged between
   client and server during connection setup.)

4) In the client, can I wait till after the connect() to
   set the buffer sizes or do I have to do it right after
   the socket() call ?

Any help will be greatly appreciated. Thanks in advance.

Aninda
-- 
Aninda DasGupta   (add@philabs.philips.com)
Ph : (914) 945-6071    Fax : (914) 945-6552
Philips (No, we don't produce Gas, we make lightbulbs) Labs.
345 Scarborough Road\n  Briarcliff Manor\n NY 10510

-----------[000103][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1992 23:41:00 -0400
From:      emv@msen.com (Edward Vielmetti)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

vjs@rhyolite.wpd.sgi.com (Vernon Schryver) writes:
: > MIPSos 5.0 has traceroute, which it didn't in the past. I don't
: > know of any other major OS vendor who ships traceroute...
 
: Really?  I would have guessed everyone ships it.  
: What about BSDI?  I bet it's there.

yup, it's in BSDI.  In the TGV Multinet software too (*knew* there was a
reason we recommended that) which made finger-pointing on a network
performance problem relatively easy.  

traceroute not only collects information about where the packets go, but
also how long it too to get each step of the way.  a day's worth of traces
on a given path is very instructive, & no doubt could be used to good
effect to help people figure out where network re-engineering would make
sense, much more so than static packet counts.

--
Edward Vielmetti, vice president for research, MSEN Inc. emv@msen.com
      MSEN Inc., 628 Brooks, Ann Arbor MI  48103 +1 313 998 4562
 A distributed system is one in which the failure of a computer you didn't
 even know existed can render your own computer unusable. (Leslie Lamport)


-----------[000104][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 17:43:34 GMT
From:      joel@skater.GSFC.NASA.Gov ("Joel K. Gallun")
To:        comp.protocols.tcp-ip
Subject:   Problem with NCSA Telnet and Orchid SVGA

Hi,

I have been using NCSA Telnet for some time with great success.  It is
an excellent package.

However, I recently upgraded my video controller to an Orchid ProDesigner
IIs with 512K VRAM and I started having a problem.  When I exit from
Telnet, the screen goes blank and stays that way until I give the 3
finger salute.

I have tried setting the CONFIG.TEL option to use BIOS with no change
in the symptoms.  Here are the details:

	Compaq deskpro 386/20
	4 Meg RAM
	NEC Multisync 3/D monitor
	DOS 5.0
	3COM 3C503 ethernet card + packet driver using
		interrupt 0x7e
	No drivers or TSRs loaded
	NCSA Telnet 2.3

I can provide my AUTOEXEC, CONFIG.SYS and CONFIG.TEL to anyone thinks
that they might contain a clue as to what the problem might be.

Anyone have any ideas or sympathy?

--
Joel Gallun                        Internet: nnjkg@robots.gsfc.nasa.gov
NASA Goddard Space Flight Center      Voice: (301) 805-4545
#include <std_disclaimer>             Snail: Code 735, GSFC, Greenbelt MD 20771

-----------[000105][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 17:53:31 GMT
From:      scoggin@opus.ee.udel.edu (John K Scoggin)
To:        comp.protocols.tcp-ip
Subject:   Re: Recommended seminars,conferences?

In article <don-090992112635@cis_staff-kntx24.cis.brown.edu> don@bodovipa.cis.brown.edu (Donald Wright) writes:
>Can anyone recommend a seminar series or conference that teaches
>intermediate      	
>to advanced TCP-IP networking, routing, SNMP, etc.  I have received a
>number of flyers from various training companies, but nothing that makes
>anyone stand
>out from the others.  Also, if there are any UNIX systems administration 
>classes that contain the above topics, I would be interseted in those as
>well.	

I had a taught on site by InterOp Inc. (the same folks that run the 2
InterOp conferences).  In our case, we had Dr. David L. Mills teach a
course on TCP/IP - he was outstanding.  They have VERY QUALIFIED teachers -
folks like Comer for TCP/IP, Case for SNMP.  Well worth the money.

	- John

-------------------------------------------------------------------------
John K. Scoggin, Jr.			Email:   scoggin@udel.edu
Supervisor, Network Operations			 scoggin@delmarva.com
Delmarva Power & Light Co.		Voice:	 302-451-5200
P.O. Box 6066				Fax:	 302-451-5321
Newark, DE 19714-6066
-------------------------------------------------------------------------
Your milage may vary.  Void where prohibited by law (or common sense).

-----------[000106][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 18:03:10 GMT
From:      arons@ash.eecs.ucdavis.edu (Tom Arons)
To:        comp.protocols.tcp-ip
Subject:   Re: Broadcasting and Subnets


   >> What I have observed in this situation (multiple subnets on same physical
   >> cable) is that *many* hosts don't recognise x.y.z.255 as a broadcast on
   >> subnet x.y.z when they are in subnet x.y.w, and respond by resending the
   >> packet to MAC address 00FFBAD1DEAD, and an ICMP redirect to the original
   >> sender! Lovely.
   >Wow, I thought I'd seen all the bad side effects of mixing subnets on the
   >same cable.  What kind of systems are these?

   - RS6000 (AIX 3.1 and 3.2)  It's my gateway machine, and is correctly
   configured, as far as I am aware. I can stop this behaviour in the other
   non-gateway ones by turning ip forwarding & redirects off, (which are on by
   default, in a deliberate design decision by IBM, despite rfc 1122). AIX
   claims to run BSD 4.3 & now 4.4 networking code.
   - Acorn running unix
   - From memory, I think one was a Sun clone
   - Many were PC's (inferred from the MAC addresses), OS unknown

Add to this:
HP hp-ux prior to 8.0
HP Xstations B.03.00, fixed in next release

--
Tom Arons			Internet: arons@eecs.ucdavis.edu
EE/CS Department		Bitnet:   tgarons@ucdavis
University of California	UUCP:     {lll-crg, ucbvax}!ucdavis!iris!arons
Davis, CA 95616			Phone:    (916) 752-1750

-----------[000107][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 18:21:28 GMT
From:      ronf@panther3.panther.mot.com (Ron Feigen)
To:        comp.protocols.tcp-ip
Subject:   Hardware from Protocol Engines Inc.

Hi,

  I hope someone in Netland has the info I need.

  I heard there is an outfit called Protocol Engines that makes  a H/W based
protocol processing device.  How do I get a hold of them?  Are there any other
outfits that make a similar product?

  Please e-mail repsonses I am not a regular to many of these threads.

  Thanks in advance,


-- 

>
Ron Feigen
ronf@panther.mot.com

-----------[000108][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 18:49:43 GMT
From:      johnk@gordian.com (John Kalucki)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

In article <2190004@otter.hpl.hp.com>, jrg@otter.hpl.hp.com (John Grinham) writes:
>Is traceroute a public domain utility, or is it a vendor specific program ?
>If it's PD does anyone know where I could pick up a copy ? 
>
>
>regards,
>John.
>
>-----------------------------------------------------------------------------
>John Grinham.  HP Labs ISC, Filton Road, Stoke Gifford, Bristol BS12 6QZ, UK.
>Tel: (0)272 799910	Fax: (0)272 790554 	Telex: 449206
>jrg@hplb.hpl.hp.com
>-----------------------------------------------------------------------------
>

Traceroute can be difficult to compile for certain OS types. Ultrix
used to be shipped with it by default, but it doesn't appear to be
on the latest release.

MIPSos 5.0 has traceroute, which it didn't in the past. I don't
know of any other major OS vendor who ships traceroute...

		-John Kalucki
		johnk@gordian.com


-----------[000109][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 19:17:18 GMT
From:      E2311DAA@vm.univie.ac.at
To:        comp.protocols.tcp-ip
Subject:   FAQ ? RFC - where to get ?



escuses, knows anyone, where to ftp or order RFCs ??

supershort mails to      e2311daa@vm.univie.ac.at     are welcome ...
supershort mails would be appreciated ...



-----------[000110][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 21:39:29 GMT
From:      edholz@ac.wfunet.wfu.edu (Ed Holzwarth)
To:        comp.protocols.tcp-ip
Subject:   SLIP/ASLIP/PPP for HPUX (PPL)

Has anyone using HPUX gotten PPL to work?  I am trying to use it, but it
always gives the error that it can't find the remote IP address in
ppl.remotes.    I have set it up according to the manual (and the sample)
but no matter what I put, it always gives the same error.  Has anyone
gotten it to work, and if so, could you please send some help?  Perhaps
someone knows the name of an hpux specific newsgroup?

Thanks,
Ed Holzwarth



-----------[000111][next][prev][last][first]----------------------------------------------------
Date:      Wed, 9 Sep 1992 22:15:58 GMT
From:      craig@sics.se (Craig Partridge)
To:        comp.protocols.tcp-ip
Subject:   re: NFS Over Satellite Link


While at SUN, Bill Nowicki built an experimental NFS that adapted to
bandwidth changes and long delays.  He successfully ran it over the
old SATNET over the Atlantic to University College London.

There's a brief description of the work in ACM SIGCOMM Computer Communication
Review of April 1989.  I seem to recall Bill also sent a detailed note to
some mailing list at the time but I cannot seem to find my copy of his note.

Craig Partridge
craig@bbn.com

-----------[000112][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 92 23:38:57 GMT
From:      noord@prism.CS.ORST.EDU (David Noor)
To:        comp.sys.mac.comm,comp.protocols.tcp-ip,comp.sys.mac.programmer,comp.sys.mac.misc,comp.sys.mac.wanted
Subject:   Needed: Refrernces for programming MacTCP

I am *trying* to design a program for the Macintosh which will end up
using MacTCP. I need help, because the Inside Macintosh books do not have
anything about MacTCP in them. So can anyone out there tell me where to
maybe ftp, or some other way to get programming information for MacTCP?
I know about the APDA MacTCP developers kit, but the CS dept. wont loan it out
to students and i dont have $600+ free in my account or anything :).

Any help would be GREATLY appreciated...

I would prefer E-Mail responses, not replys on news...
mail to either:

noord@prism.cs.orst.edu
or
noord@ucs.orst.edu

ive posted this to way too many newsgroups to check all of em so it would be
best if people could email me.... :)

Thanx again...                  -noord

-----------[000113][next][prev][last][first]----------------------------------------------------
Date:      Thu, 10 Sep 1992 00:15:09 GMT
From:      vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?

In article <1992Sep9.184943.26790@gordian.com>, johnk@gordian.com (John Kalucki) writes:
> ...
> MIPSos 5.0 has traceroute, which it didn't in the past. I don't
> know of any other major OS vendor who ships traceroute...


Really?  I would have guessed everyone ships it.  

What about BSDI?  I bet it's there.

Of course, Silicon Graphics ships it, although I think it's not
installed by default.  Could that sort of thing be why it seems rare?


Vernon Schryver,  vjs@sgi.com

-----------[000114][next][prev][last][first]----------------------------------------------------
Date:      Thu, 10 Sep 1992 04:06:48 GMT
From:      marc@dumbcat.sf.ca.us (Marco S Hyman)
To:        comp.protocols.tcp-ip,comp.dcom.isdn
Subject:   Re: TCP/IP over ISDN

In article <1992Sep8.183030.330@gandalf.ca> djarrett@gandalf.ca (Dave Jarrett) writes:
 > Don't get hung up on this user-user information thing. All we require 
 > to do a poll, is 1 available bit, anywhere in the connect packet, which
 > will tell the receiving node that this connect request is really a poll.
 > Surely even ITR6 has 1 available bit which can be used for this purpose?

Hmmmm,  I'm looking at the AT&T PRI spec (TR 41449/TR 41459) and the only
Information Elements that are mandatory are:

	Bearer Capability
	Channel Identification
	Called party number

Experience tells me that the Bearer Capability for an incoming call often
bears no relation to the one in the originators setup message.

The channel ID IE is node-to-node, not end-to-end.  All its bits are accounted
for anyway.

The "mandatory" called party number isn't in setups that come from a Definity
Generic I PBX.

Ok, on to BRI and the 5ESS specifically (I'm not picking on AT&T, it's just
that I happen to have their manuals at home).  Only the Bearer capability is
required.

Maybe in the future when the world is SS7 you'll be able to trust the bearer
cap IE.  Until that time
* as your call goes from switch to switch it can (and will) be munged
* it sometimes lies.  I've received calls with a 64K Bearer Cap IE that I
  *know* went over at least one line/network that only had 56K capabilities.

Now where is that bit?

// marc
-- 
marc@dumbcat.sf.ca.us -- pacbell!dumbcat!marc

-----------[000115][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 08:49:28 GMT
From:      Thomas.Tornblom@nexus.comm.se (Thomas Tornblom)
To:        comp.protocols.tcp-ip
Subject:   Using TCP on TLI, how to get own IP address?


How do I get my own IP address of a TCP over TLI connection? I've RTFM
but can't find anything relevant. The portnumber can be returned from
t_bind() and the remote IP address and portnumber with t_listen().

Thomas
--
Real life:      Thomas Tornblom           Email:  Thomas.Tornblom@nexus.comm.se
Snail mail:     Communicator Nexus AB     Phone:  +46 18 171814
                Box 857                   Fax:    +46 18 696516
                S - 751 08 Uppsala, Sweden

-----------[000116][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 09:21:49 GMT
From:      antonis@helios.ntua.gr (Antonis Kyriazis)
To:        comp.protocols.tcp-ip
Subject:   OSPF question


Hi

I'm trying to plan an OSPF configuration for 3 company's buildings:

+---------------+                  +---------------+
|               |     2 Mbps       |               |
|  building B   |------------------|  building A   |
|  router       |                  |  router       |
+---------------+                  +---------------+
        |
        |
        | 2 Mbps
        |
+---------------+       This will be the whole autonomous system
|               |       and obviously the buildings B, A and D
|  building D   |       will represent three different areas.
|  router       |       The questions are: Should I define a
+---------------+       (virtual?) backbone?
                        Routers D and A have to be ABR or just
                        DR?

Notice that D and A are equipped with a LAN and a WAN interface
and B with 4 WANs and 4 LANS, that means B might act as an ASBR
and as an internal router at the same time.

thanks
+-------------------------------------------------------------------+
|     Antonis Kyriazis                                              |
| Networks & Communications       e-mail: antonis@intranet.gr       |
| INTRACOM sa                                                       |
| 19.5 km Marcopoulo Ave.          fax:   +30-1- 66 44 379          |
| Peania 190 02                           +30-1- 66 43 718          |
| GREECE                          phone:  +30-1- 66 44 961-5        |
|                                         +30-1- 88 43 715          |
+-------------------------------------------------------------------+


-----------[000117][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 11:37:31 GMT
From:      J.Crowcroft@cs.ucl.ac.uk (Jon Crowcroft)
To:        comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link



 >In article <1992Sep9.121006.14055@udel.edu>, pahountis@ncf.al.alcoa.com (Julia Pahountis) writes:
 >|> Has anyone, or do you know of anyone, who has done NFS over a satellite link?  (Can it be done?)
 >

yes - we did it years ago - Bill Nowicki ran some tests of getting the
parameters right from Sun to UCL

you need ~1 sec retransmit timers, small blocksizes etc etc...

it worked ok (on a 2*64kbps 5 site satellite system)

remember if you dont have forward error correction on yr satellitew
link, your error rates will be massive compared to what you are used
to on terestrial links (e.g. 5%)

 jon

-----------[000118][next][prev][last][first]----------------------------------------------------
Date:      Thu, 10 Sep 1992 12:48:29 GMT
From:      cole@etonic.gsg.dec.com (Larry Cole)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a Internet distance measure?


	Traceroute source is available by anon ftp
	on gatekeeper.dec.com in /pub/bsd-sources-2/usr.sbin/traceroute/

	Traceroute is also included with version 2.0 of
	"TCP/IP Services for VMS", aka UCX.


-----------[000119][next][prev][last][first]----------------------------------------------------
Date:      Thu, 10 Sep 1992 14:38:36 GMT
From:      deej@cbnewsf.cb.att.com (david.g.lewis)
To:        comp.protocols.tcp-ip,comp.dcom.isdn
Subject:   Re: TCP/IP over ISDN

In article <1992Sep10.040648.20853@dumbcat.sf.ca.us> marc@ascend.com writes:
>In article <1992Sep8.183030.330@gandalf.ca> djarrett@gandalf.ca (Dave Jarrett) writes:
> > Don't get hung up on this user-user information thing. All we require 
> > to do a poll, is 1 available bit, anywhere in the connect packet, which
> > will tell the receiving node that this connect request is really a poll.
> > Surely even ITR6 has 1 available bit which can be used for this purpose?
>
>Hmmmm,  I'm looking at the AT&T PRI spec (TR 41449/TR 41459) and the only
>Information Elements that are mandatory are:
>
>	Bearer Capability
>	Channel Identification
>	Called party number

Protocol Discriminator, Message Type, and Call Reference are also mandatory.

>Experience tells me that the Bearer Capability for an incoming call often
>bears no relation to the one in the originators setup message.

I feel I must take some exception to this...

First, TR 41459 specifies the AT&T network ISDN PRI - the protocol that you
use when you purchase a PRI from AT&T Communications Services, or whatever
we're calling ourselves these days.

If the call both originates and terminates on AT&T network PRIs, the BCs
should be very closely related.  In fact, they should be identical.

IF the call originates or terminates on another network, BRI or PRI, the
situation may be different.  If it's a voice call, an Information Transfer
Capability of "speech" may be mapped to "3.1kHz audio", or vice-versa; as I
recall, Bellcore TRs treat them pretty much interchangeably.

If it's a data call, and it interworks with CSDC access ("switched 56"),
things may be a little more complicated.  For example, a call may be
originated from an AT&T PRI, destined for an Accunet Switched Digital
Service customer served by a LEC BRI, with access to Accunet SDS through a
digital switched access arrangement using 56kb/s CSDC trunks.  In this case,
the BC at the originating PRI must be coded 64kb/s circuit mode, with octet
5 of the BC indicating a user information layer 1 protocol of V.110 rate
adaption to 56kb/s.  Because this detailed information can not be
transferred across the CSDC (in-band signaled) trunks, the LEC switch may
not code the BC in the SETUP message to the BRI in this way.  Similarly, if a
call is delivered to the AT&T network over CSDC trunks for delivery to a
PRI, the Information Transfer Rate field of the BC IE will be coded 64kb/s.
Octet 5 should show rate adaption to 56kb/s.  The access BRI may not require
the rate adaption to be explicitly coded.

>The "mandatory" called party number isn't in setups that come from a Definity
>Generic I PBX.

Over a PRI it certainly is.  If a SETUP message is sent by a PBX to the AT&T
network and does not include a Called Party Number IE, the call attempt will
be cleared.  On the terminal side of the Definity, the terminal may receive
information via Keypad IEs.

>Ok, on to BRI and the 5ESS specifically (I'm not picking on AT&T, it's just
>that I happen to have their manuals at home).  Only the Bearer capability is
>required.

Called Party Number is not mandatory over BRI because BRI may use Keypad IEs
with overlap signaling procedures.  Channel ID, I'm guessing, may not be
mandatory if the switch chooses the channel and returns the selected channel
in the first Call Proceeding or Setup Ack message.

>Maybe in the future when the world is SS7 you'll be able to trust the bearer
>cap IE. 
> Until that time
>* as your call goes from switch to switch it can (and will) be munged
>* it sometimes lies.  I've received calls with a 64K Bearer Cap IE that I
>  *know* went over at least one line/network that only had 56K capabilities.
>
>Now where is that bit?

Again, to get back to the original thread, certain IEs are designated as
usable for user to user information transfer.  These include UUI, High Layer
Compatibility, Low Layer Compatibility, subaddress IEs, and Codeset 7 IEs.
If you attempt to use anything else to transfer user to user information,
there is no guarantee, explicit or implied, that it will be delivered the
way you want it - or that it will be delivered at all.


Disclaimer: Anyplace I say "may" I'm hypothesizing based on
generally-available information.  Anyplace I say "does" or "will" or "must"
I'm pulling info from TR 41459 - you could look it up...


David G Lewis                              AT&T Bell Laboratories
david.g.lewis@att.com or !att!houxa!deej     Switching & ISDN Implementation

-----------[000120][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 20:19:31
From:      Steinar.Haug@delab.sintef.no (Steinar Haug)
To:        comp.protocols.tcp-ip,comp.protocols.misc,comp.protocols.iso
Subject:   Why sequence number on bytes instead of packets?

I have sort of grown up with TCP/IP, and I feel comfortable with the notion
of having sequence numbers on bytes, instead of on packets. I have always
assumed this was to make the fragmentation/reassembly process easier. (Though
I notice that XTP version 3.6, which doesn't have fragmentation, still uses
sequence numbers on bytes - for backwards compatibility?)

Anyway, I'm now facing the possibility of designing another (lightweight)
transport protocol. (Please don't ask why - we're still in the evaluation
stages and we *may* end up using an existing protocol). So my question is:

What are the arguments for sequence numbers on bytes instead of packets,
given that the protocol in question will *not* have fragmentation?

Steinar Haug, system/networks administrator
SINTEF DELAB, University of Trondheim, NORWAY
Email: Steinar.Haug@delab.sintef.no, 
	sthaug@idt.unit.no, steinar@tosca.er.sintef.no

-----------[000121][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 17:24:13 GMT
From:      jim@cs.strath.ac.uk (Jim Reid)
To:        comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

In article <1992Sep9.221558.26816@sics.se> craig@sics.se (Craig Partridge) writes:

   While at SUN, Bill Nowicki built an experimental NFS that adapted to
   bandwidth changes and long delays.  He successfully ran it over the
   old SATNET over the Atlantic to University College London.

I believe that this work was picked up by Sun and went into SunOS4.
This has stuff in it to keep track of NFS round trip times so that it
can better handle flow control and limited bandwidth when NFS is used
on things other than high-bandwith, low delay local area networks.

		Jim

-----------[000122][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 20:20:53 GMT
From:      giri@uts.amdahl.com (Giri Padmanabhan)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   DNS question....


 I have a question here about DNS implementation. Any thoughts would be
 appreciated.

     If we have primary and secondary servers on our domain, the primary
     nameserver would reload if it receives a SIGHUP. Furthermore, if an
     intentional error is introduced in a RR record in the primary's database,
     the secondary nameserver does not get reloaded. Should IT?


Giri.

Standard DISCLAIMER APPLIES:


-----------[000123][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1992 21:09:16 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.misc,comp.protocols.iso
Subject:   Re: Why sequence number on bytes instead of packets?

In article <STEINAR.HAUG.92Sep10201931@delab.sintef.no> Steinar.Haug@delab.sintef.no (Steinar Haug) writes:
>I have sort of grown up with TCP/IP, and I feel comfortable with the notion
>of having sequence numbers on bytes, instead of on packets. I have always
>assumed this was to make the fragmentation/reassembly process easier. 

Fragmentation/reassembly takes place in IP, but sequence numbers are in
TCP, so they don't have anything to do with each other.

>What are the arguments for sequence numbers on bytes instead of packets,
>given that the protocol in question will *not* have fragmentation?

When building streams out of datagrams, using byte-oriented sequence
numbers means that you don't have to retransmit exactly the same packet.
You can buffer up packets waiting to be retransmitted, and retransmit them
as a whole rather than in the original separate packets.  This reduces the
number of packets that have to be sent during retransmission, although it
may increase the total amount of retransmitted data.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000124][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 21:51:07 GMT
From:      feoh@hal.gnu.ai.mit.edu (Chris Patti)
To:        comp.unix.bsd,comp.protocols.tcp-ip
Subject:   nameserver resolution by adress (HELP!!!)

We have a nameserver for a subdomain, and the nameserver one level
above us knows enough about us to be able to correctly resolve names,
but not addresses.  (i.e. the server above us has an NS record with
our nameserver listed to resolve all names with our domain postfix).
Does anyone happen to know the syntax necessary to make the 
nameserver above us send ADDRESS requests to our nameserver as well?
(A gethostbyname on a foreign host is successful, but then taking
that addr and trying to gethostbyaddr is NOT successful.)

Posting here is cool, or if you feel like it, please reply to:
bild@ncm.interlan.com

Thanks bunches...
bild
-- 
-------------------------------------------------------------------------------
Chris Patti | feoh@gnu.ai.mit.edu: A guest of the Free Software Foundation
=========#include <gratuitous_witty_quotes_section_to_waste_bandwidth>=========
"War is a symphony of destruction, orchestrated by few, paid for by many."

-----------[000125][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 23:02:40 GMT
From:      feoh@hal.gnu.ai.mit.edu (Chris Patti)
To:        comp.unix.bsd,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Adress correction (for previous article: nameserver problems (HELP!!!)

Hello, an incorrect adress was given in the previous article regarding name 
server resolution problems. the correct adress to reply to is: 
bild@ncm.ncm.interlan.com (we can't get our sendmail to realize that 
ncm.interlan.com is equivalent to ncm.ncm.interlan.com :)

Thanks in advance,
feoh@gnu.ai.mit.edu
posting for bild@ncm.ncm.interlan.com

-----------[000126][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 92 23:28:15 GMT
From:      johnk@gordian.com (John Kalucki)
To:        comp.sys.mac.comm,comp.protocols.tcp-ip,comp.sys.mac.programmer,comp.sys.mac.misc,comp.sys.mac.wanted
Subject:   Re: Needed: Refrernces for programming MacTCP

In article <1992Sep09.233857.15170@CS.ORST.EDU>, noord@prism.CS.ORST.EDU (David Noor) writes:
>I am *trying* to design a program for the Macintosh which will end up
>using MacTCP. I need help, because the Inside Macintosh books do not have
>anything about MacTCP in them. So can anyone out there tell me where to
>maybe ftp, or some other way to get programming information for MacTCP?
>I know about the APDA MacTCP developers kit, but the CS dept. wont loan it out
>to students and i dont have $600+ free in my account or anything :).
>
>Any help would be GREATLY appreciated...
>
>I would prefer E-Mail responses, not replys on news...
>mail to either:
>
>noord@prism.cs.orst.edu
>or
>noord@ucs.orst.edu
>
>ive posted this to way too many newsgroups to check all of em so it would be
>best if people could email me.... :)
>
>Thanx again...                  -noord

Just so the general newsreading public gets an answer: When you
buy the MacTCP developers kit from APDA, they send along a little
white three ring binder with all of the public procedure calls and
whatnot. From what I remember it's all you'll need if you already
are familiar with TCP/IP.


		-John Kalucki
		johnk@gordian.com

-----------[000127][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 06:46:20 GMT
From:      aps@world.std.com (Armando P. Stettner)
To:        comp.protocols.tcp-ip
Subject:   Req. for info on Layer 3 protocol overhead

Hi.

I am interested in the protocol overhead associated with the various
layer 3 (or there abouts) protocols and/or interconnects such as
Frame Relay, SMDS, ATM, etc.  I would appreciate referenc to texts,
papers, etc.

Please mail to me at aps@world.std.com and I will summerize if there
is interest.

An anticipated thanks.

	armando.

-----------[000128][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 12:31:27 GMT
From:      mckenzie@bbn.com (Alex McKenzie)
To:        comp.protocols.tcp-ip,comp.protocols.misc,comp.protocols.iso
Subject:   Re: Why sequence number on bytes instead of packets?

There is no need to sequence number any unit smaller than the smallest
individual unit which can be a fragment boundary.  Other transport
protocols have been designed and implemented which attach sequence
numbers to units larger than a byte (eg 256 bytes) and allow
fragmentation only in terms of these units.  The unit you pick is one of
the parameters which is involved in determining the Minimum Packet Size
which MUST be accepted by EVERY (packet) network through which packets
of your protocol will travel.  The byte was picked for TCP/IP in order
to be as all-inclusive as possible. 

    __
   /  | /\                  Alex McKenzie
  /   | \/                  mckenzie@bbn.com
  \__/|_/\_&_/~X_           617/873-2962
 

-----------[000129][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 13:20:55 GMT
From:      srm@dimacs.rutgers.edu (Scott R. Myers)
To:        comp.protocols.tcp-ip
Subject:   Book Recommendations?


I'm looking for some suggestions on books to read that provide
beginning concepts in bridge/router/protocol technology.  I'm taking
on a great challenge to teach a startup course on this and need to
find a book that is complete, accurate and appropriate for classroom
study.  Please e-mail any responses since I don't frequent this news
group.  Thanx in advance.

srm
-- 

				Scott R. Myers

Snail:	8544 Temple Road			Phone:215.247.2551
        Philadelphia, PA 19150	
				       		Voice Mail:215.440.6066

Arpa:	srm@dimacs.rutgers.edu			Uucp: ..!dimacs!srm

	"Every time I think I'm wrong I realize that I was mistaken!"

-----------[000130][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 16:55:40 GMT
From:      giri@uts.amdahl.com (Giri Padmanabhan)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Receiving SIGHUP in named....DNS question...


Question:

If the nameserver were to receive a SIGHUP, it would reload the database for
itself as well as for any secondaries through a zone file transfer. If for
argument sakes, an invalid record was encountered ( any incorrect RR ) during
the reload, that RR is skipped by the primary and the rest of the records will
be loaded. My question is, how should the secondary be updated? Should it'
database be refreshed also by the primary by a zone file transfer or not. Shouldit not also be refreshed and have the ame contents as the primary's database.
I need suggestions for implementation:

Thanks,


Giri.

Standard DISCLAIMER APPLIES:

-----------[000131][next][prev][last][first]----------------------------------------------------
Date:      Fri, 11 Sep 1992 16:56:29 GMT
From:      culbert@ender.bloom-picayune (Jim Culbert)
To:        comp.protocols.tcp-ip
Subject:   Momentum Software

Hi.

I'm looking for information about a product called Extended IPC from 
Momentum Software Corp. Has anyone heard of the product or know anything
about the company (i.e location)? Any pointers would be helpful.
I don't monitor this list so if post the response, please also e-mail
me a copy.

			Thanks,


			  -Jim

===========================================================================
> Jim Culbert                                                             <
> Research Engineer							  <
> M.I.T Intelligent Engineering Systems Laboratory                        <
> Department of Civil Engineering 					  <
> Room 1-270                                                              <
> Cambridge, Ma. 02139.                                                   <
>									  <
> Phone (617) 253-7134                                                    <
> e-mail: culbert@iesl.mit.edu                                            <
 ===========================================================================

-----------[000132][next][prev][last][first]----------------------------------------------------
Date:      Fri, 11 Sep 1992 17:18:09 GMT
From:      stritzi@Physik.TU-Muenchen.DE (Peer Stritzinger)
To:        comp.unix.aix,comp.sys.novell,comp.protocols.tcp-ip
Subject:   AIX 3.2 and LAN Workplace for DOS telnet

I use Novell's LAN Workplace for DOS telnet to log into a RS/6000.

On a new machine running AIX 3.2 i get garbage on my screen. To be more
precise, I get newlines but no carriage returns.

I'm using the 8 bit vt200 emulation of tnvt.

Any ideas?

--
Peer Stritzinger
Phone: +(49)8141-90613
Fax: +(49)8141-95235
Internet: stritzi@physik.tu-muenchen.de

-----------[000133][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 17:56:05 GMT
From:      twb0@chili.cc.lehigh.edu (Thomas Brown)
To:        comp.protocols.tcp-ip
Subject:   SLIP without root perms or OS mod.

I'm looking for a way to establish a SLIP connection across a
dialup line (19,200 baud) on one of the following machines:

   Sun w/SunOs 4.0   or   AT&T 3b15   or   IBM RS/6000 (AIX 3.2.2)

There are several packages available, but all of them seem to
require that the kernel be modified to include a SLIP daemon
or that the slip host program be run under root.  Is it possible
to run a slip host program under your own login ID so that your
current terminal session would be converted to a SLIP session?

Thanks much,
-Tom
-- 
--=--
Thomas Brown, KA2UGQ          Internet: twb0@lehigh.edu
Lehigh University UC Box 855  UUCP: ..!uunet!lehigh.edu!twb0
Bethlehem, PA  18015          AX.25: ka2ugq@ka2ugq.nj.usa.na

-----------[000134][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 18:45:56 GMT
From:      marc@aria.Ascend.COM (Marco S Hyman)
To:        comp.protocols.tcp-ip,comp.dcom.isdn
Subject:   Re: TCP/IP over ISDN

In article <1992Sep10.143836.11131@cbfsb.cb.att.com> deej@cbnewsf.cb.att.com
(david.g.lewis) writes:
 > In a previous article I wrote:
 > >The "mandatory" called party number isn't in setups that come from a
 > > Definity Generic I PBX.
 > 
 > Over a PRI it certainly is.  If a SETUP message is sent by a PBX to the AT&T
 > network and does not include a Called Party Number IE, the call attempt will
 > be cleared.  On the terminal side of the Definity, the terminal may receive
 > information via Keypad IEs.

Uhh, this may be a function of the funny way we use our Definity.  It's the
best T1/PRI test took we have (in addition to being a great PBX).  We have
PRI on both the Network and Terminal side, an AT&T PRI from the network and
several (6 or 8) internal T1 lines, some configured as PRI, with the PBX
told they are tie lines.  Each internal line is a trunk group of its
own and is dialed by using the trunk group.  When dialing from one PRI trunk
group to another PRI trunk group the Called Party Number IE on the answering
side exists but has no number.

 > Again, to get back to the original thread, certain IEs are designated as
 > usable for user to user information transfer.  These include UUI, High Layer
 > Compatibility, Low Layer Compatibility, subaddress IEs, and Codeset 7 IEs.
 > If you attempt to use anything else to transfer user to user information,
 > there is no guarantee, explicit or implied, that it will be delivered the
 > way you want it - or that it will be delivered at all.

Agreed.  However, given the heterogeneous nature of the network there is
currently no guarantee that any of the above will work between any two
arbitrary endpoints.

I guess I'm coming down on the side of making a connection and performing
control inband.  TCP/IP over xxx over ISDN.  PPP is overkill for xxx but
will certainly work.

// marc
-- 
// work: marc@ascend.com          uunet!aria!marc
// home: marc@dumbcat.sf.ca.us    pacbell!dumbcat!marc    

-----------[000135][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 19:18:27 GMT
From:      ciarfela@nu.ece.ucsb.EDU (Paul Ciarfella)
To:        comp.protocols.tcp-ip
Subject:   Does IP fragment broad/multicasts


We've been having an argument here concerning whether or not IP
fragments broadcast and multicast packets.

Could someone please answer this question, or point to the 
relevant RFC's.  I've been searching through Comer's book and
all RFC's related to IP and multicast/broadcast but have not
found the answer (I probably passed over it).

Thanks.

Paul C

-----------[000136][next][prev][last][first]----------------------------------------------------
Date:      Fri, 11 Sep 1992 20:47:29 GMT
From:      wayne@ita.lgc.com (Wayne McCormick)
To:        alt.sources.wanted,comp.sources.wanted,comp.protocols.tcp-ip,comp.protocols.misc
Subject:   Software router/bridge getty needed


This is posted for a friend, post or send replies to wmccormick@ita.lgc.com.

 Subject: Software router/bridge getty needed


   Situation: You know someone who has a direct connect to the Internet,
              and is willing to give you a mail/newsfeed. But what about
              ftp/telnet capabilities. Unless you have a bridge or router
              of some sort you cannot get ftp/telnet capabilites to remote
              sites.

   Theory: Given that you and your feed both have modems, could the feed
           run a special 'getty' on a dialup line? This getty would
           listen to the line and if a special sequence is sent to it,
           it would act as the router for you. It would connect
           to the remote site and pass though all input/output back and
           forth to your machine.
              Ftp transfers would end up on your machine, not your feed's
           machine.

     Is this possible? Does such a beastie exist?

Thanks.

-----------------------
Wayne
--
Wayne McCormick                  | Landmark/ITA, Inverse Theory & Applications
wmccormick@ita.lgc.com           | Calgary, Alberta, Canada  
                                 | Phone:(403) 269-4669

-----------[000137][next][prev][last][first]----------------------------------------------------
Date:      Fri, 11 Sep 1992 20:55:42 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Does IP fragment broad/multicasts

>We've been having an argument here concerning whether or not IP
>fragments broadcast and multicast packets.

All the BSD kernel code that I've seen refuses to fragment a broadcast
packet.  You get EMSGSIZE instead.  I don't know about multicasting.

	Rich Stevens  (rstevens@noao.edu)

-----------[000138][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 92 22:11:49 GMT
From:      neerma@nosc.mil (Merle A. Neer)
To:        comp.dcom.lans.ethernet,comp.dcom.lans.fddi,comp.protocols.tcp-ip
Subject:   NDIS drivers

I am trying to get PCNFS to work with an NDIS driver supplied by
Network Peripherals for their FDDI board (the EISA version).

The Network Peripheral documentation references a "NP-NDIS driver
for Microsoft LAN Manager Ver 2.0" and a "NP-PCNFS driver for SUN
NDIS". The "for" clauses have me confused:I thought the idea behind
NDIS (and, for that matter, ODI and packet driver) was that once
an NDIS-compliant device driver was produced for a particular net
card, it could interface with any NDIS-compliant protocol stack (like
PCNFS). It appears to me that in my particular situation, NP has
produced device drivers that talk to particular protocol stacks (and
yet they mysteriously use the NDIS label!). 

Of course, I dont get the PCNFS version. I get the other one (of
course, since I want the device driver to talk to PCNFS). No. It
doesnt work.

Perhaps, I misunderstand the NDIS business...but, I was excited to
find a device driver for the FDDI card that claimed "NDIS-ness" since
I have seen my PCNFS talk NDIS to my ether card...and, assumed, that
I had found a way to TCP over my NP card. By the way, if anyone has
a solution for this more general problem, feel free to offer it up.

Thanks in advance,
Merle
neerma at cod.nosc.mil
(619)553-4135
.

-----------[000139][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 1992 22:58:11 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Does IP fragment broad/multicasts

In article <1992Sep11.205542.24508@noao.edu> rstevens@noao.edu (W. Richard Stevens) writes:
>>We've been having an argument here concerning whether or not IP
>>fragments broadcast and multicast packets.
>
>All the BSD kernel code that I've seen refuses to fragment a broadcast
>packet.  You get EMSGSIZE instead.  I don't know about multicasting.

I don't think there's anything in the IP specification that requires that
fragmentation of broadcasts be refused.  However, neither is there anything
that requires that an IP implementation be able to fragment (they *are*
required to be able to reassemble).  So, an implementation may refuse to
send a datagram larger than the interface MTU for any reason
(non-implementation of fragmentation, destination address, configuration
options, phase of moon, etc.).

-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000140][next][prev][last][first]----------------------------------------------------
Date:      Fri, 11 Sep 1992 23:09:54 GMT
From:      ole@Csli.Stanford.EDU (Ole Jacobsen)
To:        comp.protocols.tcp-ip
Subject:   INTEROP volunteers

                        Call for Volunteers:

           The INTEROP Conference Assessment Team (CAT)

Interop Company is seeking student volunteers to serve as quality
control monitors for INTEROP 92 Fall, to be held in San Francisco,
California, October 28-30, 1992.  This is a unique opportunity for
students to attend the industry's premier networking conference and
tradeshow, while helping us improve the quality and consistency of the
conference.

As a CAT member you will receive:

* Complimentary conference registration for all three conference days

* Complimentary conference notes (for your assigned conference, there
  are actually 4 conferences within a conference this year).

* Complimentary lunch all three days

* Special INTEROP CAT T-shirts


As a CAT member you will be asked to:

* Monitor preassigned conference sessions on 1 of the 3 conference
  days, by submitting written reports and acting as the "eyes and ears"
  of the conference organizers. We will provide you with a basic
  evaluation form to aid the preparation of the reports.

(You will be free to attend any conference session and the INTEROP
exhibition on your "days off.")

* Provide an accurate count of the number of people attending the
  sessions you are assigned to. ("Clickers" will be provided!)

Successful CAT candidates will be students currently enrolled in a
computer science or electrical engineering course at graduate or
post-graduate level. Applicants should have some understanding of (and
interest in) computer networking issues. All applications must be
received by October 1, 1992.

To apply, send e-mail to: ole@interop.com with a brief biography and
relevant contact information.

Ole
-- 
Ole J Jacobsen, Editor & Publisher ConneXions--The Interoperability Report
Interop Company, 480 San Antonio Road, Suite 100, Mountain View, CA 94040,
Phone: (415) 962-2515  FAX: (415) 949-1779  Email: ole@csli.stanford.edu

-----------[000141][next][prev][last][first]----------------------------------------------------
Date:      Sat, 12 Sep 1992 12:29:43 GMT
From:      dhuber@autelca.ascom.ch (Daniel Huber)
To:        comp.sys.hp,comp.protocols.tcp-ip
Subject:   traceroute for HP's 800 series (HP-UX 8.0) ??

Hi there,

Could please somebody point me to a traceroute working on a 825 and on
a 835 with HP-UX 8.0?

There are *many* traceroute to find on archie, but which one will work
for HP?

At least the traceroute version 4.0 runs only on Sun's because of the
NIT interface.... (this version I have and use it on our Sun's...)

Thanks for any help

Daniel

-- 
Daniel J. Huber, AIN1 / Ascom Autelca AG / CH-3073 Guemligen / Switzerland
SMTP: dhuber@autelca.ascom.ch / Voice: +41 31 9996664 / Fax: +41 31 9996751 
X.400: /G=Daniel/S=Huber/OU=Autelca/O=Ascom/P=EUnet/A=Arcom/C=ch/
UNIX is the answer....what was the question?

-----------[000142][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 92 18:09:06 GMT
From:      avalon@coombs.anu.edu.au (Darren Reed)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

pahountis@ncf.al.alcoa.com (Julia Pahountis) writes:

>Has anyone, or do you know of anyone, who has done NFS over a satellite link?  (Can it be done?)
>I would appreciate any information on how it was accomplished and what
>kind of problems were
>encountered.

Well Australia is connected to the USA via a fairly fast satellite link
(512kbit/s) and I've never seen any problems when using an NFS disk
mounted from the USA.

How was it done ?

# mount wuarchive.wustle.edu:/archive/mirrors /mirror

Was much easier than having ftp running (although slower too).

Darren.

-----------[000143][next][prev][last][first]----------------------------------------------------
Date:      Sun, 13 Sep 1992 00:28:52 GMT
From:      raob@mullian.ee.mu.oz.au (richard oxbrow)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

In article <avalon.716321346@coombs> avalon@coombs.anu.edu.au (Darren Reed) writes:
>pahountis@ncf.al.alcoa.com (Julia Pahountis) writes:
>
>>Has anyone, or do you know of anyone, who has done NFS over a satellite link?  (Can it be done?)
>>I would appreciate any information on how it was accomplished and what
>>kind of problems were
>>encountered.
>
>Well Australia is connected to the USA via a fairly fast satellite link
>(512kbit/s) and I've never seen any problems when using an NFS disk
>mounted from the USA.

mounting over a 57kbit/s link isn't all that much fun ;-) but it works
provided your link doesn't died..

I would suggest you have a look at transarc's/CMU AFS (andrew files system)
which performs a similar function to NFS but has a few other nifty
features, like local file caching, added security etc..  (to cut down
on the link traffic)

	richard/..
 
richard oxbrow			   |internet    raob@mullian.ee.mu.OZ.AU
dept. ee eng,  uni of melbourne    |uunet       ..!uunet!munnari!mullian!raob
parkville, victoria         3052   |fax         +[613] 344 6678   	   
australia               	   |phone       +[613] 344 6782

-----------[000144][next][prev][last][first]----------------------------------------------------
Date:      Sun, 13 Sep 1992 02:35:40 GMT
From:      gtoal@ibmpcug.co.uk (Graham Toal)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

In article <9225710.24475@mulga.cs.mu.OZ.AU> raob@mullian.ee.mu.oz.au (richard oxbrow) writes:
>I would suggest you have a look at transarc's/CMU AFS (andrew files system)
>which performs a similar function to NFS but has a few other nifty
>features, like local file caching, added security etc..  (to cut down
>on the link traffic)

While on the subject, could someone explain to me please about this
new filesystem which uses ftp protocols.  For those of us on 14.4kbit
ppp links it seems like a really efficient way to access places.
I assume it has some local cacheing or buffering too...?

G
-- 

-----------[000145][next][prev][last][first]----------------------------------------------------
Date:      Sun, 13 Sep 1992 13:58:32 +0000
From:      vadim@cix.compulink.co.uk (Vadim Lebedev)
To:        comp.protocols.tcp-ip
Subject:   No subject


Hello,

I'm working for Kortex International.
We are located in Paris, France. We're considering to get Internet
connection. The question is: How does one obtains Internet 
connection in France? 
-----------------------------------------------------------------------
Vadim Lebedev                      |   Kortex International
vadim@cix.compulink.co.uk          |   139-147 av. Paul-Vaillant Couturier
                                   |   93126 La Courneuve - CEDEX, France



-----------[000146][next][prev][last][first]----------------------------------------------------
Date:      Sun, 13 Sep 92 18:39:48 EDT
From:      dos@major.panix.com (Dave O'Shea)
To:        comp.protocols.tcp-ip
Subject:   Re: Problem with NCSA Telnet and Orchid SVGA

joel@skater.GSFC.NASA.Gov ("Joel K. Gallun") writes:

> Hi,
> 
> I have been using NCSA Telnet for some time with great success.  It is
> an excellent package.
> 
> However, I recently upgraded my video controller to an Orchid ProDesigner
> IIs with 512K VRAM and I started having a problem.  When I exit from
> Telnet, the screen goes blank and stays that way until I give the 3
> finger salute.
> 
> I have tried setting the CONFIG.TEL option to use BIOS with no change
> in the symptoms.  Here are the details:
> 
> 	Compaq deskpro 386/20
> 	4 Meg RAM
> 	NEC Multisync 3/D monitor
> 	DOS 5.0
> 	3COM 3C503 ethernet card + packet driver using
> 		interrupt 0x7e
> 	No drivers or TSRs loaded
> 	NCSA Telnet 2.3

Just for the heck of it, have you tried using interrupt 0x66 instead? It
seems to work better for me...


--
"A leopard never changes his stripes" 
           -- Al Gore, candidate for vice-president

-----------[000147][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 05:27:15 GMT
From:      paul@atlas.abccomp.oz.au (Paul Brooks)
To:        comp.protocols.tcp-ip
Subject:   RARP/BOOTP server for small IPnumber pool wanted

One or two have been talked about a while ago, but I didn't write them down.
Now a client needs exactly what I remember reading about!

I need a BOOTP or RARP server that can dole out IP numbers from a small pool
for a much larger collection of potential hosts (PCs), of which only a 
small fraction will be using TCP/IP connections at any one time. When all
available IP numbers have been allocated, subsequent requests are refused.
I would guess that periodic PINGs are also required to determine when a host
has finished with its IP number.

Does anybody have a pointer to such a package/program for any platform?
Thanks.

-- 
Paul Brooks               |paul@abccomp.oz.au      |LIFE is a bowl of cherries:
TurboSoft Pty Ltd         |pwb@newt.phys.unsw.oz.au|  sweet at first, until you
248 Johnston St. Annandale|                        |  reach the pits.
Sydney Australia 2038     |ph: +61 2 552 3088      |  

-----------[000148][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 11:52:59 GMT
From:      aronsson@sectra.se (Lars Aronsson)
To:        comp.protocols.tcp-ip
Subject:   Rcp -- is that rlogin or what?

What hides under the hood of the "rcp" command? These UNIX (HP-UX and
SunOS) manuals don't tell the whole story. I have seen an Internet RFC
on Rlogin, and Richard Stevens' excellent book (UNIX Network
Programming) also mentions that protocol. However, I fail to find any
references to rcp.

I thought rlogin, rsh/remsh, and rcp were all the same thing in
different shapes. They all use the same system of .rhost files. Now I
have seen people comparing the performance of rcp with "cat | rsh",
which suggests they are different, and I am not so sure any longer.
-- 
Lars Aronsson, SECTRA AB, Teknikringen 2, 583 30 Linkoping, Sweden

-----------[000149][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 13:06:59 GMT
From:      jpj@jet.uk (Jean-Paul V Jeral)
To:        comp.protocols.nfs,comp.protocols.tcp-ip
Subject:   Re: NFS Over Satellite Link

In <9225710.24475@mulga.cs.mu.OZ.AU> raob@mullian.ee.mu.oz.au (richard oxbrow) writes:
 (stuff deleted...)
>I would suggest you have a look at transarc's/CMU AFS (andrew files system)
>which performs a similar function to NFS but has a few other nifty
>features, like local file caching, added security etc..  (to cut down
>on the link traffic)
 
>	richard/..

Where can I get informations on this AFS? Sorry if it is a FAQ.
Thanks.
           Jean-Paul Jeral    (jpj@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.

-----------[000150][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 15:47:31 GMT
From:      vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Rcp -- is that rlogin or what?

In article <aronsson.716470900@jeeves>, aronsson@sectra.se (Lars Aronsson) writes:
> What hides under the hood of the "rcp" command? These UNIX (HP-UX and
> SunOS) manuals don't tell the whole story. I have seen an Internet RFC
> on Rlogin, and Richard Stevens' excellent book (UNIX Network
> Programming) also mentions that protocol. However, I fail to find any
> references to rcp.
> 
> I thought rlogin, rsh/remsh, and rcp were all the same thing in
> different shapes. They all use the same system of .rhost files. Now I
> have seen people comparing the performance of rcp with "cat | rsh",
> which suggests they are different, and I am not so sure any longer.


The only real and true documentation for rcp, rlogin, rsh, and so on are
the BSD sources.  The new RFC's and the various text books do their best,
but they are not authoratative.

The source for the r-commands, as well as ruserok() (the library code
that cracks /etc/hosts.equiv and ~/.rhosts) should be in your
favorite BSD source source.  Try uunet. 

Rcp is not the same as `cat | rsh`, although rcp does use the rsh
protocol and (presumably) rshd on the remote machine.  Besides details
like the implementation of `rcp -r`, there are things you can do that
affect performance.


Vernon Schryver,  vjs@sgi.com

-----------[000151][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 16:35:13 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Rcp -- is that rlogin or what?

In article <aronsson.716470900@jeeves> aronsson@sectra.se (Lars Aronsson) writes:
>What hides under the hood of the "rcp" command?

rcp and rdump are built on top of rsh.

>I thought rlogin, rsh/remsh, and rcp were all the same thing in
>different shapes. They all use the same system of .rhost files.

rlogin is not built on the same protocol as rsh/rcp/rdump, but both rlogind
and rshd perform hosts.equiv/.rhosts authentication.  The rlogin protocol
is designed for interactive use, and provides mechanisms for transmitting
the terminal type, terminal speed, window size changes, and cooked/raw mode
changes.  Rshd just runs the given command with stdin/stdout/stderr
connected to the socket, and provides no special processing.  Rlogind
starts up a login shell, while rshd just runs an ordinary subshell.

>Now I
>have seen people comparing the performance of rcp with "cat | rsh",
>which suggests they are different, and I am not so sure any longer.

They should be similar in performance.  rcp has options to transmit file
dates and permissions, but for a large file the actual data transfer should
be the limiting factor.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000152][next][prev][last][first]----------------------------------------------------
Date:      Mon, 14 Sep 1992 17:00:36 GMT
From:      chatel_m@annecy.enet.dec.com (Marc Chatel)
To:        comp.protocols.tcp-ip
Subject:   Looking for 5250 emulation software on Telnet...


Hello all,

   I am looking for some source that would help me out here. I am trying
to get 5250 (not the 5150 Van Halen album... :-)) emulation source to plug on
a Telnet package. With luck, the stuff may even be called "tn5250".

   Anybody has an idea of where I may find such a beast???

   Special bonus to the first chap who gives me pointers to both client AND
daemon source: a free postcard of the French Alps!!!

				Regards,

				Marc Chatel,
				Digital Equipment,
				Annecy, France

		E-mail to:	chatel_m@annecy.enet.dec.com
		Phone:		(33) 50.09.42.56

Disclaimer: Digital Equipment Corporation is not in any way responsible for
            what I write, say, do, or drink. By the way, what time is it???

-----------[000153][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 92 18:47:08 GMT
From:      zjhc05@hou.amoco.com (John H. Caldwell)
To:        comp.protocols.tcp-ip
Subject:   Internetworking with TCP/IP Volume III

I just received my copy of Comer and Stevens' new book, but it doesn't
seem to indicate where one may retrieve from the network the source
code listed in the book.  I tried archie and the usual suspects like
/published/books on ftp.uu.net.  Can anyone tell me where I can pick
up the source code from this book?  


---
John H. Caldwell  reply to: zjhc05@hou.amoco.com
Amoco Corp. ITD   Technology Introduction Group
Chicago, IL       (312) 856-3524

-----------[000154][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1992 21:15:23 GMT
From:      cew6@po.CWRU.Edu (Carlin E. Wiegner)
To:        comp.protocols.tcp-ip
Subject:   Speed of different network archi...


Can anyone email me the MB/sec for the following hardware 
architectures....
10Base-T
Thick Ethernet
Thin Ethernet
Fiber Optic 
Appetalk (LocalTalk)

Thanks for the replies...
CW

-----------[000155][next][prev][last][first]----------------------------------------------------
Date:      Mon, 14 Sep 1992 23:16:35 GMT
From:      vthrc@brolga.cc.uq.oz.au (Danny Thomas)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   (Q) service aliases; pros/cons


Fellow Internetters,
                    In the near future we'll be getting our first UN*X 
system (currently only PCs/Macs) and intend it to offer various services 
- NTP, NNTP, etc. I've never really had to administer domain names 
previously, but my preference is to define an alias for each service 
offered rather than configure each client to bind to a particular host. 
This is a prejudice I've developed from dealing with email addresses - I 
just don't like the idea of addresses tied to a host, a person's name @ 
organizational-unit name seems more logical. At a guess we'd be looking 
at:
    ftp.vthrc.uq.oz.au (ie the primary/preferred FTP site in the domain)
    ntp.vthrc.uq.oz.au
    nntp.vthrc.uq.oz.au
    x500.vthrc.uq.oz.au  (or quipu.vthrc.uq.oz.au)
    ph.vthrc.uq.oz.au
    pop3.vthrc.uq.oz.au
    imap.vthrc.uq.oz.au
    smtp-relay.vthrc.uq.oz.au

Those services are ones I consider centralized within a domain in 
comparison to say NFS where connections are made to specific hosts. I 
guess this division begins to breakdown as a service becomes distributed, 
eg for redundancy (NTP?), but in these cases I assume each client is 
configured with a list of machines (or aliases) to try, probably ordered 
by preference.

    Anyway I'd to hear what other people think about this area or 
pointers to FAQs where this has been previously covered. I'd appreciate 
email copies of any netnews replies as the news-server is not as reliable 
as I'd like.


cheers,
X500: @c=AU [I'm about to update VTHRC entries, soon...]
       @o=University of Queensland
        @ou=Physiology and Pharmacology Department
         @ou=Vision Touch and Hearing
          @cn=Danny Thomas
           @rfc822Mailbox=vthrc@cc.uq.oz.au


-----------[000156][next][prev][last][first]----------------------------------------------------
Date:      Tue, 15 Sep 92 13:24:42 GMT
From:      berg@physik.tu-muenchen.de (Stephen R. van den Berg)
To:        comp.protocols.tcp-ip
Subject:   getprotobyname(), does it ever change?

I was wondering, has anyone ever seen an implementation where the protocolno.
for UDP changed without exchanging all the libraries as well?

Or, could anyone imagine a situation where it would be helpful to change the
protocolno. for UDP?

Or, perhaps more direct, has anyone ever edited the /etc/protocols file and
changed the number for TCP or UDP?

Just curious about how volatile these numbers actually are.
-- 
Sincerely,                                  berg@pool.informatik.rwth-aachen.de
           Stephen R. van den Berg (AKA BuGless).    berg@physik.tu-muenchen.de

"I have a *cunning* plan!"

-----------[000157][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 92 13:36:18 GMT
From:      markl@spuddy.uucp (Mark Liversedge)
To:        comp.protocols.tcp-ip
Subject:   DDN Handbooks

I am trying to get hold of the DDN Protocols Handbook from SRI. I have
sent a message to service@nic.ddn.mil but I think it is an autonomous
mail server and will discard my wordy request.

I would appreciate an e-mail/phone number to call. I am based in the UK
and cannot dial US-800 numbers. (My phone 44-932-571212 Druid Systems Ltd)

Any responses by email please.
-- 

 * Meeeow ! Call Spuddy on (0203) 638780/638693 for FREE mail & Usenet access *

-----------[000158][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 92 19:00:12 GMT
From:      jordan@IMSI.COM (Jordan Hayes)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP reliability

Nicolas Droux <droux@hei.unige.ch> writes:

	Our Problem is, that a receiver implemented on a NeXT receives
	much more packets than the sender implemented on a Sun
	workstation sends (1.3 : 1).

Weird.  I've never seen this before.

By the way, in your connection code:

	do
	{                               /* connect to the power server */
	  connect_status = connect(socket_fd,
	    (struct sockaddr *)&sock_addr, servlen);  
	  connect_retries--;
	} while((connect_status < 0) && (connect_retries));

If connect() fails, the socket is now invalid.  You must close() it and
make another one.

	The data exchange is made by normal read() and write() calls
	followed by an acknowledgement. We transmitt packets of 5000
	bytes of unstructered data.  If a sender send 300 packets (as a
	concrete example) the receiver receives over 400 packets and
	about 120 of them are invalid.

Are you sure the read() and write() part is correct?

What do you mean by "invalid" ...?

/jordan

-----------[000159][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1992 22:11:36 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: getprotobyname(), does it ever change?

In article <1992Sep15.132442.2149@Urmel.Informatik.RWTH-Aachen.DE> berg@physik.tu-muenchen.de (Stephen R. van den Berg) writes:
>I was wondering, has anyone ever seen an implementation where the protocolno.
>for UDP changed without exchanging all the libraries as well?

I suspect that the only reason one would ever change /etc/protocols would
be for experimental purposes.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000160][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1992 23:22:22 GMT
From:      trier@slc6.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: getprotobyname(), does it ever change?

In article <195n2pINN91b@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
>I suspect that the only reason one would ever change /etc/protocols would
>be for experimental purposes.

It would seem to me that the /etc/protocols lookups would also be useful
for packet-dump programs.  I see very little use for being able to change
the numbers of standard protocols, but the ability to add numbers for new
protocols might be handy when using a network diagnostic tool of some sort.

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000161][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 92 02:16:17 GMT
From:      dave@hiauly1.hia.nrc.ca (John David Anglin)
To:        comp.sys.hp,comp.protocols.tcp-ip
Subject:   Re: traceroute for HP's 800 series (HP-UX 8.0) ??


Try version on radar.actc.uiuc.edu.  It works on my 700.

Dave Anglin

-----------[000162][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 02:38:37 GMT
From:      osmana@well.sf.ca.us (Osman Arslaner)
To:        comp.protocols.tcp-ip
Subject:   Questions About SLIP.


We are planning to use SLIP for our remote PCs which are not
on the corporate backbone.
These PCs are in the yard or shop floor and connected to a terminal
server which runs SLIP through dedicated circuits (19.2kbs).
They will be running applications and also will access hosts on
the corporate backbone.

My questions are: is SLIP a reliable protocol to use ? Our slip program
does not support flow control, could that be a problem ? Does a SLIP
connection pass through broadcasts which are sent on the Ethernet network ?

Also, we are planning to use dial-up SLIP. So far it works fine
only with same type of modems on both ends. We were not able
to make it work with different modems on both ends. We used 9600 baud
modems for testing.

Would there be any problems associated with this set-up ?
Any help will be appreciated.
Thanks..


--------------------------------------------------------------------
|  Osman Arslaner, CEng.         |   Tel. (514) 399 7305           | 
|  Communications Engineer,      |   Fax. (514) 399 7062           | 
|  ATS, Information Systems.     |   E-Mail: arslaner@cnmtl.cn.ca  |
|  Canadian National Railways.   |           osmana@well.sf.ca.us  |
|  Montreal, Quebec H3B 2M9      |                                 |
 --------------------------------------------------------------------
|  "Truth is silent; It's views and opinions that make all the     |
|   noise". - ZEN Graffiti.                                        |
 --------------------------------------------------------------------

-----------[000163][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 04:05:51 GMT
From:      paul@uxc.cso.uiuc.edu (Paul Pomes - UofIllinois CSO)
To:        comp.protocols.tcp-ip
Subject:   who/what still uses TCP port 105, csnet-ns?

[ Sent also as email to iana@isi.edu and to the IETF m-list. -pbp]

The University of Illinois Computing and Communications Services Office
has been running a descendent of the CSNet central nameserver for several
years now.  The code has proven popular and is now running publicly at
40-odd sites and privately at several more.  The private sites either 
don't advertise the server or restrict queries to local hosts.

The original CSNet server ran on port 105/tcp.  I'm not sure if 105/udp,
also assigned to "csnet-ns" in RFC-1340 (Assigned Numbers), was ever used.

The present code does not use anything like the original CSNet nameserver
protocol.  About the only part left of the original server is the database
engine and even that has been extended.  The overall intent is still much
the same: provide a way to locate and register email addresses based on
human information such as a person's name, their office address, etc.  In
addition it functions as an electronic white pages for traditional phone
book type information.

This leads to two questions: Is anyone still using the original CSNet
nameserver protocol on port 105/tcp?  We did not think so back in late
1989 and so stuck with the same port number even while the protocol
was changed to fit our needs.  Second, can 105/tcp be re-assigned to
the new name "ph" as defined by the ph protocol paper?  The protocol
definition is available for anon-FTP from uxc.cso.uiuc.edu in
net/qi/doc/protocol.{me,ps}.

/pbp
-- 
"A well-schooled electorate, being necessary to the security of a free State,
the right of the People to keep and read Books shall not be infringed."
	-- J. Neil Schulman

-----------[000164][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 92 05:30:33 GMT
From:      gah@trc.mew.mei.co.jp (Gary A. Hildebrand)
To:        comp.dcom.lans.fddi,comp.protocols.tcp-ip
Subject:   Re: NDIS drivers

In article <1992Sep11.221149.15137@nosc.mil> neerma@nosc.mil (Merle A. Neer) writes:
>>I am trying to get PCNFS to work with an NDIS driver supplied by
>>Network Peripherals for their FDDI board (the EISA version).
>>
>>The Network Peripheral documentation references a "NP-NDIS driver
>>for Microsoft LAN Manager Ver 2.0" and a "NP-PCNFS driver for SUN
>>NDIS". The "for" clauses have me confused:I thought the idea behind
>>NDIS (and, for that matter, ODI and packet driver) was that once
>>an NDIS-compliant device driver was produced for a particular net
>>card, it could interface with any NDIS-compliant protocol stack (like
>>PCNFS). It appears to me that in my particular situation, NP has
>>produced device drivers that talk to particular protocol stacks (and
>>yet they mysteriously use the NDIS label!). 
>>
>>Perhaps, I misunderstand the NDIS business...but, I was excited to
>>find a device driver for the FDDI card that claimed "NDIS-ness" since
>>I have seen my PCNFS talk NDIS to my ether card...and, assumed, that
>>I had found a way to TCP over my NP card. By the way, if anyone has
>>a solution for this more general problem, feel free to offer it up.

I am not sure about the documentation that came with your NP board, since
it may be different from what came with our boards, depending on when you
got it.  But there is only one NDIS driver from NP for their board.  In
that sense, you are right about the "idea behind NDIS".  The problem
actually lies with Sun's PC-NFS.

Although it would be nice to connect any NDIS driver to any NDIS-compatible
protocol stack, there are limitations, just as with packet drivers or ODI.
Usually it's not a problem to connect any NDIS driver for a particular
media "type" (such as classic Ethernet, IEEE 802.2/3, Token Ring, FDDI,
etc.) to any NDIS-compatible protocol stack which supports that type.  But
not all protocol stacks support all media types.  In particular, the latest
version of Sun's PC-NFS will support Ethernet and Token Ring media, but not
FDDI.  This has to do mostly with the different format of frame/packet
headers and node/station addresses for different types of media.

Alas, there is a "quick fix" in the case of the driver for the NP board.
If you enable (set to "Yes") the ETHERNETEMULATE field in the NDIS
configuration file PROTOCOL.INI (usually found in the LANMAN subdirectory
of the root of your boot drive), the NP NDIS driver will work with PC-NFS,
since it then presents itself with a media type of IEEE 802.2/3.  The down
side is, maximum packet size will be limited to the Ethernet maximum of
around 1500 bytes, rather than the FDDI maximum of around 4500 bytes,
limiting performance.  Specifically, I have seen no difference in
performance between using an Ethernet interface (with an 8-bit bus, no
less) and using the NP-EISA!!!  Of course, there are many things that can
limit performance, besides the board/media.

If Sun is listening, when will FDDI support be added to the PC-NFS driver
NFS-NDIS.SYS?

Gary
--
/ Gary A. Hildebrand                 Internet: gah@mew.mei.co.jp       \
/  Matsushita Electric Works, Ltd.   UUCP:     uunet!mew.mei.co.jp!gah \
/   13-2, Mita 5-chome, Minato-ku    Fax:      03-3451-0793            \
/    Tokyo 108, JAPAN                Tel:      03-3452-4941            \

-----------[000165][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 14:01:43
From:      rodin@vax.ftp.com  (Jon Rodin)
To:        comp.dcom.lans,comp.protocols.tcpip,comp.dcom.lans.misc
Subject:   Does anyone use Chaosnet any more?

Does anyone know of any operational chaosnets still in existence?

Jon Rodin                  FTP Software, Inc.            voice: (508) 659-6261
rodin@ftp.com              2 High Street                 fax:   (508) 794-4488
                           North Andover, MA  01845


-----------[000166][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 1992 00:57:06 -0700
From:      kfan@aludra.usc.edu (Kan F. Fan)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Need help of MTP (Mutlicast Transport Protocol) RFC1301

Hi folks,
	
	I'm gonna to implement MTP on SunOS and I'm wandering if there is anyone out there who have the experience in implementation or developed any software above it. I really need to know more about it in addition to RFC. Would you have any source of information or any precious idea, pls send me by email. Thanks in advance.

Kan F. Fan
kfan@usc.edu


-----------[000167][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 13:46:15 GMT
From:      deej@cbnewsf.cb.att.com (david.g.lewis)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP over ISDN

In article <3302@aria.Ascend.COM> marc@aria.Ascend.COM (Marco S Hyman) writes:
In article <1992Sep10.143836.11131@cbfsb.cb.att.com> I write:
> > In a previous article I wrote:
> > >The "mandatory" called party number isn't in setups that come from a
> > > Definity Generic I PBX.
> > 
> > Over a PRI it certainly is....
>
>Uhh, this may be a function of the funny way we use our Definity... We have
>PRI on both the Network and Terminal side, an AT&T PRI from the network and
>several (6 or 8) internal T1 lines, some configured as PRI, with the PBX
>told they are tie lines.  Each internal line is a trunk group of its
>own and is dialed by using the trunk group.  When dialing from one PRI trunk
>group to another PRI trunk group the Called Party Number IE on the answering
>side exists but has no number.

The "called party number" is *not* mandatory, and is nowhere that I can find
claimed to be mandatory.  The Called Party Number Information Element, even,
is optional in a SETUP message, although either the Keypad IE or the CdPN IE
must be included, so if the Keypad IE is not included, the CdPN IE should be
considered mandatory.  But the CdPN IE has a minimum length in Q.931 of 2
octets, which means a SETUP message containing a Called Party Number IE
consisting of an IE identifier and a length is a "valid" message from the
protocol perspective - it should not be considered to be missing a mandatory
IE.

Such a SETUP message, if sent by a user to a network to set up a call, may
result in the setup request being rejected by the network because
insufficient information is provided, but should not result in the network
clearing the call because a mandatory IE is missing.  In the other
direction, one could envision a network providing an ISDN DNIS (Dialed
Number Identification Service) type service, where the called party number
is only delivered if the service is subscribed to; in this case, it would be
appropriate for the network to send a CdPN IE with no digits if the service
is not subscribed.

> > Again, to get back to the original thread, certain IEs are designated as
> > usable for user to user information transfer.  These include UUI, High Layer
> > Compatibility, Low Layer Compatibility, subaddress IEs, and Codeset 7 IEs.
> > If you attempt to use anything else to transfer user to user information,
> > there is no guarantee, explicit or implied, that it will be delivered the
> > way you want it - or that it will be delivered at all.
>
>Agreed.  However, given the heterogeneous nature of the network there is
>currently no guarantee that any of the above will work between any two
>arbitrary endpoints.

Also agreed; even with end-to-end SS7 on a connection, there's no guarantee
that any UUI won't get dropped somewhere along the way.

David G Lewis                              AT&T Bell Laboratories
david.g.lewis@att.com or !att!goofy!deej     Switching & ISDN Implementation

-----------[000168][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 15:03:33 GMT
From:      glenr@aruba.UUCP (Glen Reesor)
To:        comp.protocols.tcp-ip,comp.misc
Subject:   Dial-up IP vs. dedicated leased line

I currently investigating a direct IP connection to the Internet.  It appears
that there are two major ways I can accomplish this--dial-up IP using SLIP
and dedicated leased line with a router.  The dial-up IP would be 19.2 kbps
and the leased line would be 56 kbps.

Besides the lower bandwidth, SLIP seems to give all the functionality of
the leased line (assuming I remain connected all the time) for a significantly
lower cost.

What do people out there who are using SLIP think?  What experiences/horror
stories do you have?  Any comments or suggestions about these two alternatives?

FYI my LAN consists of 7 HP9000/730's and 30 X-terminals.

Please e-mail.  I will post a summary if there is interest.
-- 
Glen Reesor                       |Internet Style: aruba!glenr@uu2.psi.com
Systems Administrator, Project Zed|UUCP     : ...!uunet!uu2.psi.com!aruba!glenr

-----------[000169][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 92 15:05:00 GMT
From:      berman@nlm.nih.gov (Lew Berman)
To:        comp.protocols.tcp-ip
Subject:   RPC/XDR vs Pure Sockets


In the development of a client/server application that will
run over the Internet, can someone please discuss the trade-offs
between RPC/XDR and pure socket calls.  I've read in Steven's
UNIX book that the overhead associated with RPC calls can be as high
as a factor of 100.  If this is true, than this would be a reason
not to use RPC; however when passing structures and arrays of data
(images for example) would this be simpler with RPC's and XDR?


Please reply directly to me at berman@ceb.nlm.nih.gov

I'll repost a summary of replies.

Thanks,

Lew Berman
National Library of Medicine

-----------[000170][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 17:14:36 GMT
From:      k@hprnd.rose.hp.com (Steve Kao)
To:        comp.protocols.tcp-ip
Subject:   Re: Speed of different network archi...

In comp.protocols.tcp-ip, cew6@po.CWRU.Edu (Carlin E. Wiegner) writes:
> Can anyone email me the MB/sec for the following hardware 
> architectures....
> 10Base-T
> Thick Ethernet
> Thin Ethernet
> Fiber Optic 

These are all 10Mbps networks.  For Fiber, I'm assuming FOIRL instead of
FDDI.  FDDI is much faster.  The Thick Ethernet is capable of greater
than 10Mbps.

> Appetalk (LocalTalk)

I don't know, but I think this is also 10Mbps

-----------[000171][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 19:53:49 GMT
From:      mbg@cbnewsi.cb.att.com (mitchell.b.germansky)
To:        comp.protocols.tcp-ip
Subject:   ---<<< Wollongong 3B2 TCP select() bug?? >>>---

my project has encountered what seems to be a problem in wollongong's
tcp/ip on a 3b2.  if you have any experience in resolving this,
please get in touch with me.

we have a 3b2 connected over ethernet to a sharebase database
machine (aka britton-lee idm, teradata, ncr).  we are using
Wollongong Integrated Networking WIN/3B: Release 3.2A on 3b2
unix 3.2.3

we have noticed a predictable 2 second delay every 20 database
transactions.  examining the process stack trace during this 2
second interval revealed that the process was asleep in a tcp/ip
library select() call.  we believe that this problem is really
independent of the database and will write some tcp code to attempt
to reproduce the select() delay.  the sharebase performance is
constant even for the transaction that blocks for the 2 seconds.

we ran the same tests from a sun with no delays.  i also wrote some
raw tcp test code that showed no delays.

we have heard rumors that other people have had similar problems.
we are wondering if there are any workarounds or perhaps even a
patch available.

if you have any suggestions or tips on what to look at  or do next,
please contact me via email or phone.

thx!!

--
Mitch Germansky                           908-805-7841
att!suntoss!mbg

-----------[000172][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 92 21:57:25 GMT
From:      atul@yoko.rutgers.edu (Atul Sharma)
To:        comp.protocols.tcp-ip
Subject:   Anomalous behaviour of TCP over high speed network.


I have written a network interface driver for a high speed network(an
ATM switch) on a sparc 2 (basically an S-Bus driver)workstation. The 
performance seems satisfactory when I use UDP, or ICMP. But if I use 
some TCP utility like ftp, or telnet the performance is abysmal 
(sometimes .15 kbytes/sec !!). 

If I introduce some kind of delay in my driver, currently I am putting
some log statements, then the performance seems respectable.

What is the reason of this anomaly? 

First I thought TCP is doing more retransmissions than required, i.e. packets
were being retransmitted even before the retransmit timer goes off, because
of some other reason. I am not sure, but say TCP window update timer goes 
off(?)

But then I started "netstat -Imy_net 1", and saw the following behaviour:

With no delay in the driver, there were bursts of activity for a second and
then there was no activity for a long time. When I introduced the delay
there were less no of packets per second but the activity was well
ditributed as there were some packets transmitted always.

If any Guru has observed similar behaviour for high speed networks, could
he/she please let me know how to go about solving this problem?

Thanks in advance.

--Atul
--------------
atul@paul.rutgers.edu
{backbone}!rutgers!yoko.rutgers.edu!atul

-----------[000173][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 92 22:38:08 GMT
From:      mike@gordian.com (Michael A. Thomas)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP break - terminal server?

In article <1992Sep1.182505.25892@mprgate.mpr.ca>, parker@mprgate.mpr.ca (Ross Parker) writes:

  The short of this is that you may be screwed.
 
> I am currently connecting a number of systems in our computer room via
> serial cables to a Xyplex terminal server. Connections to individual ports
> on this server are made using the 'xyp_ptyd' daemon supplied by Xyplex,
> running on a central console 'host' system. This daemon essentially makes
> a connection between each port on the serial side of the terminal server and
> a pseudo-tty on the UNIX host system. I can then connect to these individual
> pseudo-ttys using the 'console' package from Ohio State, but that's irrelevant
> to this problem...
> 
> The problem is that I need to be able to generate a break signal on one of
> the serial lines on the terminal server. I have no problem in modifying the
> Xyplex xyp_ptyd code in order to do so, but I don't know what/how to send
> to the terminal server (over the network connection) to tell it that I want
> it to generate a break over the serial line.

  I'm not familiar with the specifics of the Xyplex pty daemon, but if
it is implemented in a similar fashion to the Lantronix pty daemon
(which I suspect is the case) there is no real way of transporting
the break condition across the pty pair.
  The main problem with the pty implementations is that they were
not implemented with the intent of emulating a *real* tty driver.
Things like break, modem control, parity, baud rate etc. are accepted
by the slave side of the pty and silently ignored. The master side
of the pair is completely clueless that an ioctl() of this form
has been performed by the slave side. 
  Unlike parity/baudrate (which you might be able to hack up on 
the master side by reading the parameters in the slave pty), break
is even worse since it a transient condition. There is no way
the master side could know when the slave requested the break.
  The closest suggestion that I could make is to send a signal
(say SIGUSR1) to the pty daemon process which in turn will
cause it to send the Telnet break IAC sequence. Unfortunately
this would require non-transparency on the slave side which
is probably not acceptable to you.
  The only other solution that I can think of is to hack on the
PTY code to have it transport the break condition to the master
side of the pty. The BSD pty sources are available to hack on
and it may not be too hard (just send a signal off to the master
process) though I've not tried it myself.

> 
> Can anyone help? Is there a 'normal' way of passing this kind of thing through
> TCP/IP? The Xyplex manual seems to indicate that it will do the same thing
> in the opposite direction - i.e. a break on one of the serial lines can be
> propogated through the terminal server and over TCP/IP to a host system... how?
> 

  This scenario makes some amount of sense. The master side, which 
receives the break IAC from the box can potentially send of a signal
(or whatever an incoming break is supposed to do in the driver) to
the slave process. Unfortunately this is not the direction that
you need.
  Sorry to be sender of despair...

-- 

		Michael Thomas	(mike@gordian.com)
	"I don't think Bambi Eyes will get you that flame thrower..."  
		-- Hobbes to Calvin
		USnail: 20361 Irvine Ave Santa Ana Heights, Ca,	92707-5637
		PaBell: (714) 850-0205 (714) 850-0533 (fax)

-----------[000174][next][prev][last][first]----------------------------------------------------
Date:      Wed, 16 Sep 1992 22:54:18 GMT
From:      jjensen@convex.com (James Jensen)
To:        comp.protocols.tcp-ip
Subject:   Re: Anomalous behaviour of TCP over high speed network.

atul@yoko.rutgers.edu (Atul Sharma) writes:
-
-
- With no delay in the driver, there were bursts of activity for a
second and
- then there was no activity for a long time. When I introduced the
delay
- there were less no of packets per second but the activity was well
- ditributed as there were some packets transmitted always.
-
It sounds a lot like your congestion control code isn't working.  If
you send to fast the packets get dropped and 
tcp has to time out, then retransmit.  congestion control reduces
the send window when a packet gets dropped.  Ideally the window
gets adjusted to just below the point that packets get dropped.

This problem occures most frequently when sending from a fast machine
to a slow machine.  Have you tried sending in the opposite direction?

When you say that UDP perfermance is ok, how are you measuring it?
If you measure it on the send side it can be very misleading because 
of dropped packets.

I don't know if the Sun os has congestion control in it.

Jim Jensen - jjensen@convex.com

-----------[000175][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 92 23:52:50 GMT
From:      ric@updike.sri.com (Richard Steinberger)
To:        comp.protocols.tcp-ip
Subject:   Preventing socket from lingering


	I am writing some client-server applications on SUN and VAX
systems; the server lives on the SUN.  Sometimes, during debugging,
I need to kill of the server (and usually the client as well), without
properly closing the socket.  After I do this, I sometimes cannot restart the
server because, apparently, the socket is still in use.  This is a 
common output form "netstat -a":

udp   0   0  *.2049        *.*
tcp   0   0  tolstoy.2049  rml1.1063

Until the second line disappears (usually within a few minutes), I can't
restart the server.  I do not set any options with a setsockopt() call
at the moment.

Can anyone tell me what I can do to insure that the connection always
dies when the process dies (not a few minutes later), or else, how to kill
the "old" connection once the process no longer exists?  Thanks for
all comments and suggestions.

regards,

	ric steinberger
	ric@updike.sri.com

BTW - I am using a SUN 4/670 with SUNos 4.1.2.  For what it's worth, the
client is a VAX running VMS and using the MultiNet TCP/IP socket library.

-----------[000176][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 92 00:11:12 GMT
From:      mogul@pa.dec.com (Jeffrey Mogul)
To:        comp.protocols.tcp-ip
Subject:   Re: Anomalous behaviour of TCP over high speed network.

In article <Sep.16.17.57.25.1992.26364@yoko.rutgers.edu> atul@yoko.rutgers.edu (Atul Sharma) writes:
>
>I have written a network interface driver for a high speed network(an
>ATM switch) on a sparc 2 (basically an S-Bus driver)workstation. The 
>performance seems satisfactory when I use UDP, or ICMP. But if I use 
>some TCP utility like ftp, or telnet the performance is abysmal 
>(sometimes .15 kbytes/sec !!). 
>
>If I introduce some kind of delay in my driver, currently I am putting
>some log statements, then the performance seems respectable.
>
>With no delay in the driver, there were bursts of activity for a second and
>then there was no activity for a long time. When I introduced the delay
>there were less no of packets per second but the activity was well
>ditributed as there were some packets transmitted always.

This could be the result of an interaction between the sender's Silly
Window Syndrome avoidance algorithm and the receiver's delayed-ACK
mechanism.  Make sure that the sender's buffer size is at least
twice the MTU of the network (i.e., twice the size of the IP
datagrams you are sending).

OK, actually you want it to N*MSS, where N is an integer >= 2 and
MSS is the MTU less 40, but anything bigger than that should give
reasonable results.  Anything smaller than that can give lousy
results.

-Jeff

-----------[000177][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 1992 01:11:38 GMT
From:      Dean.Roth <Dean.Roth@mixcom.mixcom.com>
To:        comp.protocols.tcp-ip,comp.unix.sysv386
Subject:   ka9q & SCO: Can it work?


I got a copy of ka9q from wuarchive, built it and installed it.

It works. (At least I connected to a PC running NCSA through
a serial port.)

Where is documentation of the domain.txt and startup.nos files?
I'm not confident I have them set correctly (or optimally).

Other questions:

	Does domain.txt have to be in the current directory
	when a client (ftp) is started?  initroot() would seem
	to not require this, but nothing calls initroot().

	What do I need to do to use numeric "names" with
	ftp rather than a domain name? In other words,
	why cann't I use "ftp 2.0.0.2" rather than "ftp foo.com"?

Thank you.

Dean

-----------[000178][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 1992 02:00:07 GMT
From:      Dean.Roth <Dean.Roth@mixcom.mixcom.com>
To:        comp.protocols.tcp-ip
Subject:   ka9q for sys V


ka9q on my SCO 3.2.2 system appears to work using SLIP.

When netserv is started, however, nostimer always quits
immediately with the message:  poll: no more processes

Is this something to worry about?

Dean

-----------[000179][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 13:11:57 PDT
From:      Michel.Davidoff@sonoma.edu (Michel Davidoff)
To:        comp.protocols.tcp-ip
Subject:   Re: Getting Internet Access

In article <jdsmith.26.716737028@novell.com> jdsmith@novell.com (Doug Smith) writes:
>Subject: Getting Internet Access
>From: jdsmith@novell.com (Doug Smith)
>Date: Thu, 17 Sep 1992 13:37:08 GMT
>Keywords: internet
>Could anyone tell me how a small company would go about linking up to the 
>Internet? 
>
>Doug Smith
>jdsmith@novell.com
There are a few companies that specialize in connecting organizations into 
the Internet. I also heared rummers the Pacific Bell may be getting into 
this.


Michel Davidoff
E-Mail Michel.Davidoff@Sonoma.edu


        PLEASE VOTE
               PLEASE VOTE
                   PLEASE VOTE
                      PLEASE VOTE

-----------[000180][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 12:19:51 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Preventing socket from lingering

>	I am writing some client-server applications on SUN and VAX
>systems; the server lives on the SUN.  Sometimes, during debugging,
>I need to kill of the server (and usually the client as well), without
>properly closing the socket.
>
>Can anyone tell me what I can do to insure that the connection always
>dies when the process dies (not a few minutes later), or else, how to kill
>the "old" connection once the process no longer exists?

The rule of TCP is that the side that does the "active close" can't reuse
that same port number until 2*MSL have passed.  MSL is the maximum segment
lifetime and RFC 1122 calls for a default value of 2 minutes (but I think
many systems still use values of 30 seconds or 1 minute?).  Normally it's
the client that does the active close, and clients use ephemeral ports, so
this isn't a problem.  The problem occurs when a server is killed, forcing
it to actively close its socket (i.e., the server is the first to send a
FIN to close the connection).  Then you try and restart the server before
2*MSL has expired, the server tries to bind its well-known port, and you
get the address-already-in-use error.

You can't change the rules of TCP, but on many systems you can specify the
SO_REUSEADDR socket option before the bind to get around this.

	Rich Stevens  (rstevens@noao.edu)

-----------[000181][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 1992 13:37:08 GMT
From:      jdsmith@novell.com (Doug Smith)
To:        comp.protocols.tcp-ip
Subject:   Getting Internet Access

Could anyone tell me how a small company would go about linking up to the 
Internet? 

Doug Smith
jdsmith@novell.com

-----------[000182][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 15:28:05 GMT
From:      gsa@easyaspi.udev.cdc.com (gary s anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: Anomalous behaviour of TCP over high speed network.

In article <1992Sep17.001112.9879@PA.dec.com>, mogul@pa.dec.com (Jeffrey Mogul) writes:
|> In article <Sep.16.17.57.25.1992.26364@yoko.rutgers.edu> atul@yoko.rutgers.edu (Atul Sharma) writes:
|> >
|> >I have written a network interface driver for a high speed network(an
|> >ATM switch) on a sparc 2 (basically an S-Bus driver)workstation. The 
|> >performance seems satisfactory when I use UDP, or ICMP. But if I use 
|> >some TCP utility like ftp, or telnet the performance is abysmal 
|> >(sometimes .15 kbytes/sec !!). 
|> >
|> >If I introduce some kind of delay in my driver, currently I am putting
|> >some log statements, then the performance seems respectable.
|> >
|> >With no delay in the driver, there were bursts of activity for a second and
|> >then there was no activity for a long time. When I introduced the delay
|> >there were less no of packets per second but the activity was well
|> >ditributed as there were some packets transmitted always.
|> 
|> This could be the result of an interaction between the sender's Silly
|> Window Syndrome avoidance algorithm and the receiver's delayed-ACK
|> mechanism.  Make sure that the sender's buffer size is at least
|> twice the MTU of the network (i.e., twice the size of the IP
|> datagrams you are sending).
|> 
|> OK, actually you want it to N*MSS, where N is an integer >= 2 and
|> MSS is the MTU less 40, but anything bigger than that should give
|> reasonable results.  Anything smaller than that can give lousy
|> results.
|> 
|> -Jeff

First, if this is a "silly window avoidance" problem then one might
try the "TCP_NODELAY" socket option (assuming your system supports this
concept) to see if the problem disappears.

My guess is that your "high performance switch" has a packet loss
problem.  You might look at the "netstat -s" (tcp protocol stats)
or other equivalent network management service to see if you are
having packet retransmissions, etc.

Another possibility is integrity errors caused by the switch fragmentation/
reassembly process (i.e. going to/from IP packets from/to ATM cells).
NOTE - ICMP doesn't checksum data and SUN'S don't, by default,
turn on UDP checksums (I know, in some recent release of one of
their OS's they actually have changed this default).  This may mean
that packets could be corrupted all the time, however, only TCP
would flag these errors.  Once again, the tcp protocol stats should
show these errors.

-----------[000183][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 15:38:59 GMT
From:      gsa@easyaspi.udev.cdc.com (gary s anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: Preventing socket from lingering

In article <38626@unix.SRI.COM>, ric@updike.sri.com (Richard Steinberger) writes:
|> 
|> 	I am writing some client-server applications on SUN and VAX
|> systems; the server lives on the SUN.  Sometimes, during debugging,
|> I need to kill of the server (and usually the client as well), without
|> properly closing the socket.  After I do this, I sometimes cannot restart the
|> server because, apparently, the socket is still in use.  This is a 
|> common output form "netstat -a":
|> 
|> udp   0   0  *.2049        *.*
|> tcp   0   0  tolstoy.2049  rml1.1063
|> 
|> Until the second line disappears (usually within a few minutes), I can't
|> restart the server.  I do not set any options with a setsockopt() call
|> at the moment.
|> 
|> Can anyone tell me what I can do to insure that the connection always
|> dies when the process dies (not a few minutes later), or else, how to kill
|> the "old" connection once the process no longer exists?  Thanks for
|> all comments and suggestions.
|> 
|> regards,
|> 
|> 	ric steinberger
|> 	ric@updike.sri.com
|> 
|> BTW - I am using a SUN 4/670 with SUNos 4.1.2.  For what it's worth, the
|> client is a VAX running VMS and using the MultiNet TCP/IP socket library.

Applications required to deal with this issue use the the
SO_REUSEADDR option.  Unless someone has added a new widget to 
"eradicate" connections in the time_wait state there isn't much else
you can do.

One possibility is to always terminate the connection in the client
first (e.g. kill the client first).  Thus, leaving the client in the
time_wait state and not the server.  NOTE - this may not work in all
cases, because of issues like current window states, etc.

-----------[000184][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 16:20:57 GMT
From:      dat91jre@ludat.lth.se (Regmyr Jonas)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP *.c ??


Ok, I'm a newbie here and this might be a very simple thing to do
but since I haven't got the foggiest about TCP/IP I'm typing my
question here. If this is in FAQ please direct me to it or send it
to me.
Thanx.

What I'm looking for is a c program that connects to a computer
in the same way as telnet does on port 23. Nothing extra just
a simple connection on which I can send and receive data.
I think this should be a simple thing to do if you understand
the protocol, but i might be wrong and in that case please tell
me so.
However if there is someone, out there, who's got some piece of c code
that does just this, pleas mail it to me.

                    /Jonas..........

PS. Don't bother to edit out the comments from the code.
    I can do that later. :) DS.

Mail: Karl Jonas Regmyr  Tunav. 39 C:411  S-223 63 Lund  Sweden
Phone 1: +46 (0)46 39 06 60         Phone 2: +46 (0)46 13 33 01
                  Email: dat91jre@ludat.lth.se

-----------[000185][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 92 17:12:35 GMT
From:      J045820@LMSC5.IS.LMSC.LOCKHEED.COM
To:        comp.protocols.tcp-ip
Subject:   IP Addressing

We are looking into the TCP-IP world and on this subject I have a quick
question concerning IP address assignments. In most of the documentation
I have read concerning IP addresses the following is stated " generally
it is a bad idea  to use 0 255 for subnet numbers or addresses." I understand
the reason for not using 255. I am interpreting this to say that it is not
a good idea to implement the following addressing scheme 155.189.0.XXX,
my question is, what is the problem with this?
                                              Thanks BLSMITH

-----------[000186][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 17:37:05 GMT
From:      harlick@prototype.uucp (Ryan Harlicker)
To:        comp.protocols.tcp-ip
Subject:   KA9Q on a UNIX System


	Due to the lack of multi-connect terminal software for unix I would
like to run a slip connection to our local server system.  Unfortunately, our
local computer center doesn't feel the need to add slip services at this
time.  Has anyone compiled up KA9Q fon a unix system?  This would appear to
allow a person to open multiple connections into a unix system and would
solve my immediate needs.

Thanks,
Ryan Scott Harlicker
harlick@plains.nodak.edu
harlick%prototype@plains.nodak.edu

-----------[000187][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 1992 18:40:25 GMT
From:      greg@carnivore.tamu.edu (Greg Economides)
To:        comp.protocols.tcp-ip
Subject:   Re: IP Addressing

In article <92261.36182.J045820@LMSC5.IS.LMSC.LOCKHEED.COM> J045820@LMSC5.IS.LMSC.LOCKHEED.COM writes:
>We are looking into the TCP-IP world and on this subject I have a quick
>question concerning IP address assignments. In most of the documentation
>I have read concerning IP addresses the following is stated " generally
>it is a bad idea  to use 0 255 for subnet numbers or addresses." I understand
>the reason for not using 255. I am interpreting this to say that it is not
>a good idea to implement the following addressing scheme 155.189.0.XXX,
>my question is, what is the problem with this?
>                                              Thanks BLSMITH

I would imagine that there are two reasons for this.  I am not sure if and
when these would actually be problems:

1) each network has a unique number that refers to that "network."  For 
   instance, with the number that you gave, if it were a class B address,
   you would have IP address for machines that look like: 155.189.XXX.YYY
   where XXX & YYY are between 1 & 254, inclusive.  Some network operations
   refer to a whole physical network (eg: the network that all the computers 
   with IP addresses that start with 155.189).  The number 155.189.0.0 refers
   to the network, in this case.  So, assigning IP addresses for machines
   with zeroes in them could cause some problems with this scheme.

2) Broadcasting:  Some networking operations broadcast information (packets)
   to all of the machines on a particular network.  I think that an old
   version of SunOS defaulted this address to be XXX.YYY.0.0 (where XXX &
   YYY are the net numbers of a class B network).  These days, the IP
   broadcast address uses 1s instead of 0s, making the broadcast address
   XXX.YYY.255.255 (or XXX.YYY.ff.ff, for you hex nuts).  So, I imagine that
   giving a machine an IP address with a 255 in it might cause some 
   broadcast-related problems.


Just one other comment:  I am writing all of this from info that I have not
looked at in a long time -- I would expect some of my reasoning or facts to be
incorrect.  But, I think that these explanations are pretty close to being
correct.

Peace,
--
Greg Economides,   Systems Administrator, Center for Biosystems Modelling
Texas A&M University
Internet: econ@tamu.edu
Fuzzy logic -- "Everything is not always something"

-----------[000188][next][prev][last][first]----------------------------------------------------
Date:      Thu, 17 Sep 1992 21:39:00 GMT
From:      pensak@eplrx7.es.duPont.com (Dave Pensak)
To:        comp.protocols.tcp-ip
Subject:   network flooding

 I know that this is a wierd question, but if I wanted
to come as close to saturation of a network (a LAN) running
TCP/IP, how close can I come.  I have tried running things
like tcpblast, which does a good job of loading a net,
but I'd like to see if there is anything out there which
will come even closer to saturation.  It has to be a job
that runs on a single node, not something that depends
on multiple processes cooperating with one another.

-----------[000189][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 92 21:59:03 GMT
From:      Thomas.Tornblom@nexus.comm.se (Thomas Tornblom)
To:        comp.protocols.tcp-ip
Subject:   More TCP over TLI woes, expert advice sought


I'm doing some research/prototyping work on security enhancement of
TCP over TLI. What I have at the moment is a (loadable) streams module
that gets auto-pushed on top of the tcp/tli driver in SunOS 4.1.2.
This module communicates with a daemon that authenticates the
connection endpoints and either accepts (with or without encryption)
or denies the connection.

I've got it mostly working but are having problems when it comes to a
daemon doing t_listen() -> t_accept(). I can authenticate the
t_listen() but when the daemon does a t_accept() I lose control over
the connection. t_accept() takes two file descriptors, the one
t_listened on and a fresh descriptor that is to "take over" the
connection. t_accept() results in a T_CONN_RES message flowing
down the fresh stream. My problem is that I need to get hold of the
old stream while processing this message. How can I do this? To me it
seems that the t_accept() in some way or another bypasses the stream
and applies some magic on the multiplexing tcp/tli driver.

I'm prepared to take a pragmatic approach if anyone has a solution.

The messages pertaining to the tcp/tli driver is of course completely
undocumented in the standard doc kit. We don't even have THE real
documentation (the source code) ;-)

Anyway, help from anyone that can give me some hint on how to proceed
is greatly appreciated.

Thomas
--
Real life:      Thomas Tornblom           Email:  Thomas.Tornblom@nexus.comm.se
Snail mail:     Communicator Nexus AB     Phone:  +46 18 171814
                Box 857                   Fax:    +46 18 696516
                S - 751 08 Uppsala, Sweden

-----------[000190][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 00:09:56 GMT
From:      mogul@pa.dec.com (Jeffrey Mogul)
To:        comp.protocols.tcp-ip
Subject:   Re: Anomalous behaviour of TCP over high speed network.

In article <47671@shamash.cdc.com> gsa@easyaspi.udev.cdc.com (gary s anderson) writes:
>First, if this is a "silly window avoidance" problem then one might
>try the "TCP_NODELAY" socket option (assuming your system supports this
>concept) to see if the problem disappears.

Actually, this has no effect on the SWS-avoidance code (at least in
BSD-based systems).  However, if the receiving host has a way of
turning off the TCP acknowledgement delay, you could try doing that.
Only for testing purposes, of course.

-Jeff

-----------[000191][next][prev][last][first]----------------------------------------------------
Date:      Fri, 18 Sep 1992 02:28:15 GMT
From:      vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: network flooding

In article <1992Sep17.213900.18357@eplrx7.es.duPont.com>, pensak@eplrx7.es.duPont.com (Dave Pensak) writes:
> 
>  I know that this is a wierd question, but if I wanted
> to come as close to saturation of a network (a LAN) running
> TCP/IP, how close can I come.  I have tried running things
> like tcpblast, which does a good job of loading a net,
> but I'd like to see if there is anything out there which
> will come even closer to saturation.  It has to be a job
> that runs on a single node, not something that depends
> on multiple processes cooperating with one another.


That depends on the hardware, medium, and the packet size.

Any reasonable modern UNIX workstation should be able to fill an
ethernet with 1500 byte IP or UDP/IP packets.  (ie. 9.6 usec
interpacket gaps for as many hours as you want, assuming no
collisions)  At least some UNIX workstations can do it with
minimum sized ethernet packets.

At least some UNIX workstations can fill an FDDI ring with as many
4KByte UDP frames as will fit (not quite 100Mbit/sec because of tokens,
gaps, and such, and variable because of varying TTRT and ring
latency).

One process on a single station cannot do TCP, because you must have
two to talk to each other.

Filling an ethernet to capacity with TCP/IP between a pair of machines
is not hard.  The ACK's tend to cause collisions and slow things down.
It is possible to get pretty close to 100Mbit/sec TCP/IP on an FDDI
ring, but releasing the token for ACKS tends to slow things down.  In
both cases, the maximum speed varies a lot, depending on how you
arrange things to mimimize collisions or token-passes, and depending on
the individual ethernet or FDDI installation.  (Ring length, ethernet
cable length, repeaters, number of MACs in the token path, etc.)

Given that Cray can do something like 1GByte/sec (or was it 2?) through
TCP/IP, you should be able to do a reasonable job on an ATM or HIPPI
switch if you have the money.


Vernon Schryver,  vjs@sgi.com

-----------[000192][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 09:09:44 GMT
From:      Q.Li@cs.ucl.ac.uk (Qin Li)
To:        comp.protocols.tcp-ip
Subject:   Bootp answer in braodcast.

Hi, does any one know if a version (or an option) of bootp server which will
braodcast the answer.

Qin

-----------[000193][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 10:06:11 GMT
From:      henk@cs.vu.nl (Henk Smit)
To:        comp.protocols.tcp-ip
Subject:   Re: Speed of different network archi...

k@hprnd.rose.hp.com (Steve Kao) writes:

>In comp.protocols.tcp-ip, cew6@po.CWRU.Edu (Carlin E. Wiegner) writes:
>> Can anyone email me the MB/sec for the following hardware 
>> architectures....
>> 10Base-T
>> Thick Ethernet
>> Thin Ethernet
>> Fiber Optic 
 
>These are all 10Mbps networks.  For Fiber, I'm assuming FOIRL instead of
>FDDI.  FDDI is much faster.  The Thick Ethernet is capable of greater
>than 10Mbps.
 
>> Appetalk (LocalTalk)
 
>I don't know, but I think this is also 10Mbps

  AppleTalk is a protocol. It can run over Ethernet (then called EtherTalk),
 or over Apple's custome cabeling system, called LocalTalk. LocalTalk
 has a throughput of only 230.4 Kbps (*Kilo*, not Mega).

           Henk.

-- 
Henk Smit                               Vrije Universiteit     Amsterdam
Internet: henk@cs.vu.nl                 Phone:    +31 20 548 6218

-----------[000194][next][prev][last][first]----------------------------------------------------
Date:      Fri, 18 Sep 92 11:30:10 GMT
From:      maw@netcom.com (Martin Walker)
To:        comp.mail.misc,comp.sys.novell,comp.protocols.tcp-ip
Subject:   PD MHS <-> SMTP gateway?

Hi all,

As the subject line says, I am looking for a public domain MHS to SMTP
gateway that will allow me to shove mail around transparently between
an MHS based mail system and our Unix machines, which are all using SMTP 
as a transport.

Please mail replies to me and I will summarise shortly....

Thanks in advance for any help,
Martin.
-- 
Martin Walker               |  "One way to stop a runaway horse
++27-12-344-3973 (w)        |   is to bet on him"
++27-12-998-7263 (h)        |   maw@netcom.com
Pretoria, South Africa      |   martin@paradigm.CO.ZA

-----------[000195][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1992 15:03:02 GMT
From:      trier@slc6.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: Silly window avoidance and ACK-only packets

What is the proper interaction between ACK-only packets and the Sender SWS 
algorithms?  If a TCP gets into a situation where it has data buffered
that it can't send because of sender silly window avoidance, and at the
same time needs to send an ACK (for instance, if the delayed ACK timer
expired), should it send an ACK with no data, holding the buffered data
for later, or should it ignore the SW avoidance and piggyback the ACK
and data in one packet?

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000196][next][prev][last][first]----------------------------------------------------
Date:      Fri, 18 Sep 1992 17:45:58 GMT
From:      gam@sun_jimh.mentorg.com (Gary Merrick)
To:        comp.protocols.tcp-ip
Subject:   Re: Speed of different network archi...

In article <2480013@hprnd.rose.hp.com>, k@hprnd.rose.hp.com (Steve Kao) writes:
|> In comp.protocols.tcp-ip, cew6@po.CWRU.Edu (Carlin E. Wiegner) writes:
|> > Can anyone email me the MB/sec for the following hardware 
|> > architectures....
|> > 10Base-T
|> > Thick Ethernet
|> > Thin Ethernet
|> > Fiber Optic 
|> 
|> These are all 10Mbps networks.  For Fiber, I'm assuming FOIRL instead of
|> FDDI.  FDDI is much faster.  The Thick Ethernet is capable of greater
|> than 10Mbps.
|> 
|> > Appetalk (LocalTalk)
|> 
|> I don't know, but I think this is also 10Mbps


FDDI: 100 Mbps
Appletalk (LocalTalk) : 230 Kbps


--


-----------[000197][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 17:57:17 GMT
From:      zjhc05@hou.amoco.com (John H. Caldwell)
To:        comp.protocols.tcp-ip
Subject:   Source Code for "Internetworking with TCP/IP, Volume III"

Recently, I posted a request for the subject item.  One of the authors (DLS),	
was kind enough to reply and let me know that he hadn't yet posted it on 
the net, but would soon.

I checked this morning and it is available by anonymous ftp from "ftp.uu.net".	
It is in the directory "published/books" and the file name is	
"comer.internetworking3.src.tar.Z"

-----------[000198][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1992 18:30:36 GMT
From:      axa12@po.CWRU.Edu (Ashok Aiyar)
To:        comp.protocols.tcp-ip
Subject:   KA9Q on MSDOS system


I just installed KA9Q on a DOS PC on our campus TCP/IP
network (thin-wire).  It seems to be working.  I want
to know if there is a POP-2 server that I can use for KA9Q
(i.e use this PC as a POP-2 server)?

Please post or reply to axa12@po.cwru.edu

Ashok
-- 
                          Ashok Aiyar
   		       axa12@po.cwru.edu
                Visit the IBM-PC Sig on Freenet

-----------[000199][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 18:45:33 GMT
From:      gsa@easyaspi.udev.cdc.com (gary s anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: Anomalous behaviour of TCP over high speed network.

In article <1992Sep18.000956.4647@PA.dec.com>, mogul@pa.dec.com (Jeffrey Mogul) writes:
|> In article <47671@shamash.cdc.com> gsa@easyaspi.udev.cdc.com (gary s anderson) writes:
|> >First, if this is a "silly window avoidance" problem then one might
|> >try the "TCP_NODELAY" socket option (assuming your system supports this
|> >concept) to see if the problem disappears.
|> 
|> Actually, this has no effect on the SWS-avoidance code (at least in
|> BSD-based systems).  However, if the receiving host has a way of
|> turning off the TCP acknowledgement delay, you could try doing that.
|> Only for testing purposes, of course.
|> 
|> -Jeff

I'm not sure about which BSD variant you are looking at, but mine is
certainly programmed to disable the SWS-avoidance code.  Maybe we
aren't discussing the same problem.

Here's some high level pseudo code to illustrate:

IF  we do not have a full MSS worth of data to send  AND
    we have already sent data which has not been ACK'ed   THEN

	don't send any data....

IFEND


The actual BSD code immediately follows the comment:  "Sender silly
window avoidance".  It is slightly more convoluted then my pseudo
code but conceptually the same.

Normally TCP will coalesce data until it can send a full MSS packet
or there is no outstanding un-ACK'ed data or some anomaly occurs (FIN, etc.).
The use of the TCP_NODELAY option (internally mapped to TF_NODELAY) means
that TCP stops coalescing data (unless the window is closed).  In this
case, you should be able to send a steady stream of small packets without
any coalescing delay (obviously this won't circumvent closed windows, but
this did not appear to be the problem).

NOTE - the effect of the delayed ACK is generally negated (unless
the ACK doesn't arrive until the window is closed) by the use of the
nodelay option.  

NOTE - the proposal to remove the delayed ACK would only provide a 
partial solution because there is still the effect of "end-to-end" delay. 

-----------[000200][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 20:15:01 GMT
From:      gsa@easyaspi.udev.cdc.com (gary s anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: Silly window avoidance and ACK-only packets

In article <19cr36INNc98@usenet.INS.CWRU.Edu>, trier@slc6.ins.cwru.edu (Stephen C. Trier) writes:
|> What is the proper interaction between ACK-only packets and the Sender SWS 
|> algorithms?  If a TCP gets into a situation where it has data buffered
|> that it can't send because of sender silly window avoidance, and at the
|> same time needs to send an ACK (for instance, if the delayed ACK timer
|> expired), should it send an ACK with no data, holding the buffered data
|> for later, or should it ignore the SW avoidance and piggyback the ACK
|> and data in one packet?
|> 
|> -- 
|> "Beware of programmers who     Stephen Trier
|>  carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
|>        Leonard Brandwein       Case Western Reserve University
|>                                trier@ins.cwru.edu

The main goal of SWS-avoidance is to maximize the amount of data in each
network packet.  If you are going to send an ACK to your peer, you
certainly want to piggy-back data (if possible) to maximize the size
of the network packet.  In fact, the BSD code will does this.

Think of SWS-avoidance as "coalescing until you need to send a packet"
(e.g. full MSS, FIN, time to ACK, retransmit, empty pipe, etc.).

-----------[000201][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 20:44:58 GMT
From:      dls@mentor.cc.purdue.edu (David L Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Internetworking with TCP/IP Volume III


	The sources for volume III of "Internetworking with TCP/IP" are
now available via anonymous FTP to "ftp.uu.net" in the file:

	published/books/comer.internetworking3.src.tar.Z


-- 
					+-DLS  (dls@mentor.cc.purdue.edu)

-----------[000202][next][prev][last][first]----------------------------------------------------
Date:      Fri, 18 Sep 1992 21:39:24 GMT
From:      bwolfe@rad.rpslmc.edu (Brian A. Wolfe)
To:        comp.protocols.tcp-ip
Subject:   Xyplex MX1500 Terminal Server tcp/ip configuration question


Hi There,

I have the following configuration here:

-------------------  router   -----------------------           -----------------------
other networks                  net 144.74.26.0        Bridge       net 144.74.31.0
 (Internet etc.)



Problem:  The xyplex MX1500 Terminal Servers on  144.74.26.0 and 144.74.31.0 can 
          only talk to hosts in their own subnet. Is there a way to configure MX1500's 
          to use different ip numbers with nets separated by a bridge instead of a router?

This isn't a problem with any of the other tcp/ip gear we own, with a unix machine 
I can use route add default `hostname` 0 and everything works just fine. I've 
tried to do a 'set server internet primary gateway [addr]' using the MX1500's 
address, but that doesn't seem to work.

Any ideas?

regards,

Brian


P.S.   We have to use separate ip net numbers because the nets are administered
       and owned by different companies, we are bridging so that we can use LAT 
       (and IP) between them. A better solution would be to use a brouter, 
       but we don't have the equipment to do that right now.


-------------------------------------------------------------------------
Brian Wolfe						
Associate Director
Dept of Diagnostic Radiology and Nuclear Medicine
Rush Presbyterian-St. Lukes Medical Center 
Chicago,  IL   USA 
Internet: bwolfe@rad.rpslmc.edu    
Voice:    (312)-942-2141          FAX:  (312)-942-2114


-----------[000203][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1992 22:30:35 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Xyplex MX1500 Terminal Server tcp/ip configuration question

In article <1992Sep18.213924.15954@rpslmc.edu> bwolfe@rad.rpslmc.edu (Brian A. Wolfe) writes:
>Problem:  The xyplex MX1500 Terminal Servers on  144.74.26.0 and 144.74.31.0 can 
>          only talk to hosts in their own subnet. Is there a way to configure MX1500's 
>          to use different ip numbers with nets separated by a bridge instead of a router?

Set the subnet mask to 255.255.0.0.  The Xyplex will then think that these
two subnets (and all other 144.74.XXX subnets) are actually the same
network.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000204][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 92 22:51:50 GMT
From:      jessea@homecare.com (Jesse W. Asher)
To:        comp.protocols.time.ntp,comp.protocols.tcp-ip
Subject:   Configuring a stratum 1 server (using ntp w/o being on the Internet).

We need to use ntp as our time protocol because we have networks
spanning the country (meaning we have an internet) and we need to use
Universal Time.  So, I'd like to set it up so that we have our own top
level time servers and then the next layer servers (stratum 2).
Basically, I'd like to fool it into thinking its on the Internet.  Can
anyone suggest how to do this?  What should my /etc/ntp.conf look like
on the stratum 1 server?

Another reason for doing this is that we plan to be on the Internet
within the year and this will enable us to be ready to go.   Thanks
for any suggestions here.


-- 
      Jesse W. Asher                                   Phone: (901)762-6000
                         Varco-Pruden Buildings
	      6000 Poplar Ave., Suite 400, Memphis, TN  38119
 Internet: jessea@homecare.COM                 UUCP: ...!banana!homecare!jessea

-----------[000205][next][prev][last][first]----------------------------------------------------
Date:      Fri, 18 Sep 1992 23:19:34 GMT
From:      magus!jai (Jaisingh  Pawar)
To:        comp.protocols.tcp-ip
Subject:   pcnfsd  version 2  for NeXT wanted


Hi,

Could someone mail me pcnfsd  version 2  for  the NeXT  ?
Thanks !

--

Jai

-----------[000206][next][prev][last][first]----------------------------------------------------
Date:      Sat, 19 Sep 1992 06:32:34 GMT
From:      rustomji@sol.cs.wmich.edu (Eric Rustomji)
To:        comp.protocols.tcp-ip
Subject:   MULTIPLE FD'S FOR CONNECTIONS

Hi there,

I am trying to implement a client/server model, wherein
for every child connection, the server should get a new 
fd from the accept() call and store it for future use.

In this way, any child should be able to send data 
to/from another child without going thru the server.

Is this possible? Questions/Comments?

Appreciate if anyone could provide good text references
for the above.

Thanks a bunch!

-- Eric
 
==================================================================
"The World is full of Surprises!"

Eric  Rustomji
Department of Computer Science
Western Michigan University

Home Phone: (616) 387 - 5566
       UCC: (616) 387 - 6426


Email:  INTERNET: rustomji@cs.wmich.edu
	UUNET :  ....!wmichgw!sol.cs.wmich.edu!rustomji
==================================================================

-- 

==================================================================
"The World is full of Surprises!"


-----------[000207][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 92 12:55:18 GMT
From:      dreblow@vax.muskingum.edu (Lewis M. Dreblow)
To:        comp.protocols.tcp-ip
Subject:   Re: Xyplex MX1500 Terminal Server tcp/ip configuration question

{ Various stuff deleted }
> 
> This isn't a problem with any of the other tcp/ip gear we own, with a unix machine 
> I can use route add default `hostname` 0 and everything works just fine. I've 
> tried to do a 'set server internet primary gateway [addr]' using the MX1500's 
> address, but that doesn't seem to work.
> 
Brian,

What you are doing above is exactly your "problem".  A gateway address is
the IP address the MX will look to if it can't find the address on the local
net.  You neet to SET and DEFINE server internet primary gateway address to
the address of your router.  If you have a local name server, setting this
into the server allows you to avoid multiple domain name settings.  Always
remember to do both SETs and DEFines after you have verified that a setting
is appropriate or else it will lose the settings after re-load.

Lewis Dreblow, Muskingum College, DREBLOW@VAX.CNS.MUSKINGUM.EDU

-----------[000208][next][prev][last][first]----------------------------------------------------
Date:      Sat, 19 Sep 1992 16:14:10 GMT
From:      pensak@eplrx7.es.duPont.com (Dave Pensak)
To:        comp.protocols.tcp-ip,comp.unix.aix
Subject:   tcpdump


Does anyone know if tcpdump has been successfully ported to
AIX 3.2 ?

If not, can anyone tell me about land-mines that I will likely
hit if I try to do it myself ?

-----------[000209][next][prev][last][first]----------------------------------------------------
Date:      Sat, 19 Sep 92 18:31:26 GMT
From:      shm@syl.syl.nj.nec.com (Shailendra Majmundar)
To:        comp.periphs.printers,comp.protocols.tcp-ip
Subject:   TCP/IP on printers

I recently tripped onto this group and have been lurking for
the last few days.  Reading the thread on network printers
triggered a thought process on how TCP/IP can/is supported on
network printers and what it really means to a user/client.

When people say a printer supports TCP/IP (or complain about the
lack of support), what do they really mean?  Is it that the
printer has a full tcp/ip stack running on it along with some
application on top of it?  Assuming that the printer is stand-alone
(ie., it is not attached to a computer), what does it take to
support such a set up?  Do the printers have a spooling space
(which I guess, implies a hard disk) where requests from various
network clients get queued and later get printed, or do they
accept only one request at a time?

I guess, I am really wondering about the software architecture
on the printer and the method(s) of communicating with it.

Can someone enlighten me?  E-mail is welcome, but I'm sure many
people will be interested on this topic, so posting on the net
is preferred.  It would be nice if this leads to a discussion
on the pros/cons of various architectures.

Thank you all in advance.

Sam (shm@syl.nj.nec.com)

-----------[000210][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 92 03:20:16 GMT
From:      spedpr@thor.cf.ac.uk (Paul Richards)
To:        comp.protocols.tcp-ip
Subject:   Routing problems

I've recently had to put a router machine between our campus backbone
and some rather specialised research hardware. The problem is that the
campus is a class B network and my machines are all on a single subnet.
Is there any way to create a sub-sub-net so that the two cards in the
router can be on different nets, otherwise packets don't get passed
between the cards. I've tried all sorts of combinations of netmasks and
I'm now beginning to wonder if this is even possible?

The campus is 131.251 and all my machines are 131.251.122. What I've
tried is to have one card address as 131.251.122.1 and the other as
131.251.122.129 using a netmask of 0xffffff80. This doesn't work, I can
access the nets connected to both cards from the router, but machines on
either side can't pass packets to each other, i.e. the router doesn't
pass them between cards. Anyone know how I can do this?

If it isn't possible to do this using the above methods then does anyone
know how I could hack the kernel source (386BSD) to get packets
transferred from one interface to another regardless of addresses.
-- 
  Paul Richards at Cardiff university, UK.

  spedpr@uk.ac.cf.thor	Internet: spedpr%thor.cf.ac.uk@nsfnet-relay.ac.uk
  UUCP:     spedpr@cf-thor.UUCP or ...!uunet!mcsun!uknet!cf!thor!spedpr
+++

-----------[000211][next][prev][last][first]----------------------------------------------------
Date:      Sun, 20 Sep 1992 19:17:38 GMT
From:      yost@adobe.com (David Yost)
To:        comp.protocols.ppp,comp.protocols.tcp-ip,comp.protocols.appletalk
Subject:   SCSI as physical layer

Can anyone relate or point to experience using
point-to-point SCSI as the physical layer
beneath any mainstream protocol stack?
If there are any such SCSI drivers floating
around, that would be especially interesting.

Thanks.

 --dave yost

-----------[000212][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 92 11:58:42 GMT
From:      rcjoep@rwc.urc.tue.nl (Joep Brand)
To:        comp.protocols.tcp-ip,bit.listserv.3com-l,comp.dcom.lans.ethernet
Subject:   NETBuilder II question


Hello,

Is it possible to forward BOOTP packets on a 3Com NETBuilder II, while IP 
routing is turned on. The NETBuilder is running software version BR5.0 ?

I'm using the NETBuilder II to route IP, IPX and DECnet, and to bridge other
proprietary protocols like LAT and MOP.

Thanks in advance,

-Joep
--
Joep Brand, Eindhoven University of Technology
Computing Center 1.64, +31 40 474394
P.O.Box 513, 5600 MB Eindhoven, The Netherlands

-----------[000213][next][prev][last][first]----------------------------------------------------
Date:      Mon, 21 Sep 1992 12:44:31 GMT
From:      zawada@ncsa.uiuc.edu (Paul Zawada)
To:        comp.protocols.tcp-ip
Subject:   Re: IP Addressing

In article <92261.36182.J045820@LMSC5.IS.LMSC.LOCKHEED.COM>, J045820@LMSC5.IS.LMSC.LOCKHEED.COM writes:
|> We are looking into the TCP-IP world and on this subject I have a quick
|> question concerning IP address assignments. In most of the documentation
|> I have read concerning IP addresses the following is stated " generally
|> it is a bad idea  to use 0 255 for subnet numbers or addresses." I understand
|> the reason for not using 255. I am interpreting this to say that it is not
|> a good idea to implement the following addressing scheme 155.189.0.XXX,
|> my question is, what is the problem with this?
|>                                               Thanks BLSMITH

Most IP routers look at 155.189.0.0 as the entire network, not just
subnet 0 of 155.189...  When an entry for 155.189.0.0 ends up in a
routing table, the router may get confused...  Most routers will
ignore such an entry if they have other subnets like 155.189.1.0,
155.189.2.0, etc. in their routing tables.  This results in a router
somewhere, most likely a boarder gateway, sending ICMP host unreachable
messages. 

--zawada

-- 
Paul J. Zawada                                    KB9FMN
NCSAnet Network Engineer                          zawada@ncsa.uiuc.edu
National Center for Supercomputing Applications   ..!pur-ee!zawada

-----------[000214][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 92 13:15:00 GMT
From:      jonkar@rsv.svskt.se (Jonas Karlsson)
To:        comp.protocols.tcp-ip,comp.benchmarks
Subject:   TCP/IP Benchmarks

I will soon have to evaluate the TCP/IP performance of at least 5 different
wendors UN*X boxes (don't know which yet). I would very much appreciate any
pointers to benchmark programs and what to look for when testing.

1K Thanks in adwance!

/Jonas Karlsson
-- 
Jonas Karlsson |                                   | Tel:   +46 8 7648444
RSV   S-171 94 |        This space for rent!       | Fax:   +46 8 294975
Solna, Sweden  |                                   | Email: jonkar@rsv.svskt.se

-----------[000215][next][prev][last][first]----------------------------------------------------
Date:      Mon, 21 Sep 1992 14:09:10 GMT
From:      brent@uis.com (Brent Engelbert)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.unix.internals
Subject:   Ethernet Collisions (?)

We are testing some code on an IBM RS/6000 that looks at ethernet
collisions, however we are having some problems generating any collisions.

1) How do we generate collisions on the ethernet?

2) We are currently looking at the ifnet structure in the kernel to
   gather network statistics.  Is this the proper place to be looking?  On
   our machine the if_collisions value is always 0, but the rest of the
   fields seem to have the correct information in them.

We would appreciate any help that any one can give us.  Please E-mail
responses to us.

----
Brent Engelbert
Unix Integration Services                                Phone: (515) 254-3074 
11323 Aurora Ave Urbandale, IA  50322                    brent@uis.com

-----------[000216][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 92 14:32:06 GMT
From:      barker@wd0gol.WD0GOL.MN.ORG (Bob Barker)
To:        comp.unix.bsd,comp.protocols.tcp-ip
Subject:   BSD Networking Software - Available via ftp?

I understand that the "BSD Networking Software" is available via anon. ftp
or from the UUNET archives and that this software includes source for
TCP/IP and XNS drivers and interfaces.

I am in desperate need of this or related software to assist me in a time
critical effort to develop an XNS driver.

If anybody can point me to where this software can be found for anon. ftp
or how I could get it off of UUNET (I'd gladly pay the $.50/min for the
900 #) I would be most greatful.  Also, if anybody knows of any other
sources I'd love to hear about them!

Thank you very much!

-Bob Barker
------------------------------------------------------------------------------
Bob Barker                                ...!uunet!umn-cs!tcnet!wd0gol!barker
Robert Barker & Associates                                barker@wd0gol.MN.ORG
(612) 949-0140

-----------[000217][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1992 16:17:21 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: MULTIPLE FD'S FOR CONNECTIONS

In article <1992Sep19.063234.8595@sol.cs.wmich.edu> rustomji@sol.cs.wmich.edu (Eric Rustomji) writes:
>In this way, any child should be able to send data 
>to/from another child without going thru the server.

By "child", do you mean "client"?  In the Unix world, "child" usually means
"child process"; in reference to client/server networking, it often refers
to the child process that is spawned by the server to handle a particular
client connection.

>Is this possible? Questions/Comments?

If you're only using Unix domain sockets, you can pass file descriptors
through the socket to clients, and then they can use them.  But if you're
going over a network, there's no way to pass file descriptors around.  In
order for two machines to talk directly to each other, they have to
establish a connection in the usual way.

>Appreciate if anyone could provide good text references
>for the above.

"Unix Network Programming" by W. Richard Stevens, section 6.10, "Passing
File Descriptors".
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000218][next][prev][last][first]----------------------------------------------------
Date:      Mon, 21 Sep 1992 21:40:39 GMT
From:      ronf@panther3.panther.mot.com (Ron Feigen)
To:        comp.protocols.tcp-ip
Subject:   TCP messages

Could somebody comment if the followig approach to TCP/IP (INET domain) 
communication is reasonable?

I need to send "messages" across a TCP/IP net.  The messages are variable length.
Each message has a header made up of: Length, Type, and ID, or 12 bytes.  I know
that by definition an INET socket deleivers a "stream" of data.  However can I 
not assume that the entire 12 bytes will be delivered "atomically"?  Especially 
since the sender has TCP_NODELAY set. 


-- 

>
Ron Feigen
ronf@panther.mot.com

-----------[000219][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 00:19:57 GMT
From:      susan@moo.org (Susan Dean)
To:        comp.databases.sybase,comp.databases,comp.sys.novell,comp.databases.theory,comp.unix.aix,comp.windows.misc,comp.protocols.tcp-ip
Subject:   Sysbase/Aix/TCP/IP/Windows

 
I have a question that maybe someone can help me with. We are working in a
client/server environment with Novell 3.1, Novell LAN WP,TCP/IP (ethernet), 
Sybase and Power Builder. Sybase is on
a RISC 6000 V 3.1.6 of AIX. After running a Power-Builder Application on a
386 running windows 3.1, we sometimes receive the following error messages:

Write to SQL server fail
Db process dead or not enable

The errors are being reported from the Sybase Netlib DLL's on the client.

 The RISC Sybase server will sometimes report a error # 1608
"Network error was encountered while sending results to front end"

But then at other times
the program will run without errors. In trying to find the problem we took
and put only two RISC 6000's and a PC on their own network. Still the same
problem. So it appears network traffic is not the problem.  The connection
appears to fine and then the connection just gets dropped by someone.

Does anyone have any ideals that may be the cause or something else we can
try doing/looking at? Right now ANY ideas will be helpful, as we are near
the end of the line. A post or email replies are ok, as I check this group
daily. Thanks!!!

susan@moo.org

-- 
"susan@moo.org"

-----------[000220][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 00:41:05 GMT
From:      jallen@oat2.standard.com (John Allen)
To:        comp.protocols.tcp-ip
Subject:   Re: Rcp -- is that rlogin or what?

Lets set the record straight.

rsh,or sometimes remsh is the remote shell program
it lets you execute a program on a remote machine
quite often this appears as a pipe implementation

cat file |rsh remotemachine nroff -mn >file.out

Beware not all remote shells work the same, some have a bad 
time trying to redirect remote i/o.
 
rlogin is for a remote network login, it allows you to login
to another machine and operate just as if you came through a 
serial port and executed your .profile as part of the normal
login procedure.  

Your .rhosts protects your login from being used
by another user on another machine who is using the
same logid that you have.  

rcp is remote copy and not some crazy notion as explained before.
why use FTP to do a simple file transfer.  I have found in the past
that FTP lends it self to the classic stupid mistake when a sysadm
is not paying attention, or that some bozo copys the entire remote /tmp
directory over their good local data.

rcp is only on native unix TCP/ip implementations.
Try this 

rcp myfile remotehost:/usr/local/newfile
                  need the colon seperating the path name

or try copying from two different hosts.

rcp machine1:/usr/lib/sendmail.cf machine2:/usr/backup/sendmail.cf

a lot of people get hung up with FTP and its really not to good in
batch files, where RCP lends it self to all sorts of unix to unix
applications in a single command line. 

You do not need berkley sources to understand all of this.
For big sites I recommend that you take FTP away from all non
qualified users if you do not use FTP from site to site.  Give them RCP
for all command line unix work, it is just much cleaner.  

-- 
===============================================================================
John Allen  - Network Executive                   E-mail: jallen@standard.com
Network Services                               Telephone: (503) 243-6189
Standard Insurance, Portland Oregon                  FAX: (503) 321-3313

-----------[000221][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1992 01:49:35 GMT
From:      trier@odin.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP on printers

In article <1992Sep19.183126.9079@research.nj.nec.com> shm@syl.nj.nec.com writes:
>Is it that the printer has a full tcp/ip stack running on it along with some
>application on top of it?

Yes.

>Assuming that the printer is stand-alone (ie., it is not attached to a
>computer), what does it take to support such a set up?

A network interface, a CPU big enough to run TCP/IP, and sufficient ROM
and RAM to support the protocol stack(s).  It doesn't take much, really,
especially when one considers the CPU resources already on board modern
300 dpi printers.

>Do the printers have a spooling space (which I guess, implies a hard
>disk) where requests from various network clients get queued and later
>get printed, or do they accept only one request at a time?

A few high-end printers have the ability to do local spooling, but I
think most assume that there is a computer somewhere that does spooling
for them.

>I guess, I am really wondering about the software architecture
>on the printer and the method(s) of communicating with it.

That is a real can of worms.  There is today no standard way to talk to
a relatively dumb printer.  Some simple ways have been tried, like simply
opening a TCP connection to the right port and sending the data.  A protocol
as simple as TFTP might work acceptably, but I have not heard of it used
to submit print jobs to a printer.

The traditional rendition of the Unix lpr protocol is not useful, since it
effectively requires the printer to spool the file before printing.  That
rules out using it with most printers.

The RFC-documented version of the lpr protocol solves this problem, in a
way not entirely compatible with the traditional lprs.  However, neither
version can accept data back from the printer.  Neither can send and receive
the out-of-band queries needed in order to ask about things like paper and
ink supplies, pages printed, and the like.

DEC came up with its own protocol for talking to a printers over TCP/IP.
That protocol is available for their LPS-20 and LPS-40 PrintServers, and
it appears to work pretty well.

There is or was an IETF working group working to straighten out this mess
by developing a standard printing protocol.  Last I heard, they were far
behind schedule.  :-(

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000222][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 01:55:09 GMT
From:      mjr@hussar.dco.dec.com (Marcus J. "Buddy can you spare a clue?" Ranum)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP on printers

>That is a real can of worms.  There is today no standard way to talk to
>a relatively dumb printer.  Some simple ways have been tried, like simply
>opening a TCP connection to the right port and sending the data.  A protocol
>as simple as TFTP might work acceptably, but I have not heard of it used
>to submit print jobs to a printer.

	QMS makes some nifty PostScript printers that support an FTP
interface. It's frightfully cool. You FTP to the printer, and do a "dir"
or "ls" to examine the queue, and can queue up new files by using the
"put" command.

mjr.

-----------[000223][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 02:14:34 GMT
From:      fitz@wang.com (Tom Fitzgerald)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.unix.internals
Subject:   Re: Ethernet Collisions (?)

brent@uis.com (Brent Engelbert) writes:

> 1) How do we generate collisions on the ethernet?

One possible scheme is to run a program on several systems, that waits for
a broadcast and sends out a response.  The systems should be as far apart
on the net as possible, preferably separated by repeaters, on the ends of
long cables.  Your program could listen for a broadcast on a UDP port and
send anything at all when it gets it; other non-IP tricks are possible.
You'll also need a system (preferably equidistant from the others) to send
out the triggering broadcasts when you want collisions.  The broadcast
synchronizes things so a collision is pretty likely.  This won't always
give you a collision, but the odds aren't bad.

> 2) We are currently looking at the ifnet structure in the kernel to
>    gather network statistics.  Is this the proper place to be looking?  On
>    our machine the if_collisions value is always 0, but the rest of the
>    fields seem to have the correct information in them.

Around here, RS/6000's always report 0 collisions, even when they've
definitely experienced them.  I don't know any more than this, but it seems
like either the interface hardware doesn't see them, or the OS ignores
them.

-- 
Tom Fitzgerald   Wang Labs       fitz@wang.com   "I went to the universe today;
1-508-967-5278   Lowell MA, USA                   It was closed...."

-----------[000224][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 02:55:47 GMT
From:      trier@odin.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: Rcp -- is that rlogin or what?

In article <BuyF8I.KwK@oat2.standard.com> jallen@oat2.standard.com (John Allen) writes:
>rcp is only on native unix TCP/ip implementations.

This is incorrect.  Rcp is available in many small-machine TCP/IP suites.
For example, FTP Software's PC/TCP has an rcp client, and CUTCP and NCSA
Telnet include rcp servers.

Your statement about the .rhosts "protecting" one's account from people
with the same login names on other machines surprises me.  .rhosts is the
source of many evils.  I have never before seen someone claim that it
_improved_ security!  :-)

>For big sites I recommend that you take FTP away from all non
>qualified users if you do not use FTP from site to site.  Give them RCP
>for all command line unix work, it is just much cleaner.  

<Sigh>

Does anyone else remember the last time an rcp vs. FTP flame war started
here?  I, for one, hope we don't have to go through that again.

Suffice it to say that FTP and rcp are good for different things.  On our
campus net, we emphasize FTP because of its cross-platform flexibility.

      Stephen

P.S. You may find the xtp utility, posted to alt.sources about two years
     ago, to be of interest.  It provides a command-line interface to the
     FTP protocol.
-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000225][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1992 03:29:44 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP messages

In article <1992Sep21.214039.24610@panther.mot.com> ronf@panther3.panther.mot.com (Ron Feigen) writes:
>Each message has a header made up of: Length, Type, and ID, or 12 bytes.  I know
>that by definition an INET socket deleivers a "stream" of data.  However can I 
>not assume that the entire 12 bytes will be delivered "atomically"?  Especially 
>since the sender has TCP_NODELAY set.

No, you can't assume this.  TCP_NODELAY causes the sender to transmit as
soon as possible, but it doesn't force the receiver to deliver all 12 bytes
to the user process immediately.  Also, the window could be less than 12
bytes, so the sender would only be able to send a part of the message, and
it will send the rest when the window opens.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000226][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 04:13:00 GMT
From:      hlevine@igc.apc.org (Howard Levine)
To:        comp.protocols.tcp-ip
Subject:   FTP via TCP/IP dial-up info wanted


I am looking for some sort of TCP/IP dial-up where I can have an ftp
session over the modem.  Does anyone know of any such software that I can
use in conjunction with MacTCP to dial-up an IP host and then switch
to a program like TelNet and simply Open a connection to an ftp server
as if I were connected serially to the InterNet?

In addition to the software, I would also need the phone number to such an
IP dial-up host.  I know the software exists to have a ppp session via
modem but I do not know where to get such software, how to configure it
or where to call to access an IP dial-up host.

Thanks!
Howard

Please post responses here or send e-mail to Howard.Levine@um.cc.umich.edu

-----------[000227][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 06:21:00 GMT
From:      mgj@franklin.ee.umr.edu (Mahesh Govind Joshi)
To:        comp.unix.questions,comp.protocols.tcp-ip
Subject:   "Daemon" refuses to connect to terminal the second time !

Hi,

I am novice  in this field and require some advise/help in a
problem I am facing.

I have developed  a chat application that works real neat!(it
provides a mix between a write & and talk (features of both))
the daemon uses stream sockets and runs at port 4050.[client
-server model)

Here's the problem:

I log in (say at ttypa) and start the daemon. Evrything works
fine till I am on this tty (ttypa). 
When I logoff and login again on a different tty (say ttypb),
and a chat request from a remote site arrives, the daemon
forks() and starts the echo server which is supposed to display
the messages on the screen. But nothing comes on the screen !!

I think that since the process looses the environment it started
under, it needs to be re-attached to the terminal.
THe FAQ says that its not possible unless you use "screen" or 
"pty". 

Is there any way, I can overcome this by changing my design? 
How does "talkd"  echo a message on a tty?  Do I need to pass the tty 
and open for writing everytime ?

Please advice !!
thanks in advance ....

Mahesh Joshi.
________________________________________________________________________
Mahesh Joshi 			|   Bura duniya ko hai kehta
Help Desk Consultant		|   Aisa bhola to na ban 
Snail Mail:- 1300 N. Oak Street	|   Jo hai karta,woh hai bharta	
#17, Rolla MO 65401.		|   Hai yahan ka hai chalan
Voice:(314)341-8197.		|   Takdeer nahi chalnegi yahan
Email : mgj@ee.umr.edu		|	---- Majrooh Sultanpuri---- 
------------------------------------------------------------------------

-----------[000228][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 09:46:03 GMT
From:      st@starlab.UUCP (Stefan Taxhet)
To:        comp.sys.novell,comp.protocols.tcp-ip
Subject:   SUMMARY: Help: Subnetting in Class C TCP/IP networks

Hello again,
some days ago I posted questions about routing in Class C IP-nets
with subnetting. See the replies appended to this posting.
Thanks to all who answered:
	Graeme Moffat (g.moffat@aukuni.ac.nz)
	Ehud Gavron (gavron@vesta.sunquest.com)
	Jack Thomasson (Jack_Thomasson@Novell.COM)
	Dan Ellison (dan@c-mols.siu.edu)
	Jon Bresler (jmb@ideas.com)
	Marty Collins (mcollins@novell.com)


st> We have only 1 class C network address (192.109.83)
st> but 2 physical networks.
st> To be prepared for more phys. nets we use 3 Bits for subnetting,
st> leaving 5 bits per subnet for hosts.
st> Today we have no connection to the Internet, but we would like to :-)

st> 1) Is there any way to let a Novell 3.11 server do the
st>    routing between these two phys. nets (ethernet)?
mcollins: Yes but you must have differnet network numbers (SUBNETS)
          when there are different Physical networks.
jmb:      Novell can route between to different IP networks ONLY!
dan:      I believe that there is a TCP/IP router for
          Novell servers that should be able to do what you want.
jack:     Very informative answer with novell examples.
          The answer for Novell seems to be YES.
gavron:   Very informative answer about routing in subnets in general.
          The answer in general is YES.

WHAT'S THE ULTIMATIVE ANSWER?
Does Novell understand subnet routing in Class C networks?
Do I need a special router?

How must the binding in the Novell Server be done?
What must the routing tables look like?
 (Is there only a typo in Ehud's answer: all 93 -> 83?)

Further details about our configuration:

1 Class C IP network address (192.109.83)
3 Bits for subnetting (netmask 192.109.83.224)
2 physical nets (192.109.83.64, 192.109.83.128)
(Internet connection in the future)

   Internet Router (NeXT)
   192.109.83.65  (64+1)
     |
  ---o-------[Physical net 1, 192.109.83.010hhhhh]----o---------
                                                      |
                                                    192.109.83.66  (64+2)
                                                     Novell Server
                                                    192.109.83.130 (128+2)
                                                      |
  -----------[Physical net 2, 192.109.83.100hhhhh]----o---------


st> This is exercise 16.5 in Douglas E. Comer's book
st> "Internetworking With TCP/IP". Is there an answerbook?
There doesn't seem to be one. 

Stefan Taxhet                                    FAX:   +49 40 23646 550
STAR Division GmbH                               Phone: +49 40 23646 952
Sachsenfeld 4                                    uucp: ...!unido!starlab!st
D-W-2000 Hamburg 1                               domain: st@starlab.uucp

++++++++++++++++++ Here are the original answers ++++++++++++++++++++++

From: unido!ccu1.aukuni.ac.nz!graeme ( Graeme Moffat)

In comp.sys.novell you write:

>I have a (novice?) question about subnetting in class C networks:
>Does subnetting in class C networks make sense?

Depends on your size & layout: it's fairly restrictive eg
2 bits mask =  2 subnets @ 62 hosts
3  "    "   =  6   "     @ 30   "
4  "    "   = 14   "     @ 14   "

>We have only 1 class C network address (192.109.83) but 2 physical networks.
>To be prepared for more phys. nets we use 3 Bits for subnetting,
>leaving 5 bits per subnet for hosts.
>Today we have no connection to the Internet, but we would like to :-)
 
>1) Is there any way to let a Novell 3.11 server do the
>   routing between these two phys. nets (ethernet)?

Yes, just connect it to both of them :)  Well, at least for IPX & IP

>2) How must the gateway to the Internet be configured (SLIP)?

Depends on how it's connected, & to what :) A leased line could use slip or
ppp, while a connection to a packet network could use X25 or frame relay

 -- 
Graeme Moffat                g.moffat@aukuni.ac.nz \ Time wastes us all, 
Computer Aided Design Centre,  Fax: +64-9-366-0702 /  our bodies & our wits
School of Engineering,    Ph: +64-9-737-999 x8384 /  But we waste time,
University of Auckland, Private Bag, Auckland, NZ \   so time & we are quits

 --------------------------------------------------------------------------

From: unido!ALPHA.SUNQUEST.COM!GAVRON (Ehud Gavron 602-570-2000 x. 2546)

In article <3950@starlab.UUCP>, you write...
#Hello,
# 
#I have a (novice?) question about subnetting in class C networks:
#Does subnetting in class C networks make sense?

	It does if you wish to use subnetworks :-)

#BTW:
#This is exercise 16.5 in Douglas E. Comer's book "Internetworking With TCP/IP".
#Is there an answerbook?
# 
#We have only 1 class C network address (192.109.83) but 2 physical networks.
#To be prepared for more phys. nets we use 3 Bits for subnetting,
#leaving 5 bits per subnet for hosts.

	So far so good.

#Today we have no connection to the Internet, but we would like to :-)
# 
#1) Is there any way to let a Novell 3.11 server do the
#   routing between these two phys. nets (ethernet)?

	Theoretically that depends on Novell.  You'd have to have
	two ethernet controllers, and therefore two IP addresses,
	one on each card, one on each network... Your novell
	server would have to understand subnets for class c.

#2) How must the gateway to the Internet be configured (SLIP)?

	Let's say you have class C 192.109.83, and your subnet mask
	is [nnnhhhh].  That means you have nets:
	192.109.83.000hhhhh
		   001hhhhh
		   010hhhhh
		   011hhhhh
		   100hhhhh
		   101hhhhh
		   110hhhhh
		   111hhhhh

	Let's say your topology is as follows:

	[Physical network 1, 192.109.83.000hhhhh]
                              |
                              |                        
                              |
                          [Novell]
                              |
                              |
                              |
	[Physical network 2, 192.109.83.001hhhhh]
                              |
Place your Internet           |
router on one of the          |
nets.  Let's choose 2:    [Internet]


	Now let's label interfaces to be convenient.  I'm going to
	pretend your novell server uses host address 42 in all subnets
	just for convenience sake.  I'm also going to pretend that your
	Internet router has a local address ending in 50.  For last
	assumption, your Internet router's wide area network interface
	address is 129.129.129.129
                                                   
                                                   
 ---[Physical network 1, 192.109.83.000hhhhh]-----o------------------
                                                  |
                                             [192.109.93.42]
					     (Novell card 1)
                                              [Novell Box]
                                             (Novell card 2)
					     [192.109.93.58]  (42+16=58)
						  |
 ---[Physical network 2, 192.109.83.001hhhhh]-----o-----------o------
                                                              |
                                                       [192.109.93.66] (50+16)
                                                       (Local Interface)
 						       [Internet Router]
							(WAN interface)
						       [129.129.129.129]

The routes should look like this:

INTERNET ROUTER:
	127.0.0.1	127.0.0.1	lo0	(localhost internal loopback)
	192.109.83	192.109.93.66	se0	(ethernet routing of 192... net)
	0.0.0		129.129.129.129	se0	(default route for offnet stuff)

Novell Router: (must understand class C subnets)
	127.0.0.1	127.0.0.1	lo0	(localhost internal loopback)
	192.109.93.0	192.109.93.42	se0	(phys net 1)
	192.109.93.32	192.109.93.58	se1	(phys net 2)
	192.109.93.64	192.109.93.58	se1	(route net 3 to phys net 2)
	192.109.93.96	192.109.93.58	se1	(route net 4 to phys net 2)
	192.109.93.128	192.109.93.58	se1	(route net 5 to phys net 2)
	192.109.93.150	192.109.93.58	se1	(route net 6 to phys net 2)
	192.109.93.182	192.109.93.58	se1	(route net 7 to phys net 2)
	192.109.93.224	192.109.93.58	se1	(route net 8 to phys net 2)	

#Thanks in Advance
#	Stefan Taxhet
#-- 
#Stefan Taxhet                                    FAX:   +49 40 23646 550
#STAR Division GmbH                               Phone: +49 40 23646 952
#Sachsenfeld 4                                    uucp: ...!unido!starlab!st
#D-W-2000 Hamburg 1                               domain: st@starlab.uucp


	I hope I answered what you asked instead of what I thought
	you asked :)
 
	Ehud
 
 --
Ehud Gavron        (EG76)     
gavron@vesta.sunquest.com
I don't know WHY I said that..  I think it came from the FILLINGS
 in my rear molars..

 --------------------------------------------------------------------------

From: unido!SED.Provo.Novell.COM!jack_thomasson (Jack Thomasson)

>>>>> On 17 Sep 92 23:19:47 GMT, st@starlab.UUCP (Stefan Taxhet) said:

ST> Hello,

ST> I have a (novice?) question about subnetting in class C networks:
ST> Does subnetting in class C networks make sense?

sure.  in small sites that have multiple networks with not many hosts
on each network.  especially when different topologies must be used,
e.g., ethernet and token ring.

ST> BTW:
ST> This is exercise 16.5 in Douglas E. Comer's book "Internetworking With TCP/IP".
ST> Is there an answerbook?

ST> We have only 1 class C network address (192.109.83) but 2 physical networks.
ST> To be prepared for more phys. nets we use 3 Bits for subnetting,
ST> leaving 5 bits per subnet for hosts.
ST> Today we have no connection to the Internet, but we would like to :-)

sounds like you match the requirements!

ST> 1) Is there any way to let a Novell 3.11 server do the
ST>    routing between these two phys. nets (ethernet)?

sure.

install a NIC for each network to be routed in the file server.

load appropriate drivers for the nics.  you'll have to use
frame=ethernet_ii or frame=ethernet_snap for your configuration.
depending on what frame type you use for ipx you'll have to load those
as well.  make sure to use name= intelligently as it will save you
frustration in the end.  if i'm dealing with multiple frame types
(which i often do!) i like to append _frametype to a name that
describes the network, e.g., name=backbone_ii, name=ENet_ii,
name=TR_snap, name=classroomA_802.3, ....

bind ip to <name> addr=192.109.83.0x3e mask=255.255.255.0xe0
bind ip to <name> addr=192.109.83.0x5e mask=255.255.255.0xe0
bind ip to <name> addr=192.109.83.0x6e mask=255.255.255.0xe0
and so on for each network.

i normally like to use the highest available address in a subnet for
the router closest to the internet (and so do lots of other people).
plan ahead.

ST> 2) How must the gateway to the Internet be configured (SLIP)?

depends entirely on how you get to the internet.  slip (ppp is better)
would make sense if you can only obtain a phone line to a nearby and
willing university or some such.  there are many ways to get into the
internet.  good luck :{)}
 ---------------------------------------------------------------------
Jack Thomasson                   :{)}               (the bearded one)
Internet:Jack_Thomasson@Novell.COM                     MHS:JKT@NOVELL
Novell Consulting Services\E-23-2\122 East 1700 South\Provo, UT 84606
Phone: (800)453-1267x7604 | (801)429-7604          FAX: (801)429-5511
"WARNING: the comments do not necessarily reflect the implementation"

 --------------------------------------------------------------------------

From: Dan Ellison <unido!c-mols.siu.edu!dan>

In comp.protocols.tcp-ip you write:

>Hello,
 
>I have a (novice?) question about subnetting in class C networks:
>Does subnetting in class C networks make sense?
 
>BTW:
>This is exercise 16.5 in Douglas E. Comer's book "Internetworking With TCP/IP".
>Is there an answerbook?
 
>We have only 1 class C network address (192.109.83) but 2 physical networks.
>To be prepared for more phys. nets we use 3 Bits for subnetting,
>leaving 5 bits per subnet for hosts.
>Today we have no connection to the Internet, but we would like to :-)
 
>1) Is there any way to let a Novell 3.11 server do the
>   routing between these two phys. nets (ethernet)?

I assume that what you are saying here is that you  are using a netmask
of 255.255.255.224 and a broadcast of 192.109.83?  This tells the protocol
that you want to logically manage your class C net as 224 subnets and
have up to 31 hosts per subnet.  Be cautioned that some TCP implementations
will not break the boundary anywhere except octet aligned.  This could
cause you some problems.  I believe that there is a TCP/IP router for
Novell servers that should be able to do what you want.

BTW, 224 subnets sounds a bit high for a class C setup.  Class B
defines 255 subnets with 255 hosts each....  Not too much difference.

>2) How must the gateway to the Internet be configured (SLIP)?

Unless you have access to some physical wire that connects via
some other media then it looks like some form of point to point
link is the best choice.  Other possibilities might be remote
bridging via token ring or ethernet to a site that is willing
to accomodate you.

 --------------------------------------------------------------------------

From: unido!ideas.com!jmb (Jonathan M. Bresler)

Stephan,
	Novell can route between to different IP networks only!  Novell can't 
deal with subnetting.  our systems administrator worked for quite a while to 
get Novell to IP tunnel from one B class subnet to another, only to be told
by Novell that it not possible.  so either get yourself another C class net
for the other side of the Novell box, or get a real router, gateway, or bridge.

Jon Bresler
jmb@ideas.com

 --------------------------------------------------------------------------

From: unido!novell.com!mcollins (Marty Collins)

In article <3950@starlab.UUCP> you write:
: Hello,
: 
: I have a (novice?) question about subnetting in class C networks:
: Does subnetting in class C networks make sense?

SURE

: 
: BTW:
: This is exercise 16.5 in Douglas E. Comer's book "Internetworking With TCP/IP".
: Is there an answerbook?

Don't think so.

: 
: We have only 1 class C network address (192.109.83) but 2 physical networks.
: To be prepared for more phys. nets we use 3 Bits for subnetting,
: leaving 5 bits per subnet for hosts.
: Today we have no connection to the Internet, but we would like to :-)
: 
: 1) Is there any way to let a Novell 3.11 server do the
:    routing between these two phys. nets (ethernet)?

Yes but you must have differnet network numbers (SUBNETS) when there are
differnet Physical networks.

: 2) How must the gateway to the Internet be configured (SLIP)?

That will depend on the router you are using to attach to the internet.
Either SLIP or PPP can be used. Novell Netware 3.11 has no SLIP or
PPP capabilities.

: 
: Thanks in Advance
: 	Stefan Taxhet
: -- 
: Stefan Taxhet                                    FAX:   +49 40 23646 550
: STAR Division GmbH                               Phone: +49 40 23646 952
: Sachsenfeld 4                                    uucp: ...!unido!starlab!st
: D-W-2000 Hamburg 1                               domain: st@starlab.uucp

 --
		
	   A woman needs a man like a fish needs a bicycle.

	                				U2 
							Tryin to throw your
							arms around the world.

   Marty Collins
   Product Support
   NOVELL, INC.  San Jose, CA.        
   mcollins @novell.com 
   1-800-NETWARE  
-- 
Stefan Taxhet                                    FAX:   +49 40 23646 550
STAR Division GmbH                               Phone: +49 40 23646 952
Sachsenfeld 4                                    uucp: ...!unido!starlab!st
D-W-2000 Hamburg 1                               domain: st@starlab.uucp

-----------[000229][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 15:16:28 EDT
From:      glazer@ohstpy.mps.ohio-state.edu (Jon Glazer)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,comp.sys.novell
Subject:   Problems with SMC/WD Cards and SUN TCP/IP HELP!?

 
We are having some serious problems with some SMC cards.
 
First a bit of background.  We have about 70 Hosts of various types, all of
which have TCP/IP abilities of some sort.  We also have a Novell Network
version 3.11 with workstations running ODI.  All of the workstations (about
75) have SMC/WD Cards (old and new).  About 95% of these have the 16bit
10Base-T cards (8003 series).  The remaining 5% have 8 bit SMC/WD cards
(8003 series again).  I am unsure of the model cards but the 8 bit cards
have the LEDs on the back, visible froum the outside (4 around the 10Base-T
port) and the 16 bit cards range from the original Combo cards to Elite 16s.
Our network consists of some 16 David Systems hubs and 2 David concentrators
arranged in sub-networks through a Kalpana etherswitch.
 
Now the problem:
 
All of the 8 bit SMC cards are registering Frame Errors and colisions on
the David Hubs and Concentrators. No real problem since I can fairly easily
upgrade them to the 16 bit cards for speed anyhow. The real problem is that
non of the 16bit cards can Telnet to any SUN/OS based boxes. We have about
any flavor of Unix imaginable, VMS and Prime *YUCK*. The PCs can all connect
to any of the boxes that are not SUN instantly but with a sun, it can tak
1-10 minutes before a connection is made.
 
Your first instinct is to say, "well there must be a wiring problem from the
PCs to the suns".  Explain this, for the PCs with the 8bit "erroring" cards
int them, the connections are instant...again, until I upgrade them to the
16 bit cards.
 
I called compuserve and got the latest ODI drivers so I don't think that can
be the problem.  My question is what physically makes the 16bit cards
different than the 8 bit cards when it comes to network communications?
What can I do to resolve this problem???
 
Thanks for any help,
Jon
 


-----------[000230][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 92 11:52:59 GMT
From:      doug@happy.vf.ge.com (Doug Hughes)
To:        comp.protocols.tcp-ip,comp.protocols.misc
Subject:   IPC performance not on same machine


I have read Richard Stevens book on UNIX Network Programming and have
found it extremely useful (gross understatement).  Especially useful
are the graphs on the performance of the various IPC mechanisms.
  However, It would like to see a graph on performance of sockets
vs. TLI vs. Streams vs. XNS.  Has anybody out there done such a performance
characterization?  I would really like to see the results.

-- 
_____________________________________________________________
Doug Hughes
Software Developer - GE Aerospace (M&DSO), Valley Forge, PA
doug@happy.vf.ge.com	or	hughes@sde.mdso.vf.ge.com

-----------[000231][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 92 13:19:31 GMT
From:      berman@moguls.nlm.nih.gov (Lew Berman)
To:        comp.protocols.tcp-ip
Subject:   rpc/xdr summary

This posting is a compilation of all the responses I received on
my posting concerning rpc's and xdr.  If any more come in I will 
repost.

Lew Berman
National Library of Medicine
Bethesda, MD

------------------------------------------------------------------

Date: Tue, 15 Sep 92 18:02:21 CDT
From: fcw@telecom.ti.com (Fred Wedemeier)
To: berman@nlm.nih.gov
Subject: Re: structures
Content-Length: 1718
Status: RO
X-Lines: 48

Your items will come out in the correct order - e.g. first min, then
max, then mean, then variance from your struct image_stat. You may have
a problem if you send data between machines of different architecture.

skip if "little-endian" and "big-endian" mean something to you.
--------------------------------------------------------------
Machines differ in the order they store bytes in memory. Say you have:

union
{
  short int i;
  char c[2];
} foo;

... and do an assignment foo.i = 0xdead;

On a "big-endian" machine (SPARC and Moto 68000, for example) the
bytes are assigned:
  c[0] = 0xde;
  c[1] = 0xad;

On a "little-endian" machine ( VAXes and Intel x86,for example) the
bytes are assigned:

  c[0] = 0xad;
  c[1] = 0xde;

Thus things can get confused in a big way, and if you want to re-ignite
a holy war, just post some indignation about one or the other -endians
(doesn't matter which) and watch the blood fly.

Start reading again.
-------------------
There are some standards for resolving this, one of the foremost being
xdr (external data representation) that's pretty widely used,
especially with intermachine-communication using RPC (remote procedure
calls). As structures are being passed between machines, xdr converts
stuff to "network order." xdr on each kind of machine is
machine-specific, and each knows how to convert from network order to
the kind of ordering using inside the machine.

dunno if this helps - the answer to your specific question was simple,
but the preceeding may help keep you out of trouble if you haven't
worked with this stuff before.

Fred Wedemeier    pho: 214-997-3213     fax: 214-997-3639
                timsg: fcw             inet: fcw@pioneer.telecom.ti.com


----------------------------------------------------------------------

Date: Wed, 16 Sep 1992 09:14:47 +0200
From: casper@fwi.uva.nl (Casper H.S. Dik)
X-Organisation: Faculty of Mathematics & Computer Science
                University of Amsterdam
                Kruislaan 403
                NL-1098 SJ Amsterdam
                The Netherlands
X-Phone:        +31 20 525 7463
X-Telex:        10262 hef nl
X-Fax:          +31 20 525 7490
To: berman@nlm.nih.gov
Subject: Re: structures
Newsgroups: comp.lang.c
Content-Length: 1032
Status: RO
X-Lines: 26

In comp.lang.c you write:

>The reason I ask this question is that I would like to be able to transmit 
>the data in this structure using sockets.  If I can use the pointer to cHeader,
>as defined in the union, than all I need to do is send the header first
>rather than sending each component of the header separately.  But this is dependent
>on whether or not the items in the structure are guaranteed to be in their 
>correct order since on the receiving end the software will expect them
>in this order.
 
>Please respond by e-mail directly to me.

If you want to transmit non-byte stream data through sockets,
you must make sure that both ends can understand it.

You should try to use XDR (with or without RPC) to package
the structure data. XDR was invented by Sun and is implemented
on most current version of Unix. The source code is freely available.
(Look for SunRPC on an archive site near you)

The answer to your question: no, structures aren't contiguous.

Casper
-- 
						|	Casper H.S. Dik
						|	casper@fwi.uva.nl

---------------------------------------------------------------------------

From berman Wed Sep 16 14:17:19 1992
Date: Wed, 16 Sep 92 14:17:18 EDT
From: berman (Lewis Berman)
To: fcw@telecom.ti.com
Subject: Re: structures
Cc: berman
Content-Length: 468
Status: RO
X-Lines: 16

Fred,

> Does your target machine have RPC and the rpcgen utility? If so you
> might want to consider using it - it takes care of these problems for
> you.

I've considered this, however my understanding (from Stevens' Unix
Book) is that the overhead associated with RPC/XDR can be a factor
of around 100.  This seems like a terribly high overhead to incur
when transferring over the internet and things are already extremely 
slow.  What do you think?

Thanks,

Lew

------------------------------------------------------------------

From beyer@sceng.UB.com Wed Sep 16 15:17:31 1992
Date: Wed, 16 Sep 92 12:17:02 PST
From: Beyer_Mark <beyer@sceng.UB.com>
To: berman@nlm.nih.gov
Subject: Re: Re: RPC/XDR vs Pure Sockets
Content-Length: 399
Status: RO
X-Lines: 12

------------   ORIGINAL ATTACHMENT   -------- 
SENT 09-16-92 FROM SMTPGATE (berman@nlm.nih.gov)

but what about overhead issues.  Is RPC/XDR that much slower?

        I haven't done lots of measurements, probably because I
        haven't felt it's a problem.  I am much more concerned 
        with having to re-invent something that RPC/XDR already
        does, and has debugged.

        Mark



------------------------------------------------------------------

From mbeyer@ub-gate.UB.com Wed Sep 16 14:47:56 1992
Date: Wed, 16 Sep 92 11:48:14 PDT
To: berman@nlm.nih.gov
From: mbeyer@ub-gate.UB.com (Mark Beyer)
Subject: Re: RPC/XDR vs Pure Sockets
Cc: mbeyer@ub-gate.UB.com
Content-Length: 897
Status: RO
X-Lines: 21

In article <1992Sep16.150500.2409@nlm.nih.gov> you write:
>From: berman@nlm.nih.gov (Lew Berman)
>Subject: RPC/XDR vs Pure Sockets
>Date: 16 Sep 92 15:05:00 GMT
 
>In the development of a client/server application that will
>run over the Internet, can someone please discuss the trade-offs
>between RPC/XDR and pure socket calls.  I've read in Steven's
>UNIX book that the overhead associated with RPC calls can be as high
>as a factor of 100.  If this is true, than this would be a reason
>not to use RPC; however when passing structures and arrays of data
>(images for example) would this be simpler with RPC's and XDR?

Probably, if you want to operate in a heterogeneous environment, where you 
don't have to worry about understanding each machine's physical 
representation of data types.  RPC/XDR handles the ugly "endian" type 
issues, among other things.

--
Mark Beyer
mbeyer@sceng.ub.com


------------------------------------------------------------------

From casper@fwi.uva.nl Wed Sep 16 16:06:11 1992
Date: Wed, 16 Sep 1992 22:06:36 +0200
From: casper@fwi.uva.nl (Casper H.S. Dik)
X-Organisation: Faculty of Mathematics & Computer Science
                University of Amsterdam
                Kruislaan 403
                NL-1098 SJ Amsterdam
                The Netherlands
X-Phone:        +31 20 525 7463
X-Telex:        10262 hef nl
X-Fax:          +31 20 525 7490
To: berman@nlm.nih.gov
Subject: Re: RPC/XDR vs Pure Sockets
Newsgroups: comp.protocols.tcp-ip
Content-Length: 936
Status: RO
X-Lines: 23

In comp.protocols.tcp-ip you write:


>In the development of a client/server application that will
>run over the Internet, can someone please discuss the trade-offs
>between RPC/XDR and pure socket calls.  I've read in Steven's
>UNIX book that the overhead associated with RPC calls can be as high
>as a factor of 100.  If this is true, than this would be a reason
>not to use RPC; however when passing structures and arrays of data
>(images for example) would this be simpler with RPC's and XDR?

RPC can give an some extra overhead. But, your choice isn't
limited to sockets vs RPC/XDR. You can just use XDR on top of
sockets. This elliminates the RPC overhead and allows you to
continue to program with sockets. The only extra overhead
taht is incurred is the translation of your data to external
representation and back. This is well worth the portability
gained, IMHO.

Casper
-- 
						|	Casper H.S. Dik
						|	casper@fwi.uva.nl


------------------------------------------------------------------

From fcw@telecom.ti.com Wed Sep 16 18:59:02 1992
Date: Wed, 16 Sep 92 18:01:49 CDT
From: fcw@telecom.ti.com (Fred Wedemeier)
To: berman@nlm.nih.gov
Subject: Re: structures
Content-Length: 3394
Status: RO
X-Lines: 92

Lew:

> 
> I've considered this, however my understanding (from Stevens' Unix
> Book) is that the overhead associated with RPC/XDR can be a factor
> of around 100.  
> 

If your program is going to be used by a lot of people for a long time
to transfer a lot of data (how's _that_ for a disclaimer?) it's
probably worth your while to do something non-RPC - it's not the
fastest interface in the world. The neat thing about RPC is that it's
so easy to bring up. You need to trade off your program/debug/maintenance
time against comm thruput requirements.

Following are some results of a test program I just happened to write a
couple of weeks ago addressing a similar issue. We're using RPC for an
inter-processor link, and the question was if it would be worthwhile to
build an interface based on System V message queues to implement an
intra-processor link that did the same kind of things as the RPC
interface.

The test program RPC server receives RPCs which contain a read/write
command and a file name of a 1024-byte file. A 'read' RPC contains a
read command and file name. The return from the RPC is a status code
and the 1024-byte content of the file. A 'write' RPC contains a write
command, file name, and the 1024 bytes that are to be written on the
server. The return is a status code.

The test program RPC client issues a bunch of read/write commands and
reports the transfer time.

The machines identifed in the memo are:

hubble   - a Sun4/330 server
pioneer  - a 25 mhz SPARCStation
tcom2    - a TI model 1500 (Moto 68030-based UNIX box)
spacelab - a TI model 1500

The times are averages for a large number of transactions, and note
that it includes a disk read/write operation on the server for each
transaction. Also, these times are for tcp transport. udp results are
maybe ~ 2/3 of what's reported here.

If you've not done any RPC programming, I'd be happy to mail you a copy
of the test program code as an example of how it's done.

-----------------------------------------------------------------------

>From fcw Wed Sep  2 01:20:04 1992
Return-Path: <fcw>
Date: Wed, 2 Sep 92 01:20:02 CDT
From: fcw (Fred Wedemeier)
To: killian
Subject: rpc vs. queues for the call manager interface

I don't think there's any question...

--------------------------------------------------------------
Average time to do a 1024-byte rpc from client tcom2 to server
spacelab:

 read: 130 milliseconds
write: 120 milliseconds
--------------------------------------------------------------
Average time to do a 1024-byte rpc from client spacelab to server
spacelab:

 read: 84 milliseconds
write: 85 milliseconds
--------------------------------------------------------------
Average time to do a 1024-byte rpc from client hubble to server
hubble:

 read: 132 milliseconds
write: 180 milliseconds (I don't understand this one...)
--------------------------------------------------------------
Average time to do a 1024-byte queue transfer with reply, on
pioneer:

 read: ~2 milliseconds
write: ~2 milliseconds
--------------------------------------------------------------
Average time to do a 1024-byte queue transfer with reply, on
spacelab:

 read: ~5 milliseconds
write: ~5 milliseconds
--------------------------------------------------------------

Fred Wedemeier    pho: 214-997-3213     fax: 214-997-3639
                timsg: fcw             inet: fcw@pioneer.telecom.ti.com


------------------------------------------------------------------

From casper@fwi.uva.nl Wed Sep 16 16:06:11 1992
Date: Wed, 16 Sep 1992 22:06:36 +0200
From: casper@fwi.uva.nl (Casper H.S. Dik)
X-Organisation: Faculty of Mathematics & Computer Science
                University of Amsterdam
                Kruislaan 403
                NL-1098 SJ Amsterdam
                The Netherlands
X-Phone:        +31 20 525 7463
X-Telex:        10262 hef nl
X-Fax:          +31 20 525 7490
To: berman@nlm.nih.gov
Subject: Re: RPC/XDR vs Pure Sockets
Newsgroups: comp.protocols.tcp-ip
Status: RO
Content-Length: 936
X-Lines: 23

In comp.protocols.tcp-ip you write:


>In the development of a client/server application that will
>run over the Internet, can someone please discuss the trade-offs
>between RPC/XDR and pure socket calls.  I've read in Steven's
>UNIX book that the overhead associated with RPC calls can be as high
>as a factor of 100.  If this is true, than this would be a reason
>not to use RPC; however when passing structures and arrays of data
>(images for example) would this be simpler with RPC's and XDR?

RPC can give an some extra overhead. But, your choice isn't
limited to sockets vs RPC/XDR. You can just use XDR on top of
sockets. This elliminates the RPC overhead and allows you to
continue to program with sockets. The only extra overhead
taht is incurred is the translation of your data to external
representation and back. This is well worth the portability
gained, IMHO.

Casper
-- 
						|	Casper H.S. Dik
						|	casper@fwi.uva.nl

------------------------------------------------------------------


From fcw@telecom.ti.com Thu Sep 17 11:06:15 1992
Date: Thu, 17 Sep 92 10:09:03 CDT
From: fcw@telecom.ti.com (Fred Wedemeier)
To: berman@nlm.nih.gov
Subject: Re: structures
Content-Length: 264
X-Lines: 6
Status: RO

> run are doing over ethernet.  It seems that we would be better off NOT using RPC's/XDR.
> 
yup - I agree. Good luck w/the project!

Fred Wedemeier    pho: 214-997-3213     fax: 214-997-3639
                timsg: fcw             inet: fcw@pioneer.telecom.ti.com


------------------------------------------------------------------

From berman Thu Sep 17 09:19:05 1992
Date: Thu, 17 Sep 92 09:19:02 EDT
From: berman (Lewis Berman)
To: fcw@telecom.ti.com
Subject: Re: structures
Cc: berman
Content-Length: 1480
Status: RO
X-Lines: 28

Thanks Fred for the timings.  This is very informative.  I suspect your
times were very good (even for the RPC based ones) because you are running
over ethernet (rather than internet routing, etc.).  With our project
going over internet (routing, bridges, microwave, etc.) and having users
request files that can be 5MBytes or 10MBytes in size it would
seem that XDR transfers could extrapolate to the following (over ethernet?):

	rpc read/write time: 84 milliseconds   <--- took the best case 
	rpc transfer rate:  1024 bytes/84 milliseconds = 12,190bytes/sec
	image size (cervical xrays): 5,135,130 bytes
	rpc transfer rate (extrapolated from Fred's results): 
			5,135,130bytes/(12,190bytes/sec) = 421seconds = 7.02 minutes

This transfer rate, I'm assuming, is over the local ethernet and is using tcp.  The
problem is that this is extremely slow over local ethernet as you already know.
We've been running a test for the last month that transfers a 1Mbyte file over the
internet every 15 minutes.  The client/server/sockets program takes about 132 seconds
on average to transfer this file (7.943 Kbytes/sec).  We have also tested ftp and find that
ftp averages about 16Kbytes/sec.  So we know that we can improve our performance to
at least 16Kbytes/sec which would be greater than what the rpc/xdr tests you have
run are doing over ethernet.  It seems that we would be better off NOT using RPC's/XDR.

Thanks for all the info.

Lew Berman
National Library of Medicine
		

------------------------------------------------------------------

	

-----------[000232][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 15:18:32 GMT
From:      johna@mobdig.ncal.mobdig.com (John Antypas)
To:        comp.protocols.tcp-ip
Subject:   News logging and ISC system V.3.2

This is probably a very easy question...

We've got CNEWS 1.5.11 runnig on ISC 2.2.1.  I'd like news logging on
to trace some odd link problems we've had with NNTP.  (That's NNTP 
1.5.11 actually).  

I enabled logging and set log to LOG_LOCAL0.  Now what exactly do
I put in Syslog.conf to pick up these messages.

-- 
John Antypas / Senior Software Engineer	johna@mobdig.com
MobileDigital Corp -- Wireless Network Wizards 
Remember, the shortest distance between two points is a straight line, but it
gets even shorter when you don't have to string a line between those two
points.

-----------[000233][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 15:25:45 GMT
From:      rob@jdsny.com (Rob Yampolsky)
To:        comp.protocols.tcp-ip,comp.unix.questions
Subject:   SLIP-based communications hub - can it be done?

  As a low-cost, high-functionality means of connecting arbitrary numbers of
terminals at an arbitrary number of remote sites to a central unix system, I
have come up with the following configuration:

	Remote:  386/486 PC running SCO or Dell unix
	Question 1:   (high speed async port)
			  	|
			  	|
	Question 2:    	 SLIP (or similar)
			  	|
			  	|
	Question 3:    	  (SLIP server)
	Central:       arbitrary unix system

Advantages I see over ethernet routers or async multiplexors are:
  1) cheaper.
  2) more expandable than multiplexor option.
  3) ability to support LAN-based PC's as well as dumb terminals with the
     same piece of hardware.
  4) possibility of writing client-server transaction-based applications to
     reduce line speed requirements (such applications wouldn't need high-
     speed lines simply to insure instantaneous character echo). 

  Is anyone out there doing such a thing, and if so, do you have recommenda-
tions for hardware to accomplish it?  Any pitfalls I should look out for?
Any completely different (presumably better) means of accomplishing the same
kind of connectivity?  Thanks for your input.
-- 
================================================================
=     Mail replies appreciated if possible (who has time to    =
=     read this stuff?)               Thanks, rob@jdsny.com    =
================================================================

-----------[000234][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 15:49:47 GMT
From:      glenr@ruby.aruba.UUCP (Glen Reesor)
To:        comp.protocols.tcp-ip,comp.misc
Subject:   SUMMARY: Re: Dial-up IP vs. dedicated leased line

In article <1992Sep16.150333.21437@aruba.uucp>, glenr@aruba.UUCP (Glen Reesor) writes:
|> I currently investigating a direct IP connection to the Internet.  It appears
|> that there are two major ways I can accomplish this--dial-up IP using SLIP
|> and dedicated leased line with a router.  The dial-up IP would be 19.2 kbps
|> and the leased line would be 56 kbps.
|> 
|> Besides the lower bandwidth, SLIP seems to give all the functionality of
|> the leased line (assuming I remain connected all the time) for a significantly
|> lower cost.
|> 
|> What do people out there who are using SLIP think?  What experiences/horror
|> stories do you have?  Any comments or suggestions about these two alternatives?
|> 

Below are the responses I received.  The last was garbled from the original
sender.


From wjk@Phonogenic.COM (Bill Keenan)

> First you must ask yourself what can you afford. There are several components
> to cost: base phone line cost (9.6K, 14.4K, 56K leased; T1; dail-up), usage
> charges (line and service provider), equipment (one-time and recuring). I
> think you need to ask your organization what can we afford. Then you will
> know what subset of possible internet connection choices are available for
> you to choose. I have enclosed a list of service providers. If you are in
> CA, I suggest you look at CERFnet's 14.4K leased line option. I believe this
> will cost you about $1000/mo ($700 for CERF and $300 to PacBell [cost depends
> on the distance between you and the nearest point-of-presence (POH)]).
> 
> If I will be happy to answer specific questions if you want to mail them to
> me.
> 
> ******************************************************************************
> NSF Network Service Center (NNSC)                     Hotline:    617/873-3400
> 10 Moulton Street                                     Email: nnsc@nnsc.nsf.net
> Cambridge, MA 02138          Info-Server requests to: info-server@nnsc.nsf.net
> ******************************************************************************
> 
>  
> ******************************************************************************
> NSF Network Service Center (NNSC)                     Hotline:    617/873-3400
> 10 Moulton Street                                     Email: nnsc@nnsc.nsf.net
> Cambridge, MA 02138          Info-Server requests to: info-server@nnsc.nsf.net
> ******************************************************************************
> 
> 
>                      NETWORK PROVIDER REFERRAL LIST
>                     NSF Network Service Center (NNSC)
>                               26 August 1992
> 
> The NSF Network Service Center (NNSC) often receives the initial calls from 
> sites wishing to connect to the Internet.  The NNSC refers such callers to 
> the appropriate contacts for any networks which might be able to serve them.
> The current list of contacts used by NNSC for this purpose follows. If you
> have any updates or corrections to this information, please send a message
> to nnsc@nnsc.nsf.net.
> 
> -------------------------------------------------------------------------
> Network		Service Area
> 	Contact Name	Phone Number	Mail Address
> -------------------------------------------------------------------------
> Providers Based in the United States of America
> -------------------------------------------------------------------------
> 
> Alternet	US and International
> 	UUNET 		(800) 4UUNET3	alternet-info@uunet.uu.net
> 
> ANS		US and International
> 	Joel Maloff	(313) 663-7610	info@ans.net
> 
> BARRNet		Northern/Central California (CA)
> 	Paul Baer    	(415) 723-7520 	info@nic.barrnet.net
> 
> CERFnet		Western US and International
> 	CERFnet Hotline	(800) 876-2373	help@cerf.net
> 			(619) 455-3900
> 
> CICnet		Midwest US (MN, WI, IA, IN, IL, MI, OH)
> 	John Hankins	(313) 998-6102	hankins@cic.net
> 
> CO Supernet	Colorado (CO)
> 	Ken Harmon 	(303) 273-3475	kharmon@csn.org
> 
> CONCERT		North Carolina (NC)
> 	Joe Ragland	(919) 248-1404	jrr@concert.net
> 
> International Connections Manager (ICM) International
> 	Robert Collet	(703) 904-2230	rcollet@icm1.icp.net 
> 
> INet		Indiana (IN)
> 	Dick Ellis	(812) 855-4240	ellis@ucs.indiana.edu
> 
> JVNCnet		US and International
> 	Sergio Heker	(800) 35TIGER	market@jvnc.net
>         Allison Pihl
> 
> Los Nettos	Los Angeles Area (CA)
> 	Ann Westine Cooper  (310) 822-1511  los-nettos-request@isi.edu
> 
> MichNet/Merit	Michigan (MI)
> 	Jeff Ogden	(313) 764-9430	jogden@merit.edu
> 
> MIDnet		Mid US (NE, OK, AR, MO, IA, KS, SD)
> 	Dale Finkelson	(402) 472-5032	dmf@westie.unl.edu
> 
> MRnet		Minnesota (MN)
> 	Dennis Fazio	(612) 342-2570	dfazio@mr.net
> 
> MSEN            Michigan (MI)
> 	Owen Medd	(313) 998-4562	info@msen.com
> 
> NEARnet		Northeastern US (ME NH VT CT RI MA)
> 	John Curran 	(617) 873-8730	nearnet-staff@nic.near.net
> 
> netILLINOIS	Illinois (IL)
> 	Joel L. Hartman	(309) 677-3100	joel@bradley.bradley.edu
> 
> NevadaNet	Nevada (NV)
> 	Don Zitter	(702) 784-6133	zitter@nevada.edu
> 
> NorthwestNet	Northwestern US (WA OR ID MT ND WY AK)
> 	Eric Hood	(206) 562-3000	ehood@nwnet.net
> 
> NYSERnet	New York (NY)
> 	Jim Luckett	(315) 443-4120	info@nysernet.org
> 
> OARnet		Ohio (OH)
> 	Alison Brown	(614) 292-8100	alison@oar.net
> 
> PACCOM		Hawaii (HI) and Australia, Japan, Korea, New Zealand, Hong Kong
> 	Torben Nielsen  (808) 956-3499  torben@hawaii.edu
> 
> PREPnet		Pennsylvania (PA)
> 	Thomas Bajzek	(412) 268-7870	twb+@andrew.cmu.edu
> 
> PSCNET		Eastern US (PA, OH, WV)
> 	Eugene Hastings	(412) 268-4960	pscnet-admin@psc.edu
> 
> PSINet		US and International
> 	PSI, Inc.	(800) 82PSI82	info@psi.com
> 			(703) 620-6651
> 
> SDSCnet		San Diego Area (CA)
> 	Paul Love	(619) 534-5043	loveep@sds.sdsc.edu
> 
> Sesquinet	Texas (TX)
> 	Farrell Gerbode	(713) 527-4988	farrell@rice.edu
> 
> SprintLink      US and International  
>         Bob Doyle       (703) 904-2230  bdoyle@icm1.icp.net
> 
> SURAnet		Southeastern US (WV, VA, SC, NC, TN, KY, LA, MS, AL, GA, FL)
> 	Jack Hahn	(301) 982-4600	hahn@sura.net
> 
> THEnet		Texas (TX)
> 	Tracy LaQuey Parker  (512) 471-2444  tracy@utexas.edu
> 
> VERnet		Virginia (VA)
> 	James Jokl	(804) 924-0616	jaj@virginia.edu
> 
> Westnet		Western US (AZ, CO, ID, NM, UT, WY)
> 	Pat Burns	(303) 491-7260	pburns@yuma.acns.colostate.edu
> 
> WiscNet		Wisconsin (WI)
> 	Tad Pinkerton	(608) 262-8874	tad@cs.wisc.edu
> 
> WVNET		West Virginia (WV)
> 	Harper Grimm	(304) 293-5192	cc011041@wvnvms.wvnet.edu
> 
> ------------------------------------------------------------------------
> Providers Based in Canada
> ------------------------------------------------------------------------
> 
> ARnet           Alberta     
>         Walter Neilson  (403) 450-5188    neilson@TITAN.arc.ab.ca
> 
> BCnet           British Columbia          
>         Mike Patterson  (604) 822-3932    Mike_Patterson@mtsg.ubc.ca
> 
> MBnet           Manitoba    
>         Gerry Miller    (204) 474-8230    miller@ccm.UManitoba.ca
> 
> AccessNB        New Brunswick
>         David MacNeil   (506) 453-4573    DGM@unb.ca 
> 
> NLnet           Newfoundland and Labrador
>         Wilf Bussey     (709) 737-8329    wilf@kean.ucs.mun.ca  
> 
> NSTN            Nova Scotia
>         Michael Martineau (902) 468-NSTN  martinea@hawk.nstn.ns.ca 
> 
> ONet            Ontario
>         Andy Bjerring   (519) 661-2151    bjerring@uwovax.uwo.ca   
> 
> PEINet          Prince Edward Island
>         Jim Hancock     (902) 566-0450    hancock@upei.ca
> 
> RISQ            Quebec
>         Bernard Turcotte (514) 340-5700   turcotte@crim.ca
> 
> SASK#net        Saskatchewan
>         Dean C. Jones   (306) 966-4860    jonesdc@admin.usask.ca
> 
> -------------------------------------------------------------------------
> Other Providers
> -------------------------------------------------------------------------
> 
> AARNet		Australia
> 	AARNet Support	+61 6 249 3385	aarnet@aarnet.edu.au
> 
> UKnet		United Kingdom of Great Britain and Northern Ireland
> 	UKnet Support	+44-227-475497	postmaster@uknet.ac.uk
> 
> EUnet           Europe, CIS-region, and Northern Africa
> 	EUnet Support   +31 20 592-5124 glenn@eu.net
> 
> Pipex           United Kingdom
>         Richard Nuttall (RN131) +44 223 424616  sales@pipex.net
> -----------------------------------------------------------------------------


From Craig Partridge <craig@aland.bbn.com>

>     With v.42bis compression, my dial-up SLIP gets around 26Kbit/s throughput
> for ASCII transfers.  Some experiments I did when I got slip performance
> suggested that achieving at least 20Kbit/s was essential for me to be
> able to use the line (by myself).  20Kbit/s was about where an FTP +
> a single telnet session didn't interfere with each other badly.
> 
>     Note, CSLIP will likely give better performance. Dunno if you can get
> it -- and, so far as I know, it doesn't work for routers, just hosts.
> 
>     With your site, I'd suspect 56Kbit may be better, just because you
> have more than one user.
> 
>     Other thing to look at is cost.  Dialup phone charges can be steep,
> and a 56Kbit leased line can pay for itself if your dial-up phone bill
> gets very large.
> -----------------------------------------------------------------------------

From jmb@ideas.com (Jonathan M. Bresler)

> 	we h been using slip as an internet connection for over a year now.
> our bandwidth is only 9600 baud, half of yours, yet it has proven to be useful.
> i would not like to see, wait :) for X rnning over a slip connection, the band-
> withd requierd is to high for slip to a dialup line. 
> 
> 	file transfers from 4kB to 2Kb per second.  telnet not bad, cusee   rses ud pa datae s, are a ltt  ittle slow.
> 
> 	it is useable, but not from   or heavy traffic.  amil    msail      ail is good, no po roblems
> encountered, but if we miss the news for a couple of days runiung    ning, trying to 
> cahrt   tch up is trule a     y a night  htmare.
 
-- 
Glen Reesor                       | Internet Style:  aruba!glenr@uu2.psi.com 
Systems Administrator, Project Zed| UUCP :  ...!uunet!uu2.psi.com!aruba!glenr

-----------[000235][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 17:50:59 GMT
From:      ronf@panther3.panther.mot.com (Ron Feigen)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP messages

In article <19m3v8INNec9@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
>In article <1992Sep21.214039.24610@panther.mot.com> ronf@panther3.panther.mot.com (Ron Feigen) writes:
>>Each message has a header made up of: Length, Type, and ID, or 12 bytes.  I know
>>that by definition an INET socket deleivers a "stream" of data.  However can I 
>>not assume that the entire 12 bytes will be delivered "atomically"?  Especially 
>>since the sender has TCP_NODELAY set.
>
>No, you can't assume this.  TCP_NODELAY causes the sender to transmit as
>soon as possible, but it doesn't force the receiver to deliver all 12 bytes
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>to the user process immediately.  Also, the window could be less than 12
>bytes, so the sender would only be able to send a part of the message, and
>it will send the rest when the window opens.
>-- 
>Barry Margolin
>System Manager, Thinking Machines Corp.
>
>barmar@think.com          {uunet,harvard}!think!barmar

If a CPU recieves a TCP packet it does not present the entire packet to the user
process?

-- 

>
Ron Feigen
ronf@panther.mot.com

-----------[000236][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 18:01:27 GMT
From:      rstevens@noao.edu (W. Richard Stevens)
To:        comp.protocols.tcp-ip,comp.protocols.misc
Subject:   Re: IPC performance not on same machine

>  It would like to see a graph on performance of sockets
>vs. TLI vs. Streams vs. XNS.  Has anybody out there done such a performance
>characterization?  I would really like to see the results.

The comparison I'd like to see is SunOS 5 (aka Solaris 2.0) versus
SunOS 4.1.x, on the same piece of hardware.  Are streams inherently
slower than BSD sockets?  Some quick tests that I've done on two
identical 386 boxes show that SVR4's TCP is about 20-35% slower than
BSD/386.  My guess would be more overhead in the SVR4 streams subsystem.

	Rich Stevens  (rstevens@noao.edu)

-----------[000237][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1992 21:11:44 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP messages

In article <1992Sep22.175059.29859@panther.mot.com> ronf@panther3.panther.mot.com (Ron Feigen) writes:
>In article <19m3v8INNec9@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
>>No, you can't assume this.  TCP_NODELAY causes the sender to transmit as
>>soon as possible, but it doesn't force the receiver to deliver all 12 bytes
>                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>If a CPU recieves a TCP packet it does not present the entire packet to the user
>process?

Yeah, I guess it usually will.  But it might decide to split it up due to
kernel buffering constraints.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000238][next][prev][last][first]----------------------------------------------------
Date:      Tue, 22 Sep 1992 23:57:45 GMT
From:      jjensen@convex.com (James Jensen)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP messages

barmar@think.com (Barry Margolin) writes:
- Yeah, I guess it usually will.  But it might decide to split it up due to

usually  is the key word here.   Most of the time a small packet will
get delivered as one piece.  But there are so many ways for it not
to be delivered in one piece that it just isn't worth thinking about.

Code that assumes that tcp data will be delivered in one piece is just
plain broken, end of discussion.  It will eventually cause hard to debug
problems that will require going back and rereleasing the program.
Fix it now, while it's easy.  Here's some code for free, and
worth every penny. :-)

/* this assumes that bcopy is cheaper than read, and works on
   overlapping buffers. if reads are cheaper then change the 
   read sizes to the exact size - headersize, and end_packet and 
   change the static declaration.  If bcopy doesn't do overlapping 
   copies do it by hand. */

static char *bufaddr = startofbuffer;

while (bufaddr< startofbuffer +headersize;) {
   if (tlen = read(fd,bufaddr,endofbuffer-bufaddr))< 0)
        handle EOF /* you are doing this right? :-) */
   else 
      bufaddr +=tlen;
}
end_packet = startofbuffaddr +size_from_header +size_of_header;
while  (bufaddr< end_packet) {
   if (tlen = read(fd,bufaddr,enddofbuffer-bufaddr))< 0)
        handle EOF 
   else
      bufaddr +=tlen;
        handle EOF
}

bcopy(endpacket,startofbuffer,bufaddr-end_packet);
bufaddr=startofbuffer+bufaddr-end_packet;

Jim Jensen - jjensen@convex.com, who has seen this mistake one to many
times.

-----------[000239][next][prev][last][first]----------------------------------------------------
Date:      Wed, 23 Sep 1992 09:15:22 GMT
From:      ga@rdg.dec.com (Giles Atkinson)
To:        comp.os.msdos.misc,comp.protocols.tcp-ip
Subject:   Remote Boot ROMs for PC network adaptors

I am interested in compiling a list of suppliers of ROMs for popular
PC network adapors which implement TCP/IP protocols (presumably BOOTP/RARP/TFTP)
to allow the PC to be remote booted.   Please send mail to the address above;
if I get a significant list I will post a summary.

				thanks,
					Giles

-----------[000240][next][prev][last][first]----------------------------------------------------
Date:      Wed, 23 Sep 1992 09:29:19 GMT
From:      bzs@ares.std.com (Barry Shein)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.unix.internals
Subject:   Re: Ethernet Collisions (?)


>>1) How do we generate collisions on the ethernet?
>
>The easiest way is to set several processes up that are each transfering
>a large block of data repeatedly.  For example, on our DOS ethernet LAN, we
>start up a .BAT file that copies a 2MB file back and forth between
>workstation and file server over and over again.  Run three of these and you'll
>see plenty of collisions!

Actually, you're better off staying away from the disks as it will
only slow things down.

Something like this would work:

	rsh othersystem cat \> /dev/null < /dev/zero

on systems which have /dev/zero (RS6000's and Sun's do.)



-----------[000241][next][prev][last][first]----------------------------------------------------
Date:      Wed, 23 Sep 1992 10:50:42 GMT
From:      clfung@uxmail.ust.hk (Fung Chi Leung)
To:        comp.protocols.tcp-ip
Subject:   Adaptive Flow control

  I heard about the FTP is using a adaptive flow control mechanism, can
someone out there can briefly explain how the adaptive flow control works?   
Since I am doing some experiments on a flow control mechanism for UDP, I
would like to know more is there any other existing adaptive flow control
that is using now?

Thanks in advance


Leung

-------------------------------------------------
Hong Kong University of Science and Technology
Fung Chi Leung [ clfung@uxmail.ust.hk ]
-------------------------------------------------

-----------[000242][next][prev][last][first]----------------------------------------------------
Date:      Wed, 23 Sep 1992 12:21:34 +0000
From:      john@elmail.co.uk (John K A Sewart)
To:        comp.protocols.tcp-ip
Cc:        john@uknet.ac.uk
Subject:   Mail user agents for SMTP/POP etc.



Really the subject line says it all: I want to find out what is out there
in terms of user agents for Unix style mail for Unix, DOS, Windows and
OS/2. Also mail readers. Public domain, shareware or commercial products are
all equally of interest. I will summarise these for the 'net.

Please send E-mail rather than waste net bandwidth with duplicate information

Thanks in advance

John Stewart

ElectricMail Ltd.

-----------[000243][next][prev][last][first]----------------------------------------------------
Date:      23 Sep 92 12:54:50 GMT
From:      jessea@homecare.com (Jesse W. Asher)
To:        comp.protocols.time.ntp,comp.protocols.tcp-ip,comp.unix.sysv386
Subject:   Anyone running local_clock in ntp with ODT 2.0?

I've been trying to get ntp set up running the local clock because we
are not connected to the Internet.  I've got networking spanning the
country and we need some way of Universal Time to bypass the time
zone changes.  Anyway, in the man pages for xnptd it mentions the fact
that you can specify a server line "server 127.127.1.0" to use a
stratum 0 local clock in order to set the rest of your network.
Unfortunately, I can't seem to get this to work.  I've only got two
hosts set up.  One I've got the following:

---

server 127.127.1.0

driftfile /etc/ntp.drift

---
This host's address is x.x.145.15

The next host is set up with:
---

server x.x.145.15

driftfile /etc/ntp.drift

---

The problem is that if I run xntpd on the second host, the time never
seems to sync up.  If I run /etc/ntpdate to explicitly set the date on
the second host I get:
/etc/ntpdate: No hosts suitable for synchronization found.

Can anyone make some suggestions here as to why this isn't working?
-- 
      Jesse W. Asher                                   Phone: (901)762-6000
                         Varco-Pruden Buildings
	      6000 Poplar Ave., Suite 400, Memphis, TN  38119
 Internet: jessea@homecare.COM                 UUCP: ...!banana!homecare!jessea

-----------[000244][next][prev][last][first]----------------------------------------------------
Date:      23 Sep 92 12:57:12 GMT
From:      jessea@homecare.com (Jesse W. Asher)
To:        comp.protocols.time.npt,comp.protocols.tcp-ip
Subject:   Radio clocks for use with ntp - where to get them?

We aren't hooked up to the Internet yet and I'd like to examine
putting a clock radio on our network to get better time than just a PC
will allow.  Does anyone know where I can get such an animal, how much
they run, and how to set them up on an ethernet network (what a mouth
full!).  Thanks for help with time sanity!
-- 
      Jesse W. Asher                                   Phone: (901)762-6000
                         Varco-Pruden Buildings
	      6000 Poplar Ave., Suite 400, Memphis, TN  38119
 Internet: jessea@homecare.COM                 UUCP: ...!banana!homecare!jessea

-----------[000245][next][prev][last][first]----------------------------------------------------
Date:      Wed, 23 Sep 1992 16:22:40 GMT
From:      jnm@ornl.gov (Jamey Maze)
To:        comp.protocols.tcp-ip
Subject:   X.400 vs. Internet mail

I'm trying to convince folks here that X.400 is no longer the obvious
ultimate destination of any e-mail architecture strategy. I want to show
that the Internet alternative (RFC-822, SMTP) is growing to meet changing
requirements for e-mail. I'll be pointing to new advances like PEM and
MIME. My question is, what else is there that is worth mentioning in this
context? Someone told me this morning that there's an IETF group developing
a new protocol to eventually replace SMTP. Is that true? If so, can someone
e-mail me some references? Anything else I'm overlooking?

Thanks!

--
Jamey Maze, Oak Ridge National Lab     

-----------[000246][next][prev][last][first]----------------------------------------------------
Date:      23 Sep 92 20:26:42 GMT
From:      das@voodoo.boeing.com (Deb Schwartz)
To:        comp.periphs.printers,comp.protocols.tcp-ip
Subject:   Re: TCP/IP on printers

In article <avalon.717184160@coombs> avalon@coombs.anu.edu.au (Darren Reed) writes:
>
>Well the QMS PS-2000 has a hard disk (and when you turn it on,
>it boots, just like a real computer :-) with about 120MB space
>for spooling, virtual memory, fonts, etc. (hope thats right).
>It also has some tcp-ip in it, not a full stack, however, since
>if you try connect to (say) the finger port on the printer,
>your connection will timeout rather than get connection refused.
>But you can ftp to it :)
>
A couple of small corrections:  the hard drive on a QMS PS-2000 is optional
if you're just going to use one of the standard interfaces (serial, parallel
or Appletalk).  The printer will boot just fine without the hard drive in it -
all the system software's in firmware.  If you order the TCP/IP option, you
will have to also order a hard drive in order for it to work properly.
The hard drives are available in either 40 or 120 MB.   Also, I don't know
what release of the firmware you're using, but in release 4 and above you can
also telnet to the printer, and finger works just fine. (We have a 40MB disk
ethernet-connected machine running release 4.1.).  

Debbie Schwartz // das@voodoo.boeing.com  // or uunet!bcstec!voodoo!das 

-----------[000247][next][prev][last][first]----------------------------------------------------
Date:      23 Sep 92 20:50:55 GMT
From:      Greenwald@cs.stanford.edu
To:        comp.protocols.tcp-ip
Subject:   SLIP implementations

A friend asked me the following question.  I have no idea about the
answer, but I agreed to ask the net.

He works for a company that has a small number of employees (1 or 2)
working in other cities.  The company itself is served adequately by
a local ethernet (actually several).  They want to provide SLIP access
to these employee's homes.  

The employee stays at one location.  They want a box that will
dial-out to the employees home; they do not want any random person to
be able to SLIP in.

They are willing to allocate a permanent IP address to each employee
served by such a scheme.  (1 now, probably never more than 3).

Upon receipt of a packet from the local net addressed to the
employee's IP address, if the line isn't currently connected, they
want the box to dial the home number.

If you dial in to the box (or issue a command over the ethernet,
whatever), it will also initiate the dial-out to the fixed employee home
phone number.

It's acceptable to have one of these boxes per employee (if it's cheap
enough), but they'd prefer having all N (N less than three or four)
employees serviced by one machine having N serial lines.

If a dial-out command isn't possible, they'd be willing to have the
machine constantly dialing the employee's homes, whenever the connection
dies. 

I guess there are three questions here:
Are there SLIP implementations that implement this dial-out-only
feature? 

What's the cheapest way of purchasing such a setup?  (Can you just get
a PC-clone with serial ports and an Ethernet Interface and some
off-the-shelf SLIP product, and use it as a SLIP gateway?)

How reliable are these products?

Email responses, please.
    Greenwald@cs.stanford.edu

Thanks.

-----------[000248][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 08:57:24 PDT
From:      Michel.Davidoff@sonoma.edu (Michel Davidoff)
To:        comp.protocols.tcp-ip
Subject:   Re: Mail user agents for SMTP/POP etc.

In article <9209231321.aa06839@cheyenne.elmail.co.uk> john@elmail.co.uk (John K A Sewart) writes:
>Subject: Mail user agents for SMTP/POP etc.
>From: john@elmail.co.uk (John K A Sewart)
>Date: 23 Sep 92 12:21:34 GMT
>
>
>Really the subject line says it all: I want to find out what is out there
>in terms of user agents for Unix style mail for Unix, DOS, Windows and
>OS/2. Also mail readers. Public domain, shareware or commercial products are
>all equally of interest. I will summarise these for the 'net.
>
>Please send E-mail rather than waste net bandwidth with duplicate information
>
>Thanks in advance
>
>John Stewart
>
>ElectricMail Ltd.
WinQVt is POP 3
NUpop is POP 3


Michel Davidoff
E-Mail Michel.Davidoff@Sonoma.edu

Don't let the future
Happen without you
Please VOTE
       PLEASE VOTE

-----------[000249][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 1992 11:10:42 GMT
From:      scoggin@opus.ee.udel.edu (John K Scoggin)
To:        comp.protocols.time.npt,comp.protocols.tcp-ip
Subject:   Re: Radio clocks for use with ntp - where to get them?

In article <1992Sep23.125712.13747@homecare.com> jessea@homecare.com (Jesse W. Asher) writes:
>We aren't hooked up to the Internet yet and I'd like to examine
>putting a clock radio on our network to get better time than just a PC
>will allow.  Does anyone know where I can get such an animal, how much
>they run, and how to set them up on an ethernet network (what a mouth
>full!).  Thanks for help with time sanity!
>-- 

Look on louie.udel.edu for a file called clock.txt - it describes the
various clock options and their sources.  I think its in the pub/ntp
directory.

	- John

-------------------------------------------------------------------------
John K. Scoggin, Jr.			Email:   scoggin@udel.edu
Supervisor, Network Operations			 scoggin@delmarva.com
Delmarva Power & Light Co.		Voice:	 302-451-5200
P.O. Box 6066				Fax:	 302-451-5321
Newark, DE 19714-6066
-------------------------------------------------------------------------
Your milage may vary.  Void where prohibited by law (or common sense).

-----------[000250][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 1992 13:14:14 GMT
From:      tkevans@eplrx7.es.duPont.com (Tim Evans)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.mail.sendmail
Subject:   Sendmail and/or DNS Problems/Questions (AIX)

I'm having some problems under AIX that seem to be
related to sendmail and/or DNS setup--hence the cross-posting.

My RS6000's (running AIX 3.2) function almost perfectly as NIS
clients (without a DNS /etc/resolv.conf file present), except for
sendmail not being able to route mail to
other hosts in the domain if they are not addressed with
FQDN's.  If I put in an /etc/resolv.conf file,
sendmail routes ok, but then equivalence problems occur unless
FQDN's are used in r-commands, user .rhosts files, and NIS maps.

Nonetheless, since my Sun machines in the same NIS domain function
without these problems, I'd like to have the RS6000's behave in the
same way if possible.

I'm not sure if this a misconfigured sendmail.cf, bad NIS maps,
or what.  Suggestions appreciated.  Thanks.
-- 
Tim Evans                     |    E.I. du Pont de Nemours & Co.
tkevans@eplrx7.es.dupont.com  |    Experimental Station
(302) 695-9353/7395           |    P.O. Box 80357
EVANSTK AT A1 AT ESVAX        |    Wilmington, Delaware 19880-0357

-----------[000251][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 1992 15:39:45 GMT
From:      pvg@po.CWRU.Edu (Paul V. Gerwe)
To:        comp.protocols.tcp-ip
Subject:   PC as host running Dos Software?


Greetings!

I am looking for software that would allow a PC to function as a host
supporting 4 or 5 simultaneous telnet connections and be able to run
DOS software.  If you have any information or know of any sources that 
might be of use, could you please send information to

		pvg@po.cwru.edu

Thanks!

Hat

-----------[000252][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 1992 16:44:55 GMT
From:      erick@sunee.uwaterloo.ca (Erick Engelke)
To:        comp.protocols.tcp-ip
Subject:   Re: PC as host running Dos Software?

 pvg@po.CWRU.Edu (Paul V. Gerwe) writes:
>
>Greetings!
>
>I am looking for software that would allow a PC to function as a host
>supporting 4 or 5 simultaneous telnet connections and be able to run
>DOS software.  If you have any information or know of any sources that 
>might be of use, could you please send information to
>

There are three solutions I know of currently:

Beame&Whiteside BW/TCP 3.0 includes a multisession DOS TELNET server 
under Windows.  Contact beame@bws.com

Essex TCP/2-DOS includes a multisession DOS TELNET server under Windows
too if you run their TCP.  Contact ddl@harvard.edu 

Superior Network Software Intl. offers a multisession DOS TELNET server
under Windows or Desqview if you are using FTP's PC/TCP.  Contact
snsi!info@uunet.ca


You may wish to post future items to comp.protocols.tcp-ip.ibmpc which
is more pc-centric. 

Erick
-- 
----------------------------------------------------------------------------
Erick Engelke				 	            WATTCP Architect
erick@development.uwaterloo.ca     TCP/IP was easy but i still can't work VI

-----------[000253][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 17:02:57 GMT
From:      howie@voodoo.boeing.com (howie)
To:        comp.protocols.tcp-ip
Subject:   Re: Source Code for "Internetworking with TCP/IP, Volume III"

In article <1992Sep18.175717.17352@hou.amoco.com> zjhc05@hou.amoco.com writes:
...
>
>I checked this morning and it is available by anonymous ftp from "ftp.uu.net".	
>It is in the directory "published/books" and the file name is	
>"comer.internetworking3.src.tar.Z"


How does this book compare to Steven's _UNIX Network Programming_?


-- 
				       howie              

				       howie@voodoo.boeing.com

-----------[000254][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 17:41:54 GMT
From:      egonzale@netmon.mty.itesm.mx (Ing. Ernesto Gonzalez)
To:        comp.protocols.tcp-ip
Subject:   Is there a way to encapsulate X.25 over TCP/IP ????

Again, how can i encapsulate X.25 traffic over TCP/IP????

Thaks in advance......

-----------[000255][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 1992 18:06:50 GMT
From:      ddk@beta.lanl.gov (David D Kaas)
To:        comp.protocols.tcp-ip
Subject:   DNS and .rhosts files

We are currently switching our network over to DNS and do
not have much experience with DNS yet.  Before DNS our hosts.equiv
and .rhosts files contained just the host name now the 
hostname.domain is required.  Is this working correctly
or have we done something wrong?
Our netwok contains SGI, SUN, Stardent and Cray XMP.

Thanks in advance..

Dave Kaas
Boeing Computer Services Richland
D. O. E. Richland
ddk@lanl.gov
-- 
Dave Kaas 	- Boeing Computer Services Richland
		- D. O. E. Richland
ddk@beta.lanl.gov

-----------[000256][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 92 18:14:23 GMT
From:      reece@eco.twg.com (Reece R. Pollack)
To:        comp.protocols.tcp-ip
Subject:   RE: ---<<< Wollongong 3B2 TCP select() bug?? >>>---


In article <1992Sep16.195349.16010@cbnewsi.cb.att.com>, mbg@cbnewsi.cb.att.com (mitchell.b.germansky) writes:
|>my project has encountered what seems to be a problem in wollongong's
|>tcp/ip on a 3b2.  if you have any experience in resolving this,
|>please get in touch with me.
|>
|>we have a 3b2 connected over ethernet to a sharebase database
|>machine (aka britton-lee idm, teradata, ncr).  we are using
|>Wollongong Integrated Networking WIN/3B: Release 3.2A on 3b2
|>unix 3.2.3
|>
|>we have noticed a predictable 2 second delay every 20 database
|>transactions.  examining the process stack trace during this 2
|>second interval revealed that the process was asleep in a tcp/ip
|>library select() call.  we believe that this problem is really
|>independent of the database and will write some tcp code to attempt
|>to reproduce the select() delay.  the sharebase performance is
|>constant even for the transaction that blocks for the 2 seconds.
|>
|>we ran the same tests from a sun with no delays.  i also wrote some
|>raw tcp test code that showed no delays.
|>
|>we have heard rumors that other people have had similar problems.
|>we are wondering if there are any workarounds or perhaps even a
|>patch available.
|>
|>if you have any suggestions or tips on what to look at  or do next,
|>please contact me via email or phone.
|>
|>thx!!

There is a newsgroup in the VMSNET hierarchy dedicated to Wollongong
products called 'vmsnet.networks.tcp-ip.wintcp'. Although it is primarily
for the VMS product line, questions and comments about all Wollongong
products are welcome.

Now on to the answer! There is a known bug in some versions of select()
in which in a timeout value of less than 2 seconds results in a timeout
of 2 seconds. A fix for this problem is available through the normal AT&T
support channels. I believe the phone number is 1-800-543-9935.

--
Reece R. Pollack
Senior Software Engineer
The Wollongong Group, Inc.

-----------[000257][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 18:41:06 GMT
From:      mr3@ukc.ac.uk (M.Rizzo)
To:        comp.protocols.tcp-ip
Subject:   Internet routing software for Sun

I am posting this on behalf of the Computing Dept at the University
of Malta. We hope to join the Internet soon and to start off we're
looking for routing software to run on a SUN Sparcstation II. Can
anybody suggest/recommend anything? Thanks.

Mike
--
Michael Rizzo                           email mr3@ukc.ac.uk
Rm109B, Computing Laboratory
University of Kent
Canterbury CT2 7NF  U.K.                Tel. (0227) 764000 Ext 7684

-----------[000258][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 19:19:53 GMT
From:      moore@cs.utk.edu (Keith Moore)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.mail.sendmail
Subject:   Re: Sendmail and/or DNS Problems/Questions (AIX)

In article <tkevans.717340454@eplrx7.es.dupont.com>, tkevans@eplrx7.es.duPont.com (Tim Evans) writes:
> I'm having some problems under AIX that seem to be
> related to sendmail and/or DNS setup--hence the cross-posting.
> 
> My RS6000's (running AIX 3.2) function almost perfectly as NIS
> clients (without a DNS /etc/resolv.conf file present), except for
> sendmail not being able to route mail to
> other hosts in the domain if they are not addressed with
> FQDN's.  If I put in an /etc/resolv.conf file,
> sendmail routes ok, but then equivalence problems occur unless
> FQDN's are used in r-commands, user .rhosts files, and NIS maps.
> 
> Nonetheless, since my Sun machines in the same NIS domain function
> without these problems, I'd like to have the RS6000's behave in the
> same way if possible.
> 
> I'm not sure if this a misconfigured sendmail.cf, bad NIS maps,
> or what.  Suggestions appreciated.  Thanks.

I added the following lines to our standard sendmail.cf file to make
it run correctly on AIX.  

We do have /etc/resolv.conf files on each of our systems, so I don't
know whether AIX sendmail will require this or not.  I recommend using
DNS over YP for host name lookups, even with the name equivalence
problems.  (Our Suns run a modified libc.so file that uses the
resolver instead of YP also.)

# Add the following for AIX v3.1
# (or invoke sendmail with -oI -oKMX)
# Better yet, get a real version of sendmail.
# Even better still, get a real operating system.
#
## treat name server lookup failure as a temporary error
OI
## use MX records, but not MB, MG or MR records
OKMX

I'll note that our mail-client sendmail.cf file runs without special
modifications on all of our other systems (SunOS, NeXT, HPUX, Ultrix,
4.3BSD, Dynix), but has to be hacked on the AIX boxes to get it to
work.

--
Keith Moore / U.Tenn CS Dept / 107 Ayres Hall / Knoxville TN  37996-1301
Internet: moore@cs.utk.edu      BITNET: moore@utkvx

-----------[000259][next][prev][last][first]----------------------------------------------------
Date:      24 Sep 92 20:28:21 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Internet routing software for Sun

In article <1694@eagle.ukc.ac.uk> mr3@ukc.ac.uk (M.Rizzo) writes:
>I am posting this on behalf of the Computing Dept at the University
>of Malta. We hope to join the Internet soon and to start off we're
>looking for routing software to run on a SUN Sparcstation II. Can
>anybody suggest/recommend anything? Thanks.

Routed is a standard component of SunOS, and implements the RIP routing
protocol.  If that's not sufficient, you'll have to ask a more specific
question.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000260][next][prev][last][first]----------------------------------------------------
Date:      Thu, 24 Sep 1992 22:14:00 GMT
From:      epang@fraser.sfu.ca (Eugene Pang)
To:        comp.protocols.tcp-ip
Subject:   Re: Source Code for "Internetworking with TCP/IP, Volume III"

>>It is in the directory "published/books" and the file name is	
>>"comer.internetworking3.src.tar.Z"

Is there an ISBN # yet?

-----------[000261][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 1992 00:34:50 GMT
From:      wendell@medsys.rn.com (Wendell Dingus)
To:        comp.protocols.tcp-ip
Subject:   Help: Unix <-> Novell via TCP/IP

Subject line says it all. I commonly setup and administer a number
of Unix systems my company sells as turn-key systems and am suddenly
in need of connecting one in some fashion to an existing Netware 386
system. I'm only moderately familiar with Novell, but have skimmed
through the TCP/IP Transport manual that comes with Netware 386. What
I want to know is what is the easist and 'best' way to link two machines
in this fashion so that PC's on the network can have telnet(?) sessions
to the Unix box, maybe have it's disk mounted via NFS, and maybe even
have the Netware box's files accessible by the Unix box. I think Novell's
product Lan Workplace for DOS does most(?) of these things, but I haven't
seen the complete literature for it yet. Anyway, I figure if Novell's
already done it, someone else has done it better :)

Points: Do I need 2 Ethernet cards in the Netware box or can one card
run both IPX/SPX and TCP/IP simultaneously?

Would there be any advantage to running 2 cards if the first answer is no?

E-Mail any thoughts or ideas you can give me unless you think it's of
sufficient general interest. Thanks!


-- 
-- Wendell Dingus    Domain: wendell@medsys.rn.com     //  
-- UUCP:    ...!gator!medsys!wendell   Only AMIGA!   \X/

-----------[000262][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 01:17:27 GMT
From:      dls@mentor.cc.purdue.edu (David L Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Source Code for "Internetworking with TCP/IP, Volume III"

	The ISBN is 0-13-474222-2.
-- 
					+-DLS  (dls@mentor.cc.purdue.edu)

-----------[000263][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 01:21:11 GMT
From:      add@is.rice.edu (Arthur Darren Dunham)
To:        comp.protocols.tcp-ip
Subject:   tn3270 over TCP/IP with local printing wanted

Recently some of our users moved from serial connections on Macs to
TCP/IP connections.  However, since the macs are now reading a 3270
stream rather than a vt102 stream, they no longer have the printer
escape capability.  Having grown used to it they would rather not leave
it.  Does anyone know of a product (preferably for the mac, SW or
commercial) that would allow a VM/CMS user to print to a local mac
printer.  Sorry but the WMAC/RMAC stuff for Brown tn3270 is not workable
because the CMS programs only run in 370 mode; we use XA mode.
-- 
Darren Dunham                 		          add@is.rice.edu
MicroConsultant                       		  Rice University
(What is that? A small consultant?)	              Houston, TX
Any resemblance between real opinions and my post is coincidental

-----------[000264][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 03:00:30 GMT
From:      romig@cse.unl.edu (Phil Romig)
To:        comp.protocols.tcp-ip
Subject:   Nonblocking socket blocking on write


A question about writing to non-blocking sockets;

I'm working on a fairly straight forward single server/multiple client
system.  I'm using the socket layer with the tcp/ip protocol; stream mode.
All I/O needs to be done asynchronously. 

To this end I open the sockets in a very straight forward manor, make
them non-blocking with fcntl flags (FNDELAY | FASYNC), and then
trap for SIGIO.

The trouble seems to come when I make several send() calls in quick
succession.  Because the sockets are set for no-blocking I use select()
to see if the socket I want is ready for writing.  When I call send()
several times very quickly with small (aprox 4096 bytes) buffers, after the
fourth or fifth send the socket select() statement reports that the
socket I want is blocked.  It remains blocked for between 5 and 10
seconds, and then allows four or five more quick sends.

Currently I have select() set to wait up to 10 seconds and then report a
failure to me, and that seems to work 99% of the time, but I do have very
long delays.

My questions: Is this normal behavior?  I assume not, where are some common
bottle necks?

If it matters I'm using a sun4 and g++ version 2.2.2.

I apologize for bothering you and I thank you very much for your time. 
-PHIL-

romig@cse.unl.edu
University Of Nebraska, Lincoln
Lincoln, NE   68588
USA

-----------[000265][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 1992 15:51:30 -0700
From:      mcitron@phad.hsc.usc.edu (Mark Citron)
To:        comp.protocols.tcp-ip
Subject:   tektronix 4014 emulation

Can anyone suggest a source (pd would be best) for a Tektronix 4014 
emulation that runs under DOS and connects via FTP's tcp/ip and also
also async over a serial line (not at the same time of course).

Next best would be a 4014 emulation that runs under SCO unix over 
tcp/ip (I suspect that must be pretty rare).

Thanks for any suggestions.
Mark Citron
Childrens Hospital Los Angeles

-- 
Mark Citron
mark@neurosci.usc.edu


-----------[000266][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 04:58:35 GMT
From:      fff@microplex.com (Fred Fierling)
To:        comp.periphs.printers,comp.protocols.tcp-ip
Subject:   Re: TCP/IP on printers

In article <1992Sep19.183126.9079@research.nj.nec.com> shm@syl.nj.nec.com writes:
>When people say a printer supports TCP/IP (or complain about the
>lack of support), what do they really mean?  Is it that the
>printer has a full tcp/ip stack running on it along with some
>application on top of it?  Assuming that the printer is stand-alone

Our TCP/IP Printer Adapters provide a tcp/ip stack including lpd, rsh, telnet,
ping arp, and rarpd services.  Although the unit isn't capable of spooling
jobs it can maintain a queue of blocked tcp connections.  This allows it
to service multiple clients on a first come first serve basis without a hard
disk.  Three methods of sending jobs are possible:

1) edit the /etc/printcap file on a BSDish system to treat the adapter like
   any other BSDish machine.

2) execute a remote shell command on any host:

     rsh m200hostname lp -dqprn < filename

3) open a socket to a port number which corresponds to the port you wish the
   data to go to.  For instance you could:

    telnet m200hostname 4400

   and everything you typed would come out the parallel port.
-- 
Fred Fierling    fff@microplex.com      Tel: 604 875-1461   Fax: 604 875-9029
Microplex Systems Ltd   265 East 1st Avenue   Vancouver, BC  V5T 1A7,  Canada

-----------[000267][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 13:21:40 GMT
From:      dcarr@gandalf.ca (Dave Carr)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a way to encapsulate X.25 over TCP/IP ????

In <6051@mtecv2.mty.itesm.mx> egonzale@netmon.mty.itesm.mx (Ing. Ernesto Gonzalez) writes:

>Again, how can i encapsulate X.25 traffic over TCP/IP????

Do you want incapsulation or routing of X.25 traffic over TCP.  We
make a protocol converter X.25-TCP-LAT...
-- 
Dave Carr               | dcarr@gandalf.ca       | It's what you learn, 
Principal Designer      | TEL (613) 723-6500     | after you know it all,
Gandalf Data Limited    | FAX (613) 226-1717     | that counts. 

-----------[000268][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 13:32:55 GMT
From:      paul@csir.co.za (Paul Nash)
To:        comp.protocols.ibm,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   TN/3270 with HLLAPI interface


We are about to change our IBM - LAN gateway from Novell's SNA Gateway
software to using a pair of Sun Sparc 10 machines, running Sunlink/3270.
This, using CUTCP or TN3270 will cover most of our needs.  However, we
have a few PC applications that require access to HLLAPI on the PC,
most notably Software AG's "Natural Connection".

Is there any commercial product that will offer us such a facility?  
Ideally, we would want something that we can load, in the same sort of
way as PC/TCP, and ignore until needed bythe application.

We thought that another way to do this would be to hack CUTCP's microTN
to support HLLAPI when it has been TSR-ified.  The problem is just that
to do this, we need the source code.  Is source available?  If it is, we
will give the hacked HLLAPI-ified source back to Clarkson for use by the
net.community on the whole.

If there is any other way to achieve this that anyone knows about,
please let me know aswell.
--
 ---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---
 Paul Nash                                   (voice) +27-12-8413050
 Network Services, CSIR Infotek                (fax) +27-12-8414109

        "OS/2 will remain the operating system of the future"

-----------[000269][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 14:00:21 GMT
From:      jhelberg@nl.oracle.com (Joost Helberg)
To:        comp.protocols.tcp-ip,comp.unix.aix,comp.unix.internals
Subject:   Re: Ethernet Collisions (?)

X-Disclaimer: This message was written by an unauthenticated user
              at Oracle Corporation.  The opinions expressed are those
              of the user and not necessarily those of Oracle.

brent@uis.com (Brent Engelbert) writes:
: We are testing some code on an IBM RS/6000 that looks at ethernet
: collisions, however we are having some problems generating any collisions.
: 
: 1) How do we generate collisions on the ethernet?
: 
: 2) We are currently looking at the ifnet structure in the kernel to
:    gather network statistics.  Is this the proper place to be looking?  On
:    our machine the if_collisions value is always 0, but the rest of the
:    fields seem to have the correct information in them.
: 
: We would appreciate any help that any one can give us.  Please E-mail
: responses to us.

Just remove the terminator from the ethernet cable for 2 or threee seconds and
put it back not too direct (i.e. be sure to make contact and loose it a few
hundred times during fitting it).

You can also add or remove a station to/from the network, that might cause 
collisions too.

Using a generator which generates a lot of 1 bytes packets on the net causes
collisions too.

Due to the algorithms used in a ethernetprocessor it is not easy to 
generate collisions on demand.

--  
   Joost Helberg                                Rijnzathe 6
   jhelberg@oracle.nl                           NL-3454 PV De Meern
   jhelberg@nl.oracle.com                       The Netherlands

   Oracle Europe BV                             Product Line Development	
   Phone: +31 3406 94211                        Fax:   +31 3406 65609

-----------[000270][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 14:43:29 GMT
From:      mgic <mgic@mixcom.mixcom.com>
To:        comp.protocols.tcp-ip
Subject:   what does completion of write() really mean?


When a process sends data to another process using TCP,
what does the completion of a blocking write() on a socket mean?
The two processes are running on different machines.

1) Data has been transferred from one TCP kernel to the other
   but may not have been read by the other process

2) Data has been transferred from one TCP kernel to the other
   *and* read by the other process

3) Data has been queued for transfer by the kernel, but
   all data may not have been moved to the other kernel yet

Or what? Thank you.

Dean
-- 

-----------[000271][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 16:24:47 GMT
From:      mgic <mgic@mixcom.mixcom.com>
To:        comp.protocols.tcp-ip
Subject:   SLIP driver Vendors Wanted for SPARC and HP machines


I've been asked to quickly find and recommend a UNIX machine
on which to run many concurrent SLIP connections. The connections
will not be through physical serial ports, but ptys (connected
to X.25).

The SLIP driver must have the ability to support up to 
128 concurrent sessions (maybe more).

The SLIP driver must have the ability to be started when
a session is needed (when a login occurs) and to go down
when a session ends so that a pty can be used for purposes
other than SLIP.

The SLIP driver must have the ability to be assigned an
IP address when it is started. (I also then need to deallocate
the address when the driver ends.)

The systems I am considering are SPARC and the new HP machines.
(I don't yet know if they already come with a SLIP driver or,
if so, if it can handle my needs.) The machine also requires
an X.25 interface.

Recommendations on UNIX machines or SLIP vendors anyone?
Thank you for your help.

Dean
-- 

-----------[000272][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 17:17:58 GMT
From:      erick@demorgan.uwaterloo.ca (Erick Engelke)
To:        comp.protocols.ibm,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: TN/3270 with HLLAPI interface

> paul@csir.co.za (Paul Nash) writes:
>
>We are about to change our IBM - LAN gateway from Novell's SNA Gateway
>software to using a pair of Sun Sparc 10 machines, running Sunlink/3270.
>This, using CUTCP or TN3270 will cover most of our needs.  However, we
>have a few PC applications that require access to HLLAPI on the PC,
>most notably Software AG's "Natural Connection".
>
>Is there any commercial product that will offer us such a facility?  
>Ideally, we would want something that we can load, in the same sort of
>way as PC/TCP, and ignore until needed bythe application.
>
>We thought that another way to do this would be to hack CUTCP's microTN
>to support HLLAPI when it has been TSR-ified.  The problem is just that
>to do this, we need the source code.  Is source available?  If it is, we
>will give the hacked HLLAPI-ified source back to Clarkson for use by the
>net.community on the whole.
>
> ---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---
> Paul Nash                                   (voice) +27-12-8413050
> Network Services, CSIR Infotek                (fax) +27-12-8414109

I can't pass up this opportunity.  Kent Fitch, also of CSIR, has just
written a TSR based tn3270 with HLLAPI using WATTCP.  He's probably
looking for testers, and he's in your timezone as well as your organization.


Erick
-- 
----------------------------------------------------------------------------
Erick Engelke				 	            WATTCP Architect
erick@development.uwaterloo.ca     TCP/IP was easy but i still can't work VI

-----------[000273][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 18:22:32 GMT
From:      pascoe@rocky.gte.com (Dave Pascoe)
To:        comp.protocols.tcp-ip
Subject:   BOOTP server for SunOS?


Is there a working BOOTP server out there for SunOS?  I'm looking for
a BOOTP server that will work with the various PC client SLIP
implementations.  Specifically, I want to be able to dial in from the
PC (using SLIPDIAL or some other package) and obtain an IP address
automatically.  I have SLIP-4.1 running on a SPARC-2 but everything is
manual (e.g., log in, run slip-attach, etc.) and I'm looking to
automate things as much as possible.  

I'm eager to learn how others are accomplishing this on Sun machines.

--
Dave Pascoe
Internet: pascoe@rocky.gte.com
GTE/MSED - Needham Heights, MA
(617) 455-5704

-----------[000274][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 18:45:05 GMT
From:      carlson@steam.Xylogics.COM (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: what does completion of write() really mean?

[...]
>3) Data has been queued for transfer by the kernel, but
>   all data may not have been moved to the other kernel yet
[...]

#3 is close.  Closer would be:

4.  All data have been enqueued for transfer by the kernel, and may
actually have been sent out the network, but this is not guaranteed. 
All that is guaranteed is that sending will commence "soon".  Nothing is
guaranteed about when this will happen, or about the peer's reception or
delivery of the data.

.//.

-----------[000275][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 1992 19:52:56 GMT
From:      pvg@po.CWRU.Edu (Paul V. Gerwe)
To:        comp.protocols.tcp-ip
Subject:   Telnet server that runs on a PC?


Greetings.

I am looking for a telnet server that will run on a PC and that is 
capable of running Dos applications simultaneously.  If you have 
any information, please send me mail at pvg@po.cwru.edu.  Thanks!

Hat

-----------[000276][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 20:19:46 GMT
From:      bharvey@gandalf.ca (Brian Harvey)
To:        comp.protocols.tcp-ip
Subject:   genric print spooler wanted

Can anybody help me find a source for a generic print spooler which
will redirect a print job to a network connected printer. Specifically a 
printer connected via a terminal server which does not have a resident 
spooler itself.

Thanks,

Brian Harvey 
harvey@gandalf.ca

-----------[000277][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 20:41:04 GMT
From:      richb@kronos.com (Rich Braun)
To:        comp.protocols.tcp-ip
Subject:   CMU/MIT PC/IP source code

[I'm posting this on behalf of someone who does not have access to
netnews.]

Where might one find the CMU PC/IP source code for MS-DOS?  I know it
used to be on husc6.harvard.edu, but that machine has been decommissioned.

-rich

-----------[000278][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 21:28:30 GMT
From:      nick@toro.MTS.ML.COM (Nicholas Jacobs)
To:        comp.protocols.tcp-ip
Subject:   Question on using Wellfleet and NetBlazer together

Hi. We're looking to use some Telebit NetBlazers as backups for Wellfleet LN's
and FN's. As a test, we are trying to set up the following. Please note that
this is on a completely internal IP network. My main question is whether or
not other people have set up something similiar or is this one of those 
situations that requires radical changes to implement (i.e., addressing
changes?). 


            | 146.125.XX.XX (255.255.255.192)
            |
            |
	+--------+
	| WF LN  | 146.125.50.XX     +------------+
	|        +-------------------+ Enet conc. |
	|        | (255.255.255.0)   +--+---------+
	+---+----+                      |
            | 44.XX.XX.XX               |
            | (255.0.0.0)               |
            |                 +---------+--+
            |                 + NetBlazer  +
            |                /+------------+
            |               /
	+------------+     /
	| Enet conc. |    / Dialup connection using WorldBlazer modems
	+------------+   /
            |           /
            |          /
	+------------+/
	| NetBlazer  +
	+------------+

The WF LN, using RIP, broadcasts that it know 146.125.XX.XX. It can 
successfully route packets from 44.XX.XX.XX to 146.125.50.XX. But, if
we disable the 146.125.50.XX interface on the WF LN (using the disable
command), a ping from 44.XX.XX.XX to 146.125.50.XX gets ICMP re-directs.
The WF LN does not send out a RIP update indicating that its 146.125.50.XX
metric has gone to infinity so that NetBlazer attached to 44.XX.XX.XX
never tries to dial up the other NetBlazer attached to 146.125.50.XX.


The software on the WF LN is 5.72. The NetBlazer software is 1.51x11-test.
Thanks in advance, please email responses. I'll summarize if there is
sufficient interest.

Thanks,

Nicholas Jacobs
Merrill Lynch Municipal Analytics & Systems, NY
nick@MAS.ML.COM, uunet!toro!nick, (212) 449-1455
I have something to say! It's better to burn out than to fade away!
My statements are my own (not Merrill Lynch's). Any resemblance to a reality
that you know is purely coincidental.

-----------[000279][next][prev][last][first]----------------------------------------------------
Date:      25 Sep 92 22:09:33 GMT
From:      donp@novell.com (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: what does completion of write() really mean?

In article <1992Sep25.144329.17680@mixcom.com> mgic <mgic@mixcom.mixcom.com> writes:
>When a process sends data to another process using TCP,
>what does the completion of a blocking write() on a socket mean?
>...
>Or what? Thank you.

It means, "I'm done using the data buffer you just gave me.  You may
reuse it for something else now."  No more, no less.  If the write
returns successfully, it also implies that TCP will make its best
effort to deliver that data to the other end of the pipe or die
trying.  It gives you no information whatsoever about where the data
is at the moment.
						don provan
						donp@novell.com

-----------[000280][next][prev][last][first]----------------------------------------------------
Date:      Fri, 25 Sep 1992 23:27:35 GMT
From:      Tom Davies <tomd@horse.demon.co.uk>
To:        comp.protocols.tcp-ip
Subject:   What RFCs describe NNTP?

I'm trying to write an NNTP server. I have RFC977, which seems to be the 
definitive definition of NNTP. However on connecting to the news reader 
(Nuntius) it issues an XHDR command, which I can't find in RFC977.

Is there another RFC that I should be reading?

Thanks in advance

Tom

Tom Davies - tomd@horse.demon.co.uk

-----------[000281][next][prev][last][first]----------------------------------------------------
Date:      26 Sep 92 13:28:59 GMT
From:      paul@tivoli.UUCP (Paul Greenspan)
To:        comp.protocols.tcp-ip
Subject:   SVR4 TLI: Bug or Feature?

Is the following a bug in the TLI implementation on SVR4 or just something
that I haven't found in the documentation?

When trying to t_bind() to a port that is already used (bound) instead
of the call failing and setting errno to EADDRINUSE, it succeeds but gives
you a different port than the one that you requested.  In addiditon this new
port is not a "reserved" port like the one that I requested.

I didn't expect this and the man page for t_bind() does not document a
t_error value that looks like it would describe this situation.

Any assistance would be greatly appreciated.

Paul

-- 
Paul Greenspan              paul@tivoli.com 
Tivoli Systems              6034 West Courtyard, Suite 210
Austin, Texas  78730        (512) 794-9070  /  FAX (512) 794-0623

-----------[000282][next][prev][last][first]----------------------------------------------------
Date:      Sat, 26 Sep 1992 19:53:06 GMT
From:      mike@uunet!tellab5!odgate (Mike J. Kelly)
To:        comp.databases.sybase,comp.databases,comp.sys.novell,comp.databases.theory,comp.unix.aix,comp.windows.misc,comp.protocols.tcp-ip
Subject:   Re: Sysbase/Aix/TCP/IP/Windows

susan@moo.org (Susan Dean) writes:

> 
>I have a question that maybe someone can help me with. We are working in a
>client/server environment with Novell 3.1, Novell LAN WP,TCP/IP (ethernet), 
>Sybase and Power Builder. Sybase is on
>a RISC 6000 V 3.1.6 of AIX. After running a Power-Builder Application on a
>386 running windows 3.1, we sometimes receive the following error messages:
 
>Write to SQL server fail
>Db process dead or not enable
 
>The errors are being reported from the Sybase Netlib DLL's on the client.
 
> The RISC Sybase server will sometimes report a error # 1608
>"Network error was encountered while sending results to front end"
 
>But then at other times
>the program will run without errors.

Sounds like a Net-Library bug in the Novell LAN WorkPlace Net-Library for
Windows.  We have found similar problems in both the FTP TCP/IP and Wollongong
PATHWAY Net-Library and gotten patches from Sybase.  What seems to happen is
some TCP/IP timeout or error condition causes the windows net-library to 
shut down (hence the Write to SQL Server failed).

Call Sybase Tech Support with this problem.  Be insistent, we had to be to
get the other Net-Libraries patched.  We eventually had to send them a test
program demonstrating the problem which they could run in Emeryville to 
reproduce the problem there.
-- 
--
Mike Kelly               Odesta Corporation, Northbrook, Illinois, USA
...!clout!odgate!mike    - Until odesta.com is registered.
odgate!mike@clout.uucp   - From the Internet.

-----------[000283][next][prev][last][first]----------------------------------------------------
Date:      26 Sep 92 22:13:18 GMT
From:      tching@target.water.ca.gov (Tracy Ching <SysAdmin>)
To:        comp.protocols.tcp-ip
Subject:   --- Help... Source code for TCP/IP... Tech specs ---


	I need help with a lot of things about TCP/IP.  I am quite
new to this IP world but would like to get up to speed.  Please help.
I want to embed TCP/IP into a small board with telnet and a serial
port for the dumb term.
1) Where can I find code to "burn" into ROM for the TCP/IP stuff?
	(packet drivers, etc... whatever else I will need.  please
	tell me)
2) Where can I find code for telnet?
3) Is there an ftp site with all the tech specs for TCP/IP (RFC's)?
4) Anything else I am overlooking?  Please let me know...
tching@water.ca.gov

-----------[000284][next][prev][last][first]----------------------------------------------------
Date:      Sun, 27 Sep 1992 04:46:46 GMT
From:      csehm@knuth.mtsu.edu (Mr. Erik Moe)
To:        comp.protocols.tcp-ip
Subject:   Fusion and MS-Kermit (TCP/IP)

Has anyone experinced trouble or even tried to get MS-Kermit to talk TCP/IP
with Network Research Corp. "Fusion" product.  I'm tring to get Kermit
to talk to a Hitachi V90/25 running System III.  Fusion has been ported to
the V90.  I tried, and finally gave up, switching to NCSA Telnet.
However, I would like to get Kermit to work because Kermit has a script
language and NCSA does not.  I found two sentences in the Kermit
documentation that reported that there was a problem with Kermit and
Fusion for VAX/VMS.  I called Columbia university, but they were unable
to elaborate.  I also talked to Network Research Corp. They informed me
that the copy of Fusion I was running had been licenced to a third party
in Japan, and as far as I can tell isn't even a complete port.  I believe
that any chances of doing anything with the Fusion side are nil.  If
something could be done it would have to be with MS-Kermit.

Thanks in advance!

Erik Moe
csehm@knuth.mtsu.edu

-----------[000285][next][prev][last][first]----------------------------------------------------
Date:      27 Sep 92 08:08:10 GMT
From:      paul@frcs.Alt.ZA (Paul Nash)
To:        comp.protocols.ibm,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   TN3270 with HLLAPI needed for DOS machines on IP network


We are busy changing our LAN - MVS interconnection from Novell SNA gateway
products (version 1.x is very buggy and unstable, and Netware for SAA
requires a mega-back upgrade to Netware 3.x on 50-odd fileservers).  We 
have decided to use a pair of Sun Sparc-10 machines, with token-ring
adapters into a Mohawk MC400 (3174 clone), and thence the machine.

However, the MIS boyos want to use a product (Software AG's Natural
Connection) on the users' PCs that requires HLLAPI to talk to the host.
Does anyone know of a commercially available (or shareware/freeware)
TN3270 that will TSR-ify itself and present an HLLAPI v3 interface to
the DOS box?  Alternatively, does anyone know whether the source code 
for CUTCP's "microTN" is available for hacking?  Or the e-mail address
of anyone at Clarkson who might be able to make it available?

If we have to, we will write our own interface from scratch (or based
on NCSA's telnet, plus the Unix tn3270).  If we have to do this, it'll
end up being published in alt.sources or some such forum, but I daresay
that this will take more than five minutes -- say three days :-) :-).

Any info welcome, by mail, or FAX.  I fear that snail-mail will take too
long (the Sun machines should be arriving next week).  Please reply to

	Paul Nash	<paul@tantrum.csir.co.za>
	Fax:	intl + 27 + 12 + 841-4109
	Voice:	intl + 27 + 12 + 841-3050

Unfortunately, our newsfeed gets a trifle erratic at times (like goes
down for a few days at a time, as the entire country is at the far end
of a single 9600bps voice-grade line), so I would appreciate mail.  I
_will_ summarise, though, if there is any interest at all.
 
 ---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---=---
 Paul Nash                                         paul@frcs.Alt.ZA
 Box 12475, Onderstepoort, 0110 South Africa         +27-12-5611879

 "Standards are great -- everyone should have one" (Kent Landfield)

-----------[000286][next][prev][last][first]----------------------------------------------------
Date:      27 Sep 92 08:47:30 GMT
From:      nelson@crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   CMU/MIT PC/IP source code

In article <1992Sep25.204104.21108@kronos.com> richb@kronos.com writes:

   Where might one find the CMU PC/IP source code for MS-DOS?  I know it
   used to be on husc6.harvard.edu, but that machine has been decommissioned.

hsdndev.harvard.edu

-russ <nelson@crynwr.com>  I'm proud to be a humble Quaker!
Crynwr Software            Crynwr Software sells packet driver support.
11 Grant St.               315-268-1925 Voice
Potsdam, NY 13676          315-268-9201 FAX

-----------[000287][next][prev][last][first]----------------------------------------------------
Date:      Sun, 27 Sep 1992 12:41:15 GMT
From:      pwilson@world.std.com (Pete Wilson)
To:        comp.protocols.tcp-ip
Subject:   New Product: Universal SNMP Agent

Of possible interest: Paul Freeman Assoc. announces the Universal SNMP Agent: 
a portable, compact, cost-effective, fully-compliant, multi-MIB Agent for 
installation under any modern OS on any platform. For a product description, 
send e-mail to pwilson@world.std.com or call 508-692-4436.

-----------[000288][next][prev][last][first]----------------------------------------------------
Date:      Sun, 27 Sep 1992 15:20:10 GMT
From:      mgic <mgic@mixcom.mixcom.com>
To:        comp.protocols.tcp-ip
Subject:   TCP and reliable transaction processing?


How can TCP be used for *reliable* transaction processing?
I don't see anything in TCP that will inform the sender of data that
the receiver did not receive the data. For example, the completion
of a write() statement does not mean data was transferred through
a network and delivered to the remote process. (The data might be
in the other process' TCP buffers, but not read by the process.)

Using TCP, how can an application-level protocol 
be designed for synchronous communications? 

	The client must know that the server has received a record.

	The server must know that the client knows that it has 
	received the record.

Simply having the server send an application-level "ACK" does not completely
solve the problem, because the server does not know if the ACK was
read by the client process.  Thank you for the help you may provide.

Dean
-- 

-----------[000289][next][prev][last][first]----------------------------------------------------
Date:      Sun, 27 Sep 1992 21:33:28 GMT
From:      stevea@i88.isc.com (Steve Alexander)
To:        comp.protocols.tcp-ip
Subject:   Re: SVR4 TLI: Bug or Feature?

In article <4519@tivoli.UUCP> paul@tivoli.com (Paul Greenspan) writes:
>When trying to t_bind() to a port that is already used (bound) instead
>of the call failing and setting errno to EADDRINUSE, it succeeds but gives
>you a different port than the one that you requested.  In addition this new
>port is not a "reserved" port like the one that I requested.

This is the specified behavior for t_bind(3N).  The man page states:

"If the requested address is not available, or if no address is specified in
req (the len field of addr in req is zero) the transport provider may assign
an appropriate address to be bound, and will return that address in the addr
field of ret.  The user can compare the addresses in req and ret to determine
whether the transport provider bound the transport endpoint to a different 
address than that requested."

So, you're stuck with this behavior if you use TLI.

-- 
Steve Alexander, Software Technologies Group    | stevea@i88.isc.com
INTERACTIVE Systems Corporation, Naperville, IL | ...!{sun,ico}!laidbak!stevea

-----------[000290][next][prev][last][first]----------------------------------------------------
Date:      Mon, 28 Sep 1992 00:44:22 GMT
From:      bob@MorningStar.Com (Bob Sutterfield)
To:        comp.protocols.tcp-ip
Subject:   Re: Internet routing software for Sun

In article <1694@eagle.ukc.ac.uk> mr3@ukc.ac.uk (M.Rizzo) writes:
   We hope to join the Internet soon and to start off we're looking
   for routing software to run on a SUN Sparcstation II.

It's not always necessary for hosts to participate in the routing
protocol fabric of the network.  You may be able to get by with static
routes pointing at the nearest router along the path outward from your
Sun to the rest of the world.

If you're using RIP between the routers on your network, then you can
use the Route Daemon (in.routed), which ships with SunOS.

If you're using some other routing topology protocol, then you'll
probably do well running the Gate Daemon on your Sun.  Get
gated.cornell.edu:pub/gated/gated-2.1.tar.Z and the three accompanying
patches, or get gated-alpha.tar.Z.

-----------[000291][next][prev][last][first]----------------------------------------------------
Date:      Mon, 28 Sep 1992 08:50:59 GMT
From:      pang@Xenon.Stanford.EDU (Swee-Chee Pang)
To:        comp.protocols.tcp-ip,comp.sys.sun.admin
Subject:   CISCO CGS routing problem/question

I have several CISCO CGS routers that joins a few Sun networks
together.
For several of those networks, I require a netmask of
255.255.255.192 (128+64) so I can divide the network further by those 2 bits.
Unfortunately, I dont seem to be able to do that with the CISCO CGS.
The hardware version is 8.3(3), and software is 4.4(2).
Anyone ever used a CISCO router with the required routing changes?

Thanks very much.

--
Pang SweeChee
pang@xenon.stanford.edu

-----------[000292][next][prev][last][first]----------------------------------------------------
Date:      Mon, 28 Sep 1992 16:24:13
From:      jbvb@vax.ftp.com  (James B. VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: what does completion of write() really mean?

In article <1992Sep25.144329.17680@mixcom.com> mgic@mixcom.mixcom.com (mgic) writes:

    When a process sends data to another process using TCP,
    what does the completion of a blocking write() on a socket mean?

The answer depends on the implementation.  In PC/TCP for DOS, it means
that all the data has been transmitted, but not necessarily acknowleged
by the receiving TCP.  If you really need to know when the other process
has actually read (and processed) the data, you need to build a session-
like protocol layer on top of TCP...

James B. VanBokkelen		2 High St., North Andover, MA  01845
FTP Software Inc.		voice: (508) 685-4000  fax: (508) 794-4488


-----------[000293][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 1992 23:33:08 -0700
From:      tli@skat.usc.edu (Tony Li)
To:        comp.protocols.tcp-ip,comp.sys.sun.admin
Subject:   Re: CISCO CGS routing problem/question

In article <1992Sep28.085059.28725@CSD-NewsHost.Stanford.EDU> pang@Xenon.Stanford.EDU (Swee-Chee Pang) writes:
    For several of those networks, I require a netmask of
    255.255.255.192 (128+64) so I can divide the network further by those 2 bits.
    Unfortunately, I dont seem to be able to do that with the CISCO CGS.
    The hardware version is 8.3(3), and software is 4.4(2).
    Anyone ever used a CISCO router with the required routing changes?
    
Works fine for me:

!
interface Ethernet 1
ip address 128.125.1.1 255.255.255.192

How about some more details?

-- 
Tony Li - Escapee from the USC Computer Science Department          tli@usc.edu
		       The net is not what it seems.

-----------[000294][next][prev][last][first]----------------------------------------------------
Date:      Mon, 28 Sep 1992 15:32:17 GMT
From:      vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
To:        comp.protocols.tcp-ip,comp.benchmarks
Subject:   Re: TCP/IP Benchmarks

In article <1992Sep28.101206.9283@uni-paderborn.de>, caro@uni-paderborn.de (Carsten Rose) writes:
> ...
> I don't know if 'ttcp' is a real 'Benchmark' , but we use it often
> to test our performance on Ethernet and FDDI. It use a socket connection
> (STREAM or DGRAM) to test.
> 
> You can get it from:
> 	
> 1) sgi.sgi.com  : look in ~ftp/sgi/src/ttcp/
> 2) ugle.unit.no : look in ~ftp/pub/unix/network/ttcp
> 
> The second source can open sockets with user defineable buffer sizes.


The primary source for ttcp is brl.mil.  BRL evidently does not
consider it a benchmark, but a file transfer protocol.  That's
why they insisted on keeping the crazy -s default.

There is not much more you can or should do in a single-virtual-circuit
TCP benchmark than is done by ttcp.  Some vendors in the past cooked up
some real doozies of so called benchmarks, which is why I like ttcp.
Ttcp at least has non-commercial origins.

Other vendors like other derivatives of ttcp.  One is `nettest`, which
runs several TCP virtual circuits at once.  I do not like them because
it is a lot easier to get bigger numbers with them.  They hide the
effects of the scheduling (including interrupt) latency in the
systems.  (For various reasons, this effect is especially evident on
half-duplex media like FDDI.)  I think Cray likes nettest.  UltraNet,
ultra.com, likes another varient, but I don't remember if it single- or
multiple-virtual-circuit.

All reasonable versions of ttcp all you to specifiy the user buffer
size, and many, perhaps most, allow you to specify the "socket buffer"
(i.e. TCP window) size.  The brl.mil and sgi.com versions do that
with -l and -b.  There is also a way to change the alignment of the
buffers, which is very important in most systems.

We ship source for ttcp; anyone with an IRIS has a copy.  I think we're
going to ship source for nettest in a future release.  I think all
machines should come with source for a reasonable set of benchmarks.
Let the buyers do their own tests, and give them enough information to
make their own judgements about the benchmarks.  Avoid benchmarks from
benchmark companies or system vendors that magically produce mystical
numbers, but are accompanied with a salescritter who waves sacred
insense burners around the benchmark.


Vernon Schryver,  vjs@sgi.com

-----------[000295][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 15:46:13 GMT
From:      ken@comanche.sbi.COM (Ken Jones - Consultant)
To:        comp.protocols.tcp-ip
Subject:   Re: Nonblocking socket blocking on write

In article <1992Sep25.030030.8755@unlinfo.unl.edu>, romig@cse.unl.edu (Phil Romig) writes:
|> 
|> A question about writing to non-blocking sockets;
|> 
|> I'm working on a fairly straight forward single server/multiple client
|> system.  I'm using the socket layer with the tcp/ip protocol; stream mode.
|> All I/O needs to be done asynchronously. 
|> 
|> To this end I open the sockets in a very straight forward manor, make
|> them non-blocking with fcntl flags (FNDELAY | FASYNC), and then
|> trap for SIGIO.
|> 
|> The trouble seems to come when I make several send() calls in quick
|> succession.  Because the sockets are set for no-blocking I use select()
|> to see if the socket I want is ready for writing.  When I call send()
|> several times very quickly with small (aprox 4096 bytes) buffers, after the
|> fourth or fifth send the socket select() statement reports that the
|> socket I want is blocked.  It remains blocked for between 5 and 10
|> seconds, and then allows four or five more quick sends.

We had to build a similar system for sending real-time data around
trading floors on wall street.  Delays of 5 - 10 seconds are all that
unreasonable, although we avoided the FASYNC mode.

If you need your server to keep going, while one of your clients is
blocked you need to add another layer of buffering.  When your client
socket blocks, send the message to a buffer (file, memory, whatever is
most appropriate).  Then add that client to your main select for writing.
When it wakes up, continue writing out of your buffer.  If you have a new
message while the client is still blocked, write it to the end of the buffer.
When you get to the end of buffer, you can clean up.

This works for us, since all of our information is bursty.
-- 
+-----------------------------------------------------------------------+
|   Ken Jones                                                           |
|   Expert Database Systems, Inc.                                       |
|   ken@mr_magoo.sbi.com              "Where the future begins tomorrow |
+-----------------------------------------------------------------------+

-----------[000296][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 17:15:37 GMT
From:      rlk@VisiCom.COM (Bob Kitzberger)
To:        comp.realtime,comp.protocols.tcp-ip
Subject:   List of embedded TCP/IP products?

Several months ago, one of these two newsgroups had a posting
of available TCP/IP implementations for embedded systems.

Can anyone email me the list?

Thanks,

	.Bob.
----------------
Bob Kitzberger          VisiCom Laboratories, Inc.
rlk@visicom.com         10052 Mesa Ridge Court, San Diego CA 92121 USA
                        +1 619 457 2111    FAX +1 619 457 0888

-----------[000297][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 17:35:50 GMT
From:      jordan@IMSI.COM (Jordan Hayes)
To:        comp.protocols.tcp-ip
Subject:   Re: Nonblocking socket blocking on write

Phil Romig <romig@cse.unl.edu> writes:

	When I call send() several times very quickly with small (aprox
	4096 bytes) buffers, after the fourth or fifth send the socket
	select() statement reports that the socket I want is blocked.
	It remains blocked for between 5 and 10 seconds, and then
	allows four or five more quick sends.

The other end is not reading fast enough.

/jordan

-----------[000298][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 18:15:17 GMT
From:      tep@tots.Logicon.COM (Tom Perrine)
To:        comp.protocols.tcp-ip
Subject:   Where is the "Big-Internet" mailing list?


I have heard that there is a mailing list devoted to the problems with
"exhausting" the current 32-bit IP address space, as well as other
issues concerning a global Internet.

I have been unable to find this list in the List of Publically
Available Mailing Lists, where I lookef for Big-Internet and other
variations.

Does it exist?  What is it named.  Who do I contact?

Thanks,


Tom E. Perrine (tep)     | tep@Logicon.COM       |Voice: +1 619 597 7221
Logicon, Inc.            | sun!suntan!tots!tep   |  or : +1 619 455 1330
4010 Sorrento Valley Blvd|                       |  FAX: +1 619 552 0729
San Diego CA 92121-1498  |Been there, done that, bought the T-shirt...

-----------[000299][next][prev][last][first]----------------------------------------------------
Date:      Mon, 28 Sep 1992 18:57:09 GMT
From:      GJAMES@HUSKY1.STMARYS.CA (GJames)
To:        comp.protocols.tcp-ip
Subject:   packet capture

I used to have a dos program whick allowed me to capture packet off of my 
ethernet. In haste I deleted the files during my last house cleaning.

Could someone tell me where I could find it again.

-----------[000300][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 19:58:48 GMT
From:      k@hprnd.rose.hp.com (Steve Kao)
To:        comp.protocols.tcp-ip
Subject:   Re: BOOTP server for SunOS?

Dave Pascoe:
> Is there a working BOOTP server out there for SunOS?  I'm looking for
> a BOOTP server that will work with the various PC client SLIP
> implementations.  Specifically, I want to be able to dial in from the
> PC (using SLIPDIAL or some other package) and obtain an IP address
> automatically.  I have SLIP-4.1 running on a SPARC-2 but everything is
> manual (e.g., log in, run slip-attach, etc.) and I'm looking to
> automate things as much as possible.  
 
> I'm eager to learn how others are accomplishing this on Sun machines.

The media option AAV for the HP JetDirect for unix cards have CMU's
bootp on them.  We've successfully used bootp to configure our LaserJet
printers for many months now.  If you know how to configure inetd.conf
and bootptab, then the same bootp can probably be used to boot your PC
clients.

- Steve Kao

-----------[000301][next][prev][last][first]----------------------------------------------------
Date:      28 Sep 92 20:48:31 GMT
From:      jim@applix.com (Jim Morton [ext 237])
To:        comp.protocols.tcp-ip
Subject:   need SLIP tip program

OK, I have slip-4.0 up and running in my Sun kernel, and I have my 
NetBlazer calling into it. But I need my Sun to call out to the
NetBlazer. Where is this "tip with SLIP" program that is alluded
to in the various README files???

--
Jim Morton, Applix Inc., Westboro, MA
...uunet!applix!jim    jim@applix.com

-----------[000302][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 1992 01:25:39 GMT
From:      kre@cs.mu.oz.au (Robert Elz)
To:        comp.protocols.tcp-ip
Subject:   Re: Where is the "Big-Internet" mailing list?

In <430@tots.UUCP> tep@tots.Logicon.COM (Tom Perrine) writes:

| Does it exist?  What is it named.  Who do I contact?

It exists.  Its called Big-Internet.  And Big-Internet-Request@munnari.OZ.AU
(Archives are in big-internet/list-archives on munnari.oz.au).

kre

-----------[000303][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 92 02:14:18 GMT
From:      dla@se05 (Doug Acker)
To:        comp.protocols.tcp-ip
Subject:   SLIP and the PC

I am having a heck of a time trying to get SLIP configured between my PC and
my XYPLEX terminal server.  On my XYPLEX, if its on 145.144.100.3, I can
configure my PC for 145.144.100.213  right?

On my PC, I am trying to use the SLIP8250 or ETHERSL packet drivers, but I
can't get the two machines to talk to each other.
-- 
Douglas L. Acker

Western Geophysical Exploration Products
Systems Engineer

-----------[000304][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 1992 04:35:29 GMT
From:      barmar@think.com (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP and reliable transaction processing?

In article <1992Sep27.152010.12973@mixcom.com> mgic <mgic@mixcom.mixcom.com> writes:
>	The client must know that the server has received a record.
>
>	The server must know that the client knows that it has 
>	received the record.

You can never fully solve this problem.  See any description of the "two
armies" problem.  It's usually brought up in the context of graceful
connection closing protocols (for instance, in Comer), but as you see it
has broader applicability.

>Simply having the server send an application-level "ACK" does not completely
>solve the problem, because the server does not know if the ACK was
>read by the client process.

Assuming the client doesn't send a command before receiving the server's
ACK, the receipt of a new command by the server can serve as the
acknowledgement of the ACK.  And a graceful close of the connection can
serve as a final acknowledgement.  This then degenerates to the traditional
two-army problem.
-- 
Barry Margolin
System Manager, Thinking Machines Corp.

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

-----------[000305][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 92 09:58:05 GMT
From:      dclunie@sirius.ucs.adelaide.edu.au (David Clunie)
To:        comp.protocols.tcp-ip
Subject:   talk and ntalk - RFC's anyone ?


I have been playing around with the bsd sources for talk, as I wanted
to change them around a bit to include such things as logging
converstations, providing encrypted conversations, and possible an
OpenWindows interface to talk, and I quickly discovered that there
were two talk protocols, an older one one (talk on 517/udp) and that
the sources I had talked to other machines via (ntalk on 518/udp).
Needless to say my Sun uses the older one so I can't talk to myself :).

Anyway, I am after two things -

1. the source to the older talk so that I can figure out what it does
for entertainment value and Sun compatibility,

2. An RFC that describes the talk protocol(s). I couldn't find one
of these. Maybe it doesn't exist. maybe someone wants to write one
as this protocol seems pretty universally used in its various forms.
I don't feel qualified either technically or historically to do it
myself if you know what I mean.

-- 
David Clunie (dclunie@itd.adelaide.edu.au)

-----------[000306][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 1992 12:23:38 GMT
From:      clfung@uxmail.ust.hk (Fung Chi Leung)
To:        comp.protocols.tcp-ip
Subject:   Retransmission count??

Hello,

  I am looking for some tools that are able to measure the number of
lost packets and retransmissions for a tcp connection. In additon, is 
there any program which can real-time monitor a specific tcp connection?

I have tried the 'netstat' but it doesn't help because it can't give
me the information of retransmission count and it just measure the statistic
at the instance that I invoke the command. 

Any help will be highly appreciated.

Leung

~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=

Fung Chi Leung                                 
E-Mail: clfung@uxmail.ust.hk
Phone:	358 7022
Addr:	Room 641 PG Hall, HKUST, Clear Water Bay, Hong Kong

~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=

-----------[000307][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 1992 14:22:44 GMT
From:      brady@cs.tcd.ie (Michael Brady)
To:        comp.protocols.tcp-ip
Subject:   Password capture on anonymous FTP?

Net friends, I suspect that this is a FAQ; what I need is a WKA (Well Known
Answer) :-).
I have some software being hosted on our FTP site, for access by anonymous
FTP. I'd really like to build a mailing list of the people who upload it,
and I thought that I could use the ID they are requested to enter as a
password when logging in anonymously. Does anyone know of software to do
this?

Thanks

Mike Brady
Computer Science Department, Trinity College Dublin, Ireland

-----------[000308][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 92 14:49:21 GMT
From:      PMW1@psuvm.psu.edu (Peter M. Weiss)
To:        comp.protocols.tcp-ip
Subject:   Re: Password capture on anonymous FTP?

In article <290992152140@brady.cs.tcd.ie>, brady@cs.tcd.ie (Michael Brady) says:

>Net friends, I suspect that this is a FAQ; what I need is a WKA (Well Known
>Answer) :-).
>I have some software being hosted on our FTP site, for access by anonymous
>FTP. I'd really like to build a mailing list of the people who upload it,
>and I thought that I could use the ID they are requested to enter as a
>password when logging in anonymously. Does anyone know of software to do
>this?

Don't expect that the password will be that of the requestor.  If I
were the recipient of such unsolicited mass mailing, I would be upset.

just my .02 EuroDollars

/Pete

-----------[000309][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 92 16:29:54 GMT
From:      berg@physik.tu-muenchen.de (Stephen R. van den Berg)
To:        comp.protocols.tcp-ip,comp.sys.sun.misc
Subject:   SLIP and a sample gated.conf file.

I currently have the following setup:

	A local ethernet network (two SUNs running NIS).

	A sun connected over ethernet to a CISCO router.

	A sun on the first network is connected by CSLIP to a sun on the
	 second network.

Or, to illustrate:

	le0
Sun A - 137.226.106.2 -|
		       |
		       |ethernet
	le0	       |
Sun B - 137.226.106.1 -|
  \
   \	/dev/ttya		CSLIP			/dev/ttya
    \- 137.226.218.2 ---------------------------------- 137.226.218.1 - Sun C
									 /
							 le0		/
						      |- 137.226.99.2 -/
						      |
					     ethernet |
						      |  le0
						      |- 137.226.99.1 - CISCO
									  /
							 le0		 /
						      |- 137.226.100.1 -/
						      |
				The rest of INTERNET ...

Well, so far, everything works fine, except for the SLIP line.
I have tested the serial connection and sure enough, IP-packets seem to
get across (using sliplogin on SunB and SunC).
But, anything like "ping" or telnet does not work.

Now, the question is, what would be the best way to get this working?
I thought running a gated on both SunB and SunC would do the trick.  However,
setting up gated to do this seems to be non-trivial.

So I was wondering if someone could give me some pointers or provide working
examples on how to get this running.  Someone must have done this before.

Currently, on SunB I have the following /etc/gated.conf file (this was written
by a gated novice):
-------------------------SunB /etc/gated.conf--------------
RIP yes
HELLO no
EGP no

trustedripgateways	137.226.218.1
sourceripgateways	137.226.218.1
-----------------------------------------------------------

On the other Sun (SunC), I have the following:
-------------------------SunC /etc/gated.conf--------------
RIP yes
HELLO no
EGP no

trustedripgateways      137.226.218.2 137.226.99.1
sourceripgateways       137.226.218.2 137.226.99.1
-----------------------------------------------------------

Now, shouldn't this be enough for both gateds to be able to figure out
all the rest by themselves?

Am I doing something stupid here?  Have I missed something obvious?
Something non-obvious?

Any pointers are appreciated.  I can of course figure out everything myself,
but I already noticed, there are too many permutations and the turn-around
time is way to long.

BTW, the gateds are able to communicate with each other (I have seen
traffic going across).  But apart from gated traffic, I couldn't do a thing
(i.e. ping, telnet didn't work).
-- 
Sincerely,                                  berg@pool.informatik.rwth-aachen.de
           Stephen R. van den Berg (AKA BuGless).    berg@physik.tu-muenchen.de

"And now for something *completely* different!"

-----------[000310][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 1992 17:54:09 GMT
From:      tmhoff@oogoody.Corp.Sun.COM (Todd Hoff)
To:        comp.protocols.tcp-ip
Subject:   Network Topology Deducer Needed

I'm not exactly sure what to call it, but are there any programs
out there than can determine what hosts are on a network, when
they change IP addresses, go down, come up, etc.. It was be nice
if was multi-protocol, that is work on Mac networks, etc..

Thanx for any help.



-----------[000311][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 1992 18:39:28 GMT
From:      hofer@rchland.vnet.ibm.com (Kent Hofer)
To:        comp.protocols.tcp-ip
Subject:   socket shutdown() - what's the deal

Help!

The semantics of shutdown() continue to completely elude me, even after much study/testing.  As I understand it, shutdown() allows you to shutdown the read side, the write side, or both sides of a TCP conversation.  Thus, if you shutdown the write side, a FIN will travel, and a read on the other side can complete with 0 (end of file).  If you shutdown the read side, on the other hand, no network traffic results.  Other reads on the same socket should no longer hang when no data is available, they sho




mplete with ???.  

I believe that the sematics of shutdown() also relate to multiple processes sharing the same socket.  If one of the processes has a read() or write() outstanding which is blocked, the appropriate shutdown will "break them loose" with EPIPE for write side and 0 for read side.  (BTW, do any applications actually use this particular semantic?)
  

Ok, so what's my problem?  Two things:

1) I want to know why anyone would use shutdown at all - 
 a) for the uni-directional "close" capability?
 b) for the "break em loose" capability?  (Do any socket applications actually share a socket across multiple processes?) 
 c) for some other reason that I haven't dreamed of yet?


2) What am I supposed to expect on a read(sd...) after issuing a shutdown(sd, read side)?  Please read the following two scenarios because they seem to give completely incosistent answers to this question!

   

SCENARIO 1:
	
application 1 issues shutdown(ns,read side) on its new just accepted socket descriptor (first operation on the new socket descriptor)

application 2 (the client) sends lots of data until it blocks on a send()
application 1 now issues a read(ns,...) and actually gets what seems to be a random amount of data - WHY should the read be able to get data?

SCENARIO 2:

application 2 (the client) sends lots of data until it blocks on a send()
application 1 issues a shutdown (ns, read side) on its new just accepted socket 
descriptor
application 1 now issues a read() and will get 0 back every time.  

Summary of question 2 and the scenarios:

From the results of the above scenarios, it appears to me that if data was already available to be read when the shutdown(read_side) was issued, it was discarded.  That seems reasonable.

Yet, if there is no data available to be read, a shutdown(read) is then issued, data is then received by TCP, and then the read() is issued... well then it seems that the application is actually given some data instead of the 0 for end of file.  This seems really strange to me, am I off in the outfield exporing things that no stable socket programmer would do?  Am I seeing a bug?  

THANKS for reading this, do me an even bigger favor and explain it to me!!!


-- 
Kent Hofer             hofer@rchland.vnet.ibm.com
(my opinions are my own and have nothing to do with my employer)


-----------[000312][next][prev][last][first]----------------------------------------------------
Date:      Tuesday, 29 Sep 1992 19:00:23 TUR
From:      Cevdet Dengi <CEVDET@TRMETU.BITNET>
To:        comp.dcom.lans,comp.os.msdos.misc,comp.protocols.tcp-ip
Subject:   Clarkson Token-Ring Packet Driver on PcRoute ver 2.4

I use pcroute (ver 2.4) software for ip routing between two ethernet
LAN s. Now I need also a token-ring connection. I have an IBM token-ring
card and Clarkson packet driver for token-ring cards. I installed
the packet driver and changed the pcroute configuration but
it didn't run correctly. Does anybody arround who has tried to
use Token-Ring packet driver with pcroute software. Any information will
be appreciated.


Cevdet Dengi

TUBITAK Software Center

-----------[000313][next][prev][last][first]----------------------------------------------------
Date:      29 Sep 92 22:57:57 GMT
From:      kevink@grant.ncd.com (Kevin Kelleher)
To:        comp.unix.large,comp.protocols.tcp-ip
Subject:   What do you do when you run out of ip ports?

I have recently started running out of reserved ports used
by rsh (ports < 1023).

Has anyone found a solution for this problem?

I have run into this on both a Sun running SunOS 4.1.2
and a Solbourne running OS/MP 4.1A.1

Kevin Kelleher				kevink@ncd.com uunet!lupine!kevink
Network Computing Devices (NCD)		415-691-2593


-----------[000314][next][prev][last][first]----------------------------------------------------
Date:      Tue, 29 Sep 1992 23:12:15 GMT
From:      ffuentes@tolten.puc.cl (Fernando Fuentes SCC)
To:        comp.protocols.tcp-ip
Subject:   Differences on TCP/IP and OSI


I've been an espectator in the fight of OSI poeple an d TCP/IP people,
but I'm not sure of the differences that OSI could bring compared with
TCP/IP. I know that the address scheme of OSI, will be better than the
one of TCP, solving the porblem of clase B (equivalent) networks. 
I would like to know if someone has some other differences related with
topics like performance, funtionality, scalability, interoperability,
etc.



Fernando Fuentes
SECICO 
Pontificia Universidad Catolica de Chile
Vicuna Mackenna 4860, Santiago, Chile 
ffuentes@tolten.puc.cl 

-----------[000315][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 1992 00:04:50 GMT
From:      ae302@cleveland.Freenet.Edu (Peter Haller)
To:        comp.protocols.tcp-ip
Subject:   X.25 & TCP/IP


Hello and Help:

I have been asked to investigate X.25 and TCP/IP. I thought I'd
start here, as the net is always a good place to start.

First Question: I understand that X.25 is a subset of an ISO standard,
but this says little to a novice. What IS X.25 in simple terms?

Next: What does a X.25 PAD do? I have a Hayes modem that says it can be
used as a X.25 PAD using X.32(?). What does this imply?

Last: How does TCP/IP fit in with X.25. Does it 'ride' on top of it?

If there is a good source of info - an FAQ perhaps - I would love
hearing about it.

PCH
-- 
Peter C. Haller                         |       'If you do what you
Lorain Products                         |       have always done,
ae302@cleveland.Freenet.Edu             |       expect what you have
"Are we having fun yet?"                |       always received.'

-----------[000316][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 00:45:20 GMT
From:      sklower@diva.Berkeley.EDU (Keith Sklower)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a way to encapsulate X.25 over TCP/IP ????

In article <6051@mtecv2.mty.itesm.mx> egonzale@netmon.mty.itesm.mx (Ing. Ernesto Gonzalez) writes:
>Again, how can i encapsulate X.25 traffic over TCP/IP????

Assuming you really do mean that you want to receive All X.25 traffic
(and not just have a PAD/LAT/TCP-IP translator),

CISCO systems does have a way to forward incoming X.25 connections
over a TCP-IP network.  Despite repeated pleas by phone and e-mail
I was never able to get any better information than
``we just forwarded the X.25 packet over a TCP-connection, headers and all''

This begs the question (that I was interested in the time) of how do
you assign DNICs to the recipients of the embeded traffic, or if
you have to declare yourself to be an X.75 gateway, or what tcp
port numbers you use.  I was willing to write up an internet draft
about this, but got no co-operation from the people at cisco.

{The larger motivation for me was pursuing alternative COTS-CLTS
gateways, and having a couple of universities share an X.25 connection
for testing purposes).

-----------[000317][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 01:05:24 GMT
From:      was@gdwb.oz.au (Warren Stokes)
To:        aus.wanted,comp.protocols.tcp-ip,comp.sys.sun.admin
Subject:   AS400s and Suns together

Hello,

We are running a site with around 100 Suns and are in the process of 
replacing our IBM S/38 with a new machine.  We would like that to be 
a Unix box but the incumbent would like to replace it with an AS400.

IBM are telling us all about how great the TCP/IP suite is for the
AS400.  We are sceptical.  The interconnectivity between our Sun
network and this replacement machine is essential.

Are there any sites our there which are doing this kind of thing with
Unix boxes and AS400s?  Does anyone have any comment on the quality of
the IBM AS400 TCP/IP suite?

Thanks

Warren Stokes			Geelong & District Water Board
Phone: +61 52 262598		61-67 Ryrie St Geelong
Fax:   +61 52 218236		Victoria 3220 Australia

-- 
Warren Stokes			Geelong & District Water Board
Phone: +61 52 262598		61-67 Ryrie St Geelong
Fax:   +61 52 218236		Victoria 3220 Australia

-----------[000318][next][prev][last][first]----------------------------------------------------
Date:      Wed, 30 Sep 92 01:30:08 GMT
From:      nelson@crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   SLIP and the PC

In article <1497@airgun.wg.waii.com> dla@se05 writes:

   I am having a heck of a time trying to get SLIP configured between my PC and
   my XYPLEX terminal server.

Serial links are fun to configure.  Lots of mistakes to make, lots of
things to learn.

   On my XYPLEX, if its on 145.144.100.3, I can configure my PC for
   145.144.100.213  right?

Extremely likely, but not necessarily.  It depends your netmask.
Also, the XYPLEX needs to know which port ...213 is on, and of
course, you have to be connected to that port.

   On my PC, I am trying to use the SLIP8250 or ETHERSL packet drivers, but I
   can't get the two machines to talk to each other.

Try 'ping'ing the machine from the XYPLEX and watching the modem
lights, or if it's direct connect, get a RS-232 monitor (one of those
cheapie $10 LED boxes will do).  If you don't get any characters into
the PC after pinging, then the problem is not in the PC.

After you're sure characters are being received, use pktstat to see
if any packets are being received.

Good luck!

-russ <nelson@crynwr.com>  I'm proud to be a humble Quaker!
Crynwr Software            Crynwr Software sells packet driver support.
11 Grant St.               315-268-1925 Voice
Potsdam, NY 13676          315-268-9201 FAX

-----------[000319][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 01:37:52 GMT
From:      nelson@crynwr.com (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Password capture on anonymous FTP?

In article <290992152140@brady.cs.tcd.ie> brady@cs.tcd.ie writes:

   Net friends, I suspect that this is a FAQ; what I need is a WKA (Well Known
   Answer) :-).

   I have some software being hosted on our FTP site, for access by anonymous
   FTP. I'd really like to build a mailing list of the people who upload it,
   and I thought that I could use the ID they are requested to enter as a
   password when logging in anonymously. Does anyone know of software to do
   this?

Technological fix: insist on an email address as a password.

You might do something like insist on a valid and/or valid-looking
address.  The wustl FTP server checks for valid-looking addresses.
You *could* cause the FTP server to connect to the host and verify
the username, but that doesn't work for a number of reasonable cases.

Sociological fix: Ask them for a valid email address.

For this, I suggest that you use the wustl FTP server (I believe it's
on wuarchive.wustl.edu somewhere), which is capable of logging
userids and passwords.  It also can spit out a standard signon
message.  In this signon message, say something like "We will use the
email address you give as a password to build a mailing list.  Please
use your best emailing address."  Those who want to be on the list
will comply, and those who don't will be sure to give an invalid
address.  :)

-russ <nelson@crynwr.com>  I'm proud to be a humble Quaker!
Crynwr Software            Crynwr Software sells packet driver support.
11 Grant St.               315-268-1925 Voice
Potsdam, NY 13676          315-268-9201 FAX

-----------[000320][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 03:00:19 GMT
From:      nam@bach.esl.com (Nam Nguyen)
To:        comp.protocols.tcp-ip
Subject:   SLIP usage (test)

Please help me on this matter!

I would like to use SLIP as the communication channel between
two Sparc stations (sunos 4.1).  After obtaining a tar file of 
slip-4.1 version on the net, I installed SLIP on both machines 
(and reboot,...).  I directly connect /dev/ttya of one machine 
to /dev/ttya of the other (using an RS232 cable).

I try "slip-attach..." to connect these two machines together,
but having problem communicating thereafter (e.g. ping, ftp,...)

If you have successfully used SLIP on sunos 4.1, please give 
me a hint to what I should do to test SLIP to see that it is
working under this condition.

Questions:

  1.  Was I right when I installed SLIP on both machines.
  2.  Do I have to "slip-attach" on both machines, or only one?
  3.  How do I test either "ping" or "ftp" using this setup?

Thanks very much,

Nam N.



-----------[000321][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 12:22:17 EST
From:      phillips@qut.edu.au
To:        comp.protocols.tcp-ip
Subject:   LM2.1 & WinQVT 2.X on WD8003 card????

Hi
On our current system we are trying run together
	LAN Manager 2.1
	WinQVT 2.7
	over Western Digital (WD8003) ethernet cards.

Does anyone have the above configuration working.  
Thanks in advance....

	Doug
________________________________________________
Doug Phillips
Faculty of Information Technology
Queensland University of Technology
Brisbane	QLD	Ausrtralia
E-mail: d.phillips@qut.edu.au

-----------[000322][next][prev][last][first]----------------------------------------------------
Date:      Wed, 30 Sep 1992 10:50:15 GMT
From:      jcasey@netcom.com (John Casey)
To:        comp.protocols.ibm,comp.protocols.tcp-ip,comp.protocols.nfs,comp.unix.aix,comp.unix.large,bit.listserv.ibm-main,bit.listserv.ibm-net
Subject:   IBM mainframe/Unix connectivity questions/issues

Hello,

This is my first posting on the internet.  I am an applications
programmer working for a mainframe applications software company. 
I have varied experience in the mainframe and unix environments; 
but am no way considered an expert on either.  Our company is
starting to look at the unix environment to enhance it's mainframe
products.  I am on a fact finding mission to deal with the
connectivity issues between a unix machine and a mainframe (VM/CMS,
MVS/CICS,...).  I've read some of IBM's brochures on the RS6000 and
their TCP/IP for the mainframe.  Well, I trust sales brochures just
about as much as I trust any of the Presidential candidates.  I
would like to get the scoop from you people, the experts, who have
actually connected the two machines and are working in a mixed
environment. 

I am not necessarily concerned about using an RS6000 or any
particular vendor's hardware.  I don't want to be locked into a
proprietary solution.  I would like to know however, real hardware
configurations (mainframe and unix), network hardware (tokenring,
eithernet, sneakerNet) and software (TCP, NFS, SNA, APPC, ?)
configurations,  and any other special software configurations.  

I am also interested in connectivity concepts; ways of sharing
data.   The way I see it, distributing data across a multi-machine
network can be categorized in three groups,  application packet
passing, full file transfer, & network packet passing.  Do you
argee with this or have I been smoking to much nuclear waste?

Application Packet Passing (client/server technology) -
     The application bundles its data and sends it to another
     application.  These two application use a specific protocol,
     i.e. APPC, NFS/RPC, TCP-SOCKETS.  

Full file transfer -
     FTP, TAPE, ?  Transferring the data files from one machine two
another.  

Network Packet Passing -
     NFS, ?  This is effectively like having a disk on another
     machine mounted to your local machine.  Your operating system
     thinks that this is a local disk.  This means that a program
     on one machine can perform data access to a file that is on
     another machine.  


I know this is a big subject.  If you can help out in any one or
all of these areas, I would greatly appreciate it.  

Thank you in advance,

John Casey

-----------[000323][next][prev][last][first]----------------------------------------------------
Date:      Wed, 30 Sep 1992 11:19:10 GMT
From:      brady@cs.tcd.ie (Michael Brady)
To:        comp.protocols.tcp-ip
Subject:   Re: Password capture on anonymous FTP?

In article <92273.104922PMW1@psuvm.psu.edu>, PMW1@psuvm.psu.edu (Peter M.
Weiss) wrote:
> 
> Don't expect that the password will be that of the requestor.  If I
> were the recipient of such unsolicited mass mailing, I would be upset.
> 

Peter has a point. I guess what would be preferable is the ability to
modify the remote 'get' so that the getter is asked if she wants to be
included in an update-notification mailing list.
Once again, has anyone any pointers as to who might have code for this
already that they might share?
Thanks
Mike Brady
Computer Science Department, Trinity College Dublin, Ireland

-----------[000324][next][prev][last][first]----------------------------------------------------
Date:      Wed, 30 Sep 1992 12:54:20 GMT
From:      dv@welch.jhu.edu (Deepak Vaidya)
To:        comp.protocols.tcp-ip
Subject:   Re: SLIP and the PC

Also check if you are using compressed slip on the xyplex, I had the same
problem when I tried to run slip between a pc and portmaster terminal server.

- Deepak
dv@welchgate.welch.jhu.edu
[128.220.59.13]


-----------[000325][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 13:15:20 GMT
From:      ciacovel@telesciences.com (Chris D Iacovelli)
To:        comp.protocols.tcp-ip
Subject:   FTAM Supplier?

Does anybody in net-land know of an FTAM supplier? Please post or e-mail.

Thanks in advance.
Rick Kemp
rkemp@telesciences.com

-----------[000326][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 14:08:27 GMT
From:      shafer@CS.ColoState.EDU (spencer shafer)
To:        comp.protocols.tcp-ip
Subject:   How to write TCP/IP specs in a procurement


  1. Does it make sense to reference the RFC's, i.e. does it do
     anything?
     2. What RFC's should I use?

     Also, if I want to specify the application gateway between OSI
     and TCP is there a way to do that?

     Would appreciate any help.....


-- 
/\__________________________________________________________________/\
||                     SPENCER SHAFER                               ||
|| email:shafer@handel.cs.colostate.edu    Work Phone: 303-498-0901 ||
\/------------------------------------------------------------------\/

-----------[000327][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 92 15:44:19 GMT
From:      tim@mentor.cc.purdue.edu (Timothy Lange)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP Printers.

I'm looking for a TCP/IP ready printer to connect via Ethernet.  I
have seen lots of IPX and Appletalk ready printers but none for TCP/IP
(lpd).  Has anyone seen such a machine?

-- 
Tim Lange.

Purdue University Computing Center      1408 MATH Building
West Lafayette, IN  47907-1408          317-494-1787

-----------[000328][next][prev][last][first]----------------------------------------------------
Date:      Wed, 30 Sep 1992 16:32:24 GMT
From:      erick@demorgan.uwaterloo.ca (Erick Engelke)
To:        comp.protocols.tcp-ip
Subject:   Re: what does completion of write() really mean?

In article <920928162413@cream.ftp.com> jbvb@ftp.com writes:
>In article <1992Sep25.144329.17680@mixcom.com> mgic@mixcom.mixcom.com (mgic) writes:
>
>    When a process sends data to another process using TCP,
>    what does the completion of a blocking write() on a socket mean?
>
>The answer depends on the implementation.  In PC/TCP for DOS, it means
>that all the data has been transmitted, but not necessarily acknowleged
>by the receiving TCP.  

So does that mean you satisfy Nagle by blocking on subsequent writes?

Erick

-- 
----------------------------------------------------------------------------
Erick Engelke				 	            WATTCP Architect
erick@development.uwaterloo.ca     TCP/IP was easy but i still can't work VI

-----------[000329][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 1992 20:34:12 GMT
From:      trier@slc6.ins.cwru.edu (Stephen C. Trier)
To:        comp.protocols.tcp-ip
Subject:   Re: Password capture on anonymous FTP?

In article <717817072snx@crynwr.com> nelson@crynwr.com (Russell Nelson) writes:
>Those who want to be on the list will comply, and those who don't will
>be sure to give an invalid address.  :)
 
:-)  There will be lots of bounces on that mailing list!

You could also offer a SITE subcommand for the mailing list, either for
signing up or for not signing up.  Users would access it (on most clients)
with the command "quote site <blah>".

This should be pretty easy to add to many of the fancy FTP servers out
there now, since at least some of them already support various SITE
subcommands.

-- 
"Beware of programmers who     Stephen Trier
 carry screwdrivers."          Network Services Engineering, IRIS/INS/Telecom
       Leonard Brandwein       Case Western Reserve University
                               trier@ins.cwru.edu

-----------[000330][next][prev][last][first]----------------------------------------------------
Date:      30 Sep 1992 21:43:26 GMT
From:      terry@eece.ksu.edu (Terry Hull)
To:        comp.protocols.tcp-ip
Subject:   Sockets library

I am looking for a public domain sockets library available for SysV
R3.2.  At this point, it does not need to talk to a network card, it
just needs to provide a sockets ingerface on the local machine.  Any
information would be greatly appreciated.

--
Terry Hull  -  terry@eece.ksu.edu
Department of Electrical and Computer Engineering, Kansas State University

END OF DOCUMENT