The 'Security Digest' Archives (TM)

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

ARCHIVE: TCP-IP Distribution List - Archives (1994)
DOCUMENT: TCP-IP Distribution List for September 1994 (711 messages, 452132 bytes)
SOURCE: http://securitydigest.org/exec/display?f=tcp-ip/archive/1994/09.txt&t=text/plain
NOTICE: securitydigest.org recognises the rights of all third-party works.

START OF DOCUMENT

-----------[000000][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 00:03:52 GMT
From:      [email protected] (Thomas Skibo)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP ACK's: every 2 packets?
In article <[email protected]>,
Erich M. Nahum <[email protected]> wrote:
>Howdy.  I've got a question about ACKing policy in Net/2 and 4.4 TCP.
>
>Both RFC1323 and Stevens' TCP/IP Illustrated (page 277) say that a 
>"reasonable" TCP will ACK every Nth packet, where N is usually 2.  
>I understand *why* you'd want to ACK every 2 packets, in that it provides 
>feedback to the sender frequently, prevents the congestion window from
>closing, etc, but don't see *how* this could happen in 4.4 or Net/2.  In fact, 
>in a version of Net/2 I did for the x-Kernel, I know it's not
>happening, but I may have done something wrong.


The ACK is actually generated by tcp_output().  tcp_output()
goes through a series of tests to decide if anything should
be sent (an ACK in this case).  One of the tests is these two lines...

line 182, tcp_output.c, Net/2 release:

		if (adv >= (long) (2 * tp->t_maxseg))
			goto send;

where adv is the difference between locally available receive window
and the last window advertised to the peer.  tcp_output()
gets called every time data is read out of the socket by the
application so an ACK is generated after every Nth maxsegs worth of data
are consumed by the application.

I think N was changed to 3 for 4.4 if I recall (haven't seen the code
in ages).


-- 
---
Thomas Skibo 				Networking Hardware Group
[email protected]				Silicon Graphics, Inc.

-----------[000001][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 00:45:13 GMT
From:      [email protected] (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP ACK's: every 2 packets?
> I understand *why* you'd want to ACK every 2 packets, in that it provides 
> feedback to the sender frequently, prevents the congestion window from
> closing, etc, but don't see *how* this could happen in 4.4 or Net/2. 
> ...
> So what seems to happen is, if everything arrives in-order on a simplex
> connection, the data is only ACK'ed when the fast timer goes off
> (every 200 ms).  Given that Stevens and Jan Jacobson et al say that
> one should ACK every 2 packets, either Net/2 and 4.4 are not "reasonable," 
> or I am missing something.  Can anyone elaborate?

The code you're looking for is (hidden) in tcp_output():

 	/*
         * Compare available window to amount of window
         * known to peer (as advertised window less
         * next expected input).  If the difference is at least two
         * max size segments, or at least 50% of the maximum possible
         * window, then want to send a window update to peer.
         */
        if (win > 0) {
		...
                if (2 * adv >= (long) so->so_rcv.sb_hiwat)
                        goto send;
        }
 
The ACK is really a window update :-)  But why is tcp_output() even
being called since as you noticed all tcp_input() does is set the
delayed-ACK flag ... because the receiver is reading the data out
of the receive buffer, causing a PRU_RCVD request.  So the ACK
every other policy is only when the receiving process is reading
from its receive buffer.

If you want to see ACKs triggered only by the 200 ms timeout, run
the sock program from Appendix C of "TCP/IP Illustrated".  Run the
server with the -P option (so it doesn't read) and with a big receive
buffer (-R).  Then send that server data from another system and
watch with tcpdump.  All you'll see are the 200 ms acks.

	Rich Stevens

-----------[000002][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 00:53:57 GMT
From:      [email protected] (Erich M. Nahum)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP ACK's: every 2 packets?
In <[email protected]sd.sgi.com> Thomas Skibo ([email protected]) 
writes:

> where adv is the difference between locally available receive window
> and the last window advertised to the peer.  tcp_output()
> gets called every time data is read out of the socket by the
> application so an ACK is generated after every Nth maxsegs worth of data
> are consumed by the application.

This is what I missed.  This tcp_output() call is in tcp_usrreq(), for 
case PRU_RCVD.  The x-Kernel doesn't use sockets, so it works
differently.  Thanks for setting me straight.

> I think N was changed to 3 for 4.4 if I recall (haven't seen the code
> in ages).

Looking at the 4.4BSD-Lite source taken from cdrom.com, it's still 2.

-Erich

--
Erich Nahum                            Department of Computer Science     
Networks and Performance Group         University of Massachusetts at Amherst
[email protected]                     Amherst, MA 01003                      

-----------[000003][next][prev][last][first]----------------------------------------------------
Date:      Thu, 01 Sep 94 08:37:28 CDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Packet Driver for SMC Token Ring
Can someone direct me to a packet driver for an SMC8100 Token Ring card?
 
I am trying to set up Mosaic.  I downloaded the Crynwyr Packet group, but
wasn't able to find it.
 
Thanks,
Tim :-)

-----------[000004][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 10:47:13 -0400
From:      [email protected] (Jim Brain)
To:        comp.protocols.tcp-ip,comp.protocols.ibm
Subject:   Protocol Family AF_SNA
I am hoping someone can steer me to the right group for this:

I see, in the sockets .h files, defs for more AF families, notably AF_IPX and
AF_SNA, etc.  I am especially interested in the SNA part.  DO specs on how
to lay sockets API on top of SNA exist, or am I missing something?

Why do I ask, you might ask?  I am wondering whether it is possible for
sockets API with some extensions could cover SNA LU6.2/LU0,1,2,3, and so
on, as well as TCP/UDP/RAW/IP, and NetBios.  Should I be checking out
the rfc list?  Is this just for future expeansion?  Is any work being done
on it?

Notice the crossposting.  Please direct followups accordingly, or email
me directly.  Thanks.

--
Jim Brain, Embedded Systems Designer, Brain Innovations.
[email protected]  
Dabbling in VR, Old Commodore Computers, and Good Times!
"The above views DO reflect my employer, since I am my employer" - Jim Brain


-----------[000005][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 11:09:23 -0400
From:      [email protected] (Dave Cornejo)
To:        news.announce.newgroups,news.groups,comp.dcom.cell-relay,comp.dcom.isdn,comp.dcom.lans.ethernet,comp.dcom.lans.fddi,comp.dcom.lans.misc,comp.protocols.tcp-ip,bit.listserv.novell
Subject:   CFV: comp.dcom.frame-relay
                     FIRST CALL FOR VOTES (of 2)
	       unmoderated group comp.dcom.frame-relay

Newsgroups line:
comp.dcom.frame-relay	Technology and issues regarding Frame Relay

Votes must be received by 23:59:59 UTC, 22 September 1994.

This vote is being conducted by a neutral third party. For voting
questions only contact Dave Cornejo <[email protected]>. For questions
about the proposed group please contact Allen Robel <[email protected]>

This CFV is also being sent to the cicso discussion mailing list
[email protected]


CHARTER

To provide a mechanism for the discussion of issues related to the
technology and application of frame relay in communications networks.

Topics expected to be discussed include (but are not limited to) the
following examples:

- frame relay standards and proposed standards
- applications of frame relay
- user experiences with frame relay
- relationship and comparison of frame relay to other technologies,
  including asynchronous transfer mode (ATM), cell relay, SMDS,
  X.25 packet switching, IP routing, ISDN, Broadband ISDN, etc.
- announcements of conferences, seminars, etc., related to frame relay
- news of the availability of frame relay products or services
- testing and interoperability of frame relay
- information related to performance and implementation of frame relay
- research findings
- activities of The Frame Relay Forum and other similar organizations
  with activities related to frame relay
- sources of additional information regarding frame relay

This newsgroup is not intended to be a forum in which standards are
developed, although it may be expected to relate news pertaining to
standards and to the progress of recognized standards bodies.

This list will be gatewayed to the Frame Relay Discussion list
<[email protected]>.


RATIONALE: 
 
Given that Frame Relay is beginning to emerge as a mainstream WAN service,
and given that discussions of Frame Relay technology are appearing with
increasing frequency on other newsgroups such as comp.dcom.cell-relay, it
makes sense at this time to create a newsgroup to provide a home for
discussion of this technology. 


HOW TO VOTE

Send MAIL to:   [email protected]
Just Replying should work if you are not reading this on a mailing
list.

The subject in your mail message must contain the name of the group to
allow your vote to be correctly handled by an automated vote-counter,
if you do not do this you will receive a message asking you to
resubmit your vote with the name of the group that you are voting for
in the subject line.

Please do not quote the entire CFV in your vote, it will confuse the
automatic vote counter and will earn you a duplicate vote message.

Your mail message should contain one of the following statements:
      I vote YES on comp.dcom.frame-relay
      I vote NO on comp.dcom.frame-relay

You may also ABSTAIN in place of YES/NO - this will not affect the
outcome.  Anything else may be rejected by the automatic vote counting
program.  The votetaker will respond to your received ballots with a
personal acknowledgement by mail - if you do not receive one within
several days, try again.  It's your responsibility to make sure your
vote is registered correctly.

Only one vote per person, no more than one vote per account.  No votes
will be accepted from system accounts (i.e. root, news, uucp) or
anonymous accounts.  Addresses and votes of all voters will be
published in the final voting results list.
-- 
Dave Cornejo                                There is nothing so subtle
Dogwood Media                                           as the obvious
Fremont, California

-----------[000006][next][prev][last][first]----------------------------------------------------
Date:      Thu, 01 Sep 1994 12:53:18 -0500
From:      [email protected] (Pete Resnick)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: Why does MacTCP do it's own direct DNS queries???
In article <[email protected]>,
[email protected] (John L. MacFarlane) wrote:

>Can anyone tell me why in the heck the Apple engineers put in the code to keep
>track of the authoritative answers in DNS queries and then do DNS lookups
>directly rather than going through the given local nameserver???

If the local nameserver is giving back authority records to the client
resolver, then I think the client resolver is perfectly justified in using
them. But I don't think this is your actual problem.

>I have had several people complain that their mac's don't work (DNS lookup 
>failed error) after putting in a screening router on their site which did
>not allow UDP (i.e. DNS traffic). After looking into the matter a little
>I found that the mac's (MacTCP) were compiling a table of authoritative server 
>and trying to contact them directly rather than using the site's local dns 
>server as they should (IMHO). WHAT IS THE DEAL APPLE??

It seems to me you have two problems here:

1. Your local server is giving back authority records which you don't want
your users to ever use. That seems like something you should fix.

2. Even if you are getting back such records, the MacTCP DNR should still
be configured with local server in it's server list and should still query
it even if the lookup to the authoritative server fails because of your
firewall. This makes me think that you (along with lots of other people,
so don't feel bad) have the MacTCP DNR information misconfigured. Without
getting into a long discussion about it here, suffice it so say that
checking the little "Default" button next to the IP address of your
primary local server is *not* how to specify your primary local
nameserver. You need to have a line in the DNS configuration box with "."
as the domain and the IP address of your primary nameserver. If your
configuration looks like this (a common mistake):

software.com    1.2.3.4     *
software.com    5.6.7.8

then you've got it configured wrong and I can guarantee that the only
server MacTCP is talking to is the authoritative one returned by your
local server.

I am going to post very soon (I promised last week but never got to it) a
MacTCP DNR FAQ of sorts to the comp.sys.mac.comm newsgroup. Perhaps I'll
call it "The One True MacTCP DNR Bible". That should give you a fuller
explanation. If you need specific help on how to configure the MacTCP DNR
for your site (and this is a private offer to John, not everyone), write
to me in private e-mail and I will help you out.

[For the record, I'm one of the MacTCP alpha testers, specifically *the*
alpha tester who annoyed the MacTCP engineers the most on the DNR problems
and caused MacTCP 2.0.3 to never see the light of day. I've done a lot of
disassembling of the MacTCP DNR code. I know what I'm talking about, if
only on this topic; other topics, I take the fifth. :-) ]

pr
-- 
Pete Resnick    	(...so what is a mojo, and why would one be rising?)
Doctoral Student - Philosophy Department, Gregory Hall, UIUC
System manager - Cognitive Science Group, Beckman Institute, UIUC
Internet: [email protected]

-----------[000007][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 09:42:15
From:      [email protected] (John L. MacFarlane)
To:        comp.protocols.tcp-ip
Subject:   Why does MacTCP do it's own direct DNS queries???
Hello,

Can anyone tell me why in the heck the Apple engineers put in the code to keep
track of the authoritative answers in DNS queries and then do DNS lookups
directly rather than going through the given local nameserver???

I have had several people complain that their mac's don't work (DNS lookup 
failed error) after putting in a screening router on their site which did
not allow UDP (i.e. DNS traffic). After looking into the matter a little
I found that the mac's (MacTCP) were compiling a table of authoritative server 
and trying to contact them directly rather than using the site's local dns 
server as they should (IMHO). WHAT IS THE DEAL APPLE??

This "feature" leaves companies with two options:
	1) Drop the screening rules
	2) Drop their mac's

Now I am not one who dislikes mac's, but I think option 1 is not possible.

Thanks for any help,

John L. MacFarlane ([email protected])
Software.com
203 Chapala Street                (805) 899-4274
Santa Barbara, California 93101   (805) 962-5188 Fax.

-----------[000008][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 11:22:45 -0400
From:      [email protected] (Douglas C. Pontius)
To:        comp.protocols.tcp-ip
Subject:   Need  help adding ARP entry from a program
What's wrong with this picture?
-----
add_arp_entry(ip_addr, hlen, eth_addr)
long ip_addr;
int hlen;
char *eth_addr;
{
    struct arpreq arpreq;

	memset(&arpreq, 0, sizeof(arpreq));
    	arpreq.arp_pa.sa_family = AF_INET;
	memcpy(arpreq.arp_pa.sa_data, &ip_addr, sizeof(ip_addr));
        arpreq.arp_ha.sa_family = AF_UNSPEC;
        memcpy(arpreq.arp_ha.sa_data, eth_addr, hlen);
	arpreq.arp_flags = ATF_PERM | ATF_PUBL;
        if (ioctl(s_in, SIOCSARP, (caddr_t)&arpreq) != 0)
	{
		perror("ioctl");
	}
}
-----

When I run it (as root) it gives an error which says "Network is unreachable"
I use /etc/arp to see if it created the entry, and it did not.
The IP address I gave it is on the same net as my host.

What am I doing wrong?

-Doug-


-----------[000009][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 08:18:11 GMT
From:      [email protected] (Richard D. Huff)
To:        comp.protocols.tcp-ip
Subject:   Q: ICMP causing connection lag?
I have experienced an odd form of temporary network lag between
workstations and a server on my local area network.  The same lag appears
between those workstations and random sites on the Internet.  Note that, at
the same time, lag does not appear between the next closest host and the
same sites on the Internet.

[Basic description of my network]

-10Base2 Backbone
-UNIX Server
-Six Macintosh's Using MacTCP 2.0.2 or 2.0.4
-Misc. Mac, Windows, and DOS dial-up and PPP connected computers
-IP Router

[Specific description of problem]

-Opening a telnet session between a workstation and the local server will
take 1 to 2 minutes to bring up a, "login:" prompt.
-Opening a telnet session between a workstation and a server on the
Internet will take 1 to 2 minutes to bring up a, "login:" prompt.-
-Opening a telnet session between the local server and a server on the
Internet is instant.
-Opening a telnet session between a server on the Internet and the local
server is instant.
-telnet, ftp, http connections are all affected; CuSeeme connections are
not.

[Attempts made to isolate the cause]

A).  Power down the router that connects the local network to the Internet
across a T1 Frame Relay link.

No change in the time to establish a telnet session to the local server.

B).  Power down the server/power up the router.

No change in the time to establish a telnet session to a server on the
Internet.

C).  Physically disconnect all 10Base2 wiring except a link between one
workstation and the router.

Problem persists.

D).  Physically disconnect all 10Base2 wiring except a link between one
workstation and the local server.

Problem persists.

E).  Kill in.routed and named on the UNIX server.

Problem goes away!

F).  Rebuild named db tables and restart named.

Problem seems to come and go.

[Other qualifiers]

Problem seems to be immune to the difference between a domain name and its
IP address.

The only application NOT affected was CuSeeme.  (For those of you not
knowing: CuSeeme is a client application to view video broadcast across the
Internet.)

[One Theory]

Someone has tried to explain this problem as an attack on my network via a,
"spray" of erroneous ICMP packets.  Has anyone heard of this being an
issue.  The theory is that the erroneous ICMP packets were picked up by my
server's in.routed and used to redirect my network traffic.

I personally have a very difficult time believeing this theory.  Anybody
have a better explaination?  I would appreciate any input.

Thanks in advance,
Richard D. Huff
[email protected]

-----------[000010][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 14:18:37
From:      [email protected] (Kenneth Croll)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (Johan Montald) writes:
>From: [email protected] (Johan Montald)
>Subject: Novell TCP/IP routing problem
>Date: 1 Sep 1994 16:29:50 GMT


>Hi all,
 
>I'am wondering if its possible to configure our Novell router
>to transparantly route packages between these subnets.
 
>A drawing makes this  hopefully easier:


>                   132.202.75.62          132.202.75.126
>     --------      0xff.ff.ff.c0 -------- 0xff.ff.ff.c0
>----| bridge |------------------| Novell |--------------------------
>     --------     |             | Server |               |
>                  |              --------                |
>                  |                  |                   |
>                  | 132.202.75.1     | 132.202.75.190    | 132.202.75.50
>                  | 0xff.ff.00.00    | 0xff.ff.ff.c0     | 0xff.ff.ff.c0
>              ---------              |                -------  
>             |  unix1  |             |               | unix2 |
>              ---------                               -------  
 
>The problem is that they want to divide the B class subnet number
>we have in yet 3 subnets from which one is a tokenring network.
>I cannot seem to configure the Novell server in order to route to
>net 132.202.0.0, which is logical. Changing the unix1 netmask to
>0xffffffc0 enables me to access unix2 but not what's behind the 
>bridge.
 
>ANY suggestions appreciated, except from removing the router
>which is in my opinion the best solution, but we could run into
>problems with the cable length (there are already repeaters installed)
 
>______________________________________________________________
>Johan Montald                                                /
>Customer Representative                                     ///
>  Computer Associates International Belgium                /// /
>  Woluwelaan 34 B13                                       /// / /
>  1200 Brussels                                          ///// / /
>  Belgium                                               ///// / / /
>email: [email protected]                                /////// / / /
>phone: +32-2/773.28.11  ext 865                       ///////// / / /
>fax  : +32-2/762.73.59                               ///////////// / /
>____________________________________________________///////////////////.
 
>God is real, unless declared integer

Johan,

The Novell file server only uses RIP for routing.  It does not support multiple
subnet masks for the same Class B address as you are trying to do.

 Also, your mask of ff.ff.ff.c0 will only provide 2 legal subnets.  You may
not use either a subnet or host with all bits on or all bits off.  The c0 mask
will give you hosts 65-126 and129-190.

The unix2 host also is using an IP address which would logically belong
in the same subnet with unix1.  Both of them are in subnet 0, which is not
supported.

You could use a mask of ff.ff.ff.e0 to produce 6 subnets of 30 addresses.
You could also use a mask of ff.ff.ff.f0 to get 14 subnets of 14 hosts.  You
will need to decide the maximum number of hosts that are needed on any
segment.

Hope this gets you going in the right direction.

Kenneth Croll      [email protected]

-----------[000011][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 12:56:07 GMT
From:      [email protected] (Barry Flanagan)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
Gary Sarff ([email protected]) wrote:
: [email protected] (Charles Cooley) writes:
 
: >In article <[email protected]>,
: >Aaron Leonard <[email protected]> wrote:
: >The server should be keeping a log of who connected and which IP
: >address the received, so you have all of the information you need.
 
: That sounds fine if the provider that is being called lets customers
: log in to the provider's computer's serial ports to do slip. But,
: what if the provider is using terminal servers with modems and the
: terminal servers are doing slip on their modem ports. How does the 
: provider's computer log who connected and which IP address the caller
: received, if the caller does not access the provider's computer.
: Example, I have a set up here at work.  People can call modems on
: the terminal server and once connected to the terminal server with
: slip running at their end they could telnet or ftp anywhere on the
: internet.  Is there a mechanism to 'log' what user called and what
: IP address they were given?  I sure have not been able to find one.

Change your terminal server or read the manual! Do you allow callers to
activate a SLIP connection without even a security check? 

We use Xylogics Annex3's which communicates with a logging and security
host. The logging shows ALL Annex activity, including details of every SLIP
connection, the username and address supplied.

I find it hard to believe that _any_ terminal server would not at least do
this. How on earth do you measure usage without it?!

-Barry Flanagan

--
   *********************************************************************** 
              IRELAND ON-LINE, West Wing, Furbo, Galway, Ireland
                Tel: +353 (0)91 592727 : Fax: +353 (0)91 592726
           IOL Internet Services - Dublin: 671-5185 : Galway 592711

-----------[000012][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 13:00:46 GMT
From:      [email protected] (Barry Flanagan)
To:        comp.protocols.tcp-ip,comp.protocols.snmp,comp.dcom.lans.ethernet,comp.dcom.lans.misc
Subject:   Re: WANTED: ETHERNET SERVER with RS232 OUTPUT
Rich Krinsky ([email protected]) wrote:

: We are designing a microprocessor controlled product for industrial control.
: The product will have two RS232 serial ports.
 
: We would like to attach one of the serial ports to a commercial "box" that
: takes in RS232 data (in TCP/IP format) and spits out ETHERNET packets.

You will need to implement SLIP or PPP on this serial line.

: I believe what I am searching for is an ETHERNET terminal server?


: *  Does anyone know of a company that makes such a product?

Ask [email protected] for information on their Annex terminal servers. We
have been using a pair of Annex3's here for about 18months and they are
rock-solid for us.

As long as you can implement SLIP/PPP on that serial line, the Annex will do
the job for you.


Hope this helps.

-Barry


: *  How much do they cost?



: Any help will be appreciated.  LAN interfaces are a new area for us!



: Thankyou,
 
: Richard Krinsky
: Earth Station Development Group
 
: [email protected]


: -- 
 
: [email protected]
 
: ---

--
   *********************************************************************** 
              IRELAND ON-LINE, West Wing, Furbo, Galway, Ireland
                Tel: +353 (0)91 592727 : Fax: +353 (0)91 592726
           IOL Internet Services - Dublin: 671-5185 : Galway 592711

-----------[000013][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 13:22:37 GMT
From:      [email protected] (Barry Flanagan)
To:        comp.protocols.tcp-ip
Subject:   Re: Make a telnet session to modem look like COMx?
[email protected] wrote:
: Greetings,
 
: Here's the background.  I have a 3Com AccessBuilder, formerly the
: Centrum box with modems attached.  I can configure it so that when I
: telnet to a port, it will give me the next modem in the set that is
: open.  Then I can AT wherever I want.  SO,
 
: Being greedy of course, I would now like to use these modems with
: Procomm.  Is there a TSR or something for Windows that I can run that
: will establish a telnet session to the AccessBuilder, log in, and make
: it look like a COM port?

There i a Windows driver recently released called comt that will do what you
need. Work over Winsock, and will allow any comms program to connect to
another Internet address using telnet. Using this you will be able to user
Procom to telnet to the modem pool then dial out.

I can't recall where I FTPed the demo (it's shareware), but the company,
Performance Designs has a CI$ account  [email protected]

Hope this helps.

-Barry Flanagan


--
   *********************************************************************** 
              IRELAND ON-LINE, West Wing, Furbo, Galway, Ireland
                Tel: +353 (0)91 592727 : Fax: +353 (0)91 592726
           IOL Internet Services - Dublin: 671-5185 : Galway 592711

-----------[000014][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 15:21:34 GMT
From:      [email protected] (B.V. Jagadeesh)
To:        comp.protocols.tcp-ip,comp.dcom.cell-relay,comp.dcom.lans.fddi
Subject:   Call For Papers - Engineer's Interop - Mar 1995, LasVegas

                            =====================
                           Call for Papers
                           ------------------                             
           Engineer's Conference at "NetWorld(R) + INTEROP(R) 95"           
---------------------------------------------------------------------

Interop Company is soliciting technical papers for the 2nd Engineer's
Conference to be held as part of the upcoming "NetWorld(R) + INTEROP(R)
95" event, March 27-31 1995, in Las Vegas, Nevada. The Engineer's
Conference, which will run March 29-30,  is a two-day focused event
offering approaches and solutions to practical systems and software
design for networking. All participants in the conference will be
able to attend the "NetWorld + INTEROP 95" exhibition, which will
run from Mar 28-30.

FORMAT

The conference will feature the presentation of original papers
which will have been selected by a review committee. All accepted
papers will be published in Conference Proceedings. Accepted papers
must be presented by original authors during the 2-day conference. 

The Engineer's Conference will concentrate on engineering design
problems in three areas: High Speed Networking, Internetworking,
and Network Management. This conference seeks to bring together
research scholars, engineers, and vendors to address pragmatic
engineering issues in the field of networking and distributed
systems interoperability. It is an excellent forum for Engineers
and Researchers to publish papers on solutions to today's
engineering-related problems.

Papers are solicited in the following areas:

*   High Speed Networking: ATM, Fast Ethernet, SONET, FDDI-II,
    HIPPI, SMDS, Frame Relay, Broadband ISDN, etc.
    
*   Internetworking: Design of Bridges, Routers, and Multiprotocol
    Brouters, Addressing Schemes, Routing Protocols, Application 
    Gateways, Switch LAN Technology etc.
    
*   Network Management: Bandwidth utilization, Traffic Trend
    Analysis/Capacity Planning, Automated Trouble Ticket Systems,
    Congestion detection, Network Simulation, SNMP v1 and v2, Security,
    Export considerations for secure systems, Manager-to-manager
    communications, Standardized Testing Suites, Expert Systems,
    Accounting, Distributed/Hierarchical Management architectures, etc.
    
SUBMISSION GUIDELINES

Interested authors are invited to submit an abstract (up to 100
words) clearly describing the problem and the solution offered. All
abstracts will be reviewed and authors are notified for acceptance
or rejection of the abstract. Authors of accepted abstracts must
submit the paper before the last date. These papers are reviewed by
a technical committee for technical merit of the paper before final
acceptance.

Please note the important dates for abstract and paper submission.
All abstracts must contain the authors name, address, telephone
number, Fax number and e-mail address (if available). 

Please send your abstract to:


Interop Company
303, Vintage Park Dr. Suite 201
Foster City, CA-94404

Attn: Engineer's Conference

or e-mail it (in ASCII ) to: [email protected]

** E-mail is preferred. **

===================
= Important Dates =
===================
Abstracts due:		Oct. 1, 1994 
Notification to Authors Oct. 15, 1994
Draft paper due: 	Dec. 1, 1994 
Feedback to authors:	Dec. 24, 1994 
Camera ready copy due:	Jan. 10, 1995 
Overhead slides due:	Feb. 15, 1995		

	


-----------[000015][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 15:28:47 GMT
From:      [email protected] (Mike Albaugh)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP fragmentation
	I ahve trimmed the attributions, and am not claiming that Michael
Salmon wrote the particular piece I'm responding to:

Michael Salmon ([email protected]) wrote:
: |> I have done the above, layered a packet struct over the stream of
: |> bytes TCP socket protocol.  I was sucessful, and it was very
: |> problematic and technically tough:
: |> 
: |> 1) your byte count (your 4 bytes) can be broke into 1 or more reads.
: |> Therefore you must fill a read piece part buffer prior to returning
: |> a buffer to the app.  Your next read may not get the rest.  You may
: |> have to return from this wrapper function to the app, to allow some
: |> real time to elapse.

	Just kibitzing from the sidelines, but a 32-bit length seems
a priori suspect. I know the conventional (i.e., what seems to be taught
C.S. classes) wisdom is to make such fields "big enough to never overflow".
But such reasoning just about guarantees a cock-up when they do, and a little
armchair reasoning about likely OS buffer structures would make it a dicey
proposition to assume one could postpone reading until one had potentially
4 gigabytes of kernel buffers devoted to one "packet". You _may_ know that
such can't happen in your particular case, but it seems a nightmare waiting
to happen to anyone you do the favor of passing the code on to.

	When I have such a situation, I read whatever is available at each
"wakeup" into my own (presumably user-space) buffers, and use the "length"
only to find packet boundaries when picking them apart.

: I have also done it, it was a piece of cake. All that you need is a
: state machine.

	Seems Michaels think alike :-)

					Mike
| Mike Albaugh ([email protected])
| Atari Games Corp (Arcade Games, soon to be Time Warner Interactive)
| 675 Sycamore Dr. Milpitas, CA 95035		voice: (408)434-1709
| The opinions expressed are my own (Boy, are they ever)

-----------[000016][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 15:32:11 +0000
From:      [email protected] (Adam Goodfellow)
To:        comp.protocols.tcp-ip
Subject:   Re: HELP!! binary file xfer over Telnet sessio
[email protected] (Ashok Aiyar) writes:

> 
> In article [email protected],  [email protected] (M. K. Hess) writes:
> 
> >The users have no problems doing their work on the SUN. They get into
> >trouble when they attempt to initiate a BINARY FILE DOWNLOAD/UPLOAD
> >TO FROM THEIR PC's.  They have NO problems with ASCII transfers. They have
> >no problem using Kermit. X, Y, Zmodem is lousy with CRC errors, timeouts,
> >and other problems. Most of the use sz/rz for their transfers.
> >

The basic problem appers to be that the vitual link is operating in text
mode, thus it is messing with EOL sequences, for eg striping or adding
CRs after LFs.

You need to get it switched to BINARY or RAW mode. AKIAK, sz and rz both
tell the tty to goto binary mode during the transfer - perhaps this
request isn't filtering through to telnet (if telnet does EOL
processing), or (surprise! surprise!) the PC telnet is broken.

> >If you execute a xfer straight to the serial port on the SUN...No Problem
> >If you Rlogin and execute a xfer....No Problem
> >If you try to download the same file 10 times it never breaks in the same
> >spot twice (sometimes it blows up right away, sometimes 375K into the xfer.)

That sounds very ood - the break point should be predictable, and
usually at a LF or CR - least it was when I first had this problem.

> >[.. deleted ..]
> >

Actually I guess its the PC - have heard of PC telnet implementations
insisting upon adding a CR after every LF that is seen, regardless of
whether binary mode is being used. Probably someone got hold of an old
telnetd and hacked in PC EOL handling, but didn't quite do it the
right place - have just reversed this hack on the telnet I use. Is this
a PC as in a BSD variant or Linux, or is it DOS?

> I had exactly the same problem, except that it was on a Linux machine, and
> not on a Sun.  I ended up grabbing the latest BSD telnetd source, and 
> compiling it for Linux.  Now file transfers work fine whether users telnet
> in or rlogin into my machine.
> 

Well it would - the BSD telnetd has decent options negotiation and
control mechanisms... including a working BINARY switch that tells the
tty to go to binary, and I think visa versa - just been poaching it
myself!

Adam
-- 
======================================================================
| Computech  Tel/Fax: 081 673 7817  email: [email protected] |
======================================================================


-----------[000017][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 15:46:26 GMT
From:      [email protected] (Donald L. Gover)
To:        comp.protocols.tcp-ip,comp.sys.novell,comp.sys.proteon
Subject:   Netware TCP Performace Question/Bug?
We have an application that uses bsd style sockets on a Netware server. 
The performance under certain situation seems to be less then we expected
and when monitoring the LAN the server seems to be having trouble 
handling congestion. The server is connect to an ethernet and is routed
to and IBM MVS host with a proteon router. What we see on the LAN traces 
is initially a whole bunch of packets from the server to the host with 
ack back from the host. Then over time the transfer degrades with the server 
sending a 536 byte packet followed after a timeout period by the same 536
byte packet. After the second packet we see the ack back from the host
with a completely open window and the sequance of the two timedout packets.

  Has anyone seen a Netware server behave this way? And, does anyone have
any ideas how one could improve this from the application layer?

  Don...

--
=========================================================================
Donald L. Gover                             New Era Systems Services ltd
Internet                                    #710, 425 1st Str. S.W.
office: [email protected]                Calgary, Alberta
home:   [email protected]            Canada T2P 3L8
Phone:  (403) 231-9181
=========================================================================

-----------[000018][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 15:47:42 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
In article <[email protected]> [email protected] (Charles Cooley) writes:

> ...
>>- any Internet-side initiated connections destined to the client's
>>machine will not work in any sane manner.  For example, if you have
>>1000 customers, and 100 IP addrs assigned with names 
>>"slip-00" thru "slip-99.yourdom.uk", and someone sends an SMTP 
>>reply to a message that came from "slip-47.yourdom.UK", there will
>>be a small chance that it will be correctly delivered to the sender,
>>and a very large chance that it'll be misdelivered to someone else's
>>machine.
>
>If you're connecting to a dial-up link (via SLIP, PPP, etc) you should
>NOT be receiving incoming connections like this, especially mail.  Any
>mail not immediately deliverable sits in a mail queue on the sending
>machine.  The message will remain in the queue until the periodic check
>to send old mail happens to correspond to the time you're connected again.
>
>You should be sending mail with addresses referencing your service 
>provider and they should be providing MX records to capture any mail
>for the SLIP addresses to reduce the load on the rest of the network.
>Actually, this is almost required for the Store-and-Forward model to
>work correctly.

I hope the intent of that proscription is only to limit people using
transient IP addresses.  There are otherwise too many millions of us
who get mail over dial-up PPP and SLIP lines with fixed IP addresses.
MX records can be handy for such situations, but they are certainly not
necessary.

Dynamic IP addresses for anything except transient internal links in
the network will soon be gone and forgotten.  There are simply too many
things that do not work well when IP addresses change.  Except for
transient internal links such as backups for leased lines, dynamically
assigned IP addresses are a complicated, fragile way of providing dumb
shell services.  Dynamic IP addresses can't compete such as TIA for
people who only want shells, and dynamic IP addresses are too crippled
for the rest of us.  Saying that you ought to use POP for mail with
dynamic IP addresses is an obscure way of saying that dynamic IP addresses
are painfully constricting.


Vernon Schryver    [email protected]

-----------[000019][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 16:10:20 GMT
From:      [email protected] (Gary Sarff)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
[email protected] (Barry Flanagan) writes:
>Gary Sarff ([email protected]) wrote:
>: [email protected] (Charles Cooley) writes:
>: >In article <[email protected]>,
>: >Aaron Leonard <[email protected]> wrote:
>: >The server should be keeping a log of who connected and which IP
>: >address the received, so you have all of the information you need.
>: That sounds fine if the provider that is being called lets customers
>: log in to the provider's computer's serial ports to do slip. But,
>: what if the provider is using terminal servers with modems and the
>: terminal servers are doing slip on their modem ports. How does the 
>: provider's computer log who connected and which IP address the caller
>: received, if the caller does not access the provider's computer.
>: Example, I have a set up here at work.  People can call modems on
>: the terminal server and once connected to the terminal server with
>: slip running at their end they could telnet or ftp anywhere on the
>: internet.  Is there a mechanism to 'log' what user called and what
>: IP address they were given?  I sure have not been able to find one.
>Change your terminal server or read the manual! Do you allow callers to
>activate a SLIP connection without even a security check? 

uhm, I have read the manual!  I have old Bridge terminal servers.

>We use Xylogics Annex3's which communicates with a logging and security
>host. The logging shows ALL Annex activity, including details of every SLIP
>connection, the username and address supplied.
 
>I find it hard to believe that _any_ terminal server would not at least do
>this. How on earth do you measure usage without it?!

Why on earth would I want to measure usage?  I don't care.  I don't
care if people call for 2 minutes to suck down mail or 3 hours to
ftp a 10Meg file.

I only brought this up because people were posting about this problem 
and only considering people calling and being logged, as if that was
the only way things would happen.  It is not.
-- 
  []   (the null signature)

-----------[000020][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 16:29:50 GMT
From:      [email protected] (Johan Montald)
To:        comp.protocols.tcp-ip
Subject:   Novell TCP/IP routing problem

Hi all,

I'am wondering if its possible to configure our Novell router
to transparantly route packages between these subnets.

A drawing makes this  hopefully easier:


                   132.202.75.62          132.202.75.126
     --------      0xff.ff.ff.c0 -------- 0xff.ff.ff.c0
----| bridge |------------------| Novell |--------------------------
     --------     |             | Server |               |
                  |              --------                |
                  |                  |                   |
                  | 132.202.75.1     | 132.202.75.190    | 132.202.75.50
                  | 0xff.ff.00.00    | 0xff.ff.ff.c0     | 0xff.ff.ff.c0
              ---------              |                -------  
             |  unix1  |             |               | unix2 |
              ---------                               -------  

The problem is that they want to divide the B class subnet number
we have in yet 3 subnets from which one is a tokenring network.
I cannot seem to configure the Novell server in order to route to
net 132.202.0.0, which is logical. Changing the unix1 netmask to
0xffffffc0 enables me to access unix2 but not what's behind the 
bridge.

ANY suggestions appreciated, except from removing the router
which is in my opinion the best solution, but we could run into
problems with the cable length (there are already repeaters installed)

______________________________________________________________
Johan Montald                                                /
Customer Representative                                     ///
  Computer Associates International Belgium                /// /
  Woluwelaan 34 B13                                       /// / /
  1200 Brussels                                          ///// / /
  Belgium                                               ///// / / /
email: [email protected]                                /////// / / /
phone: +32-2/773.28.11  ext 865                       ///////// / / /
fax  : +32-2/762.73.59                               ///////////// / /
____________________________________________________///////////////////.

God is real, unless declared integer


-----------[000021][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 17:07:53 GMT
From:      [email protected] (Pete Kruckenberg)
To:        comp.protocols.tcp-ip,comp.os.linux.admin
Subject:   Configure OSPF on gated 3.0.3 (Linux)?
I need to figure out how to set up OSPF routing on my Linux (1.1.18)
gateway. I'm using gated 3.0.3, with a current RIP configuration. 
Unfortunately, our new Internet access provider uses OSPF (only),
and I've never used OSPF.

What kinds of information do I need to get from them? I've got a few
sample configurations (from the gated 3.5 Alpha) that have OSPF stuff,
but I don't know if they'll even work (so far they haven't).

I have tried a few things (ospf yes; in the gated.conf), but gated -c
always gives a "parsing error at ospf", so I assume either gated
doesn't support ospf (though it says it does), or I have to give it
some more information.

Also, will I have to have my Internet provider "feed" me OSPF updates
(as I had to do with RIP)?

Actually, the best thing would be to give me a good source on gated
and OSPF configuration. The O'Reilly _TCP/IP_Network_Administration_
is pretty out of date on gated and OSPF, so a more up-to-date source
would be great.

I would appreciate any response on this, and the sooner, the better.
I'm working on this right now, and need help. If possible, email
to [email protected] or [email protected] would be best, and I 
will summarize to this group.

Thanks.
Pete.

  ------------------------------------------------------------------------
  Pete Kruckenberg                       School: [email protected]
  University of Utah                       Work: [email protected]
  Computer Engineering    For even more addresses, "finger [email protected]"

-----------[000022][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 18:06:05 GMT
From:      [email protected] (Gloria Hernandez. proyecto informatica)
To:        comp.protocols.tcp-ip
Subject:   TIMEOT PROBLEM (RPC)

Hello, my name is Gloria Hernandez.



        I am reading the book "UNIX Networking Programming", and try to
program a distributed application. The examples are very good.

But I have a problem with a application that I have programmed.
I am using the rpgen utility and my application runs correctly. BUt I have
a very important problem. I need that the client program waits forever or
for a long interval of time the server's answer when send a request of service.

        I has tried to assign a struct timeval with a big number off seconds
but it is ignored.



        Please can you help me?


                                        Gloria Hernandez Ballester
                                        Universidad Autonoma de Barcelona
                                        E-mail: [email protected]


-----------[000023][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 18:28:13 GMT
From:      [email protected] (Stefan Sharkansky)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
In article <[email protected]> [email protected] (Lance Lu) writes:
>
>I use read( s, buf, len)
>where s is unblocking TCP stream socket.
>
>How do I know if the remote socket is closed when I call read()?

read() will return 0.

This is different from the case where there is no data (on a non-blocking
socket) where read() will return -1 with errno=EWOULDBLOCK.

--

Stefan Sharkansky
Prospero Systems Research, Inc.
USMAIL	520 Frederick St. Box 19, San Francisco, CA 94117
VOICE	(415) 731-8114		FAX  (415) 731-3395
E-MAIL	[email protected]



-----------[000024][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 18:28:54 GMT
From:      [email protected] (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
In article <[email protected]>, [email protected] (Gary Sarff) writes:
[...]
|> I only brought this up because people were posting about this problem 
|> and only considering people calling and being logged, as if that was
|> the only way things would happen.  It is not.

Unfortunately, most servers (like ours) will permit the administrator
to log starting and ending points of SLIP or PPP sessions, but will
not log particular activity over those sessions.

(Our filtering capabilities include the ability to generate syslog
messages on a filter 'hit'.  This gets close to the stated issue, but
isn't a solution for it.  Doing this, of course, would be extremely
wasteful of bandwidth for all but the narrowest of uses.  And, of
course, syslog is *NOT* reliable, so one would NOT want to use it for
logging traffic if security or reliability is any kind of concern.)

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

-----------[000025][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 18:37:02 GMT
From:      [email protected] (renrick tulloch)
To:        comp.protocols.tcp-ip
Subject:   Looking for timed startup program?

        Does anyone know of a program that will allow me to start a 
program after a specfic one.
  
Problem:  I startup my tcp/ip on os/2 before my DNS software but the 
tcp/ip software doesn't finish loading before the next program starts 
loading which is the DNS software.  And the DNS software crashes due to 
the tcp/ip software which is not done loading yet.  Is such a program 
 out there that will wait till a program has finsihed loading before 
it starts another one?

Thanks in Advance

--
         _|___/v\___|_                 ////
    -====(~)=(.*.)=(~)====-            o o
              `-'                o00___(_)___00o

==================================================|
|        Rickster                                 |
|        [email protected]                            |
|        [email protected]                |
|        NSBE  '95' 617-247-4998                  |
 ==================================================|
|"Knowledge may be the key"                       |
|"But wisdom unlocks the door"                    |
==================================================|


-----------[000026][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 18:54:41 GMT
From:      [email protected] (Ripple Parikh)
To:        comp.protocols.tcp-ip
Subject:   read() hangs.
Hi,

   I am writing a TCP/IP client which reads multi-line response from the 
   client. I have a function that reads one line at a time and returns -1
   on error or returns 0 on EOF or the number of characters read.

   It reads the line pretty fine; but it hangs up (waiting for data to be
   read from the server). It just stays there. The man page on read() says
   that it should return 0 on EOF. But it does not happen in my case.

   I don't know why.

   Any pointers would be helpful.

   Thanks.
-- 

_______________________________________________________________________________
Ripple Parikh
Internet: [email protected]
_______________________________________________________________________________

-----------[000027][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 19:01:42 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: FAQ's
In article <[email protected]> [email protected] (Jim Raykowski) writes:
>Hello All,
>    I have found the TCP/IP FAQ and am loking for a FAQ on the UDP/IP 
>protocals.

In this (and most other) context, the term "TCP/IP" is intended to refer to
the entire IP protocol suite, which includes UDP, ICMP, etc.  FAQs about
UDP would be in the TCP/IP FAQ.


-- 
Barry Margolin                                                [email protected]


-----------[000028][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 94 19:38:31 GMT
From:      [email protected] (Stephen Johnson)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
[email protected] (Gregory LeBlanc) writes:

>Hi there...

<stuff deleted>

>Right now our nets are subnetted with a 2 bit (192) mask and a 3 bit 
>(224?) mask (the other address is in Texas, so I am a little removed from 
>it at the moment).
 
>We need to know what the usable range of values is in our nets, and which 
>part of the subnetis unusable.

This very subject turn into a lively debate here a couple of months ago. I'll
try to clearly state what I learned during that debate. In a subnetted 
class C network the number of usable subnets is (2 ^ n) - 2 where n is the
number of bits from the fourth octet used for the subnet mask. You might
expect to get (2 ^ n) subnets but RFC 1022 ( I think...how quickly we forget)
states that addresses with subnet mask values of all ones or all zeros
should not be used. For example:

in a class C net with a 255.255.255.192 subnet mask,
the addresses  X.X.X.0 thru X.X.X.63     (first two bits zeros) and
               X.X.X.192 thru X.X.X.255  (first two bits ones)
should not be used.

In a class C net with a 255.255.255.224 subnet mask,
the addresses  X.X.X.0 thru X.X.X.31     (first three bits zeros) and
               X.X.X.224 thru X.X.X.255  (first three bits ones)
should not be used.    

I have be told that some network equipment (Cisco I think was the vendor 
named) will not correctly handle subnets that violated that standard.


-----------[000029][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 20:12:44 GMT
From:      [email protected] (Steven Hancock)
To:        comp.protocols.tcp-ip
Subject:   Problem getting UDP client/server working on SGI
Hi all -

I'm trying to implement a UDP client/server application on an
SGI Indy platform running IRIX 5.2 and I can't get the sockets
to communicate properly.  I am just trying to get the UDP (udpcli,
udpserv) echo server code from W. Richard Stevens' book "UNIX 
Network Programming" to work and it gives me an error on the client 
end:

mightydog 35% ./udpcli
saddasdasdas
./udpcli: dg_cli: recvfrom error (Bad address)
Abort (core dumped)
mightydog 36% 

The server seems to get the data ok, but it bombs on the recvfrom()
call on the client end. I've tried it on an old 4.0.5 machine
thinking it might a bug in the OS. The TCP version works fine. Has 
anyone else seen this problem in the UNIX environment? Am I doing 
something wrong?

Thanks in advance,

-- 
+-----------------------------------------------------------+
| Steve Hancock             [email protected]     |
| Senior Analyst            Voice: (205) 544-3516           |
| New Technology, Inc.      FAX:   (205) 544-5147           |
+-----------------------------------------------------------+

-----------[000030][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 20:59:58 GMT
From:      [email protected] (David Oberst)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: MacTCP and RIP problems
In article <[email protected]> [email protected] (Garry Hornbuckle) writes:
>In article <[email protected]>,
>
>    [reference to OpenTransport [aka MacTCP 3.0]
>
>Give us a couple more months to get it finished and tested, OK?
>
>-------------------------------------------------------------------
>Garry Hornbuckle    Product Manager, Communications & Collaboration
>-------------------------------------------------------------------
>"If I told you that I   | email      [email protected]
> spoke only for myself  | applelink  HORNBUCKLE1
> would you believe me?" | fax        (408) 974-1211
>-------------------------------------------------------------------

Was this poetic license, or is OpenTransport due to be ready this
quickly?

Also, are there plans for an Inside Mac:Open Transport type of book
to help us poor small potatoes programmers? <g>

      David Oberst/GNWT Bureau of Statistics/Yellowknife, NWT, Canada
      [email protected]        [Pascalite Diehard]


-----------[000031][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 21:17:12 GMT
From:      [email protected] (Dennis Sutherland)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]>, [email protected] (Kenneth Croll) says:
>
>In article <[email protected]> [email protected] (Johan Montald) writes:
>>
>>                   132.202.75.62          132.202.75.126
>>     --------      0xff.ff.ff.c0 -------- 0xff.ff.ff.c0
>>----| bridge |------------------| Novell |--------------------------
>>     --------     |             | Server |               |
>>                  |              --------                |
>>                  |                  |                   |
>>                  | 132.202.75.1     | 132.202.75.190    | 132.202.75.50
>>                  | 0xff.ff.00.00    | 0xff.ff.ff.c0     | 0xff.ff.ff.c0
>>              ---------              |                -------  
>>             |  unix1  |             |               | unix2 |
>>              ---------                               -------  
>>
>
>Johan,
>
>The Novell file server only uses RIP for routing.  It does not support multiple
>subnet masks for the same Class B address as you are trying to do.

Logically this makes sense to me, but I have seen it work?!?  Go figger.

> Also, your mask of ff.ff.ff.c0 will only provide 2 legal subnets.  You may
>not use either a subnet or host with all bits on or all bits off.  The c0 mask
>will give you hosts 65-126 and129-190.

If this were a class C address the above statement would be true, because you
can't have subnets with all 0's or all 1's.  However, since this is a class B
address the 3rd byte is also part of the subnet address.  Thus the ranges of
1 - 62, and 193 - 254 are also available.

>The unix2 host also is using an IP address which would logically belong
>in the same subnet with unix1.

Correct.

>You could use a mask of ff.ff.ff.e0 to produce 6 subnets of 30 addresses.
>You could also use a mask of ff.ff.ff.f0 to get 14 subnets of 14 hosts.  You
>will need to decide the maximum number of hosts that are needed on any
>segment.


>Hope this gets you going in the right direction.
>
 
>Kenneth Croll      [email protected]

Good Luck!
Dennis

[email protected]
Sr. Electronics Technician
Kent State University
Kent, Ohio USA

-----------[000032][next][prev][last][first]----------------------------------------------------
Date:      Thu, 1 Sep 1994 23:22:04 GMT
From:      [email protected] (Alexander Adolf)
To:        comp.unix.sys5.r4,comp.protocols.tcp-ip,comp.protocols.ppp,comp.unix.pc-clone.32bit
Subject:   Re: SLIP verses PPP
Alan Cox ([email protected]) wrote:

: Sun machines seem to support PPP including Sync PPP for ISDN.

Serial-line dial-up PPP is built-in as of Solaris 2.3. ISDN-PPP is
payware.


  -- Alexander Adolf

-- 
                                              #include <std-disclaimer.h>
Alexander Adolf ---------------------------------- [email protected]
Georg-Simon-Ohm Polytech.Univ. Nuernberg/FRG --- Department of Electrical
Engineering -------- Computer Science and Information Technology Division

-----------[000033][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 11:34:19 -0700
From:      [email protected] (John Gulbenkian)
To:        comp.protocols.tcp-ip
Subject:   Free IEEE Tutorial



                              IEEE
               OEB Communications Society Presents
                         A Free Tutorial

        Present Day Challanges of TCP/IP Internetworking


The  proliferation of TCP/IP technology worlwide has  caused  the 
rate of assignment of IP addresses to increase dramatically  over 
the  recent  years.  To help prevent the quick  depletion  of  IP 
addresses,  the  Network Information Center (NIC)  has  began  to 
assign IP addresses at a much slower rate than in previous years. 

This has forced network engineers to use their IP addresses  more 
cautiously  and  to deploy techniques such as  non-byte  boundary 
subnet masking, Classless Inter-Domain Routing (CIDR) and  Super-
netting.

There has been very few guidelines as to how to do the masking on 
non-byte  boundary (10-bit subnet mask being the  most  popular), 
the correct calculations of broadcast addresses and the ramifica-
tions/pitfalls of using multiple IP (sub)nets on the same  physi-
cal medium (violating the basic rule of IP that assigns a network 
address to every physical network).

RFC 1597, issued on March 1994, on page 5 states "Using  multiple 
IP  (sub)nets  on the same phsical medium has many  pitfalls.  We 
recommend  to avoid it unless the operational problems  are  well 
understood  and  it is proven that all  equipment  supports  this 
properly."

To use some of the more advanced IP subnetting features,  network 
implementors and administrators are migrating to the OSPF routing 
protocol where again there is a lack of good tutorial material.

This  tutorial will focus on these specific topics  including  an 
explanation of the OSPF routing protocol, how its features relate 
to IP subnetting and how routers should deal with these issues in 
general.

*****************************************************************

Saturday  September 24

Speaker: Allan Leinwand (Cisco Systems)

Time: 9 AM - 2 PM

 8:30 -  9:00   Registration
 9:00 - 10:30   IP and IP subnetting
10:30 - 10:45   break
10:45 - 11:45   Multiple IP subnets on a wire, RIP, OSPF basics
11:45 - 12:30   break
12:30 -  2:00   Advanced RIP and OSPF (Supernetting)

Location: Pacific Bell, 2600 Camino Ramon 
          San Ramon
          (I-680 exit Bollinger Canyon going East, left at 
           Camino Ramon)

Open to IEEE members and non-members.

Reservations: Required  Voicemail (510) 945-2363
                        e-mail:  [email protected]

 

-----------[000034][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 00:05:17 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (Johan Montald) writes:
>                   132.202.75.62          132.202.75.126
>     --------      0xff.ff.ff.c0 -------- 0xff.ff.ff.c0
>----| bridge |------------------| Novell |--------------------------
>     --------     |             | Server |               |
>                  |              --------                |
>                  |                  |                   |
>                  | 132.202.75.1     | 132.202.75.190    | 132.202.75.50
>                  | 0xff.ff.00.00    | 0xff.ff.ff.c0     | 0xff.ff.ff.c0
>              ---------              |                -------  
>             |  unix1  |             |               | unix2 |
>              ---------                               -------  

First of all, the picture makes perfect sense except for the unix2
address of 132.202.75.50. I'm guessing that that's a typo, and it
really should be something like 132.202.75.70: i.e., it's in the
network range of 132.202.75.126/ff.ff.ff.c0. If 132.202.75.50 is
correct and an integral part of the problem, then my comments here
will be fairly useless.

What you need to do this is a feature known as "proxy ARP".
Unfortunately, you have to buy Novell's MultiProtocol Router product
in order to get the TCPIP.NLM which has this feature. (Or you could
just wait until "the next major release" of NetWare comes out, which
-- finally! -- includes proxy ARP support gratis.)

Assuming one way or the other you get the proxy ARP support, then your
configuration will work except that the mask on the 132.202.75.62
interface should be FF.FF.0.0 (i.e., agreeing with the other nodes on
that segment) instead of FF.FF.FF.C0: proxy ARP eliminates the
requirement that all the masks for different parts of a network agree
with one another.

By the way, another poster commented that the ff.ff.ff.c0 mask allows
only two subnets, and, in fact, you appear to have set your subnet
numbers as if you thought this was true as well. But it turns out
that, because your base network is class B, it only takes up the first
two bytes, so the FF.FF.FF.C0 mask allocates 10 bits to the subnet
instead of the two one might expect. Since, in your case, the third
byte is always 75, that means that the subnet number is already
non-zero, so the contribution to it from the last byte may be zero.
This means that, since you only have two subnets, you could use the
mask ff.ff.ff.80 for them, allowing you to have 126 nodes on each
instead of 62.

>The problem is that they want to divide the B class subnet number
>we have in yet 3 subnets from which one is a tokenring network.
>I cannot seem to configure the Novell server in order to route to
>net 132.202.0.0, which is logical. Changing the unix1 netmask to
>0xffffffc0 enables me to access unix2 but not what's behind the 
>bridge.

I assume by "behind the bridge" you mean addresses which don't start
with the three bytes 132.202.75, correct? That would be consistent
with the explanation. (From IP's point of view, the bridge should be
no more relevant than any old ethernet cable.

					don provan
					[email protected]

-----------[000035][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 00:45:18 GMT
From:      [email protected] (John Bell)
To:        comp.sys.novell,comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Winsock connectivity to Internet resources over ipx/spx
Is it possible to have an application utilize a winsock interface to commuicate
with an internet resource (tcp-ip) in such a manner that the communication 
between the pc and the (local) netware server is ipx/spx and the communication
between the netware server and the internet resource is tcp/ip

I need to avoid is running a tcp/ip stack on each client machine.  I can 
install/configure additional software on the servers if needed.

I'd like to know what software I need to install on the client and on the
servers.  Also example configuration files would be very helpful.  I will
gladly follow pointers to good information sources.  A good text would be
wonderful.

Public Doman, free, shareware, and commercial solutions are all fine.

TIA

John Bell
 'very greatful for your help, but too poor for a sig.'

 

-----------[000036][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 11:40:32 -0500
From:      [email protected] (Matt Midboe)
To:        comp.protocols.tcp-ip
Subject:   Can this configuration be put on the Internet?
Okay, we have one company who has a system that looks like this:

     ethernet interface+--------+ slip interface
    /==========/=======|Linux PC|=======[goes to Netblazer ST SLIP connection]
    |          |       +--------+
    |          |ethernet        
    |      +------------+
    |      |PC w/Windows|
    |ether +------------+
 +------------------+ arcnet
|Novell File Server|----------[A bunch of PCs with arcnet cards]
+------------------+

Okay, not the greatest drawing but I hope that it illustrates what I'm
trying to describe here. There is a PC running linux which has two
interfaces, one slip and one ethernet. On that ethernet there is a PC
running windows, and a Novell file server. Then coming off of the
Novell file server is an arcnet with a bunch of PCs. Now my question
is how can I put all of these on the Internet. Do I need one class C
for the ethernet, and one class C for the arcnet? How does the routing
work from the arcnet? 

*any* help on this is much appreciated

Matt

-----------[000037][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 12:57:49
From:      [email protected] (Anthony Jones)
To:        comp.protocols.tcp-ip
Subject:   WHAT IS REQUIRED TO PROVIDE SLIP SERVICE
Hello world !

I am writing a paper about the internet from a novice point of view.  (Yes I 
will run a spelling check on it.)  My research shows that dial up slip 
connections are, at present, the most rapidly growing type of access.   I 
should like to include a small section that explains briefly what setup is 
required in order for new providers to offer a dial up slip connection to the 
public.  Rapid demand growth causing a co-ordination amongst the existing 
elements of the net is a mojor theam of the papper and I hope to display is 
here.

If you have any insight or experience in setting up as a SLIP provider I would 
be very happy to hear from you and would be glad to drop you a copy of the 
finished product.

Thanks all.

AJ
*****
Anthony E. Jones, B.Sc., CNE
North Vancouver, British Colombia, Canada
E-Mail: [email protected]
*****

-----------[000038][next][prev][last][first]----------------------------------------------------
Date:      1 Sep 1994 15:35:02 +1000
From:      [email protected] (Colin Campbell)
To:        comp.protocols.tcp-ip
Subject:   Routing unser SunOS and Class B nets
Hi,

(First foray into this group so be gentle with me :-)

SunOS 4.1.3C or SunOS 5.3

We have a Class C and a Class B with C subnetting. I want to retain
some control over which B-subnets a machine on the C net can see. My
original guess was this: (we use static routes)

    On `C' (203.5.10.10)

	route add net 147.132.157.0 203.5.10.2 4
	route add net 147.132.176.0 203.5.10.2 2

The intention was that `C' can only see the B subnets that I specify.
The first command works. The second fails with `file exists' meaning
that the route to that net already exists. I can sort of see the logic
behind it but don't want it to behave that way.

Further, despite the first working and the second failing, I can ping
anything on 147.132.176.

/etc/netmasks contains `147.132.0.0 255.255.255.0' but this is apparently
irrelevant.

What is broken?

	SunOS? (Can it be fixed?)
	Me? (Can't be fixed.)
	Nothing? (That is how it is supposed to work.)

I do not believe the last option cos our network guru says I should be
able to do what I am trying to.

Thanks
Colin

-----------[000039][next][prev][last][first]----------------------------------------------------
Date:      Fri, 02 Sep 1994 15:48:16 -0400
From:      [email protected] (Dhawal Tyagi)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP application:a multithreaded approach !!
In article <[email protected]>, [email protected] (Kanghoon Lee) wrote:


Hi netters,

       I am writing an TCP-IP application with muti-threaded approach.
       I expect to have large number of client sessions (around 3500)

       Each client session will have two socket connections.The application
       will be running on solaris sparc1000 .I intend to run all these
       sessions on a single machine.Has anyone developed a similiar
       application for such a large number of sessions using multi threads?


      I shall appreciate if you could suggest/recommend any ideas as
       to implement this kind of application.

       Thanks in advance !!

      -manoj





Manoj Gupta                  | Internet: [email protected]
DEC India                    | 
Boxboro MA tel # 508-264-5388 | Disclaimer: This message reflects my own views.

 
 Is this another language I have learn to read this group? :)

The above was posted using ROT-13 coding, I don't know if it was intentional or
not, but I have included the correct text.

Regards
Dhawal

--
Research Associate
Communications Networks Services
(703) 231-7973
include <stddisclaimer.h>

-----------[000040][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 10:56:58 +0000
From:      [email protected] (Adam Goodfellow)
To:        comp.protocols.tcp-ip
Subject:   Re: source routing
[email protected] (Art Berggreen) writes:

[source routing info]

From what you say, am I to assume that if one were patch by hand a
strict source route, could one confuse a firewall and intermediate
routers by get packets to come back to you machine by giving the
illusion that you have a different IP address to that which you actually
have.

The reason I ask, is that someone was requesting that I should add a
patch to FTP to cause it to accept connectiuon only from pre-defined IP
addresses. A friend advised me that by using strict source routing,
a hacker can still get in quite easily by faking the route, to make it
appear to have come from a friendly machine, with the source route
taking care of the problem of ensuring returned packets find their way
back to the hacker rather than to the machine that has been allocated
the address you ae trying to fake.

The only way round that, it seemed to me, would be to ignore connection
attempts where a source route was specified...

Adam
-- 
======================================================================
| Computech  Tel/Fax: 081 673 7817  email: [email protected] |
======================================================================


-----------[000041][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 12:11:46 GMT
From:      [email protected] (Malcolm White)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
In article [email protected], [email protected] (Gregory LeBlanc) writes:

>
>We need to know what the usable range of values is in our nets, and which 
>part of the subnetis unusable (someone said that one of the netowrks is 
>not to be used).
>

With your current mask of 255.255.255.192, on a Class C address this would only provide for 2 
subnets of 62 hosts. Although two bits are available for the subnetting, an 'all ones' and 'all zeroes' 
subnet should not be used. This also applies to the host portion of the address as these addresses are
reserved for broadcasts and subnet identification.

Using the .224 mask would increase the number of subnets to 6 but reduce the number of hosts per 
subnet to 30.

The range of addresses using the .192 mask would be:

x.x.x.65 to 126
x.x.x.129 to 190

Your homework for tonight is to work out the address ranges using the .224 mask :-)

Malcolm White
Salomon Brothers, London


-----------[000042][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 12:41:16 GMT
From:      [email protected]_img.b15.ingr.com (Daniel F. Wygant)
To:        comp.protocols.tcp-ip
Subject:   Re: Do anyone have the X-terminal emulator on PC
Yes, I believe Intergraph is offering one.  I have seen it run, and it looks
very good. If you have mosaic, check:

http://www.ingr.com/

In article <[email protected]>, [email protected] (Wichai Tang) writes:
|> Hi
|> 	Do you know where I can get the X-terminal run on PC ?
|> 	Thanx
|> Wichai Tang
|> [email protected]
|> 
 
-- 
Have Fun :)
  DFW

-----------[000043][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 12:44:41 GMT
From:      [email protected]_img.b15.ingr.com (Daniel F. Wygant)
To:        comp.protocols.tcp-ip
Subject:   RPC dazd&confzd :: re-mapped & refused connections
We have an RPC server which has two problems:
  - First, there seems to be a problem with the RPC Server servicing
    a long RPC call.  If more than one more call this server, the first caller
    will wait up until the timeout, but any other callers after this
    (the third, 4th, ...) will be rejected.  This happens on the UNIX System V
    R3.1 operatoring system but not on the SunOS 4.1.3 (Solaris 1.1.1).

    Is there something I can do to prevent this.

  - Next, after working intensly for about 10 days, the RPC server on
    one of the machines appeared to become unmapped.  I did rpcinfo -p
    and my rpc showed up.  I then ran rpcinfo -t nodename # and
    it came back rejected:

rpcinfo: RPC: Remote system error - Connection refused
program # is not available

    I know this implies that it was unmapped, because the only time I've seen
    this happen (toher than this most recent time) was when I re-ran the
    RPC server (while the first was still running) and then terminated.
    The second server starts out by unmapping the RPC number.  Is there any
    way around this.  I thought of setting an alarm before the svc_run,
    and then in the SIGALRM callback checking the status of the RPC# mapping.
    Is there some way of checking or being notified of an unmapping.

-- 
Have Fun :)
  DFW




-- 
Have Fun :)
  DFW

-----------[000044][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 12:58:21 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: XTP availability?
In article <[email protected]> [email protected] (Stefan Grefen) writes:
>I don't have a problem with that, because you could add routes with bigger
>than default MTU's to routes known to be able to handle it.
>It's a kludge, but again better than nothing.

Indeed the Linux mtu/window setting actually finally got added so we could
make better use of SuperJanet (SMDS) speeds. Upping the main routes used
for the superjanet links to big windows and 1500 byte MTU's did a lot for
the performance at no cost to anything else. I'm all for kludges that do
the job.

Alan

-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000045][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 13:08:14 GMT
From:      [email protected] (Peter Howlett)
To:        comp.protocols.tcp-ip
Subject:   Grabbing IP packets
Howdy,

If I was to disable routing in the kernel, to implement my
own routing protocol (not RIP), is it possible to get the
kernel to pass IP packets up to my user level program
to check destination addr/ports etc...? Perhaps a raw
socket is in order? (Cant modify kernel!)

--------------------------------------------------------------------
Peter Howlett			Atlantic Systems Group
E-Mail: [email protected]	Fredericton, N.B. Canada

-----------[000046][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 20:10:27 -0400
From:      [email protected] (michael v mascari)
To:        comp.protocols.tcp-ip
Subject:   Routing Problem
	Let me first apologize - I am a bit new at this and I'm sure my
incompetence will show.  I have a problem I have been unable to 
fix in regards to routing packets:

|------------|                               |-------------|
| Amiga 4000 |- 168.0.0.14 ----- 168.0.0.24 -| IBM 486     |
|------------|        SLIP @ 115Kbps         |             |
                                             |             |
                                             |-------------|
   |-------------|                                  |
   |             |--- 168.0.0.1 --------------- 168.0.0.22
   | RS/6K 530H  |  Token Ring @4 Mbps
   |             |
   |-------------|

	I am running AmiTCP on my Amiga 4000 with SLIP @ 115KBps
to my 486 (both on my desk), running IBM TCP/IP 2.1.1.  The 486,
as well as ~20 other 386's and 486's are on a 4Mbps token-ring
network with an RS/6K.  Even though the number indicates a Class B
network, we have a subnet-mask of 255.255.255.0.  
	On the Amiga, I have:

	168.0.0.24 as the default gateway and the desination address
		of the SLIP interface

	On the PC, I have no default gateway defined.
	
	On th RS/6K, I have 168.0.0.14 routed to 168.0.0.22

	I can ping from the RS/6K to either of the two IP addresses
on the PC.  I can ping from my Amiga to either of the two IP addresses
on the PC.  But, for the life of me, I cannot from the RS/6K to the
Amiga.  The reason I want to go through the PC is for the 115Kbps
SLIP rate - I can only go as high as 38.4Kbps with a direct connection
to the RS/6K.

	Thanks for ANY help you could offer,

	Mike Mascari ([email protected])

	


-----------[000047][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 14:51:04 GMT
From:      [email protected] (Peter Howlett)
To:        comp.protocols.tcp-ip
Subject:   Re: source routing
Adam Goodfellow ([email protected]) wrote:
: [email protected] (Art Berggreen) writes:
 
: [source routing info]
 
: From what you say, am I to assume that if one were patch by hand a
: strict source route, could one confuse a firewall and intermediate
: routers by get packets to come back to you machine by giving the
: illusion that you have a different IP address to that which you actually
: have.
 
: The reason I ask, is that someone was requesting that I should add a
: patch to FTP to cause it to accept connectiuon only from pre-defined IP
: addresses. A friend advised me that by using strict source routing,
: a hacker can still get in quite easily by faking the route, to make it
: appear to have come from a friendly machine, with the source route
: taking care of the problem of ensuring returned packets find their way
: back to the hacker rather than to the machine that has been allocated
: the address you ae trying to fake.
 
: The only way round that, it seemed to me, would be to ignore connection
: attempts where a source route was specified...
 
: Adam
: -- 
: ======================================================================
: | Computech  Tel/Fax: 081 673 7817  email: [email protected] |
: ======================================================================

It is my understanding that any self-respecting firewall should not
by default allow any source routed packets to pass through it (Unless
you have a specific need).

A good paper on this is from D. Brent Chapman - "Network (In)Sercurity
through IP Packet Filtering", [email protected]

--------------------------------------------------------------------
Peter Howlett			Atlantic Systems Group
E-Mail: [email protected]	Fredericton, N.B. Canada

-----------[000048][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 15:39:48 GMT
From:      [email protected] (Clay Luther)
To:        comp.protocols.tcp-ip
Subject:   Question about bootp forwarding
At my site, we have a Novell machine that acts as a packet router. It
allows for forwarding of bootp packets, but only one bootp server can be
specified.

What vendors sell hardware or software that allow you to forward
bootp packets, even if they originate from multiple bootp servers?

Chris Pearce, [email protected]

-----------[000049][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 94 17:20:29 GMT
From:      [email protected] (Kanghoon Lee)
To:        comp.protocols.tcp-ip
Subject:   Re: UNIX --> SAME PRINTER <-- Novell

>I am looking for a robust solution (Software of
>Hardware) to allow laser printers to print from both
>Novell Print queues as well as Unix Print queues.
 
>We have currently standardized on Intel Netport Print
>Servers, but could change to another vendor if it
>provides a better solution.
 
>Environment -
 
>SunOS 4.1.3 using a variety of X-Terminals and
>Character based terminals.
 
>Novell 3.11 - Using mostly 386/486 based PCs running
>WordPerfect & Lotus.
 
>Thanks in advance for any suggestions.
 
>Chris Lloyd
>Manager, Production systems
>American Institute of Physics
 
>(516) 576-2313   [email protected]

I have a perfect solution for you.  We have just released two new mult-network
print systems that understands multiple protocols.  So you can print from Unix
through TCP/IP, NetWare, LanMan, EtherTalk, and VAX.  Currently it fully 
supports Novell Netware through 3.11, and works in Version 4 with 3.11 
emulation.  More info is available on http://www.qms.com/www/QMSNews.html.

Kang

---------------------------------------------------------------------------
 Kang H. Lee            ################################## ######### 
 [email protected]        #         ##   #####   ##        # ########: 
 (205) 633-4300 x-1231  #  #####  ##    ###    ##  ####### #      .:
                        #  ##  #  ##  #  #  #  ##        # ##    .::
 Systems Engineer       #  ###    ##  ##   ##  ########  # ##   .::: 
 QMS Incorporated       #      _  ##  #######  ##        # ## .::::: 
 One Magnum Pass        ################################## #.::::::: R 
 Mobile, AL 36618           ..Where Imagination Leads
---------------------------------------------------------------------------

-----------[000050][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 17:24:04 GMT
From:      [email protected] (Bruce Wollen)
To:        comp.protocols.tcp-ip
Subject:   IP bcast: is all 0's dead?
Any opinions out there?  Is there really anybody left that uses
the old broken Berkeley software that used all 0's as the IP
broadcast address (instead of all 1's like they should have done)?

I'm wondering if it's reasonable to even implement support for
this old broken broadcast address in current networking software
products.

Any input is appreciated
Bruce
--
                           Northwest Digital Systems 
    Bruce Wollen           Seattle, WA
    [email protected]          Vox: (206) 524-0014
                           Fax: (206) 524-3440

-----------[000051][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 94 17:24:28 GMT
From:      [email protected] (Kanghoon Lee)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP application:a multithreaded approach !!
[email protected] (Manoj Gupta) writes:

>Uv arggref,
 
>        V nz jevgvat na GPC-VC nccyvpngvba jvgu zhgv-guernqrq nccebnpu.
>        V rkcrpg gb unir ynetr ahzore bs pyvrag frffvbaf (nebhaq 3500)
 
>        Rnpu pyvrag frffvba jvyy unir gjb fbpxrg pbaarpgvbaf.Gur nccyvpngvba
>        jvyy or ehaavat ba fbynevf fcnep1000 .V vagraq gb eha nyy gurfr
>        frffvbaf ba n fvatyr znpuvar.Unf nalbar qrirybcrq n fvzvyvne
>        nccyvpngvba sbe fhpu n ynetr ahzore bs frffvbaf hfvat zhygv guernqf?


>        V funyy nccerpvngr vs lbh pbhyq fhttrfg/erpbzzraq nal vqrnf nf
>        gb vzcyrzrag guvf xvaq bs nccyvpngvba.
 
>        Gunaxf va nqinapr !!
 
>	-znabw


>[RBO]


>-- 
>Znabw Thcgn 		      | Vagrearg: [email protected]
>QRP Vaqvn		      | 
>Obkobeb ZN gry # 508-264-5388 | Qvfpynvzre: Guvf zrffntr ersyrpgf zl bja ivrjf.
>-- 

Is this another language I have learn to read this group? :)



-----------[000052][next][prev][last][first]----------------------------------------------------
Date:      Fri, 2 Sep 1994 17:40:18 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: ICMP's messages
In article <[email protected]> [email protected] (Jean-Sebastien Bettez) writes:
>I don't understand how you could encode in binary base format the type message
>16 (information reply), 17 and 18 (addrss mask request and reply) with a 
>8 bits field (the TYPE field in a ICMP message).

8 bits can hold 0-255.  You seem to be thinking of 4 bits, which can hold
0-15.

>I don't know where R. Stevens in is book "TCP Illustrated V1" found the code
> over 5 with the TYPE 3 in an ICMP message. The rfc 792 (the official ICMP 
>description) have only code 0, 3,4 and 5.

The additional codes were added by later RFCs.  RFC 792 is very old.  As
new protocols and conventions have developed, ICMP has been updated to
allow for them.  A good example is code 13, "communication adminitratively
prohibited by filtering."  When RFC 792 was written there were no such
things as Internet firewalls (well, there was a plan to put some kind of
email-only filter between the ARPANET and MILNET, but it was never
implemented).  Now they are commonplace, and this code was introduced to
permit them to send a more informative error message (so users can tell the
difference between a router configuration error and an intentional
firewall).
-- 
Barry Margolin                                                [email protected]


-----------[000053][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 18:58:02 GMT
From:      Thomas Schoenauer <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Re: RFCs for Class D addressing ?
Could please someone tell me the relevant RFCs for
Class D addressing (multicast)

-- 
Thomas Schoenauer                    [email protected]
Ernst-Krebs-Str. 3
D-82131 Gauting

-----------[000054][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 19:06:15 GMT
From:      [email protected] (Not On Our Customer File)
To:        comp.protocols.tcp-ip
Subject:   Broadcast/Multicast file transfer ?
Is there a way to broadcast data packets to multiple hosts simultaneously?

I'm wondering if there is a broadcast or multicast file transfer method/
protocol since TCP won't do this.
						Thanx !

-----------[000055][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 19:37:17 GMT
From:      [email protected] (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP application:a multithreaded approach !!
In article <[email protected]>, [email protected] (Kanghoon Lee) writes:
[...]
|> Is this another language I have learn to read this group? :)

Nope.  [email protected] (Manoj Gupta) just inexplicably
rot-13'ed his text:

|> >Hi netters,
 
|> >        I am writing an TCP-IP application with muti-threaded approach.
|> >        I expect to have large number of client sessions (around 3500)
 
|> >        Each client session will have two socket connections.The application
|> >        will be running on solaris sparc1000 .I intend to run all these
|> >        sessions on a single machine.Has anyone developed a similiar
|> >        application for such a large number of sessions using multi threads?
|> 
|> 
|> >        I shall appreciate if you could suggest/recommend any ideas as
|> >        to implement this kind of application.
 
|> >        Thanks in advance !!
 
|> >	-manoj
|> 
|> 
|> >[EOB]
|> 
|> 
|> >-- 
|> >Manoj Gupta 		      | Internet: [email protected]
|> >DEC India		      | 
|> >Boxboro MA tel # 508-264-5388 | Disclaimer: This message reflects my own views.

I guess it's a dirty networking joke.  :-|

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

-----------[000056][next][prev][last][first]----------------------------------------------------
Date:      2 Sep 1994 19:39:34 GMT
From:      [email protected] (Rick Watson)
To:        comp.sys.dec,comp.unix.osf.osf,comp.protocols.tcp-ip
Subject:   Direct access to ethernet
I have an application where I essentially need to tunnel IP
packets between numerous clients connected over proprietary
serial links and an ethernet. 

I had originally planed on using an HP 9000/715 with 2 ethernets, the
first for normal system IP connectivity and the 2nd ethernet would be
the interface for my IP tunnel using the LLA packet intervce.
Unfortunately, I found that the LLA ethernet interface does not allow
me to use the IP and ARP ethertypes, even though I'm not configuring
the system's IP on those interfaces. 

My next hardware choice is a DEC Alpha running OSF. My question
is what software interface to the ethernet should I use:
packet-filter, Streams or DLPI(/Streams?)? It is acceptable to
use a 2nd ethernet for my tunnel interface.

Rick Watson 
The University of Texas Computation Center, Networking Services, 512/471-8220
 [email protected]

-----------[000057][next][prev][last][first]----------------------------------------------------
Date:      02 Sep 1994 20:17:10 GMT
From:      [email protected] (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: Packet Driver for SMC Token Ring
In article <[email protected]> [email protected] writes:

   Can someone direct me to a packet driver for an SMC8100 Token Ring card?

   I am trying to set up Mosaic.  I downloaded the Crynwr Packet group, but
   wasn't able to find it.

Most packet driver client programs are looking for a type 1 (Ethernet)
packet driver.  If you have an IBM LAN Support driver, you can use the
Crynwr IBMTOKEN.COM.  If you have an ODI driver, you can use oditrpkt.

--
-russ <[email protected]>    http://www.crynwr.com/crynwr/nelson.html
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)  | What is thee doing about it?
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000058][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 00:15:34 GMT
From:      [email protected] (Jeffrey Mogul)
To:        comp.sys.dec,comp.unix.osf.osf,comp.protocols.tcp-ip
Subject:   Re: Direct access to ethernet
In article <[email protected]> [email protected] (Rick Watson) writes:
>I have an application where I essentially need to tunnel IP
>packets between numerous clients connected over proprietary
>serial links and an ethernet. 
>I had originally planed on using an HP [...]
>Unfortunately, I found that the LLA ethernet interface does not allow
>me to use the IP and ARP ethertypes, even though I'm not configuring
>the system's IP on those interfaces. 
>
>My next hardware choice is a DEC Alpha running OSF. My question
>is what software interface to the ethernet should I use:
>packet-filter, Streams or DLPI(/Streams?)? It is acceptable to
>use a 2nd ethernet for my tunnel interface.

As one of the authors of the packet filter interface, I know what
I would suggest.  Seriously, if you use it correctly (and I don't
have a lot of time now to explain what that means) you should be
able to see every packet ... run "tcpdump not ip" to see what I mean.
You should not have to use a second ethernet interface.

The packet filter is a "character special device" and probably much
faster than Streams or DLPI or any socket-based interface.  See the
paper by McCanne and Jacobson in a recent USENIX proceedings.

Look at the tcpdump sources on ftp.ee.lbl.gov for examples of how
to do things.

-Jeff

-----------[000059][next][prev][last][first]----------------------------------------------------
Date:      Sat, 3 Sep 1994 03:02:15 GMT
From:      [email protected] (mike macedonia)
To:        comp.protocols.tcp-ip
Subject:   Re: RFCs for Class D addressing ?
Thomas Schoenauer ([email protected]) wrote:
: Could please someone tell me the relevant RFCs for
: Class D addressing (multicast)

RFC1112, "Host Extensions for IP Multicasting", Steve Deering.

See also Internetworking with TCP/IP, Vol 1. by Comer. He has an excellent chapter on IP Multicast and Class D addressing.

--
Mike Macedonia | [email protected]
MAJ, USA       | CS Dept, Naval Postgraduate School,
               | Monterey, CA 93943
               | PH:(408) 656-2903  FAX:(408) 656-2814
------------------------------------------------------------

-----------[000060][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 18:35:46 -0700
From:      [email protected] (Donald Edwards)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
Dennis Sutherland ([email protected]) wrote:
: In article <[email protected]>, [email protected] (Kenneth Croll) says:
: >The Novell file server only uses RIP for routing.  It does not support multiple
: >subnet masks for the same Class B address as you are trying to do.
 
: Logically this makes sense to me, but I have seen it work?!?  Go figger.

Kenneth is correct *if* your Novell server is *not* enhanced with
Novell's Multiple Protocol Router software, an extra-cost item
which is specifically intended to increase routing capability.

: > Also, your mask of ff.ff.ff.c0 will only provide 2 legal subnets.  You may
: >not use either a subnet or host with all bits on or all bits off.  The c0 mask
: >will give you hosts 65-126 and129-190.
 
: If this were a class C address the above statement would be true, because you
: can't have subnets with all 0's or all 1's.  However, since this is a class B
: address the 3rd byte is also part of the subnet address.  Thus the ranges of
: 1 - 62, and 193 - 254 are also available.

I don't know, but I suspect that the software is only aware of octets.
The division between class B and class C is completely arbitrary.  If
my suspicion is correct, then those two ranges aren't available.

-- 
-- "Self-interest has no place in society" -- Sandy Cuney, HCI, 5/6/94


-----------[000061][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 18:40:02 -0700
From:      [email protected] (Donald Edwards)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP application:a multithreaded approach !!
Kanghoon Lee ([email protected]) wrote:
: [email protected] (Manoj Gupta) writes:
 
: >Uv arggref,
 
: >        V nz jevgvat na GPC-VC nccyvpngvba jvgu zhgv-guernqrq nccebnpu.
: >        V rkcrpg gb unir ynetr ahzore bs pyvrag frffvbaf (nebhaq 3500)
 
: Is this another language I have learn to read this group? :)

Nah, it's just rot13 encoding.  If you're using tin, just hit 'D'.
-- 
-- "Self-interest has no place in society" -- Sandy Cuney, HCI, 5/6/94


-----------[000062][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 14:47:56 -0400
From:      [email protected] (michael v mascari)
To:        comp.protocols.tcp-ip
Subject:   Routing problem
	Thank you to those who suggested solutions to my routing
problem with an Amiga 4000 running SLIP through a 486DX onto a 
Class "B" network.  After subnetting the 4000 (and playing around with
the TCPDOS.INI file on the 486), & adding a static route onto our RS/6K to 
the new subnet, I was able to get the gateway working properly.

	Again, thanks.

	Mike Mascari ([email protected])



-----------[000063][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 11:40:11 GMT
From:      [email protected] (Krzysztof Rozycki)
To:        comp.protocols.tcp-ip
Subject:   Re: Packet Driver for SMC Token Ring
[email protected] wrote:
: Can someone direct me to a packet driver for an SMC8100 Token Ring card?
:  
: I am trying to set up Mosaic.  I downloaded the Crynwyr Packet group, but
: wasn't able to find it.
:  
: Thanks,
: Tim :-)

You can find it at ftp.smc.com.

--
-----------------------------------------------------------------------
Krzysztof Rozycki
OETO Politechnika Krakowska
Krakow, Warszawska 24
tel. (12) 330300 w. 2101           e-mail: [email protected]
-----------------------------------------------------------------------

-----------[000064][next][prev][last][first]----------------------------------------------------
Date:      Sat, 3 Sep 1994 13:15:01 GMT
From:      [email protected] (Kenneth Manheimer)
To:        comp.protocols.tcp-ip
Subject:   rpc - des authentication questions
I am looking for guidance over what seems to be one last, small hurdle
in hooking up des authentication with a scripting-languge-level
rpc/xdr package.  I'm really hoping that someone can help me fill in
this one gap, so i can be on my way...

I'm working on SunOS 4.1.3, and have gotten pretty far, using the
SunOS 5.3 docs and the RFC (1057).  I can produce a des credential at
the C level, and pass whatever portions are relevant to the scripting
level.  I just need to identify the proper fields of the authdes_cred
credential to use.

The problem is that the docs seem to presume that i will not be
picking apart the credential "myself", but instead passing the entire
thing (as the cl_auth component of the CLIENT structure) to the C
level rpc routines, and the rcp routines would encode the proper parts
into the header.

I, however, need the individual pieces of the header at the scripting
level, where i can handle them using the basic scripting-level xdr and
rpc-sockets routines.  I have no problem with encoding or sending any
of the pieces, i just need to know *which* portions of the
authdes_cred i need to pack! (Am i supposed to pack the entire thing?)

Can anyone out there help me, or point me to some references which
better explain the solution??  I'm so near yet so far!

Thanks in advance for any help,

Ken Manheimer	      					301 975-3539
[email protected]				   FAX: 301 963-9137

	Computer Systems and Communications Division
	Network Engineering Group

		National Institute of Standards and Technology
		Technology A151
		Gaithersburg, MD 20899

-----------[000065][next][prev][last][first]----------------------------------------------------
Date:      Sat, 3 Sep 94 16:31:31 GMT
From:      [email protected] (Greg Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: Q: Get MAC Address by ARP command?

In article <[email protected]> [email protected] (Pisut Tranchindavong - SCCS - 3737111) writes:
> I used to work with one kind of UNIX system that come together with ARP 
> command telling about MAC addresses on the network connected. Now I'm 
> working another UNIX system, NCR series, it doesn't have a ARP command 
> with it. Are there any commands in Internet that have an ARP function?
> 
> Regards,
> Pisut T.
> 

If the Box supports SNMP, try fetching the MIB II ipNetToMediaTable.

-- 
Greg Jones
Data General Corp.

-----------[000066][next][prev][last][first]----------------------------------------------------
Date:      Sat, 03 Sep 1994 17:10:50 GMT
From:      [email protected] (Mark T. Dornfeld)
To:        comp.protocols.tcp-ip
Subject:   lpd on SysV R 4 - AT&T emulation
I am trying to get AT&T's BSD LPD emulation working on a DG/UX box.
According to the docs there is a listener daemon (not LPD) running on
the normal lpd port.  When I send a job from an LPD client, in this
case Intercon's print client for MS-Windows, the connection is made
and then lost without any further diagnostics, and the job doesn't print,
of course.

Does anyone know whether this implementation requires the setup of
/etc/printcap in addition to the normal Sys V print spooler?  I have
assumed up until now, that the BSD emulation just routes jobs to the
appropriate System V printer.  I have also setup /etc/hosts.equiv which is
usually all it takes to get a job printing.

I have set this up numerous times on Sun boxes with no problems, but have
not had the pleasure of doing it on System V before.

Thanks in advance.
-- 

Mark T. Dornfeld, Cyantic Systems Corporation       Voice: (416) 621-6166
1 Eva Road Suite 301                            Facsimile: (416) 621-6212
Etobicoke, Ontario, M9C 4Z5 CANADA                  Email: [email protected]

-----------[000067][next][prev][last][first]----------------------------------------------------
Date:      03 Sep 1994 18:29:57 GMT
From:      [email protected] (Steinar Haug)
To:        comp.protocols.tcp-ip
Subject:   Re: Applying for Class B
> A local company providing IP connectivity got a class B address after they
> couldn't figure out how to supernet their 8 class C's.  So, it seems the
> less you know about networking, the more chance that you will be able to
> get a class B.

Seems to me that the agency handing out the class B address wasn't doing
the job, in this case...

Steinar Haug, SINTEF RUNIT, University of Trondheim, NORWAY
Email: [email protected]

-----------[000068][next][prev][last][first]----------------------------------------------------
Date:      03 Sep 1994 19:05:39 GMT
From:      [email protected] (Steinar Haug)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
> This very subject turn into a lively debate here a couple of months ago. I'll
> try to clearly state what I learned during that debate. In a subnetted 
> class C network the number of usable subnets is (2 ^ n) - 2 where n is the
> number of bits from the fourth octet used for the subnet mask. You might
> expect to get (2 ^ n) subnets but RFC 1022 ( I think...how quickly we forget)
> states that addresses with subnet mask values of all ones or all zeros
> should not be used. For example:
> 
> in a class C net with a 255.255.255.192 subnet mask,
> the addresses  X.X.X.0 thru X.X.X.63     (first two bits zeros) and
>                X.X.X.192 thru X.X.X.255  (first two bits ones)
> should not be used.

Most of today's systems don't support variable length subnet masks
(VLSM), and for such systems the above is true. However, all the major
router vendors and *some* Unix systems (BSD 4.4 based ones) support
VLSMs, and in that case the situation is more complicated :-)

With VLSMs (necessary to support CIDR, see RFC 1519), you can utilize the
address space more efficiently. Routing lookups are based on *longest*
match, and this means that you can for instance subnet the class C net
with a mask of 255.255.255.224 (27 bits) in addition to the subnet mask
of 255.255.255.192 (26 bits) given above. You will then be able to use
the addresses x.x.x.33 through x.x.x.62 (first three bits 001) and the
addresses x.x.x.193 through x.x.x.222 (first three bits 110) with this
new subnet mask. And you can continue with a subnet mask of 28 bits, etc.
(Note also, by the way, that non-contiguous subnet masks are deprecated.)

This is all very nicely covered in the paper by Havard Eidnes:

  Practical Considerations for Network Address using a CIDR Block Allocation
  Proceedings of INET '93

This paper is available with anonymous FTP from

	aun.uninett.no:/pub/misc/eidnes-cidr.ps

The same paper, with minor revisions, is one of the articles in the
special Internetworking issue of Communications of the ACM (last month,
I believe).

> I have be told that some network equipment (Cisco I think was the vendor 
> named) will not correctly handle subnets that violated that standard.

As far as I know cisco is one of the router vendors that *do* handle
VLSMs correctly. Could you substantiate this claim?

Steinar Haug, SINTEF RUNIT, University of Trondheim, NORWAY
Email: [email protected]

-----------[000069][next][prev][last][first]----------------------------------------------------
Date:      Sat, 3 Sep 94 20:49:39 GMT
From:      [email protected] (Frank Lofaro)
To:        comp.protocols.tcp-ip
Subject:   Re: HELP!! binary file xfer over Telnet sessio
In article <[email protected]> [email protected] writes:
>[email protected] (Ashok Aiyar) writes:
>
>> I had exactly the same problem, except that it was on a Linux machine, and
>> not on a Sun.  I ended up grabbing the latest BSD telnetd source, and 
>> compiling it for Linux.  Now file transfers work fine whether users telnet
>> in or rlogin into my machine.
>> 
>
>Well it would - the BSD telnetd has decent options negotiation and
>control mechanisms... including a working BINARY switch that tells the
>tty to go to binary, and I think visa versa - just been poaching it
>myself!
>
>Adam

A Linux telnetd I saw did null stuffing after line feeds instead of after 
carriage returns when in ASCII mode. With that kind of brain-damage, g*d 
only knows _what_ could be wrong with it. Heck, the linecheck program 
that comes with TERM would not even run (it would hang since it never saw 
a good "packet"). It would help if people read RFC 854, etc, and actually 
heeded the advice there (and did more bug testing).


-----------[000070][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 21:33:57 GMT
From:      [email protected] (Ruediger Volk)
To:        comp.protocols.tcp-ip
Subject:   Re: Applying for Class B
In article <[email protected]>, [email protected] (Steinar Haug) writes:
  > > A local company providing IP connectivity got a class B address after they
  > > couldn't figure out how to supernet their 8 class C's.  So, it seems the
  > > less you know about networking, the more chance that you will be able to
  > > get a class B.
  > 
  > Seems to me that the agency handing out the class B address wasn't doing
  > the job, in this case...

and hopefully that "local company providing IP connectivity" does not boast
about being BIG guys showing off their power end expertise - as indicated
by having "our own class B number" :-(

--------
Ruediger Volk
Universitaet Dortmund, Informatik IRB
D-44221 Dortmund, Germany

E-Mail: [email protected]
Phone:  +49 231 755 4760                 Fax:  +49 231 755 2386


-----------[000071][next][prev][last][first]----------------------------------------------------
Date:      3 Sep 1994 22:33:46 GMT
From:      [email protected] (David Rudder)
To:        comp.protocols.tcp-ip
Subject:   Help needed on KA9Q
I have KA9Q running a router at a (somewhat) distant site.  I don't like 
going over there, but have to make frequent changes to the setup.  All of 
the stuff I do I could be doing over telnet.  I have setup telnet to run 
on it, but I can't seem to create a password file.  As it stands, I can 
telnet there, but can not login because it doesn't recognize any logins.  
The docs don't explain this.

			Thanks in Advance,
			David Rudder
			[email protected]


-----------[000072][next][prev][last][first]----------------------------------------------------
Date:      Sun, 04 Sep 1994 02:05:29 GMT
From:      [email protected] (Mark T. Dornfeld)
To:        comp.protocols.tcp-ip
Subject:   Re: UNIX --> SAME PRINTER <-- Novell
In article <[email protected]> [email protected] (Kanghoon Lee) writes:
>
>>I am looking for a robust solution (Software of
>>Hardware) to allow laser printers to print from both
>>Novell Print queues as well as Unix Print queues.
 
>>We have currently standardized on Intel Netport Print
>>Servers, but could change to another vendor if it
>>provides a better solution.
 
>>Environment -
 
>>SunOS 4.1.3 using a variety of X-Terminals and
>>Character based terminals.
 
>>Novell 3.11 - Using mostly 386/486 based PCs running
>>WordPerfect & Lotus.
 
>>Thanks in advance for any suggestions.
 
>>Chris Lloyd
>>Manager, Production systems
>>American Institute of Physics
 
>>(516) 576-2313   [email protected]
>
>I have a perfect solution for you.  We have just released two new mult-network
>print systems that understands multiple protocols.  So you can print from Unix
>through TCP/IP, NetWare, LanMan, EtherTalk, and VAX.  Currently it fully 
>supports Novell Netware through 3.11, and works in Version 4 with 3.11 
>emulation.  More info is available on http://www.qms.com/www/QMSNews.html.
>
If you don't have any printers, then QMS is the way to go.  If you need to
retrofit older printers, then a number of print servers can do the trick.

We have had excellent experiences with the Microplex print servers.  They
are very fast and reliable.
-- 

Mark T. Dornfeld, Cyantic Systems Corporation       Voice: (416) 621-6166
1 Eva Road Suite 301                            Facsimile: (416) 621-6212
Etobicoke, Ontario, M9C 4Z5 CANADA                  Email: [email protected]

-----------[000073][next][prev][last][first]----------------------------------------------------
Date:      Sun, 4 Sep 1994 02:11:07 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: IP bcast: is all 0's dead?
In article <[email protected]> [email protected] (Bruce Wollen) writes:
>Any opinions out there?  Is there really anybody left that uses
>the old broken Berkeley software that used all 0's as the IP
>broadcast address (instead of all 1's like they should have done)?
>
>I'm wondering if it's reasonable to even implement support for
>this old broken broadcast address in current networking software
>products.

Every time you add or replace a host on a network using 0's broadcast
you can either go to every existing host and router on the network and
switch it to 1's, or you can make the new box use 0's.  Which do you
supposed happens 99.99% of the time?  Which is a sure fire way to break
something for at least a few hours as you forget to change a few systems,
or cannot find the manual on how to change the odd router or system?

There will be zillions of networks using 0's broadcast long after the
last VAX running 4.2BSD has been sent to a museum.  I bet most of the
networks running 0's broadcast are now composed only of machines that
prefer 1's broadcast, but the hassles of renumbering and the odd "legacy"
system that cannot be ignored or upgraded are enough to eliminate any
enthusiasm for change.

Theorem: Never underestimate the durability of "legacy" systems.
Corollary: don't bet the company on IPv6.


Vernon Schryver    [email protected]

-----------[000074][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1994 11:11:25 -0400
From:      [email protected] (Al Berg)
To:        comp.protocols.tcp-ip
Subject:   Re: Free IEEE Tutorial


[email protected] (John Gulbenkian) wrote:


>
>
>
>
>                              IEEE
>               OEB Communications Society Presents
>                         A Free Tutorial
>
>        Present Day Challanges of TCP/IP 
>Internetworking

This sounds great!  Since Hoboken, NJ is a bit of a commute, 
I'd love to see some
of the materials used for the seminar on an FTP site 
somewhere...  Is there any
possibility of this happening?

Al

-----------------------------------------------------------------
---------
Al Berg / [email protected] / Hoboken, NJ, USA
-----------------------------------------------------------------
---------




-----------[000075][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1994 23:00:50 -0500
From:      [email protected] (Douglas C. Schmidt)
To:        comp.object,comp.client-server,comp.lang.c++,comp.protocols.tcp-ip,comp.unix.solaris
Subject:   ACE 2.15.1 now available
Hi,

Release 2.15.1 of the ADAPTIVE Communication Environment (ACE) is now
available.  The ADAPTIVE Communication Environment (ACE) is an
object-oriented network programming toolkit.  ACE encapsulates the
following user-level BSD UNIX and System V Release 4 (SVR4) UNIX IPC
facilities via type-secure, object-oriented interfaces:

	. UNIX IPC mechanisms -- Internet- and UNIX-domain sockets,
		TLI, Named pipes (FIFOs) and STREAM pipes (note
		that a Windows NT version of Internet-domain
		sockets is also now available)

	. Event multiplexing via select and poll Solaris and DCE
		pthreads 

	. SVR4 explicit dynamic linking facilities -- 
		dlopen/dlsym/dlclose

	. The mmap family of memory-mapping APIs

	. System V IPC -- shared memory, semaphores, message queues

In addition, ACE includes a set of higher-level network programming
frameworks that integrate and enhance the lower-level C++ wrappers to
support the dynamic configuration of concurrent network daemons
composed of complex distributed application services.  These
frameworks include the following:

	. Reactor Framework 

	. Service Configurator Framework

	. ADAPTIVE Service Executive Framework 

Many of the C++ wrappers and higher-level components have been
described in issues of the C++ Report, as well as in proceedings of
the following journals, conferences, and workshops:

    	. The 2nd C++ World conference, October, 1993

	. The 11th and 12th Annual Sun Users Group Conference in
		December, 1993 and June, 1994 

	. The 2nd International Workshop on Configurable Distributed
		Systems, March, 1994

	. The 6th USENIX C++ Conference, April, 1994

	. The 1st Conference on the Pattern Languages of Programs,
		August, 1994

	. the 9th OOPSLA Conference to be held in October, 1994 
	. 3rd C++ World conference in November, 1994

	. IEE Distributed Systems Engineering Journal, to appear 1995

ACE components are currently being used in a number of commercial
products including the AT&T Q.port ATM signaling software product, the
Ericsson EOS family of PBX monitoring applications, and the network
management portion of the Motorola Iridium mobile communications
system.

OBTAINING ACE

An HTML version of this README file is available at URL
http://www.cs.wustl.edu/~schmidt/ACE.html.  All software and
documentation is available via both anonymous ftp and the Web.

ACE is available for anonymous ftp from the ics.uci.edu (128.195.1.1)
host in the gnu/C++_wrappers.tar.Z file (approximately .5 meg
compressed).  This release contains contains the source code,
documentation, and example test drivers for C++ wrapper libraries and
higher-level network programming frameworks developed as part of the
ADAPTIVE project at the University of California, Irvine and
Washington University in St. Louis.

The following subdirectories are included in C++_wrappers.tar.Z file:

	. apps    -- complete applications written using the ACE wrappers

        . bin     -- utility programs for building this release

        . build   -- a separate subdirectory that keeps links into the main
                     source tree in order to facilitate multi-platform
                     build-schemes

        . include -- symbolic links to the include files for the release

        . lib     -- object archive libraries for each C++ wrapper library

	. libsrc  -- the source code for the following C++ wrappers:

		. ASX -- higher-level C++ network programming framework
		. Get_Opt -- a C++ version of the UNIX getopt utility
		. SOCK_SAP -- wrapper for BSD sockets
		. TLI_SAP -- wrapper for SVR4 TLI 
		. FIFO_SAP -- wrapper for FIFOS (named pipes)
		. SPIPE_SAP -- wrapper for SVR4 STREAM pipes and connld 
		. Log_Msg -- library API for a local/remote logging
			facility 
		. Mem_Map -- wrapper for BSD mmap() memory mapped files 
		. Message_Queues -- wrapper for SysV message queues
		. Reactor -- a framework for event demultiplexing and
			event handler dispatching 
		. Semaphores -- wrapper for SysV semaphores
		. Service Configurator -- a framework for dynamically
			linking/unlinking 
		. Shared_Memory -- wrapper for SysV shared memory
		. Shared_Malloc -- wrapper for SysV/BSD shared mallocs 
	
  	. tests -- programs that illustrate how to use the various wrappers
	. WIN32 -- contains the versions of ACE that are ported to 
		   Windows NT (currently on the SOCK_SAP C++ wrappers
		   for sockets are ported)

In addition, a relatively complete set of postscript documentation and
papers is included with the release.  The documentation is stored in
gnu/C++_wrappers_doc.tar.Z file (approximately 2.5 meg compressed).
The following directories are included along with the documentation.

	. doc     -- LaTeX documentation (in both latex and .ps format)

	. papers  -- postscript versions of various papers describing ACE 

Please note that there are companion tar files called
C++_wrappers_doc.tar.Za[a-c].  I used the UNIX "split" command to make
sure each of these files is less than 1.2 Meg in size to accommodate
ACE users who only have access to modem connections on PCs.  To
recreate the original tar file, simply to the following:

% cat C++_wrappers_doc.tar.Za[a-c] doc.tar.Z
% uncompress doc.tar.Z
% tar xvf doc.tar

BUILDING AND INSTALLING ACE

Please refer to the INSTALL file for information on how to build and
test the ACE wrappers.  The overall ACE release is very large (~2
Meg).  Therefore, I'm sorry, but I will be unable to distribute the
ACE wrappers via email.  The BIBLIOGRAPHY file contains information on
where to obtain articles that describe the ACE wrappers and the
ADAPTIVE system in more detail.

The current release has been tested fairly extensively on Sun
workstations running Sun OS 4.1.x and Solaris 2.x using GNU G++ and
Sun C++ 3.x and 4.x.  Portions of the release have also been ported to
SCO UNIX, HP-UX, OSF/1, Windows 3.1 and Windows NT.  I expect that
major portions of the release will port easily to other platforms.  If
anyone is willing to help coordinate ports to other platforms please
let me know.

ACE MAILING LIST

A mailing list is available for discussing bug fixes, enhancements,
and porting issues regarding ACE.  Please send mail to me at the
[email protected] if you'd like to join the mailing list.

COPYRIGHT INFORMATION

You are free to do anything you like with this code.  However, you may
not do anything to this code that will prevent it from being
distributed freely in its original form (such as copyrighting it,
etc.).  Moreover, if you have any improvements, suggestions, and or
comments, I'd like to hear about it!  It would be great to see this
distributed evolve into a comprehensive, robust, and well-documented
C++ class library that would be freely available to everyone.
Naturally, I am not responsible for any problems caused by using these
C++ wrappers.

        Thanks,
	
	Douglas C. Schmidt 
	[email protected] 

ACKNOWLEDGEMENTS
        
Special thanks to Paul Stephenson for devising the recursive Makefile
scheme that underlies this distribution, as well as for devoting
countless hours to discussing object-oriented techniques for
developing distributed application frameworks.

Thanks to Olaf Kruger for explaining how to instantiate templates for
shared libraries on SunOS 4.
 

-- 
Dr. Douglas C. Schmidt 			([email protected])
Department of Computer Science, Washington University
St. Louis, MO 63130. Work #: (314) 935-7538; FAX #: (314) 035-7302
http://www.cs.wustl.edu/~schmidt/

-----------[000076][next][prev][last][first]----------------------------------------------------
Date:      Sun, 4 Sep 1994 16:14:28 +0000
From:      [email protected] (Niall Teasdale)
To:        comp.protocols.tcp-ip
Subject:   Re: IPng information?

Kevin L White ([email protected]) wrote:
: I am looking for information on the proposed IPng standard.  I read
: the short FAQ here and have scanned the messages for about a week and
: haven't seen anything.  Can anybody point me to information?  E-mail
: me if you want and I will post a summary.
 
: Thanks!
 
: Kevin
 
: -- 
: Kevin White --- [email protected] --- Finger for Pub PGP Key! 
: "Where...the ENIAC is equipped with 18,000 vacuum tubes and weighs 30
: tons, computers in the future may have 1,000 vacuum tubes and perhaps
: weigh just 1-1/2 tons."           Popular Mechanics, March 1949

A lot of IPng related RFCs have just appeared. These are numbered
1167 thru 1688. I have not read them yet, though I plan to give them at
least a quick looking over soon. Perhaps they will be some use to you.

Niall.

~==========================================================================~
 Niall Teasdale - [email protected]
 Hedgehog Software (at home) - Cogsys Ltd (at work)
 PGP key available on request or from public key servers.
 Any views expressed above are mine, and mine alone.

-----------[000077][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1994 16:48:57 GMT
From:      [email protected] (William C. Fenner)
To:        comp.protocols.tcp-ip
Subject:   Re: IP bcast: is all 0's dead?
In article <[email protected]>,
Bruce Wollen <[email protected]> wrote:
>Is there really anybody left that uses
>the old broken Berkeley software that used all 0's as the IP
>broadcast address (instead of all 1's like they should have done)?

Solaris 1.x (SunOS 4.x) uses all 0's by default.  There appear to be lots
of people out there who prefer Solaris 1 to Solaris 2.

  Bill
-- 
Bill Fenner                  [email protected]

-----------[000078][next][prev][last][first]----------------------------------------------------
Date:      Mon, 05 Sep 94 02:56:15 EST
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   bind() and EADDRINUSE ?!



IF> 2. Is it possible to identify which process uses a particular port ?

Try "netstat -a".  Look for sockets in the TIME_WAIT state.


IF> 4. What is the "fastest" mechanism that can allows a program to reuse a
IF>    particular port if the previous process is already dead?

Enable the SO_REUSEADDR socket option just before bind:

   int reuse=1;
   setsockopt(socket,SOL_SOCKET,SO_REUSEADDR,&reuse,sizeof reuse))

--
==============================================================================
| ... The Galacticomm Demo System - 305/583-7808 - Home of The Major BBS ... |
==============================================================================


-----------[000079][next][prev][last][first]----------------------------------------------------
Date:      Mon, 05 Sep 94 02:59:37 EST
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Bug in non-blocking connect()s ?


IR>Does anyone have any experience putting a TCP socket into non-blocking
IR>(FNDELAY) mode and then doing a connect() ?

Funny, but in the stack I use, I set the socket to non-blocking with
ioctl(socket,FIONBIO,&1).  And you are select()ing for SEND to find when
the connection is made aren't you?  Not RECV?

-- Bob Stein

--
==============================================================================
| ... The Galacticomm Demo System - 305/583-7808 - Home of The Major BBS ... |
==============================================================================


-----------[000080][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 02:45:34 GMT
From:      [email protected] (Serguei Vodopianov)
To:        comp.protocols.tcp-ip
Subject:   WANTED: Computer Networks
Computer Networks by Andrew Tanenbaum
2nd edition

Please reply to [email protected]


-----------[000081][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 03:40:10 GMT
From:      [email protected] (S.T. Wong)
To:        comp.protocols.tcp-ip
Subject:   Help: How to close FIN_WAIT_2 connections ?


--
S.T. Wong                                | BITNET: [email protected]
Computer Services Centre                 | Internet: [email protected]
The Chinese University of Hong Kong      | Tel. No: (852) 609 8825
Shatin, N.T., Hong Kong                  | Fax  No: (852) 603 5001

-----------[000082][next][prev][last][first]----------------------------------------------------
Date:      4 Sep 1994 08:24:43 +1200
From:      [email protected] (Quentin Smart)
To:        comp.protocols.tcp-ip,comp.sys.novell,comp.sys.proteon
Subject:   Re: Netware TCP Performace Question/Bug?
What version of the TCPIP.NLM are you running? If it's the one that 
shipped with NW3.11 it would pay to upgrade. FTP TCP187.exe from a Novell 
archive.

-----------[000083][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 05:08:30 GMT
From:      [email protected] (Paul Quinn)
To:        comp.protocols.tcp-ip
Subject:   Re: UNIX --> SAME PRINTER <-- Novell
In article <[email protected]> [email protected] (Mark T. Dornfeld) writes:
>In article <[email protected]> [email protected] (Kanghoon Lee) writes:
>>
>>>I am looking for a robust solution (Software of
>>>Hardware) to allow laser printers to print from both
>>>Novell Print queues as well as Unix Print queues.
 
>>>We have currently standardized on Intel Netport Print
>>>Servers, but could change to another vendor if it
>>>provides a better solution.
 
>>>Environment -
 
>>>SunOS 4.1.3 using a variety of X-Terminals and
>>>Character based terminals.
 
>>>Novell 3.11 - Using mostly 386/486 based PCs running
>>>WordPerfect & Lotus.
 
>>>Thanks in advance for any suggestions.
 
>>>Chris Lloyd
>>>Manager, Production systems
>>>American Institute of Physics
>>


I have installed Unix and Novell where the Unix user print and the novell
queue takes the print job.

All you need is TCP/IP for Novell.  In your printcap, set the printer as
remote and to access the novell server.

Then your users print as usual, but the novell print queue takes the jobs.

--
________
Paul Quinn
[email protected]
Computer Science: Systems Architecture
Concordia University
Montreal, QC, CANADA
--------

-----------[000084][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 07:14:59 GMT
From:      [email protected] (Herr Lein)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing unser SunOS and Class B nets
Colin Campbell ([email protected]) wrote:
: Hi,
 
: (First foray into this group so be gentle with me :-)
 
: SunOS 4.1.3C or SunOS 5.3
 
: We have a Class C and a Class B with C subnetting. I want to retain
: some control over which B-subnets a machine on the C net can see. My
: original guess was this: (we use static routes)
 
:     On `C' (203.5.10.10)
 
: 	route add net 147.132.157.0 203.5.10.2 4
: 	route add net 147.132.176.0 203.5.10.2 2

[stuff deleted]

: /etc/netmasks contains `147.132.0.0 255.255.255.0' but this is apparently
: irrelevant.
 
: What is broken?
 
: 	SunOS? (Can it be fixed?)
: 	Me? (Can't be fixed.)
: 	Nothing? (That is how it is supposed to work.)
 
: I do not believe the last option cos our network guru says I should be
: able to do what I am trying to.
 
: Thanks
: Colin

As far as I know, one of the basic ideas of IP routing says, that ONE class B
net has ONE entry point. . So the routing algorithm
on SunOS reads only the first two bytes of a route entry for a class B net, all
further information are ignored. (Will say  "route add net 147.132.157.0 ..."
is equivalent to "route add net 147.132.0.0 ...") Take care: this is true
for all class B networks except those, the node has an interface to.
(Cum grano salis: Its the same for Class A or C, of cause)

You should be aware of the fact, that a subnet oriented routing is only possible,
if the system knows the subnet mask of the remote class B net. But there is no
way to tell it. "/etc/netmasks" could be the right file to store the data, but
I did not found any way to pass this data to the kernel, where the routing takes place.

Some routers (pure routers, as cisco's) gives you much more possibilities for
selecting or supressing routing. But this stuff only work for locally in the
router, it does not enable you to have a node in the net, which routes two
parts of a class B net to diferent gateways.

Yours sincerely
ln

--
Frank Lein
Sietec Systemtechnik GmbH & Co. OHG
Nonnendammalle 101
D-13629 Berlin
phone +49-30-386-27971
e-mail: [email protected]

-----------[000085][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 09:30:57 GMT
From:      [email protected] (Stefano Klett)
To:        comp.networks.noctools.wanted,comp.dcom.lans.misc,comp.protocols.tcp-ip
Subject:   Wanted dynamic IP-filter
To give more flexibility to users in business trip, we need a dynamic IP-filter.

The dynamic IP-filter are composed from a IP-filter and a programm:

- The filter can be a UNIX machine or a dedicated router. Very important is the 
  possibility to put new filters rules on-line from an other program (in the 
  same host or from an other host).

- The program must request an additional code (variable code) after the normal login. 
  The allowed user, can with this program add and delete entry in to the filter. 
  The delete is allowed only from entry added form the same user.

If you know something about filters or program with this requirement, please contact me immediately.


---


                           Stefano Klett

--
Stefano Klett              CSCS (Centro Svizzero di Calcolo Scientifico)
Tel. +41 91 50 82 15       Via Cantonale
Fax: +41 91 50 67 11       CH-6928 Manno (Switzerland)
E-mail: [email protected]     
X.400:  S=sklett O=cscs PRMD=SWITCH ADMD=ARCOM C=CH



-----------[000086][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 10:07:04 GMT
From:      [email protected] (Malcolm White)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
In article [email protected], [email protected] (Stephen Johnson) writes:
>[email protected] (Gregory LeBlanc) writes:
 
>
>I have be told that some network equipment (Cisco I think was the vendor 
>named) will not correctly handle subnets that violated that standard.
>

Cisco does allow the use of subnet zero although you must specifically tell the box to do so. It is not supported by default ( wisely ) unlike Wellfleet. The use of subnet zero is a very contentious issue. I'm not prepared to go into the (well battled) reasons for or against.

Regards, Malcolm White
Salomon Brothers, London


-----------[000087][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 10:34:27 GMT
From:      [email protected] (Dirk Braner)
To:        comp.protocols.tcp-ip
Subject:   Re: Make a telnet session to modem look like COMx?
Barry Flanagan ([email protected]) wrote:
: I can't recall where I FTPed the demo (it's shareware), but the company,
: Performance Designs has a CI$ account  [email protected]
 
: Hope this helps.
 
: -Barry Flanagan

You can find it on cica (ftp.cica.indiana.edu) in the winsock directory.

:-> Dirk Braner

-----------[000088][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 17:38:13 EST
From:      MARCO PACE <[email protected]>
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Lost UDP packets above network interface layer !!
Hello everybody,

I'm writing a couple of applications (running on SUN SPARCstations with
Solaris 2.3) communicating using UDP via an Ethernet LAN. The API used to
access the communication primitives is the socket interface.

The problem I am experiencing is a loss of UDP packets sent over the network.
But the problem is occuring not at the network level, but above that. I managed
to check this using a network 'sniffer', which was able to capture all of
the sent packets. The receiver application, on the contrary was not only
unable to get all of the packets, but missed a huge amount of them !!

I suspect the packets are dropped somewhere between the network interface
and the socket interface layer, but I have no clue where and I find it diffi-
cult to detect where the problem is occurring. The tools available on Solaris
(snoop, netstat, etc.) don't seem to provide a valid aid.

What I'd like to know is:
- did anybody have a similar problem and can provide me with some useful hint
  as to where the problem is located ?
- which tool can I use to detect packets dropped at the different software
  layers ?

Thanks,

Marco


-------------------------------------------------------------------------
Marco Pace, Software Engineer                   Tel   : +49 6151 903065
Vitrociset (Space Division)                     Fax   : +49 6151 904065
ESOC, Robert-Bosch-Strasse 5                    e-Mail: [email protected]
D-64293 Darmstadt, Germany   x400: C=DE;A=DBP;P=ESA;O=ESOC;S=PACE;G=MARCO

-----------[000089][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 14:32:08 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: HELP!!  Etherlink III hangs after heavy traffic on pentium
In article <[email protected]> [email protected] (Joe Debiso) writes:
>I have a 90mhz pentium with a 3Com Etherlink III (3c579) that 
>keeps hanging.  If I use ftp or telnet and create heavy usage, 
>such as a file transfer of several megabytes the process hangs. 
>diags provided by 3com.  SCO was no help at all.

Passing diagnostics is just a good sign. I suggest you get a DOS disk and
DOS ftp up and retry the transfer. If it works under DOS you probably ought
to ask SCO some searching questions, like 'Why did I pay all this money'. If
it doesn't then you can start checking out the card and motherboard for
compatibility.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000090][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 14:34:42 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: tcplink? (sp?) Where is it?
In article <[email protected]> [email protected] (Ron Schnell) writes:
>I remember a program a couple of years ago, called "tcplink", I think
>(but I might be wrong).  It allowed you to connect a tcp port to
>a pty, and then connect the pty to a program, so that, for example,
>someone could telnet to your site, at a known port, and this program
>would run, using the connection as stdin/out.

Its called telnetd, its freely available and you just stuff it on a
different inetd port and compile a version to run something other than
login 8)

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000091][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 16:17:35 GMT
From:      [email protected] (Juergen Schoenwaelder)
To:        comp.lang.tcl,comp.protocols.snmp,comp.protocols.tcp-ip,comp.unix.admin
Subject:   ANNOUNCE: tkined-1.0 and scotty-1.0
Release 1.0 of our network management software tkined and scotty is
now available on ftp.ibr.cs.tu-bs.de [134.169.34.15]. The files are:

	/pub/local/tkined/tkined-1.0.2.tar.gz
	/pub/local/tkined/scotty-1.0.2.tar.gz

	(This is actually patchlevel 2 due to some silly bugs.)

tkined and scotty are based on tcl 7.3 and tk 3.6 and compile at least
on LINUX, SUN OS, HP-UX, ULTRIX 4.3, IRIX, SCO, OSF1, 386BSD.


What is tkined?

	tkined is a drawing program that allows you to create maps of
	your network configuration. But the most important feature of
	tkined is its tcl-based programming interface that allows you
	to use tkined as a GUI for a network management system. tkined
	provides commands to present status information in stripcharts
	or barcharts.


And why do I need scotty?

	scotty is a tcl interpreter that has extensions to set up TCP
	and UDP connections, to submit ICMP packets, to query the
	domain name system (DNS), to check clock screw using the NTP
	protocol and to query various SUN RPC services (like rstat and
	mountd). Included are some sample scripts for troubleshooting
	your network (ping, traceroute, finger, query tcp services,
	query RPC services), for monitoring network status plus
	scripts to discover and layout the topology of your IP network.

	scotty also includes a tcl SNMP interface. There are scripts
	to dump routing tables, to query interface status and to
	monitor SNMP variables. There is also a MIB browser to inspect
	the MIB hierarchy.


Where can I get more information about tkined and scotty?

	Visit our WWW server www.ibr.cs.tu-bs.de. It contains some
	sample maps, screen dumps of a running tkined editor and links
	to some related documents. A guide to the 1.0 version is in
	preparation and will appear on this server in the next days.

	If you have problems or if you have made any changes to run
	tkined and scotty on your hardware or if you have found any
	bugs, please contact us. You are also invited to join our
	tkined mailing list.  To join, send a request to

		[email protected] 

	Messages to the list should be send to [email protected]


Why did we change to version number 1.0?

	The programming interface of tkined and scotty has been
	revised to fix some oddities. We hope that we can provide
	some sort of compatibility regarding the tcl interface
	starting with the 1.0 release. Note, this promise does not
	include the SNMP interface which will be replaced with our
	own SNMPv1/v2 stack soon.


---
Juergen Schoenwaelder   ([email protected])    (Tel. +49 531 / 391-3249)
Technical University Braunschweig, Dept. Operating Systems & Computer Networks
Bueltenweg 74/75,  D-38106 Braunschweig, Germany. (http://www.ibr.cs.tu-bs.de)


-----------[000092][next][prev][last][first]----------------------------------------------------
Date:      5 Sep 1994 18:29:24 GMT
From:      [email protected] (Joerg Vreemann)
To:        comp.protocols.tcp-ip
Subject:   Re: Where to find WINSOCK TN3270
In article <[email protected]>, [email protected] (David Brown)
wrote:
> 
> I am looking for a public domain, shareware, or even commercial TN3270
> application that uses WINSOCK/Microsoft TCP/IP stack.  Any suggestions
> would be appreciated.  Preferred commercial applications would only
> have the TN3270, I don't need or even want a bundled FTP, Gopher,
> Mosaic, Telnet, FTP solution.
> 
qws3270 - from ftp.cica.indiana.edu, for ex. (FreeWare)
I'm not sure about commercial tn3270 variants using WinSock.
> 
> tks...dkb

Joerg

Joerg Vreemann, RZ der Univ. Konstanz, Germany

-----------[000093][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 06:12:19 -0700
From:      [email protected] (Paul Smith)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
Stefan Sharkansky ([email protected]) wrote:
: In article <[email protected]> [email protected] (Lance Lu) writes:
: >
: >I use read( s, buf, len)
: >where s is unblocking TCP stream socket.
: >
: >How do I know if the remote socket is closed when I call read()?
 
: read() will return 0.
 
: This is different from the case where there is no data (on a non-blocking
: socket) where read() will return -1 with errno=EWOULDBLOCK.

Or errno=EAGAIN on some TCP/IP stacks (Lachman on SVR4.2).

But, this is one of my greatest grips about SOCK_STREAM and circuit failures! 
Why couldn't a reasonable errno be returned consistantly.  The Lachman stack
on SVR4.2 v1.1.2 returns 0, errno = 0 for the server when the client side
closes() or exits premature.  The Client side however returns -1, errno=EPIPE
if the server exits premature??

If you rumage /usr/include/errno.h you find several streams/socket 
oriented errnos like; ENOTCON that would make better sense in the above
scenarios.  And forget about poll()ing for the flags S_HANGUP or S_ERROR
or even getsockopt(fd, SOL_SOCKET, SO_ERROR) it doesn't detect the hangup
condition either???

It seems the wishy-washy return of 0, errno=0 is IT!!

...



-----------[000094][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 18:43:47 GMT
From:      [email protected] (W. Richard Stevens)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
> The problem I am experiencing is a loss of UDP packets sent over the network.
> But the problem is occuring not at the network level, but above that. I managed
> to check this using a network 'sniffer', which was able to capture all of
> the sent packets. The receiver application, on the contrary was not only
> unable to get all of the packets, but missed a huge amount of them !!
>
> I suspect the packets are dropped somewhere between the network interface
> and the socket interface layer, but I have no clue where and I find it diffi-
> cult to detect where the problem is occurring. The tools available on Solaris
> (snoop, netstat, etc.) don't seem to provide a valid aid.

Yes, it's quite possible to lose UDP datagrams between the interface
and the application, normally because the socket's receive buffer
doesn't have room for the new datagram and address.  You can change
the receiver's receive buffer with SO_RCVBUF socket option.

You can normally check for this with "netstat -s", looking at the
UDP "socket overflows" counter before and after.  I think the
Solaris 2.3 output shows this as udpInOverflows.

	Rich Stevens

-----------[000095][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 19:56:27 GMT
From:      [email protected] (Jim Raykowski)
To:        comp.protocols.tcp-ip
Subject:   Creating a SLIP/PPP driver.
Hello All,
    I am about to embark on a mission.  Create a shareware SLIP driver for 
OS/2's Work Place Shell (WPS).  Before beggining this adventure I need to get 
some information on how SLIP/PPP work.  I have the TCP/IP and PPP FAQ's are 
the anyother sources of info I should or BETTER use.  Like is there public 
domain source code I cound view and look at to give me a idea of how to start. 
 I have the interface built and working I just need to add the SLIP/PPP driver 
source code and test it out.

Thanks for you help,

Jim Raykowski          internet:  [email protected]
San Diego, CA          OS/2 BBS:  WZ02186
FAX: (619) 435-6814

-----------[000096][next][prev][last][first]----------------------------------------------------
Date:      Mon, 5 Sep 1994 23:53:28 GMT
From:      [email protected] (Al Longyear)
To:        comp.protocols.tcp-ip
Subject:   Re: tcplink? (sp?) Where is it?
[email protected] (Alan Cox) writes:

>In article <[email protected]> [email protected] (Ron Schnell) writes:
>>I remember a program a couple of years ago, called "tcplink", I think
>>(but I might be wrong).  It allowed you to connect a tcp port to
>>a pty, and then connect the pty to a program, so that, for example,
>>someone could telnet to your site, at a known port, and this program
>>would run, using the connection as stdin/out.
 
>Its called telnetd, its freely available and you just stuff it on a
>different inetd port and compile a version to run something other than
>login 8)

Strange, Alan, I thought that he was asking for 'inetd' itself. This
is one of the functions of inetd -- to bind a socket into stdin/stdout
handles and run the program with arguments.

BSD 4.4 allows RFC 1078 style extensions to inetd for the tcpmux
extension.  This may be more flexiable than the hard coding each
program in the inetd.conf file.

By the way, archie fails to find something called 'tcplink' as any substring.
-- 
Al Longyear           [email protected]

-----------[000097][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 00:41:17 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
In article <[email protected]> [email protected] (W. Richard Stevens) writes:
>> The problem I am experiencing is a loss of UDP packets sent over the network.
>> But the problem is occuring not at the network level, but above that. ...
 
>Yes, it's quite possible to lose UDP datagrams between the interface
>and the application, normally because the socket's receive buffer
>doesn't have room for the new datagram and address.  You can change
>the receiver's receive buffer with SO_RCVBUF socket option.
>
>You can normally check for this with "netstat -s", looking at the
>UDP "socket overflows" counter before and after.  I think the
>Solaris 2.3 output shows this as udpInOverflows.

There can also be losses on the transmit side, between the application
and the wire if the application tries to write more than the wire will
accept.  BSD style stacks tend to complain with ENOBUFS as the interface's
output buffer overflows.  There is absolutely no flow control with UDP
that is not implemented by the application.  That is why `ttcp` has that
funky 18 ms delay after each output UDP ENOBUFS.

Some UNIX systems count such output overruns as "drops" in the interface
structure and report it with `netstat -i`.


Vernon Schryver    [email protected]

-----------[000098][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 00:59:04 GMT
From:      [email protected] (Jim McGrady)
To:        comp.protocols.tcp-ip
Subject:   Re: RARP vs. BOOTP
Ralph E. Droms ([email protected]) wrote:
: In article <[email protected]> [email protected] (David Harding) writes:
 
: Yes, you can be guaranteed the same ethernet address.  It's an
: administrative policy decision.  DHCP provides the mechanism to
: support dynamic address reassignment, or fixed assignment (like
: BOOTP), or automatic assignment (first time use gets a new address,
: which is retained thereafter).

Ive had a look for any working DHCP packages on the net, and havent been able
to find any.  Are their any share/freeware packages available yet ?

Jim

-----------[000099][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 02:19:41 GMT
From:      [email protected] (Cheng Chang Huang)
To:        comp.protocols.tcp-ip
Subject:   [Q] Implementation of Multicast Transport Protocol Aavailable?
Hi,

I am looking for any information regarding prototype/implementation/evaluation
of rfc1301(Multicast Transport Protocol) by Armstrong et al.  

Thanks in advance!


Chengchang Huang
-- 
Chengchang Huang ([email protected])
Computer Science Dept.
Michigan State University

-----------[000100][next][prev][last][first]----------------------------------------------------
Date:      Tue, 06 Sep 1994 11:33:34 -0400
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   ftp location/ report on harvard router tests
could someone point me toward the location for the current report? thanks
very much.

-----------[000101][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 05:10:37 GMT
From:      [email protected] (000-Ryan(000))
To:        comp.protocols.tcp-ip,comp.os.linux.admin
Subject:   Re: Configure OSPF on gated 3.0.3 (Linux)?
Pete Kruckenberg ([email protected]) wrote:
: I need to figure out how to set up OSPF routing on my Linux (1.1.18)
: gateway. I'm using gated 3.0.3, with a current RIP configuration. 
: Unfortunately, our new Internet access provider uses OSPF (only),
: and I've never used OSPF.
 
: What kinds of information do I need to get from them? I've got a few
: sample configurations (from the gated 3.5 Alpha) that have OSPF stuff,
: but I don't know if they'll even work (so far they haven't).


Where can I get a copy of GATED for Linux with OSPF enabled? All the ones
I find dont have OSPF enabled.

-ryan

--
Ryan A. Whelan            Unix and the World Unixes with you. Vax and you
                                                               Vax ALONE!
[email protected]  [email protected]                      
Gopher: absolut.labs.gmu.edu:70         URL: http://absolut.labs.gmu.edu:80/

-----------[000102][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 06:21:58 GMT
From:      [email protected] (Pete A. Zaitcev)
To:        comp.protocols.tcp-ip
Subject:   Re: IP bcast: is all 0's dead?
In <[email protected]> [email protected] (Vernon Schryver) writes:

>>I'm wondering if it's reasonable to even implement support for
>>this old broken broadcast address in current networking software
>>products.
 
>Every time you add or replace a host on a network using 0's broadcast
>you can either go to every existing host and router on the network and
>switch it to 1's, or you can make the new box use 0's.  Which do you
>supposed happens 99.99% of the time?

Sun took a way when the Solaris broadcasts with 1's and accepts 0's.
So my network is "mixed-broadcasts". Old systems (PC-NFS 3.5 among
others [sic]) will survive here for a long time.

Pete

-----------[000103][next][prev][last][first]----------------------------------------------------
Date:      Tue, 06 Sep 94 16:14:26 PDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Tivoly TCP/IP Product

Does anybody have experience with a TCP/IP product called Tivoly, I'm looking
for information on what this product does and what its performance is like.

Thank you in advance for your response.


-----------[000104][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 09:16:23 GMT
From:      [email protected] (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
In article <[email protected]_dike.sbil.co.uk> [email protected]
writes: 
    
    Cisco does allow the use of subnet zero although you must specifically
    tell the box to do so. It is not supported by default ( wisely ) unlike
    Wellfleet. The use of subnet zero is a very contentious issue. I'm not
    prepared to go into the (well battled) reasons for or against.   
    
Using subnet zero with a classful routing protocol can lead to all sorts of
interesting ambiguities since the route for subnet zero is syntactically
identical to that of the network number itself.  All sorts of nasty things
can happen.

If you're using classless routing protools, then subnet zero is no longer
an issue, however, subnet zero, host zero is still considered a broadcast
address.

cisco also supports the all-ones subnet.  This has none of the ambiguity,
so is enabled by default.  Again, the all-ones subnet all-ones host is a
broadcast address.

Tony



-----------[000105][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 09:40:47 GMT
From:      [email protected] (Tony Li)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing unser SunOS and Class B nets
In article <[email protected]> [email protected] (Herr Lein) writes:

    Some routers (pure routers, as cisco's) gives you much more
    possibilities for selecting or supressing routing. But this stuff only
    work for locally in the router, it does not enable you to have a node
    in the net, which routes two parts of a class B net to diferent
    gateways.
    
Actually, that is technically possible.  However, there are a whole LOT of
reasons why it's a bad idea in the general case.

Tony

-----------[000106][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 17:06:20 -0400
From:      Mike O'Connor <[email protected]>
To:        comp.infosystems,comp.protocols.tcp-ip
Subject:   Any mirrors for the InterNIC WHOIS server?
Are there any mirrors of rs.internic.net out there?  I'm getting tired
of the following message:

*
* WELCOME to InterNIC Registration Services 
*
* Sorry, the system load is temporarily too heavy.
*
* Please wait a while and try again.  Thanks
*

						...Mike
 
--
 Mike O'Connor, [email protected]
 http://www.msen.com/~mjo/

"I always have to postpone what I WANT to do for what I HAVE to do!"  -Calvin

-----------[000107][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 10:56:45 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
In article <[email protected]> [email protected] (Vernon Schryver) writes:
>I hope the intent of that proscription is only to limit people using
>transient IP addresses.  There are otherwise too many millions of us
>who get mail over dial-up PPP and SLIP lines with fixed IP addresses.
>MX records can be handy for such situations, but they are certainly not
>necessary.

They are a very good idea. Otherwise you risk having your mail never arrive
because it times out at remote sites or goes via site that doesn't queue
retry and requeue email all day.

>for the rest of us.  Saying that you ought to use POP for mail with
>dynamic IP addresses is an obscure way of saying that dynamic IP addresses
>are painfully constricting.

Dynamic IP doesn't need POP. It needs someone to queue your email and other
incoming data that is sent on a 'must be delivered now' basis (that you
want). That can easily be mail and good mailers will happily queue for site and
deliver by smtp when you connect. Personally I hope IPng ensures there are
enough easily to get address spaces that nobody bothers with dial up IP. As
to TIA.. well its a fine way to do a job for simple users. Term is the same
sort of stuff but free. Both are very useful.

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000108][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 11:01:45 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: Can this configuration be put on the Internet?
In article <[email protected]> [email protected] (Matt Midboe) writes:
>Novell file server is an arcnet with a bunch of PCs. Now my question
>is how can I put all of these on the Internet. Do I need one class C
>for the ethernet, and one class C for the arcnet? How does the routing
>work from the arcnet? 

Arcnet works the same as anything else. If you have >126 arcnet hosts or
>126 ethernet hosts you want two networks. If you don't and are not likely
to have that you can get a class C network and subnet it on 7 bits (ie

x.y.z.1-126 are the arcnet.
x.y.z.129-254 are the ethernet.

Alan

-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000109][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 11:57:11 GMT
From:      [email protected] (Denis Lafont)
To:        comp.protocols.tcp-ip
Subject:   IPng overview
Hello,

After hours of serching, I was not able to find a document (easy to  
understand ! ;) ) describing IPng with all the different implementations  
(SIPP, TUBA and catnip)...

If someone has an idea... ;)

Thanks,


D.



-----------[000110][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 12:16:47 GMT
From:      [email protected] (Holger Meyer)
To:        comp.protocols.tcp-ip
Subject:   REQUEST: Reliable Datagram Protocol
We need a working implementation of a reliable datagram protocol.  Any
solution, IL, RDP, on-top-of-UDP are welcome.  Any hints or
suggestions?

Regards

Holger

--
Holger Meyer, Uni of Rostock, Dpt. of CS: [email protected]
-- Where'd you learn about /  Satan - out of a book /  Love? - out of a box
-- Jim Morrison

-----------[000111][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 12:32:02 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: HELP!! binary file xfer over Telnet sessio
In article <[email protected]> [email protected] (Frank Lofaro) writes:
>A Linux telnetd I saw did null stuffing after line feeds instead of after 
>carriage returns when in ASCII mode. With that kind of brain-damage, g*d 
>only knows _what_ could be wrong with it. Heck, the linecheck program 
>that comes with TERM would not even run (it would hang since it never saw 
>a good "packet"). It would help if people read RFC 854, etc, and actually 
>heeded the advice there (and did more bug testing).

Actually the old Linux telnetd was a direct port of the BSD one hacked
(badly) to use posix termios. The current ones are rather better done and
the only ones distributed with the NetKit stuff. 

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000112][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 13:15:44 GMT
From:      [email protected] (Chris Davies)
To:        comp.protocols.tcp-ip
Subject:   Re: IPng overview
Denis Lafont ([email protected]) wrote:
: After hours of serching, I was not able to find a document (easy to  
: understand ! ;) ) describing IPng with all the different implementations  
: (SIPP, TUBA and catnip)...

There are a number of draft IETF papers available for Anon FTP.  I have
a script in ftp.visionware.co.uk /pub/staff/chris/utils/rfc-draft that
will go and anon-ftp the required papers for you.

Get the index with,
	rfc-draft -index

Then search the result for IPng.  One of the matches is, for example,
draft-ghiselli-ipng-whitepaper-req-00.txt.  You would then grab this
back with,
	rfc-draft draft-ghiselli-ipng-whitepaper-req-00.txt

Chris
--
          VISIONWARE LTD, 57 Cardigan Lane, LEEDS LS4 2LE, England
  Tel +44 532 788858.  Fax +44 532 304676.  Email [email protected]
---------- "Visionware:   The home of DOS/SQL/UNIX/X integration" ----------

-----------[000113][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 94 18:28:50
From:      [email protected] (Mukesh Kacker)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!

> From: [email protected] (Rick Jones)
 
>> Vernon Schryver ([email protected]lite.com) wrote:
 ...
>>  : accept.  BSD style stacks tend to complain with ENOBUFS as the interface's
>> : output buffer overflows. 
>
> I've only seen (can recall) this desirable behaviour on IRIX and
> HP-UX, and then only on interfaces from either SGI or HP. Every Sun
> interface I've seen so far (SunOS or Solaris) does not seem to report
> ENOBUFS. This has the nasty side effect of causing netperf to report
> 3+ MB/s send rates on Ethernet :( Of course, that is why programs like
> netperf and ttcp report the receive side throughput as well.
>

Only partially true...
SunOS 4.X (part of Solaris 1.X) was BSD derived OS and indeed does
return ENOBUFS error.

SunOS 5.X (part of Solaris 2.X) does not. That is a Streams based
implementation. Solaris 2.3 (and later) versions implement
flow control on the write side of UDP (Why discard UDP packets
because your ethernet driver is backed up ...application can be
slowed down !). This can block the application in a transmit call and
slow it down. (There is no stated semantics that UDP sockets cannot
block).

-Mukesh Kacker
 Internet Engineering
 SunSoft Inc.

-----------[000114][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 14:42:48 GMT
From:      [email protected] (Nicko van Someren)
To:        comp.protocols.tcp-ip
Subject:   Extending the TELNET protocol
The TELNET protocol supports option negotiation.  How do you go about
getting a new option standardised?  For ages I have thought it would
be a really useful feature if there were a pair of options to allow
the client to tell the server the size of the window.  Then if a
client gets a SIGWINCH, for example because the xterm it is running in
is expanded, it could send the new size to the server and the server
could raise the same signal for the process group attached to the
client connection.

Does anyone out there in netland know how to get this sort of thing
discussed and adopted?

	Nicko
--
+-----------------------------------------------------------------------------+
| Nicko van Someren, [email protected], extn. 34446, Home (44) 223 358707     |
 +-----------------------------------------------------------------------------+

-----------[000115][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 04:40:12 -0700
From:      [email protected] (Paul Smith)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
Casper H.S. Dik ([email protected]) wrote:
: [email protected] (Paul Smith) writes:
 
: >But, this is one of my greatest grips about SOCK_STREAM and circuit failures! 
: >Why couldn't a reasonable errno be returned consistantly.  The Lachman stack
: >on SVR4.2 v1.1.2 returns 0, errno = 0 for the server when the client side
: >closes() or exits premature.  The Client side however returns -1, errno=EPIPE
: >if the server exits premature??
 
: You must have mixed up two observations.  A close on one end will
: generate a read of 0, (the standard Unix EOF, you're not allowed to
: inspect errno at this point, so errno == 0 is a meaningless observation).
: The EPIPE is returned when, after a close of one end, the other end
: *writes* to the socket (actually, you get a SIGPIPE and die, unless you
: catch/ignore SIGPIPE).
 
: Also, you shouldn't use client/server when talking about the SOCK_STREAM
: layer.  In the application layer, there may well be a client and a server,
: but at the SOCK_STREAM layer they are peers.
 
: >If you rumage /usr/include/errno.h you find several streams/socket 
: >oriented errnos like; ENOTCON that would make better sense in the above
: >scenarios.  And forget about poll()ing for the flags S_HANGUP or S_ERROR
: >or even getsockopt(fd, SOL_SOCKET, SO_ERROR) it doesn't detect the hangup
: >condition either???
 
: >It seems the wishy-washy return of 0, errno=0 is IT!!
 
: As with all Unix files.  EOF is a read that returns 0 (errno undefined).
: A close on one end of a connection (pipe/socket/whatever), will cause
: an EOF.  It is not an error condition at the communication layer.
 
: When you want to detect the close, you select/poll for reading and read will
: return 0.

Very well stated!  The socket API extends the Unix file oriented programming
paradigm at the expense of clarity and precision!  

No harm if errno where to be set to ENOCON or what ever in this scenario...

And yes errno is 
indeterminate, but my style is to errno=0; prior to ever system call so as
to differentiate for sure who set errno to != 0.


-----------[000116][next][prev][last][first]----------------------------------------------------
Date:      Tue, 6 Sep 1994 17:00:41 GMT
From:      [email protected] (Malcolm Sainsbury)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,alt.winsock
Subject:   Ethernet to Slip converter or repeater
I have searched the FAQ's at hand so I hope this isn't too trivial a
question for the net experts.

Is there a PC-based prog that I can use to forward ethernet packets on the
one side, up a serial line, talking SLIP on the other.

Let me explain:

I have a leased line from my home to my office running at 38400bps. (nice)
The PC in my office is connected via a WD8003E card onto the net. (great)
The PC at home is all fired up with a SLIP packet driver.  (fine)
Now how do I make the two talk without buying another PC.

I configured PCRoute v2.2 into my PC at work and it certainly will take
ethernet packets off the net and send them up the serial port,
BUT not in SLIP form (it would seem).  It wants to talk its own protocol to
ANOTHER PC running PCRoute v2.2 on the other side and from there I would
connect my home system.  (thats a waste of a PC, even if I had one)

Isn't there some simple software that I can load in my work machine that
will pluck my ethernet packets off the WD8003 and send them in proper SLIP
form directly into my COM port at home, to be gratefully received by the
packet driver waiting expectantly there for proper slip packets ?

Kind regards

Malcolm
--
Malcolm Sainsbury - Dept of Information Systems - Rhodes University
Internet: [email protected]   Phone: +27 461 318244 (25049 fax)

-----------[000117][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 17:12:55 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Wanted: TCP/IP Package

Hello,

I'm looking for a (commercial) TCP/IP package implemented in CHILL.
Can anyone give me a hint.
Please reply to   [email protected]  !

Thanks in advance.

 Janos Breiteneicher

-- 


-----------[000118][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 18:27:23 GMT
From:      [email protected] (Rick Jones)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
Vernon Schryver ([email protected]) wrote:
: There can also be losses on the transmit side, between the application
: and the wire if the application tries to write more than the wire will
: accept.  BSD style stacks tend to complain with ENOBUFS as the interface's
: output buffer overflows. 

I've only seen (can recall) this desirable behaviour on IRIX and
HP-UX, and then only on interfaces from either SGI or HP. Every Sun
interface I've seen so far (SunOS or Solaris) does not seem to report
ENOBUFS. This has the nasty side effect of causing netperf to report
3+ MB/s send rates on Ethernet :( Of course, that is why programs like
netperf and ttcp report the receive side throughput as well.

I've also not seen ENOBUFS on some third-party HP-UX interfaces (but I
think that they are fixing that :)

rick jones

It's been so long since I've run netperf on Ultrix, OSF, or AIX that I
cannot recall which way it goes wrt ENOBUFS.

-----------[000119][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 16:01:59 +0200
From:      [email protected] (Casper H.S. Dik)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
[email protected] (Paul Smith) writes:

>But, this is one of my greatest grips about SOCK_STREAM and circuit failures! 
>Why couldn't a reasonable errno be returned consistantly.  The Lachman stack
>on SVR4.2 v1.1.2 returns 0, errno = 0 for the server when the client side
>closes() or exits premature.  The Client side however returns -1, errno=EPIPE
>if the server exits premature??

You must have mixed up two observations.  A close on one end will
generate a read of 0, (the standard Unix EOF, you're not allowed to
inspect errno at this point, so errno == 0 is a meaningless observation).
The EPIPE is returned when, after a close of one end, the other end
*writes* to the socket (actually, you get a SIGPIPE and die, unless you
catch/ignore SIGPIPE).

Also, you shouldn't use client/server when talking about the SOCK_STREAM
layer.  In the application layer, there may well be a client and a server,
but at the SOCK_STREAM layer they are peers.

>If you rumage /usr/include/errno.h you find several streams/socket 
>oriented errnos like; ENOTCON that would make better sense in the above
>scenarios.  And forget about poll()ing for the flags S_HANGUP or S_ERROR
>or even getsockopt(fd, SOL_SOCKET, SO_ERROR) it doesn't detect the hangup
>condition either???
 
>It seems the wishy-washy return of 0, errno=0 is IT!!

As with all Unix files.  EOF is a read that returns 0 (errno undefined).
A close on one end of a connection (pipe/socket/whatever), will cause
an EOF.  It is not an error condition at the communication layer.

When you want to detect the close, you select/poll for reading and read will
return 0.

Casper

-----------[000120][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 20:00:56 GMT
From:      [email protected] (Tony Trehan)
To:        comp.protocols.tcp-ip
Subject:   Embedded Systems TCP/IP
I need to provide TCP/IP (FTP, TELNET etc) in embedded
industrial controllers running under real-time operating systems.....

Can anybody provide me pointers to:

1) Where can i get "general info" about TCP/IP on embedded systems?
2) Any commercial drivers available in the marketplace (small, re-entrant code
3) ANY suggestions etc....????

Thanx a lot!
-Tony

-----------[000121][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 1994 20:38:50 GMT
From:      [email protected] (Tom Straub)
To:        comp.protocols.tcp-ip
Subject:   PEER to PEER wanted!!

Has anybody got any information on any PEER-to-PEER products for
TCP/IP that would allow one client to "look in" and operate another
TCP/IP client remotely?

I am trying to use this technology for remote diagnostics and
configurations of PC's on a WAN.

Please e-mail me at [email protected] if you have any
ideas or products.  Willing to buy!


-----------[000122][next][prev][last][first]----------------------------------------------------
Date:      6 Sep 94 22:14:29 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   TCP/IP and ARCHNET
Some where I heard that TCP/IP can not run on top of ARCHNET.  
Is this true?  The only potential reason I can think of is the fact that
ARCHNET does not have unique MAC address might cause some 
confusion.

Thanks in advance.

Paul Evans
Pacific Gas & Electric
/*  The opinions may not reflect those of PG&E */

-----------[000123][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 00:04:25 GMT
From:      [email protected] (Garry Hornbuckle)
To:        comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: MacTCP and RIP problems
In article <[email protected]nt.ca>, [email protected] (David
Oberst) wrote:

> In article <[email protected]>
 [email protected] (Garry Hornbuckle) writes:
> >In article <[email protected]>,
> >
> >    [reference to OpenTransport [aka MacTCP 3.0]
> >
> >Give us a couple more months to get it finished and tested, OK?
> > 
> Was this poetic license, or is OpenTransport due to be ready this
> quickly?

Well, alpha this month, beta in this year. final code shipping in early
'95. But I am planning a wide distribution beta progam, so it is likely
that you'll get a chance to play with it in a couple of months.... 

> 
> Also, are there plans for an Inside Mac:Open Transport type of book
> to help us poor small potatoes programmers? <g>
> 

Yes. Inside Mac:Networking-Open Transport is under development. Hope to
have first drafts available by beta.

-------------------------------------------------------------------
Garry Hornbuckle    Product Manager, Communications & Collaboration
-------------------------------------------------------------------
"If I told you that I   | email      [email protected]
 spoke only for myself  | applelink  HORNBUCKLE1
 would you believe me?" | fax        (408) 974-1211
-------------------------------------------------------------------

-----------[000124][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 01:32:01 GMT
From:      [email protected] (John Hascall)
To:        comp.protocols.tcp-ip
Subject:   Re: Extending the TELNET protocol
Nicko van Someren <[email protected]> wrote:
}The TELNET protocol supports option negotiation.  How do you go about
}getting a new option standardised?  For ages I have thought it would
}be a really useful feature if there were a pair of options to allow
}the client to tell the server the size of the window.  Then if a
}client gets a SIGWINCH, for example because the xterm it is running in
}is expanded, it could send the new size to the server and the server
}could raise the same signal for the process group attached to the
}client connection.

   It's such a good idea it already exists.  See RFC1073
   (ftp://ftp.iastate.edu/pub/netinfo/rfc/rfc1073.txt.Z is one place)
   for details.

}Does anyone out there in netland know how to get this sort of thing
}discussed and adopted?

   Have a good idea and write an RFC (see RFC1543).

John
-- 
John Hascall                   ``An ill-chosen word is the fool's messenger.''
Systems Software Engineer
Project Vincent
Iowa State University Computation Center  +  Ames, IA  50011  +  515/294-9551

-----------[000125][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 01:51:34 GMT
From:      [email protected] (John Hascall)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
Rick Jones <[email protected]> wrote:
}Vernon Schryver ([email protected]) wrote:
}: There can also be losses on the transmit side, between the application
}: and the wire if the application tries to write more than the wire will
}: accept.  BSD style stacks tend to complain with ENOBUFS as the interface's
}: output buffer overflows. 
 
}I've only seen (can recall) this desirable behaviour on IRIX and
}HP-UX, and then only on interfaces from either SGI or HP. Every Sun
}interface I've seen so far (SunOS or Solaris) does not seem to report
}ENOBUFS.
 
}It's been so long since I've run netperf on Ultrix, OSF, or AIX that I
}cannot recall which way it goes wrt ENOBUFS.

    Ultrix and OSF/1 will also return ENOBUFS.
    AIX probably dials IBM and orders another 32MBs...        :)

John
-- 
John Hascall                   ``An ill-chosen word is the fool's messenger.''
Systems Software Engineer
Project Vincent
Iowa State University Computation Center  +  Ames, IA  50011  +  515/294-9551

-----------[000126][next][prev][last][first]----------------------------------------------------
Date:      Wed, 07 Sep 94 09:46:13 PDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Cross Platform RPC

I would appreciate any comments/thoughts on the current state of RPC on
the various *ix platforms (HP/SUN/IBM).  I spent significant time 3 years 
ago looking at RPC and portability issues.  In the end the company that I
worked with at the time went with Netwise RPC.  The Netwise package was
fairly good.  I now need to rethink the cross platform communications issue
for a new product and am leaning towards using the standard RPC libraries
that come on the various platforms.  Some questions I have are: 

Do most or all vendor provided RPC packages use XDR?  If not what are the
various vendors using for the underlying data portability and how do they
deal with a multi-platform environment?

Are the vendor provided RPC packages fairly bug free.  When I evaluated 
the Sun RPC several years ago it had quite a few bugs.

Finally, what are the standards bodies doing with RPC, is there a universally
accepted RPC standard in place or on the horizon.

Thanks in advance for any comments you may have.

Fred Covely.



-----------[000127][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 02:56:07 GMT
From:      [email protected] (Heather L. Nadelman)
To:        comp.protocols.tcp-ip
Subject:   Setting up MacTCP
Hi, guys!

I'm helping a friend set up MacTCP.  I am wholly unfamiliar with Macs, but I 
have gotten pretty proficient at setting up Trumpet Winsock with SLIP.  She's 
using a powerbook and wants to dial in to both her ethernet network at work 
and the internet.  What do I need?  Is there some handy-dandy FAQ that I could 
use?   Is it trickier than Winsock?

Thanks in advance!

--Heather L. Nadelman
[email protected]


-----------[000128][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 08:11:01
From:      [email protected] (Clark Bremer)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]> [email protected] (Michael Salmon) writes:
>From: [email protected] (Michael Salmon)
>Subject: Re: TN5250/TN3270 Standards?
>Date: 2 Sep 1994 09:46:39 GMT
 
>In article <[email protected]>
>[email protected] (Rich Oldroyd) writes:
>|> 
>|> We're looking for the  specifications/RFCs covering the
>|> TN3270 and/or TN5250 products. We're aware that we
>|> can buy some off the shelf products supporting these
>|> standards but would like the specs themselves. 
>|> Any pointers?
 
>To the best of my knowledge these are proprietry IBM protocols and you
>have to by the specs. from them.

I'm not sure where its written down, but for TN3270, I'm sure of the answer - 
its simply to negotiate for "binary mode" on and "send end of record" on.  CB.
===========================================================================
          _  _               Clark Bremer     [email protected]
         /  /_)              Software Engineer, NetStar Inc.
         \_/__)              10250 Valley View Road  MPLS, MN 55344

-----------[000129][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 10:23:56 -0400
From:      [email protected] (Cary B. O'Brien)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
In article <[email protected]>,
Tony Trehan <[email protected]> wrote:
>I need to provide TCP/IP (FTP, TELNET etc) in embedded
>industrial controllers running under real-time operating systems.....
>
>Can anybody provide me pointers to:
>
>1) Where can i get "general info" about TCP/IP on embedded systems?
>2) Any commercial drivers available in the marketplace (small, re-entrant code
>3) ANY suggestions etc....????
>
>Thanx a lot!
>-Tony

One place to start for a biased view is the real time kernel vendors.
both ICL/Software Components (pSOS) and Ready Systems (VRTX) (*) have
TCP/IP add-on packages.  We used the pSOS/pNA package on a rack
of VME cards for one project.  One card had an ethernet port, and
we implemented another network across the VME backplane, so the
whole rack was on the net (well... our local net).  I learned 
a LOT about TCP/IP on that project.  They supply drivers for common
VME cards.

There is also a TINYTCP in Comp.sources.unix archives Volume 7.

(*) I am not 100% sure about the company names since there have
    been mergers/splits

Cary O'Brien
[email protected]



-----------[000130][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 05:32:07 GMT
From:      [email protected] (George Neville-Neil)
To:        comp.protocols.tcp-ip
Subject:   TCP/IP FAQ for September 1994

Archive-name:tcp-ip/FAQ
Last-modified:  1994/09/01

Internet Protocol Frequently Asked Questions

Maintained by: George V. Neville-Neil ([email protected])
Contributions from:
Ran Atkinson
Stephane Bortzmeyer
Dr. Charles E. Campbell Jr.
Phill Conrad 
Alan Cox
Jon Kay 
William Manning
Barry Margolin 
Jim Muchow
W. Richard Stevens 
 
Version 1.5


************************************************************************

	The following is a list of Frequently Asked Questions, and
their answers, for people interested in the Internet Protocols,
including TCP, UDP, ICMP and others.  Please send all additions,
corrections, complaints and kudos to the above address.  This FAQ will
be posted on or about the first of every month.

	This FAQ is available for anonymous ftp from :
ftp.netcom.com:/pub/gnn/tcp-ip.faq .  You may get it from my home page at
ftp://ftp.netcom.com/pub/gnn/gnn.html

************************************************************************

Table of Contents:
Glossary
1) Are there any good books on IP?
2) Where can I find example source code for TCP/UDP/IP?
3) Are there any public domain programs to check the performance of an
IP link? 
4) Where do I find RFCs?
5) How can I detect that the other end of a TCP connection has
crashed?  Can I use "keepalives" for this?
6) Can the keepalive timeouts be configured?
7) Can I set up a gateway to the Internet that translates IP
addresses, so that I don't have to change all our internal addresses 
to an official network? 


Glossary:

I felt this should be first given the plethora of acronyms used in the
rest of this FAQ.

IP: Internet Protocol.  The lowest layer protocol defined in TCP/IP.
This is the base layer on which all other protocols mentioned herein
are built.  IP is often referred to as TCP/IP as well.

UDP: User Datagram Protocol.  This is a connectionless protocol built
on top of IP.  It does not provide any guarantees on the ordering or
delivery of messages.  This protocol is layered on top of IP.

TCP: Transmission Control Protocol.  TCP is a connection oriented
protocol that guarantees that messages are delivered in the order in
which they were sent and that all messages are delivered.  If a TCP
connection cannot deliver a message it closes the connection and
informs the entity that created it.  This protocol is layered on top
of IP.

ICMP:  Internet Control Message Protocol.  ICMP is used for
diagnostics in the network.  The Unix program, ping, uses ICMP
messages to detect the status of other hosts in the net.  ICMP
messages can either be queries (in the case of ping) or error reports,
such as when a network is unreachable.

RFC: Request For Comment.  RFCs are documents that define the
protocols used in the IP Internet.  Some are only suggestions, some
are even jokes, and others are published standards.  Several sites in
the Internet store RFCs and make them available for anonymous ftp.

SLIP:  Serial Line IP.  An implementation of IP for use over a serial
link (modem).  CSLIP is an optimized (compressed) version of SLIP that
gives better throughput.

Bandwidth:  The amount of data that can be pushed through a link in
unit time.  Usually measured in bits or bytes per second.

Latency:  The amount of time that a message spends in a network going
from point A to point B.

Jitter:  The effect seen when latency is not a constant.  That is, if
messages experience a different latencies between two points in a
network.

RPC:  Remote Procedure Call.  RPC is a method of making network access
to resource transparent to the application programmer by supplying a
"stub" routine that is called in the same way as a regular procedure
call.  The stub actually performs the call across the network to
another computer.

Marshalling:  The process of taking arbitrary data (characters,
integers, structures) and packing them up for transmission across a
network.

MBONE: A virtual network that is a Multicast backBONE.  It is still a
research prototype, but it extends through most of the core of the
Internet (including North America, Europe, and Australia).  It uses IP
Multicasting which is defined in RFC-1112.  An MBONE FAQ is available
via anonymous ftp from: ftp.isi.edu"  There are frequent broadcasts of
multimedia programs (audio and low bandwidth video) over the MBONE.


1) Are there any good books on IP?

A) Yes.  Please see the following:

Internetworking with TCP/IP Volume I
(Principles, Protocols, and Architecture)
Douglas E. Comer
Prentice Hall 1991

This volume covers all of the protocols, including IP, UDP, TCP, and
the gateway protocols.  It also includes discussions of higher level
protocols such as FTP, TELNET, and NFS.

Internetworking with TCP/IP Volume II
(Design, Implementation, and Internals)
Douglas E. Comer / David L. Stevens
Prentice Hall 1991

Discusses the implementation of the protocols and gives numerous code
examples.

Internetworking with TCP/IP Volume III (BSD Socket Version)
(Client - Server Programming and Applications)
Douglas E. Comer / David L. Stevens
Prentice Hall 1993

This book discusses programming applications that use the internet
protocols.  It includes examples of telnet, ftp clients and servers.
Discusses RPC and XDR at length.

TCP/IP Illustrated, Volume 1: The Protocols, 
W. Richard Stevens
(c) Addison-Wesley, 1994 

An excellent introduction to the entire TCP/IP protocol suite,
covering all the major protocols, plus several important applications.

Unix Network Programming
W. Richard Stevens
Prentice Hall 1990

An excellent introduction to network programming under Unix.

The Design and Implementation of the 4.3 BSD Operating System
Samuel J. Leffler, Marshall Kirk McKusick, Michael J. Karels, John S.
Quarterman 
Addison-Wesley 1989

Though this book is a reference for the entire operating system, the
eleventh and twelfth chapters completely explain how the networking
protocols are implemented in the kernel.


2)  Where can I find example source code for TCP/UDP/IP?

A)  Code from the Internetworking with TCP/IP Volume III is available
for anonymous ftp from:

arthur.cs.purdue.edu:/pub/dls

Code used in the Net-2 version of Berkeley Unix is available for
anonymous ftp from:

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

and

gatekeeper.dec.com:/pub/BSD/net2/sys/netinet

Code from Richard Steven's book is available on:
ftp.uu.net:/published/books/stevens.*

Example source code and libraries to make coding quicker is available
in the Simple Sockets Library written at NASA.  The Simple Sockets
Library makes sockets easy to use!  And, it comes as source code.  It
has been tested on: Unix (SGI, DecStation, AIX, Sun 3, Sparcstation;
version 2.02+: Solaris 2.1, SCO), VMS, and MSDOS (client only since
there's no background there).  It is provided in source code form, of
course, and sits atop Berkeley sockets and tcp/ip.

You can order the "Simple Sockets Library" from

                           Austin Code Works
                          11100 Leafwood Lane
                       Austin, TX 78750-3464 USA
                         Phone (512) 258-0785
                 
Ask for the "SSL - The Simple Sockets Library".  Last I checked, they
were asking $20 US for it.


For DOS there is WATTCP.ZIP (numerous sites): 

WATTCP is a DOS TCP/IP stack derived from the NCSA Telnet program and
much enhanced. It comes with some example programs and complete source
code. The interface isn't BSD sockets but is well suited to PC type
work. It is also written so that it can be used and memory
allocation).

3)  Are there any public domain programs to check the performance of
an IP link?

A)  

TTCP:  Available for anonymous ftp from....

Host gatekeeper.dec.com

    Location: /.0/BSD/NetBSD/NetBSD-current/othersrc
      DIRECTORY dr-xr-xr-x        512  Apr  8 09:57  ttcp
    Location: /.0/BSD/NetBSD/NetBSD-current/othersrc/ttcp
           FILE -r--r--r--       3885  Nov  7 03:35  ttcp.1
           FILE -r--r--r--      19225  Nov  7 03:35  ttcp.c

Host world.std.com

    Location: /src/wuarchive/graphics/graphics/mirrors/sgi.com/sgi/src/ttcp
           FILE -r--r--r--       3885  Oct  4 1991  ttcp.1
           FILE -r--r--r--      19170  May 17 1993  ttcp.c
           FILE -r--r--r--      13033  Sep  5 1989  ttcp.c-brl

On ftp.sgi.com are netperf (from Rick Jones at HP) and nettest
(from Dave Borman at Cray).  ttcp is also availabel at ftp.sgi.com.




There is suite of Bandwidth Measuring programs from [email protected]
Available for anonymous ftp from ftp.netcom.com in
~ftp/gnn/bwmeas-0.3.tar.Z These are several programs that meausre
bandwidth and jitter over several kinds of IPC links, including TCP
and UDP.


4) Where do I find RFCs?

A)  This is the latest info on obtaining RFCs:
Details on obtaining RFCs via FTP or EMAIL may be obtained by sending
an EMAIL message to [email protected] with the message body 
help: ways_to_get_rfcs.  For example:

        To: [email protected]
        Subject: getting rfcs

        help: ways_to_get_rfcs

The response to this mail query is quite long and has been omitted.

RFCs can be obtained via FTP from DS.INTERNIC.NET, NIS.NSF.NET,
NISC.JVNC.NET, FTP.ISI.EDU, WUARCHIVE.WUSTL.EDU, SRC.DOC.IC.AC.UK,
FTP.CONCERT.NET, or FTP.SESQUI.NET.


Using Web, WAIS, and gopher:

Web:

http://web.nexor.co.uk/rfc-index/rfc-index-search-form.html

WAIS access by keyword:

wais://wais.cnam.fr/RFC

Excellent presentation with a full-text search too:

http://www.cis.ohio-state.edu/hypertext/information/rfc.html

With Gopher:

gopher://r2d2.jvnc.net/11/Internet%20Resources/RFC
gopher://muspin.gsfc.nasa.gov:4320/1g2go4%20ds.internic.net%2070%201%201/.ds/
.internetdocs



5) How can I detect that the other end of a TCP connection has crashed?
Can I use "keepalives" for this?

A) Detecting crashed systems over TCP/IP is difficult.  TCP doesn't require
any transmission over a connection if the application isn't sending
anything, and many of the media over which TCP/IP is used (e.g. ethernet)
don't provide a reliable way to determine whether a particular host is up.
If a server doesn't hear from a client, it could be because it has nothing
to say, some network between the server and client may be down, the server
or client's network interface may be disconnected, or the client may have
crashed.  Network failures are often temporary (a thin ethernet will appear
down while someone is adding a link to the daisy chain, and it often takes
a few minutes for new routes to stabilize when a router goes down), and TCP
connections shouldn't be dropped as a result.

Keepalives are a feature of the sockets API that requests that an empty
packet be sent periodically over an idle connection; this should evoke an
acknowledgement from the remote system if it is still up, a reset if it has
rebooted, and a timeout if it is down.  These are not normally sent until
the connection has been idle for a few hours.  The purpose isn't to detect
a crash immediately, but to keep unnecessary resources from being allocated
forever.

If more rapid detection of remote failures is required, this should be
implemented in the application protocol.  There is no standard mechanism
for this, but an example is requiring clients to send a "no-op" message
every minute or two.  An example protocol that uses this is X Display
Manager Control Protocol (XDMCP), part of the X Window System, Version 11;
the XDM server managing a session periodically sends a Sync command to the
display server, which should evoke an application-level response, and
resets the session if it doesn't get a response (this is actually an
example of a poor implementation, as a timeout can occur if another client
"grabs" the server for too long).

6) Can the keepalive timeouts be configured?

A) I know they can on many systems, but I don't know the details.

7) Can I set up a gateway to the Internet that translates IP addresses, so
that I don't have to change all our internal addresses to an official
network?

A) There's no general solution to this.  Many protocols include IP
addresses in the application-level data (FTP's "PORT" command is the most
notable), so it isn't simply a matter of translating addresses in the IP
header.  Also, if the network number(s) you're using match those assigned
to another organization, your gateway won't be able to communicate with
that organization (RFC 1597 proposes network numbers that are reserved for
private use, to avoid such conflicts, but if you're already using a
different network number this won't help you).

However, if you're willing to live with limited access to the Internet from
internal hosts, the "proxy" servers developed for firewalls can be used as
a substitute for an address-translating gateway. See the firewall FAQ.
-- 
[email protected]

Law is there to clean up etiquette's failures.
					Ms. Manners

-----------[000131][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 94 14:21:11 -0500
From:      [email protected] (James Harvey)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting a Class C address?
In article <[email protected]>, [email protected] (Tony Li) writes:
	[snip]
>
> cisco also supports the all-ones subnet.  This has none of the ambiguity,
> so is enabled by default.  Again, the all-ones subnet all-ones host is a
> broadcast address.

Is it a directed broadcast to the net or to the all-ones subnet?  Just
wondering because Cisco provides an option to do something useful with
directed net broadcasts (the "ip helper-address" feature).
--
James Harvey   [email protected]   IUPUI IT Networks and Systems
Disclaimer:  These are my own opinions.  I do not speak for Indiana University.

-----------[000132][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 14:22:22 -0500
From:      [email protected] (Chad Adams)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
In article <[email protected]>,
MARCO PACE  <[email protected]> wrote:
>Hello everybody,
>
>I'm writing a couple of applications (running on SUN SPARCstations with
>Solaris 2.3) communicating using UDP via an Ethernet LAN. The API used to
>access the communication primitives is the socket interface.
>
>The problem I am experiencing is a loss of UDP packets sent over the network.
>But the problem is occuring not at the network level, but above that. I managed
>to check this using a network 'sniffer', which was able to capture all of
>the sent packets. The receiver application, on the contrary was not only
>unable to get all of the packets, but missed a huge amount of them !!

Do you have patch 101694-01.  NIS sure has dataloss without it.

I would suggest it to everyone.  Our trouble with Solaris was endless
before we installed it.


-- 
Chad Adams              __               Educational Computing Network
Systems Programmer     /  ) /         /  University Park, IL  USA
1-708-235-2200        /    /_  __, __/   ECN, networking in IL since 1975
[email protected]      (__/ / /_/_/_(_/_   "Wherever you go, there we are."

-----------[000133][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 07:04:40 GMT
From:      [email protected] (Elena Leong)
To:        comp.protocols.tcp-ip
Subject:   Getting duplicate IP addresses
Here's a little problem that I just cant figure out (neither can the network
people).

We have a SunSparc2 running SunOS 4.1.3.  It has two Ethernet cards.  le0 is
connected to our project LAN and has an IP network 192.245.13.0 happily
happening.  le1 is connected to the backbone APost LAN and we also happily
run FTAM/OSI to some of the VAXes on this LAN.

Now, our Sun starts to complain, like this:

Aug 31 17:32:35 grunge vmunix: duplicate IP address! send from ethernet 
address: aa:0:4:0:1f:2c
Sep 2 09:59:10 grunge vmunix: duplicate IP address! send from ethernet 
address: aa:0:4:0:65:28

The APost network guys tell me that these identify to be DEC Ultrixs 
running v4.2a and v4.2 resp, running TCP/IP on a network of 155.144.0.0
These arent the only ones, BTW, there appear to be 5/6 others, all Unix
based DECs....

Now I thought maybe my machine was doing something weird, so I temporarily
changed our IP address to something totally bogus, and still these messages
appear.
Can anyone out there think why this is happening?
Does anyone out b

Thanks,
Elena.

===============================
Elena Leong
Systems & Network Administrator
EDIPost, Australia Post
===============================

-----------[000134][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 12:38:10 GMT
From:      [email protected] (Ricard Wolf)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?
In article <[email protected]> [email protected] (John Antypas) writes:
>
>Recently, while shopping for an IP provider for a client, a vendor
>noted that the use of "sider" IP numbers was better because of the
>crunch on address space.
>
>I've never heard this term before.  What is a 
>"sider" number?  I assume it means a subnet of a class C address?

There is a group of class A B and C networks which are designed to
be used internally in networks that will never be directly connected
to the Internet (however, connection via a router is possible), which
are defined in RFC 1597. Could this be what the vendor meant? 
Don't know about the term "sider" though; never heard it before.

/Ricard
-- 
Ricard Wolf                   / | \  / | /-           email: [email protected]
Axis Communications AB       /__|  \/  | \__          uucp:  axisab.se!ricard
S-223 70 LUND               /   |  /\  |    \         Tel:   +46 46 19 18 63
SWEDEN                     /    | /  \ | \__/         Fax:   +46 46 13 61 30

-----------[000135][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 13:15:57 GMT
From:      [email protected] (Patrick Klos)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and ARCHNET
In article <[email protected]>,  <[email protected]> wrote:
>Some where I heard that TCP/IP can not run on top of ARCHNET.  
>Is this true?  The only potential reason I can think of is the fact that
>ARCHNET does not have unique MAC address might cause some 
>confusion.
>
>Thanks in advance.
>
>Paul Evans
>Pacific Gas & Electric
>/*  The opinions may not reflect those of PG&E */

This is NOT true.  ARCNET can support TCP/IP as well as any other network
media.  ARCNET will be prone to fragmentation (either IP or ARCNET) since
the maximum data size for any single packet is 508 bytes.
-- 
============================================================================
    Patrick Klos                           Internet: [email protected]
    Klos Technologies, Inc.                Voice: (603) 424-8300
    604 Daniel Webster Highway             FAX:   (603) 424-9300
    Merrimack, New Hampshire  03054        BBS:   (603) 429-0032
============================================================================

-----------[000136][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 94 13:24:00 GMT
From:      [email protected] (Claus T|ndering)
To:        comp.protocols.tcp-ip
Subject:   Proxy FTP server
Can anyone point me to the source code for a proxy FTP server?

---
Claus Tondering, Lyngby, Denmark
E-mail: [email protected]

-----------[000137][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 14:02:29 GMT
From:      [email protected] (Charles Ganzhorn)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?
In article <[email protected]>, [email protected] (Ricard Wolf)
wrote:
> 
> In article <[email protected]> [email protected] (John Antypas) writes:
> >
> >Recently, while shopping for an IP provider for a client, a vendor
> >noted that the use of "sider" IP numbers was better because of the
> >crunch on address space.
> >
> >I've never heard this term before.  What is a 
> >"sider" number?  I assume it means a subnet of a class C address?
> 
> There is a group of class A B and C networks which are designed to
> be used internally in networks that will never be directly connected
> to the Internet (however, connection via a router is possible), which
> are defined in RFC 1597. Could this be what the vendor meant? 
> Don't know about the term "sider" though; never heard it before.
> 

I believe that what the providers are doing is pronouncing the acronym CIDR
which stands for Classless Inter-Domain Routing.  This is a technique for
aggregating the routes of several contiguous networks into a single route.

See RFC 1519 for the details.

Charles.

--
Charles Ganzhorn                        Email:  [email protected]
Consulting Engineer                     
cisco Systems
One Appletree Square, Suite 1452        
8009 34th Avenue South                  Phone:  612-851-8310
Bloomington, MN  55425                  FAX:    612-851-8311

-----------[000138][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 14:03:52 GMT
From:      [email protected] (Michael Salmon)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?
In article <[email protected]>
[email protected] (Ricard Wolf) writes:
|> In article <[email protected]>
|> [email protected] (John Antypas) writes:
|> >
|> >Recently, while shopping for an IP provider for a client, a vendor
|> >noted that the use of "sider" IP numbers was better because of the
|> >crunch on address space.
|> >
|> >I've never heard this term before.  What is a 
|> >"sider" number?  I assume it means a subnet of a class C address?
|> 
|> There is a group of class A B and C networks which are designed to
|> be used internally in networks that will never be directly connected
|> to the Internet (however, connection via a router is possible), which
|> are defined in RFC 1597. Could this be what the vendor meant? 
|> Don't know about the term "sider" though; never heard it before.

It's written CIDR, Classless Inter-domain Routing, RFC 1519. The idea
is that you get a block of class C addresses rather than a class B.

-- 

Michael Salmon

#include	<standard.disclaimer>
#include	<witty.saying>
#include	<fancy.pseudo.graphics>

Ericsson Telecom AB
Stockholm

-----------[000139][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 14:06:47 GMT
From:      [email protected] (Michael Salmon)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]>
[email protected] (Clark Bremer) writes:
|> In article <[email protected]>
|> [email protected] (Michael Salmon) writes:
|> >In article <[email protected]>
|> >[email protected] (Rich Oldroyd) writes:
|> >|> 
|> >|> We're looking for the  specifications/RFCs covering the
|> >|> TN3270 and/or TN5250 products. We're aware that we
|> >|> can buy some off the shelf products supporting these
|> >|> standards but would like the specs themselves. 
|> >|> Any pointers?
 
|> >To the best of my knowledge these are proprietry IBM protocols and you
|> >have to by the specs. from them.
|> 
|> I'm not sure where its written down, but for TN3270, I'm sure of the answer - 
|> its simply to negotiate for "binary mode" on and "send end of record" on.  CB.

Sorry but that's how you enter 3270 mode in Telnet. Telnet then become
the carrier for the 3270 protocol.

-- 

Michael Salmon

#include	<standard.disclaimer>
#include	<witty.saying>
#include	<fancy.pseudo.graphics>

Ericsson Telecom AB
Stockholm

-----------[000140][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 14:07:17 GMT
From:      [email protected] (Dennis Newport)
To:        comp.protocols.tcp-ip
Subject:   PING
We were doing some tests recently on a dual lan configuration that we have. I have used ping
in the past but I was surprised to find that it hangs when the LAN connection(s) is physically
removed and you are PINGing an Internet address. If you PING a host name then what happens is that
after some time (this time period seems indeterminate - or isn't it ???), the PING returns and 
says, naturally, that it can't get to the NIS server because PING in this case uses gethostbyname.

Is this behaviour normal ?

Shouldn't PING time out ?

We do:

/etc/ping 'host' 16 1

and

/etc/ping aa.bb.cc.dd 16 1

on HP-UX boxes.

Has anyone any comments on this ?  All information gratefully appreciated.




-----------[000141][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 15:02:44 GMT
From:      [email protected] (Clay Luther)
To:        comp.protocols.tcp-ip
Subject:   ftp API?
I need to be able to access ftp from within a C or C++ program. Does anyone
know of an ftp API that will allow me to do so effortlessly?

--
Chris Pearce, [email protected]

-----------[000142][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 21:59:39 -0400
From:      [email protected] (Eric V. Smith)
To:        comp.protocols.tcp-ip
Subject:   DHCP server and NT
I'm writing a DHCP server to run under Linux, with the specific 
purpose of serving NT Daytona clients (and later, Chicago).  I had 
a few questions about NT's usage of DHCP, and wondered if anyone here 
knows the answers.

Specifically, NT doesn't ask for a subnet mask.  I know how to hack
the registry to make it ask, but it seems to me it ought to ask for it
by default.  Does anyone know why it doesn't?  Does it inspect the
class of the address I give it and assume something from that?

Does anyone know what happens if I don't give a subnet mask? Once I
get a little more work done, I can answer this one myself, but I
thought I'd see if anyone out there has an opinion.

Does anyone know how NT behaves with DHCP if it has multiple adapters?
If you look through the registry and how it is used to control DHCP
requests, it appears that many of the options must be the same on
every interface.  This doesn't seem correct to me.  Does anyone have a
multihomed NT client and any experience with DHCP?

Thanks.

Eric.

-----------[000143][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 16:02:13 GMT
From:      [email protected] (Mark Wuest)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
In article <[email protected]> [email protected] (Cary B. O'Brien) writes:

>In article <[email protected]>, Tony Trehan <[email protected]> wrote:
>>I need to provide TCP/IP (FTP, TELNET etc) in embedded
>>industrial controllers running under real-time operating systems.....
 
>One place to start for a biased view is the real time kernel vendors.
>both ICL/Software Components (pSOS) and Ready Systems (VRTX) (*) have
>TCP/IP add-on packages.

Well, I *thought* someone there would be watching this group, but I
was wrong. Microware Systems (Des Moines, IA) sells OS-9 and OS-9000
and has a TCP/IP package. It worked quite well for us...

Mark
--
Mark Wuest                     | "Fan her head!" the Red Queen anxiously
[email protected]         | interrupted. "She'll be feverish after
                               | so much thinking."

-----------[000144][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 16:27:33 GMT
From:      [email protected] (Wim.Holemans)
To:        comp.dcom.lans.ethernet,comp.protocols.snmp,comp.protocols.tcp-ip
Subject:   rmon(-alike) software for dos machine
Hi,

i'm looking for software that works like a network monitor and enables me
to get the information by snmp requests. I'm only interested in the following
parameters :
- number of packets/bytes seen on the net
- number of errors seen on the net, if possible with a subdivision for types
  of errors

the info should be available for snmp requests and the software should run on
a 386-DX with no harddisk and a SMC ethernet card.

I got part of this functionality with the fergie(beholder) software, which
offers me packets/bytes counts, but no error info.

Thanks,

-- 
-----------------------------------------------------------------------
Wim Holemans				phone + 32 3 820 22 03
Network/System manager			fax   + 32 3 820 22 44
U.I.A.

-----------[000145][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 16:41:04 GMT
From:      [email protected] (David Cabot)
To:        comp.protocols.tcp-ip
Subject:   How do you SLIP sco?
I'm trying to setup a SLIP between Linux and SCO.  Linux is easy with 
dip-3.3.7.  SCO 3.2v4 talks about slattach/sldetach, but doesn't talk about
what to do if the connection is via modem.  It would seem that SCO expects 
the tty to be direct to another computer.

I've tried running slattach to setup a modem that is quiet (no responses)
and DIPed to that.  Its not working.  I've set up a datahawk to see if I
can reverse engr it, but it doesn't make much sense.

I'd be grateful for any help.


-----------[000146][next][prev][last][first]----------------------------------------------------
Date:      07 Sep 94 16:12 BST
From:      Chief and Asst Chief Fire Officers Association <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Shared Modems on TCP/IP LAN ??
We are looking for a mechanism of accessing shared modem(s) on PC
networks. The networks include the following types, Novel, Netbios, TCP/ip.
We have a mechanism of achieving this on Novel and Netbios networks
but we are at present unable to access a shared modem on TCP/ip networks. 
Ideally the solution should satisfy all three network protocols, but 
we are willing ( if absolutely necessary ) to settle for multiple 
solutions for different network types.  
We are aware of NASI/NACS for Novel, and are 
already using Network Products NMP2 product for Novel/Netbios networks.

As an alternative we would be interested in examining hardware
solutions should such exist.
 
 We would appreciate any ideas on this topic.

 Kind Regards, and thanks in anticipation of replies,

 Martin Kelly,

 The Chief and Assistant Chief Fire Officers' Association
 United Kingdom 
 
 email: [email protected]

 By Post:
 C.A.C.F.O.A. ( Research Ltd ).
 10, 11 Pebble Close
 Amington 
 Tamworth 
 Staffordshire
 B77 4RD
 Great Britain.

 Telephone +441827 61516
 Fax.      +441827 61530

-----------[000147][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 17:19:16 GMT
From:      [email protected] (Rajiv Arora)
To:        comp.protocols.tcp-ip
Subject:   Re: Is MTU discovery available?
In article <[email protected]>,  <[email protected]> wrote:
>
>If not, is there any reasonable way to push the system into sending larger
>packets out to systems in other nets?
>

On AIX systems, there are at least 3 ways to do this:
  1) set "subnetsarelocal" on using "no". This causes subnets to be considered
     local for the purposes of MSS negotiation.
  2) set "tcp_mssdflt" using "no" to some value other than 512. This new value
     is used during MSS negotiation with a host on a "remote" network.
  3) Use the -mtu flag in route to set the MSS for a specific remote network.

All of these are poor cousins of the full-fledged path MTU discovery.

-Rajiv Arora
AIX Network Performance

-- 
Rajiv Arora                            IBM RISC System/6000 Division
[email protected]          	       Austin, TX

-----------[000148][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 14:33:52 +0200
From:      [email protected] (Casper H.S. Dik)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
[email protected] (Paul Smith) writes:

>No harm if errno where to be set to ENOCON or what ever in this scenario...

errno has no meaning when no error code is returned.

>And yes errno is 
>indeterminate, but my style is to errno=0; prior to ever system call so as
>to differentiate for sure who set errno to != 0.

It won't help.  There is no guarantee that errno is untouched, even
if a system call returns success.  There are a few exceptions:
system calls that can return -1 on success must guarantee that
errno is untouched when it returns a successful -1.

E.g., many of the socket calls are implemented as library calls in SVR4
systems, these library calls may call any number of system calls, some of
which may return failure and set errno, while the socket call still
returns success.

In other words:

	errno = 0;
	ioctl(fd, ....);
	if (errno != 0) {
	    /* failure */
	}

is wrong (non-portable).

The only time when "errno != 0" is meaningful is for functions like ptrace,
and even then the only proper incantation is:

	errno = 0;
	if (ptrace(..PTRACE_PEEK..) == -1 && errno != 0) {
	    /* failure */
	}

Casper

-----------[000149][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 18:45:18 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and ARCHNET
In article <[email protected]> [email protected] writes:
>Some where I heard that TCP/IP can not run on top of ARCHNET.  
>Is this true?

No, it's not. RFC-1201 documents how to send TCP/IP on Arcnet.

					don provan
					[email protected]

-----------[000150][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 21:50:32 GMT
From:      [email protected] (Marten Terpstra)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?
In <[email protected]> [email protected] (Michael Salmon) writes:

>|> >I've never heard this term before.  What is a 
>|> >"sider" number?  I assume it means a subnet of a class C address?
>|> 
>|> There is a group of class A B and C networks which are designed to
>|> be used internally in networks that will never be directly connected
>|> to the Internet (however, connection via a router is possible), which
>|> are defined in RFC 1597. Could this be what the vendor meant? 
>|> Don't know about the term "sider" though; never heard it before.
 
>It's written CIDR, Classless Inter-domain Routing, RFC 1519. The idea
>is that you get a block of class C addresses rather than a class B.

The idea is more that there are no class A, B or C any more. Addresses
space is to be handed out in bits, no longer in classes... The
Classless Routing part comes in so that you can take all of these bits
and route them as one entry, rather than one routing entry per
classful net. How's that for an oversimplicification. Perhaps RFC1466
and RFC1467 are good reading as well.

-Marten

-----------[000151][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 23:21:58 GMT
From:      [email protected] (Piercarlo Grandi)
To:        comp.protocols.tcp-ip
Subject:   Re: How to get pass fire well?
>>> On Thu, 25 Aug 1994 15:12:34 GMT, [email protected] (Jeffrey D
>>> Carter) said:

Jeffrey> I call this an alleged firewall, because the firewall admin
Jeffrey> should never allow traffic (even to WKS) to machines that he
Jeffrey> can't administer. If you can hack smtpd, then his firewall is
Jeffrey> completely useless.

Most firewalls are rather useless -- they are based on the premise that
somebody controls _all_ interconnection points to "outside". Large
organizations that want firewalls almost never can guarantee this. They
can monitor for packets from external hosts, but even then it's often
too late.

Simply put, there is no substitute for securing information on secure
machines, trying to isolate large internal networks often just gives a
false sense of security.

  in particular because if one keeps sensitive information on unsecure
  isolated hosts, if the isolation is breached, and it is as a rule
  breached *behind* the firewall, catastrophe happens, as lots and lots
  of hosts become easy to pick...

"Firewall" is nearly just a cute moniker, it's no substitute for a
serious information security setup.

On the other hand transport level gateways are going to be very
important in the future for *addressing* and *routing* reasons,
especially as the space of IPv4 addresses gets crowded.

Mark my words (and I am not the first to say it): increasingly people
and protocols will use names instead of numeric addresses, and the
latter will just be fleetingly allocated, despite, or perhaps because,
IPng having a chance to have 16 byte numeric addresses.

To be more brutal: numeric addresses are starting to be used as if they
were half numbers of a virtual circuit between the transport stations of
machines.


-----------[000152][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 23:35:30 GMT
From:      [email protected] (Piercarlo Grandi)
To:        comp.protocols.tcp-ip
Subject:   Re: Is there a self-aliasing IP address?
>>> On Thu, 1 Sep 1994 15:47:42 GMT, [email protected] (Vernon
>>> Schryver) said:

Vernon> Dynamic IP addresses for anything except transient internal
Vernon> links in the network will soon be gone and forgotten.  There are
Vernon> simply too many things that do not work well when IP addresses
Vernon> change.  Except for transient internal links such as backups for
Vernon> leased lines, dynamically assigned IP addresses are a
Vernon> complicated, fragile way of providing dumb shell services.

Uhmmm. I agree with the sentiment, but I think it's a vain hope. I
understand that Microsoft have put their considerable weight behind
dynamic addresses assignment (leases are going to be the default for
most installations using DHCP) and dynamic DNS updates (the WINS
service). This is also Microsoft's answer to the mobile computing
problem.

Simply put, manually configuring zillions of PCs out there, even if it
is just listing them in some bootptab, is something that Microsoft's
large customers don't want to do, and the customer is always right.

I still think that eventually *all* (except name servers') numeric
addresses will become dynamic, essentially fleeting things, used only to
identify a session between two machines. All permanence will be in
*names*. Perhaps it is not that bad a future, even if it will break
quite a few things.

-----------[000153][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 00:02:05 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Getting duplicate IP addresses
In article <[email protected]> [email protected] (Elena Leong) writes:
>We have a SunSparc2 running SunOS 4.1.3.  It has two Ethernet cards.  le0 is
>connected to our project LAN and has an IP network 192.245.13.0 happily
>happening.  le1 is connected to the backbone APost LAN and we also happily
>run FTAM/OSI to some of the VAXes on this LAN.
>
>Now, our Sun starts to complain, like this:
>
>Aug 31 17:32:35 grunge vmunix: duplicate IP address! send from ethernet 
>address: aa:0:4:0:1f:2c
>Sep 2 09:59:10 grunge vmunix: duplicate IP address! send from ethernet 
>address: aa:0:4:0:65:28
 
>The APost network guys tell me that these identify to be DEC Ultrixs 
>running v4.2a and v4.2 resp, running TCP/IP on a network of 155.144.0.0
>These arent the only ones, BTW, there appear to be 5/6 others, all Unix
>based DECs....

A common cause of this kind of error is Proxy ARP.  This kind of error will
occur if there's a Proxy ARP router on the network, and the Sun is
configured with an address that has an incorrect subnet portion.  When it
boots, the Sun sends out an ARP query for its own address; the Proxy ARP
router sees this, recognizes that the address isn't local, so it sends back
a reply.  When the Sun sees this reply it thinks someone else has its
address.

-- 
Barry Margolin                                                [email protected]


-----------[000154][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 00:31:27 GMT
From:      [email protected] (Michael Salmons)
To:        comp.protocols.tcp-ip
Subject:   Dialup via MacTCP? (Q)

Hello all...

I've ordered my copy of the MacTCP control panel, and I'm going
to try establishing a bootp via dialup access. I know to
use Mosaic, MacWeb, etc. for navigating, but what do I use
to dial in my connection? Is there a share/freeware control
panel or app for dialing up an IP connection?

Any replies appreciated!!

Michael Salmons

-----------[000155][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 02:39:56 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Making secondary DNS sites work
I'm having some difficulty getting my secondary DNS site to
automatically transfer the primary information after serial numbers are
updated.

I'm running BIND 4.8.3 under SunOS 4.1.3_U1 on a Sparc 1 and a Sparc
IPX as primary and secondary DNS nameservers.

I've been pretty good about updating serial numbers, but I had the
secondary site expire data today (a week since last update). Do I need
to add a crontab entry on either machine to force a transfer?

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000156][next][prev][last][first]----------------------------------------------------
Date:      Thu, 08 Sep 94 09:51:15 -0400
From:      "Kenneth Chin" <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Use of NIT filters in SunOS
Hi Folks,

I've asked this question before but I've lost the answer (it happens
when you moved from job to job too much!).   So here it goes again:

The basic problem is I want to capture packets which falls within a
set of UDP port numbers.  I'm using the /dev/nit interface in SunOS
4.1.3, the nit filtering package.  The following code fragment should
work, but it doesn't:

	for (i = 0; i < num_ports; i++) {
		*fwp++ = ENF_PUSHWORD + offset;
		*fwp++ = ENF_PUSHLIT | ENF_COR;
		*fwp++ = htons(port[i]);
	}

If num_ports = 1, this code works fine.  Anything greater than 1, it
doesn't return anything.  Of course, above this, I check to insure that
it's a UDP packet.

Any help would be appreciated!

Ken Chin
Pariclete Systems, Inc.
(212)249-0931
[email protected]


-----------[000157][next][prev][last][first]----------------------------------------------------
Date:      7 Sep 1994 10:47:43 +1000
From:      [email protected] (Colin Campbell)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing unser SunOS and Class B nets
Perhaps I could add some more information :-). SunOS will not allow me
to specify routes to subnets of the same network on the same router
but will allow me to specify routes to subnets of the same net on
different routers. ie

     route add net 147.132.157.0 203.5.10.2 4
     route add net 147.132.176.0 203.5.10.2 2

Fails, while

     route add net 147.132.157.0 203.5.10.2 4
     route add net 147.132.176.0 203.5.10.1 2

works. Figure that one out.

Colin

Herr Lein ([email protected]) wrote:

	[ my stuff deleted ]

: As far as I know, one of the basic ideas of IP routing says, that ONE class B
: net has ONE entry point. . So the routing algorithm
: on SunOS reads only the first two bytes of a route entry for a class B net, all
: further information are ignored. (Will say  "route add net 147.132.157.0 ..."
: is equivalent to "route add net 147.132.0.0 ...") Take care: this is true
: for all class B networks except those, the node has an interface to.
: (Cum grano salis: Its the same for Class A or C, of cause)
 
: You should be aware of the fact, that a subnet oriented routing is only possible,
: if the system knows the subnet mask of the remote class B net. But there is no
: way to tell it. "/etc/netmasks" could be the right file to store the data, but
: I did not found any way to pass this data to the kernel, where the routing takes place.
 
: Some routers (pure routers, as cisco's) gives you much more possibilities for
: selecting or supressing routing. But this stuff only work for locally in the
: router, it does not enable you to have a node in the net, which routes two
: parts of a class B net to diferent gateways.

-----------[000158][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 10:12:53
From:      [email protected] (Shawn Curtis)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]> [email protected] (Michael Salmon) writes:
>From: [email protected] (Michael Salmon)
>Subject: Re: TN5250/TN3270 Standards?
>Date: 7 Sep 1994 14:06:47 GMT
 
>In article <[email protected]>
>[email protected] (Clark Bremer) writes:
>|> In article <[email protected]>
>|> [email protected] (Michael Salmon) writes:
>|> >In article <[email protected]>
>|> >[email protected] (Rich Oldroyd) writes:
>|> >|> 
>|> >|> We're looking for the  specifications/RFCs covering the
>|> >|> TN3270 and/or TN5250 products. We're aware that we
>|> >|> can buy some off the shelf products supporting these
>|> >|> standards but would like the specs themselves. 
>|> >|> Any pointers?
 
>|> >To the best of my knowledge these are proprietry IBM protocols and you
>|> >have to by the specs. from them.
>|> 
>|> I'm not sure where its written down, but for TN3270, I'm sure of the answer - 
>|> its simply to negotiate for "binary mode" on and "send end of record" on.  CB.
 
>Sorry but that's how you enter 3270 mode in Telnet. Telnet then become
>the carrier for the 3270 protocol.
 
>-- 
 
>Michael Salmon
 
>#include        <standard.disclaimer>
>#include        <witty.saying>
>#include        <fancy.pseudo.graphics>
 
>Ericsson Telecom AB
>Stockholm

There are no RFCs specifically for TN5250.  In the case of TN3270, you might 
check RFC 1576 "TN3270 Common Practices" available at ds.internic.net.
This RFC discusses the Telnet options mentioned above and provides some 
additional detail regarding... (you guessed it!)... common practices in 
implementing TN3270.  As Michael mentioned Telnet is just the carrier for 3270 
datastream after binary mode is negotiated, IBM is your source for 3270 
datastream programming information.

--
Shawn Curtis
Sr. Software Engineer
OpenConnect Systems, Inc.
Dallas, TX
[email protected]
--
Shawn Curtis, Sr. Software Engineer
OpenConnect Systems, Inc.
[email protected]
(214)484-5200
Fax(214)888-0686

-----------[000159][next][prev][last][first]----------------------------------------------------
Date:      Wed, 7 Sep 1994 18:06:08 GMT+1100
From:      [email protected] (Gerrit Thomson)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip,alt.winsock
Subject:   Re: Ethernet to Slip converter or repeater
Hi Malcolm,
    I notice you are running pcroute 2.2 perhaps 2.4 may solve a problem or 
two. I have a setup using pcroute2.4 as a slip dial in. Be assured that 
pcroute uses very simple slip. Make sure that your home pc slip does not use 
any compression or fancy headers. just plain simple slip.
     ip addresses: The ip adress you give the slip side of pcroute is the 
gateway for your pc. Your pc at home will need to be in the same subnet as 
the pcroute slip gateway adress.
     ie.
        work pc ip= 103.45.23.11
        pcroute route ip= 103.45.23.11
        pcroute slip ip = 103.45.30.1
        pc at home ip=103.45.30.6
        pc at home gateway = pcroute slip ip = 103.45.30.1
   There is no need for the slip packet driver if you are using trumpet 
winsock use the built in one. If you are using PC-FTP slip packet driver 
etc. then make sure to configure it for standard slip as possible.
Cheers,
    Gerrit Thomson

In article 
<[email protected]> [email protected] (Malcolm 
Sainsbury) writes:>From: [email protected] (Malcolm Sainsbury)>Subject: 
Ethernet to Slip converter or repeater>Date: Tue, 6 Sep 1994 17:00:41 
GMT>Summary: Is there a PC based ethernet to slip repeater/converter>Keywords: 
ethernet,slip,ibmpc

>I have searched the FAQ's at hand so I hope this isn't too trivial a 
>question for the net experts.
 
>Is there a PC-based prog that I can use to forward ethernet packets on the 
>one side, up a serial line, talking SLIP on the other.
 
>Let me explain:
 
>I have a leased line from my home to my office running at 38400bps. (nice)
>The PC in my office is connected via a WD8003E card onto the net. (great)  
>The PC at home is all fired up with a SLIP packet driver.  (fine)
>Now how do I make the two talk without buying another PC.
 
>I configured PCRoute v2.2 into my PC at work and it certainly will take 
>ethernet packets off the net and send them up the serial port,
>BUT not in SLIP form (it would seem).  It wants to talk its own protocol to 
>ANOTHER PC running PCRoute v2.2 on the other side and from there I would 
>connect my home system.  (thats a waste of a PC, even if I had one)
 
>Isn't there some simple software that I can load in my work machine that 
>will pluck my ethernet packets off the WD8003 and send them in proper SLIP 
>form directly into my COM port at home, to be gratefully received by the 
>packet driver waiting expectantly there for proper slip packets ?
 
>Kind regards
 
>Malcolm
>--
>Malcolm Sainsbury - Dept of Information Systems - Rhodes University
>Internet: [email protected]   Phone: +27 461 318244 (25049 fax)


-----------[000160][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 14:24:35 GMT
From:      [email protected] (Sandy Fraser)
To:        comp.protocols.tcp-ip
Subject:   Etherman, Interman
	Hi all - I compiled a copy of these tools (wonderful!) at my last
	job, and brought the executables with me.  I forgot about the
	hershey fonts, however (/usr/local/lib/hershey), and I can't 
	remember where I found them before.  Could someone please let me
	know of an ftp site where I can pick them up?

	Thanks much,

	Sandy




-----------[000161][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 14:30:07 GMT
From:      [email protected] (Todd A Wallace)
To:        comp.protocols.tcp-ip
Subject:   PPP driver for TCP/IP32 from Microsoft
Does anyone know of a PPP driver that works with a dialup PPP account on
the one hand, and the TCP/IP32 package that works with  Windows for
Workgroups that you can get from Microsoft?

----------------------------------------------------------------
|    Todd Wallace               |  "A pessimist is surprised   |
|    [email protected]    |   as often as an optimist,   |
|-------------------------------|   but always pleasantly."    |
| Expatriate Midwesterner (tm)  |            - Robert Heinlein |
----------------------------------------------------------------

-----------[000162][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 16:36:57 GMT
From:      [email protected] (Werner Fouche )
To:        comp.protocols.tcp-ip
Subject:   Wanted: TCP/IP Developer's Kit
I am looking for a commercial/public domain developer's kit for TCP/IP
(portable source code for TCP/IP)
to be used in a micro-kernel based operating system. I've seen
an advertisement for FUSION TCP/IP by Network Research, Oxnard CA.
Unfortunately, they do not seem to exist anymore.

Any help is greatly appreciated.

	Werner

-----------[000163][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 16:58:07 GMT
From:      [email protected] (Manfred Kwiatkowski)
To:        comp.protocols.tcp-ip
Subject:   Re: Routing unser SunOS and Class B nets
In article <[email protected]>, [email protected] (Colin Campbell) writes:
> Perhaps I could add some more information :-). SunOS will not allow me
> to specify routes to subnets of the same network on the same router
> but will allow me to specify routes to subnets of the same net on
> different routers. ie
> 
>      route add net 147.132.157.0 203.5.10.2 4
>      route add net 147.132.176.0 203.5.10.2 2
> 
> Fails, while
> 
>      route add net 147.132.157.0 203.5.10.2 4
>      route add net 147.132.176.0 203.5.10.1 2
> 
> works. Figure that one out.
> 
So what? route just does not crosscheck every possibility. Anyhow,
the whole of net 147.132 is routed via the first reference to 
147.132.what.ever appearing in the routing table.

-- 
Manfred Kwiatkowski         [email protected]

-----------[000164][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 17:01:40 +0000
From:      [email protected] (Steve Cashmore)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
In article: <[email protected]>  [email protected] (Tony Trehan) writes:
> I need to provide TCP/IP (FTP, TELNET etc) in embedded
> industrial controllers running under real-time operating systems.....
 
Me too please!!
--
Steve


-----------[000165][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 17:10:45 GMT
From:      [email protected] (Manfred Kwiatkowski)
To:        comp.protocols.tcp-ip
Subject:   Re: Getting duplicate IP addresses
In article <[email protected]>, [email protected] (Barry Margolin) writes:
> In article <[email protected]> [email protected] (Elena Leong) writes:
> >We have a SunSparc2 running SunOS 4.1.3.  It has two Ethernet cards.  le0 is
> >connected to our project LAN and has an IP network 192.245.13.0 happily
> >happening.  le1 is connected to the backbone APost LAN and we also happily
> >run FTAM/OSI to some of the VAXes on this LAN.
> >
> >Now, our Sun starts to complain, like this:
> >
> >Aug 31 17:32:35 grunge vmunix: duplicate IP address! send from ethernet 
> >address: aa:0:4:0:1f:2c
> >Sep 2 09:59:10 grunge vmunix: duplicate IP address! send from ethernet 
> >address: aa:0:4:0:65:28
 
> >The APost network guys tell me that these identify to be DEC Ultrixs 
> >running v4.2a and v4.2 resp, running TCP/IP on a network of 155.144.0.0
> >These arent the only ones, BTW, there appear to be 5/6 others, all Unix
> >based DECs....
> 
> A common cause of this kind of error is Proxy ARP.  This kind of error will
> occur if there's a Proxy ARP router on the network, and the Sun is
> configured with an address that has an incorrect subnet portion.  When it
> boots, the Sun sends out an ARP query for its own address; the Proxy ARP
> router sees this, recognizes that the address isn't local, so it sends back
> a reply.  When the Sun sees this reply it thinks someone else has its
> address.

These addresses look like DECnet addresses to me. DECnet has the habit
of changing hardware adresses to node-addresses. If DECnet is started
a f t e r  the Ultrix machine has sent an arp-response with the original 
address the Suns will be puzzled.

-- 
Manfred Kwiatkowski         [email protected]

-----------[000166][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 17:34:56 GMT
From:      [email protected] (Jim Shankland)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
In article <[email protected]> [email protected] (Paul Smith) writes:
>Casper H.S. Dik ([email protected]) wrote:
>: As with all Unix files.  EOF is a read that returns 0 (errno undefined).
>: A close on one end of a connection (pipe/socket/whatever), will cause
>: an EOF.  It is not an error condition at the communication layer.
 
>: When you want to detect the close, you select/poll for reading and read will
>: return 0.
>
>Very well stated!  The socket API extends the Unix file oriented programming
>paradigm at the expense of clarity and precision!  

Of course, this is purely a statement of opinion.  I don't share it.
Normal closing of the write end of a data stream does not strike
me as an error condition for the reader.

>No harm if errno where to be set to ENOCON or what ever in this scenario...
>
>And yes errno is 
>indeterminate, but my style is to errno=0; prior to ever system call so as
>to differentiate for sure who set errno to != 0.

This, on the other hand, betrays some confusion.  errno is set as
a result of a system call if and only if the system call fails (generally,
returns -1).  Setting errno to 0 before the system call helps nothing.
The correct test is to see if the system call has failed, and check
errno only if it has.

jas


-----------[000167][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 94 18:20:17 GMT
From:      [email protected] (Lou Fernandez)
To:        comp.protocols.tcp-ip
Subject:   Re: Extending the TELNET protocol
In article <[email protected]>,
Nicko van Someren <[email protected]> wrote:
>The TELNET protocol supports option negotiation.  How do you go about
>getting a new option standardised?  For ages I have thought it would
>be a really useful feature if there were a pair of options to allow
>the client to tell the server the size of the window.

Check out RFC 1073, Telnet Window Size Option.  It should do what you
want.

...Lou
-- 
Louis F. Fernandez			Sequent Computer Systems
[email protected]			Beaverton, OR 97006-6063

-----------[000168][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 18:57:36 GMT
From:      [email protected] (Brian Pollack)
To:        comp.protocols.tcp-ip
Subject:   Re: Setting up MacTCP
Heather L. Nadelman ([email protected]) wrote:
:>Hi, guys!
 
:>I'm helping a friend set up MacTCP.  I am wholly unfamiliar with Macs, but I 
:>have gotten pretty proficient at setting up Trumpet Winsock with SLIP.  She's 
:>using a powerbook and wants to dial in to both her ethernet network at work 
:>and the internet.  What do I need?  Is there some handy-dandy FAQ that I could 
:>use?   Is it trickier than Winsock?

Check out or web server, one of our users has a great MacTCP and InterSLIP
setup page.  http://www.indirect.com/

:>Thanks in advance!
 
:>--Heather L. Nadelman
:>[email protected]


--
This is my sig file.
It is empty;

-----------[000169][next][prev][last][first]----------------------------------------------------
Date:      Thu, 8 Sep 1994 19:24:22 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
In article <[email protected]> MARCO PACE <[email protected]> writes:
>- did anybody have a similar problem and can provide me with some useful hint
>  as to where the problem is located ?
UDP is an _unreliable_ protocol (at all levels). You probably want to either
use TCP, or some flow control.

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000170][next][prev][last][first]----------------------------------------------------
Date:      Thu, 08 Sep 94 19:39:01 GMT
From:      [email protected] (Jonathan Walker)
To:        comp.protocols.tcp-ip
Subject:   IP/TCP/UDP checksum implementation in C?
Does anyone have some source, or know where I can get some source, for 
a C impementation of the checksum that is present in the various IP
headers? 

Thanks,
Jon Walker
[email protected]

-----------[000171][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 94 20:00:11 GMT
From:      [email protected] (Blair Porter)
To:        biz.sco.general,comp.protocols.tcp-ip
Subject:   dlput: no streams buffer ???
I have posted on this before, but have never received any kind of response,
so I will try once again.  I am using SCO ODT 1.0, and periodically get
into a hung state.  The entire workstation locks up, and a reset/reboot is
the only way out.  When this happens, the file /usr/adm/messages begins
to fill up remarkably fast with a repetitive message shown below:


Thu Sep 8 14:36:20
dlput: no streams buffer

dlput: no streams buffer
dlput: no streams buffer
dlput: no streams buffer

The dlput: no streams buffer message just keeps repeating, indefinitely.
This has happened at night before and the file has grown to in excess of
10 MB by the next day.  Obviously this is inconvenient and a real problem.
Not being a UNIX kernel/tuning wizard, I don't know what to do to correct
this situation.  I am darn tired of rebooting every time this happens.
Actually I am just plain tired of it happening in the first place.  Can
ANYONE help me out and point me in the right direction to correct this
situation???

Thanks for anyone willing to respond.
--
================================================================================
  Blair Porter - Process Specialist		One Bell Center  23-L-1
  Computer Operations Division			St. Louis, MO  63101
  Information Services				Work: (314) 235-3419
  Southwestern Bell Telephone Company		FAX:  (314) 235-1397
  E-mail: [email protected]
	THE RACE GOES TO THOSE WHO KEEP ON RUNNING...
================================================================================

-----------[000172][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 20:21:04 GMT
From:      [email protected] (Pat Hennessy)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
In article <[email protected]>  writes:
> I need to provide TCP/IP (FTP, TELNET etc) in embedded
> industrial controllers running under real-time operating systems.....
> 
> Can anybody provide me pointers to:
> 
> 1) Where can i get "general info" about TCP/IP on embedded systems?
> 2) Any commercial drivers available in the marketplace (small,  
 re-entrant code
> 3) ANY suggestions etc....????
> 
> Thanx a lot!
> -Tony


Get a copy of "Embedded Systems Programming" magazine.  Any month will do.   
I was investigating this same thing a couple of months ago and I must have  
seen a dozen vendors (most of them RTOS vendors) in the ads with this  
capability.  Many of the smaller kernel companies provide full source  
code.  Some examples below:

QNX (Canada)                          (613) 591-3579
U.S. Software                         (800) 356-7097
System Six                            (303) 526-1747
Intermetrics                          (800) 356-3594
Integrated Systems (pSOS)             (408) 980-1500
RTXC (very friendly, very helpful)    (713) 728-9688
JMI Software Systems                  (215) 628-0840
SMX                                   (800) 366-2491


Good luck,

Pat

-----------[000173][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 22:13:53 GMT
From:      [email protected] (Paul Hilchey)
To:        comp.protocols.tcp-ip
Subject:   Re: IP/TCP/UDP checksum implementation in C?

In article <[email protected]>, [email protected] (Jonathan Walker) writes:
|>Does anyone have some source, or know where I can get some source, for 
|>a C impementation of the checksum that is present in the various IP
|>headers? 


/* Perform a 16-bit ones-complement checksum.
   For summing non-contiguous blocks of data, make several calls
   and pass the result of the previous call as accum.  Use -1 for
   accum on the first (or only) call.
*/
unsigned short
ipsum(unsigned short accum, unsigned short *data, int bytes)
{
        unsigned long sum;

        sum = ~accum & 0xffff;

        while (bytes >= 2) {
                sum += *data++;
                bytes -=2;
        }

        if (bytes == 1)
                sum += *((unsigned char *)data) << 8;

        sum = (sum & 0xffff) + (sum >> 16);     /* add carry */
        sum = (sum & 0xffff) + (sum >> 16);     /* and again */

        return (~sum & 0xffff);                 /* complement and return */
}

-----------[000174][next][prev][last][first]----------------------------------------------------
Date:      Thu, 08 Sep 1994 07:43:07 +0900
From:      [email protected] (Tim Bigham)
To:        comp.protocols.tcp-ip
Subject:   [Q] TCP/IP MIL-STD
Which MIL-STD has the specification for tcp/ip?

-- 
tim ([email protected])

-----------[000175][next][prev][last][first]----------------------------------------------------
Date:      8 Sep 1994 23:08:02 GMT
From:      [email protected] (Urs - Rufer)
To:        comp.protocols.tcp-ip
Subject:   IP multicasting under OSF/1

When using IP multicasting on Ethernet via Berkley Socket under DEC OSF/1, 
the sending of the datagram was blocked due to "Unreachable Hoat".  The IP 
multicasting address used is "224.0.0.2". Is there anything need to set up 
in network software for IP multicasting implementation ?  If you have any 
experience in this area, please help !!

Thanks in advance.


-----------[000176][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 00:27:46 GMT
From:      [email protected] (Sam Ghandchi)
To:        comp.protocols.tcp-ip
Subject:   Are TTCP and NetPerf Available for Windows NT?
Hi,

Are TTCP and NetPerf Available for Windows NT?


TIA,
- SG

-----------[000177][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 00:30:58 GMT
From:      [email protected] (Mark Andrews)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Making secondary DNS sites work
In article <[email protected]> [email protected] (Glenn Fleishman) writes:
>I'm having some difficulty getting my secondary DNS site to
>automatically transfer the primary information after serial numbers are
>updated.
>
>I'm running BIND 4.8.3 under SunOS 4.1.3_U1 on a Sparc 1 and a Sparc
>IPX as primary and secondary DNS nameservers.
>
>I've been pretty good about updating serial numbers, but I had the
>secondary site expire data today (a week since last update). Do I need
>to add a crontab entry on either machine to force a transfer?
>
>----
>Glenn Fleishman, Point of Presence Company, Seattle, Washington
>        Send to <[email protected]> for information on our services
>        Moderator, Internet Marketing discussion list; send the message
>        INFO INET-MARKETING to the address [email protected] for info

	Are you using dot notation for youre serial numbers as in
	6.436? Did you go from 6.3xx -> 6.4 (or similar)?

	Dot notation is broken in bind 4.8.X. 

	The above translation is equivelent to 60003xx -> 60004 which is
	a counter intuative change.

	I would recomment that just encode the date as YYMMDDVV where
	YY is the current year MM the current month DD the current day
	and VV allows for 100 updates a day.

	Mark

-----------[000178][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 09:20:20 -0400
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Know of a ftpable IPng spec?

I am looking for the IPng spec.... anyone know where I could find it?

Thanks in advance...

Lutz

Lutz C. Jacob       
Internet: [email protected]   
ATT: 0-700-LCJACOB

-----------[000179][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 05:35:57 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Dialup via MacTCP? (Q)
In article <[email protected]> [email protected] (Michael Salmons) writes:
>Is there a share/freeware control panel or app for dialing up an IP connection?

InterSLIP is a free implementation of SLIP for Macintosh from Intercon, one
of the leading vendors of Macintosh TCP/IP software.  You can download it
from ftp.intercon.com.
-- 
Barry Margolin                                                [email protected]


-----------[000180][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 94 11:50:51 MDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   TCP or UDP?  That is the question.
Hello all!  I am doing some programming in TCP/UDP, and I want to confirm,
or dispell something that was told to me.  From my question, you will
see that I am a new comer to tcp/ip programming... :)

I was told that when using UDP (DGRAM), it doesn't make sure that the
information arrives in order, or that it will arrive at all, but if it
does arrive, IT IS THE CORRECT DATA THAT WAS SENT.

I was also told that when using TCP (STREAM), it makes sure that the data 
arrives in the right order, and if it doesn't arrive, you will be notified,
etc...  but IT DOES ***NOT*** MEAN THAT THE DATA THAT WAS SENT IS THE DATA
YOU GET!  

I am working over frame-relay lines, and because it is possible to lose
packets during overloads, I would like to use TCP (STREAM) so it will 
fix this undesirable feature.  But!  it is more important to me to have
the data I send be correct when it gets there... 

This idea that TCP doesn't make sure the data is the same seems a little
strange, what is is timely data if it isn't the data you requested!!!  haha

I would really appreciate clarification.  Another person said he thought that
TCP had all of the good features of UDP, plus it made sure the data arrived
in order, etc...

I am sorry if this is posted in a FAQ somewhere, the FAQ's I found didn't
mention it that I could find...  (how do you find the FAQ's anyway)  is
there a FAQ news group???).

Thanks, Jeff


-----------[000181][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 06:47:37 GMT
From:      [email protected] (Wayne Baird)
To:        comp.protocols.tcp-ip
Subject:   Connecting IP networks to Internet
I work for a large national corporation.  We have a private TCP/IP
internet built on ACC routers and consisting of almost 200 Class C
networks.

Problem:  We want to provide Internet access using our existing
backbone.  There aren't many nodes which need access but they're
scattered across the country.

I am assuming that we would not be able to get enough Class C
addresses or ANY Class B address.
I've toyed with the idea of a "router" which could satisfy our
needs.  This would be a box with two LAN interfaces, one pointing
towards the Internet and one towards our private network.  The
"router" would maintain a table mapping private IP addresses to 
"public" Internet addresses.  When receiving an outbound IP packet
it would perform a table lookup and replace the private source 
address with it's public counterpart.  For inbound packets it would
do the same for the destination address.  An added value would be
some firewall protection.  Only pre-defined private addresses could
be accessed from the Internet. Our private routers would point
to this box as their default route.  Routing within the private
network would be unaffected since we are using a dynamic routing
protocol (RIP, moving to OSPF in future).

I'd be very interested in opinions on this concept or any other ideas
you might have.

-----------[000182][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 06:51:33 GMT
From:      [email protected] (Christoph Conrad)
To:        comp.protocols.tcp-ip
Subject:   [Q] Simple Sockets Library: Functionality, E-Mail adress of company
Hello,

does anyone have a more detailed view of the Simple Sockets Library (see FAQ)?
A E-Mail address for 'Austin Code Works'?

Thanks a lot,
Christoph.
-- 
[email protected]

-----------[000183][next][prev][last][first]----------------------------------------------------
Date:      Fri, 09 Sep 1994 15:19:19 -0500
From:      [email protected] (Eric Scouten)
To:        comp.protocols.tcp-ip
Subject:   Re: Extending the TELNET protocol
In article <[email protected]>, [email protected]
(Nicko van Someren) wrote:

> The TELNET protocol supports option negotiation.  How do you go about
> getting a new option standardised?  For ages I have thought it would
> be a really useful feature if there were a pair of options to allow
> the client to tell the server the size of the window.

I don't think you need to create a new standard for this one. Find an FTP
site that carries the IETF Request For Comments. Get a copy of RFC 1073
(dated 10/88). This describes the Telnet Negotiate About Window Size
option (NAWS) which should do exactly what you're looking for.

-Eric

__________________________________________________________________________
Eric Scouten <[email protected]> * MS Comp Sci, Univ of Illinois

They're my feet and I'll put them in my mouth if I want to.

-----------[000184][next][prev][last][first]----------------------------------------------------
Date:      Fri, 09 Sep 1994 15:23:48 -0500
From:      [email protected] (Eric Scouten)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <[email protected]>, [email protected] wrote:

> Hello all!  I am doing some programming in TCP/UDP, and I want to confirm,
> or dispell something that was told to me.  From my question, you will
> see that I am a new comer to tcp/ip programming... :)
> 
> I was told that when using UDP (DGRAM), it doesn't make sure that the
> information arrives in order, or that it will arrive at all, but if it
> does arrive, IT IS THE CORRECT DATA THAT WAS SENT.
> 
> I was also told that when using TCP (STREAM), it makes sure that the data 
> arrives in the right order, and if it doesn't arrive, you will be notified,
> etc...  but IT DOES ***NOT*** MEAN THAT THE DATA THAT WAS SENT IS THE DATA
> YOU GET!

I'm not sure where you got that idea. TCP enforces data validity as well
as data sequence on the receiving end (i.e., the data you receive will be
the same data in the same order as it was sent). This means delays may
occur while the TCP daemons request retransmission. Notifications occur
only when the network completely fails to deliver data.


I think both TCP and UDP contain checksum-type information to enforce data
validity.

-Eric

__________________________________________________________________________
Eric Scouten <[email protected]> * MS Comp Sci, Univ of Illinois

They're my feet and I'll put them in my mouth if I want to.

-----------[000185][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 08:47:14 GMT
From:      [email protected] (Marten Terpstra)
To:        comp.protocols.tcp-ip
Subject:   Re: Connecting IP networks to Internet
In <[email protected]> [email protected] (Wayne Baird) writes:

[...]

>I am assuming that we would not be able to get enough Class C
>addresses or ANY Class B address.
>I've toyed with the idea of a "router" which could satisfy our
>needs.  This would be a box with two LAN interfaces, one pointing
>towards the Internet and one towards our private network.  The
>"router" would maintain a table mapping private IP addresses to 
>"public" Internet addresses.  When receiving an outbound IP packet
>it would perform a table lookup and replace the private source 
>address with it's public counterpart.  For inbound packets it would
>do the same for the destination address.  An added value would be
>some firewall protection.  Only pre-defined private addresses could
>be accessed from the Internet. Our private routers would point
>to this box as their default route.  Routing within the private
>network would be unaffected since we are using a dynamic routing
>protocol (RIP, moving to OSPF in future).
 
>I'd be very interested in opinions on this concept or any other ideas
>you might have.

This is not a new idea. You may want to have a look at RFC 1631. It is
called " The IP Network Address Translator (NAT)".

-Marten

-----------[000186][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 11:02:28 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
In article <[email protected]> [email protected] (Paul Smith) writes:
>But, this is one of my greatest grips about SOCK_STREAM and circuit failures! 
>Why couldn't a reasonable errno be returned consistantly.  The Lachman stack
>on SVR4.2 v1.1.2 returns 0, errno = 0 for the server when the client side
>closes() or exits premature.  The Client side however returns -1, errno=EPIPE
>if the server exits premature??

If you write to a closed connection you get a SIGPIPE and the client by
default dies off. It's consistent with pipes and a logical way of avoiding
processes hanging around. 

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000187][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 12:49:34 GMT
From:      [email protected] (leslie.j.williams)
To:        comp.protocols.tcp-ip
Subject:   TCP-IP Sockets left hanging -- HELP
I've been working on a program that uses TCP-IP Sockets to communicate via our
ethenet LAN.  I have a C++ class that I use to establish the link, read/write
over it and it's destructor closes the socket.  However, after the program dies
netstat still thinks my socket is active.  I have to wait about 1 minute until
it is cleared and I can run my program again.  Am I not closing the socket 
correctly?  Can I not trust my destructor?  Is something else causing the
socket to be hung???  Please help!

Also, I can use netstat to see the status of my socket.  Is there some tool
I can use to kill that socket?

Thanks for the help,

			LJ Williams

-----------[000188][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 15:36:53 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Are TTCP and NetPerf Available for Windows NT?
In article <[email protected]> [email protected] (Sam Ghandchi) writes:
>
>Are TTCP and NetPerf Available for Windows NT?

You generally get source for those benchmarks, and since Windows NT is
said to offer a socket interface, porting should be possible.  It should
be quite easy to port ttcp.c, because it is such a trivial program.

As with any benchmark, you should not trust your results if you do not
understand ttcp.c well enough to port its source to a similar system.
Simply running a program and observing the numbers it produces is almost
completely unrelated to useful benchmarking.  You must really understand
the benchmark.  All benchmarks are more or less bogus.  They are at
least somewhat and generally very different from the target applications.
Those who are unwilling to expend the effort to understand a benchmark
must not expect to it to do them much good.

About the only "science" or "scientific method" in "computer science"
is in benchmarking, where people make hypotheses, test them, report
repeatable results, replicate each others' experiments, and so on.  Like
real science, cheating in benchmarks such as doing the experiment without
a clue about the theory being tested rarely produces useful results.

That the benchmark numbers reported by the friendly salescritters from
your favorite vendor are often suspect is implied by this thereom.


Vernon Schryver    [email protected]

-----------[000189][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 16:04:08 GMT
From:      [email protected] (Steven Bellovin)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP Sockets left hanging -- HELP
In article <[email protected]>, [email protected] (leslie.j.williams) writes:
> I've been working on a program that uses TCP-IP Sockets to communicate via our
> ethenet LAN.  I have a C++ class that I use to establish the link, read/write
> over it and it's destructor closes the socket.  However, after the program dies
> netstat still thinks my socket is active.  I have to wait about 1 minute until
> it is cleared and I can run my program again.  Am I not closing the socket 
> correctly?  Can I not trust my destructor?  Is something else causing the
> socket to be hung???  Please help!
> 
> Also, I can use netstat to see the status of my socket.  Is there some tool
> I can use to kill that socket?

Exactly what does netstat say?  I suspect that it shows the socket in
TIMEWAIT state, which is a required part of TCP.  It has nothing to do
with sockets, C++, or anything being hung.  See any good book on TCP (i.e.,
Stevens or Comer) for an explanatoin.

The two usual bypasses to the problem are (a) to run your application
from inetd, so that it doesn't have to create a socket each time (but
not all applications are suited to this), or (b) use the SIOCREUSEADDR
ioct before doing the bind().  I don't know if your class library will
permit this, though.

-----------[000190][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 17:03:54 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Looking for Ipaddressing management software
I am looking for any software available to assist in the managemnt of Ip addres
ses. Things like whos got what, blocking addresses and keeping records or a sma
ll database of the addresses. The operating systems I have available to run on
are dos and Sunos....thanks any help will be appreciated.

-----------[000191][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 17:19:14 GMT
From:      [email protected] (Christopher Green)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
In article <[email protected]> [email protected] (Tim Bigham) writes:
>Which MIL-STD has the specification for tcp/ip?
>
>-- 
>tim ([email protected])

The MIL-STD specifications are obsolete.  The following should be used instead:

RFC 1600, Internet Official Protocol Standards, March 1994

RFC 1009, Requirements for Internet Gateways, June 1987

RFC 1122, Requirements for Internet Hosts -- Communication Layers, Oct 1989

RFC 1123, Requirements for Internet Hosts -- Application and Support, Oct 1989

For the record, the MIL-STDs are:

MIL-STD-1777 for IP
MIL-STD-1778 for TCP
MIL-STD-1780 for FTP
MIL-STD-1781 for SMTP
MIL-STD-1782 for TELNET

For more information on Internet standards, get RFC 1600 from any RFC
repository.

Chris Green                                          Consulting Systems Analyst

AMINTEL
-------------------------------------------------------------------------------
                                                   Applied Machine Intelligence
111 Pacifica, Suite 250                                    Phone (714) 727-2493
Irvine, CA 92718-3311                                        Fax (714) 727-2495

-----------[000192][next][prev][last][first]----------------------------------------------------
Date:      09 Sep 1994 18:01:45 GMT
From:      [email protected] (Russell Nelson)
To:        comp.protocols.tcp-ip,comp.unix.solaris
Subject:   Re: Lost UDP packets above network interface layer !!
In article <[email protected]> [email protected] (Alan Cox) writes:

   In article <[email protected]> MARCO PACE <[email protected]> writes:

   >- did anybody have a similar problem and can provide me with some
   >  useful hint as to where the problem is located ?

   UDP is an _unreliable_ protocol (at all levels). You probably want to either
   use TCP, or some flow control.

No, UDP is an unreliable *delivery* protocol.  The protocol itself is
perfectly reliable -- after all, the DNS is built on top of it.

--
-russ <[email protected]>    http://www.crynwr.com/crynwr/nelson.html
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)  | What is thee doing about it?
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000193][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 19:02:09 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: Getting duplicate IP addresses
In article <[email protected]> [email protected] (Manfred Kwiatkowski) writes:

> ...
>> >Sep 2 09:59:10 grunge vmunix: duplicate IP address! send from ethernet 
>> >address: aa:0:4:0:65:28
 
> ...
>These addresses look like DECnet addresses to me. DECnet has the habit
>of changing hardware adresses to node-addresses. If DECnet is started
>a f t e r  the Ultrix machine has sent an arp-response with the original 
>address the Suns will be puzzled.

Reasonable DECnet implementations or reasonable Ethernet or FDDI drivers
broadcast a new ARP response when they change their hardware address,
and reasonable IP implementations pay attention to such those ARP packets.
This keeps systems from being "puzzled" no matter when DECnet is started.
I have no idea whether Sun's are "reasonable" in these respects, but
one should hope they are.


Vernon Schryver    [email protected]

-----------[000194][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 19:52:58 GMT
From:      [email protected] (John Bottoms)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
From: [email protected] (Tim Bigham)
>Which MIL-STD has the specification for tcp/ip?

Check out MIL-STD 1777 and MIL-STD 1778.

 John Bottoms
 Consulting Engineer
 Avalon Systems Inc.
 Concord, MA



-----------[000195][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 20:19:30 GMT
From:      [email protected] (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <[email protected]>, [email protected] writes:
|> Hello all!  I am doing some programming in TCP/UDP, and I want to confirm,
|> or dispell something that was told to me.  From my question, you will
|> see that I am a new comer to tcp/ip programming... :)
|> 
|> I was told that when using UDP (DGRAM), it doesn't make sure that the
|> information arrives in order, or that it will arrive at all, but if it
|> does arrive, IT IS THE CORRECT DATA THAT WAS SENT.

That's moderately true.  (Misbehaving software could send you random
data to your UDP port and an operating system bug could send a corrupted
packet with no checksum, but, barring these problems, yes, that is
effectively true.)

|> I was also told that when using TCP (STREAM), it makes sure that the data 
|> arrives in the right order, and if it doesn't arrive, you will be notified,
|> etc...  but IT DOES ***NOT*** MEAN THAT THE DATA THAT WAS SENT IS THE DATA
|> YOU GET!  

That's not quite right.  TCP guarantees reliable in-order unduplicated
byte streams.  It doesn't guarantee that what you send in one operation
on one system (say by calling write(2)) shows up in a single operation
on the peer (say by calling read(2)).  Thus, you may find that several
write(2) calls result in one block of data seen by read(2) on the other
side or that one write(2) results in several read(2) calls on the other
side, or both.

It *does* mean that all bytes written will (unless the connection is
abended) arrive in the same order on the other side.  If the connection
is abended (due, say, to a crash on either side), it's not possible to
find out from TCP whether any particular bytes were actually delivered.
If you need that, you'll have to write your own protocol on top of TCP.

|> I am working over frame-relay lines, and because it is possible to lose
|> packets during overloads, I would like to use TCP (STREAM) so it will 
|> fix this undesirable feature.  But!  it is more important to me to have
|> the data I send be correct when it gets there... 

Yes, TCP will do this.

|> This idea that TCP doesn't make sure the data is the same seems a little
|> strange, what is is timely data if it isn't the data you requested!!!  haha

As an important note, you should know that TCP does *NOT* guarantee any
sort of timing at all.  It does *NOT* handle isochronous data and should
not be used in situations that require it.

|> I would really appreciate clarification.  Another person said he thought that
|> TCP had all of the good features of UDP, plus it made sure the data arrived
|> in order, etc...

TCP is missing the "packet oriented" nature of UDP and the ability to
handle broadcasts.  But, yes, it is in some ways "one better" than UDP.

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

-----------[000196][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 21:36:44 GMT
From:      [email protected] (Rick Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: Are TTCP and NetPerf Available for Windows NT?
Vernon Schryver ([email protected]) wrote:
: In article <[email protected]> [email protected] (Sam Ghandchi) writes:
: >Are TTCP and NetPerf Available for Windows NT?
: You generally get source for those benchmarks, and since Windows NT is
: said to offer a socket interface, porting should be possible.  It should
: be quite easy to port ttcp.c, because it is such a trivial program.

Netperf would be a triffle more involved than ttcp. Netperf makes use
of signals and fork (netserver uses fork when running as a standalone
daemon, signals are used in conjunction with alarm()). Apart from
that, it should not be terribly dificult to tweak the sockets calls to
winsock calls.

If someone would like to make a port of netperf to NT, the latest
source can be found at ftp.cup.hp.com under
dist/networking/benchmarks/exp/ .

rick jones

-----------[000197][next][prev][last][first]----------------------------------------------------
Date:      Fri, 9 Sep 1994 22:10:09 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (Donald Edwards) writes:
>Dennis Sutherland ([email protected]) wrote:
>: If this were a class C address the above statement would be true, because you
>: can't have subnets with all 0's or all 1's.  However, since this is a class B
>: address the 3rd byte is also part of the subnet address.  Thus the ranges of
>: 1 - 62, and 193 - 254 are also available.
>
>I don't know, but I suspect that the software is only aware of octets.
>The division between class B and class C is completely arbitrary.  If
>my suspicion is correct, then those two ranges aren't available.

Just to set the record straight, the NetWare TCP/IP implementation
has never had any such problem. In a subnetted class B address, the
subnet bits in the last byte can be anything at all so long as the
third byte of the address is neither 0 nor ff hex.

						don provan
						[email protected]

-----------[000198][next][prev][last][first]----------------------------------------------------
Date:      9 Sep 1994 22:59:14 GMT
From:      [email protected] (Ran Atkinson)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
In article <[email protected]> [email protected] (John Bottoms) writes:
>
>Check out MIL-STD 1777 and MIL-STD 1778.
>

I got a postal mail note today indicating that the gears are now turning
to formally and officially terminate 1777 and 1778.  They have known
errors in the specifications and should not be used.  The replacement
MIL-STDs have different numbers and consist mainly of pointers to the
RFCs.  Smart folks look first at the relevant RFCs, including Host
Requirements, and then at the BSD networking code.

Ran
[email protected]

employed by, but not speaking officially for the
	Naval Research Laboratory



-----------[000199][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 02:01:52 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Making secondary DNS sites work
I went from 1019 to 1020 to 1021....etc. Doesn't that work correctly?

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000200][next][prev][last][first]----------------------------------------------------
Date:      Sat, 10 Sep 1994 04:02:03 GMT
From:      [email protected] (Joseph Konarkowsk)
To:        comp.protocols.tcp-ip
Subject:   tcp shareware pd prog needed
Joe




-----------[000201][next][prev][last][first]----------------------------------------------------
Date:      Sat, 10 Sep 1994 08:58:59 +0000
From:      [email protected] (Phil Royse)
To:        comp.protocols.tcp-ip
Subject:   Re: Can this configuration be put on the Internet?
In article <[email protected]>
           [email protected] "Alan Cox" writes:

>In article <[email protected]> [email protected] (Matt
> Midboe) writes:
>>Novell file server is an arcnet with a bunch of PCs. Now my question
>>is how can I put all of these on the Internet. Do I need one class C
>>for the ethernet, and one class C for the arcnet? How does the routing
>>work from the arcnet? 
>
>Arcnet works the same as anything else. If you have >126 arcnet hosts or
>>126 ethernet hosts you want two networks. If you don't and are not likely
>to have that you can get a class C network and subnet it on 7 bits (ie
>
>x.y.z.1-126 are the arcnet.
>x.y.z.129-254 are the ethernet.

I believe this is not possible (or not advisable) because by subnetting
on the first bit of the last octet, you end up with only two subnets,
"0" and "1", both of which are illegal, for different reasons.

However, it's ok to subnet down to two bits in the last octet:

subnet 00 = illegal
subnet 01 = legal
subnet 10 = legal
subnet 11 = illegal

Therefore you can have two subnets (10, 01) with 62 hosts on each.

This is still likely to be OK for Matt Midboe's setup.

Also by variable subnet masking it's possible to recover some
of the lost address space.  For example, even though all addresses
beginning with 00 and 11 in the last octet might seem unusable,
it's perfectly OK to use 001xxxxx and 110xxxxx to give two more
usable subnets with 5-bit host addresses (30).  (etc.)

Phil
-- 

Phil Royse     Comms Consultant  |  PRA Consulting Ltd.
TUDOR HOUSE                      |
12 Woodside Road, Purley
Surrey  CR8 4LN (UK)  Tel: (+44) 81 645-9868   [email protected]

-----------[000202][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 16:09:07 -0400
From:      [email protected] (David Barr)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Registering reverse lookups (in-addr.arpa)
In article <[email protected]>,
Glenn Fleishman <[email protected]> wrote:
>I submitted the template correctly filled out to InterNIC. They said,
>"Sprintlink handles the set of Class C addresses in which you fall"
>(204.94.44 is my address) "and so we've forwarded this information to
>them." Unfortunately, no contact info was provided to me.

bosnia:~>whois 204.94.44
Object Software Development (NETBLK-SPRINT-CC5E2F) SPRINT-CC5E2F
                                                     204.94.40.0 - 204.94.47.0
Sprint (NETBLK-SPRINTLINK-BLKB) SPRINTLINK-BLKB      204.94.0.0 - 204.97.255.0

The InterNIC Registration Services Host ONLY contains Internet Information
(Networks, ASN's, Domains, and POC's).
Please use the whois server at nic.ddn.mil for MILNET Information.
bosnia:~>whois NETBLK-SPRINT-CC5E2F
Object Software Development (NETBLK-SPRINT-CC5E2F)
   701 5th Ave.
   Seattle, WA 98104
   US

   Netname: SPRINT-CC5E2F
   Netblock: 204.94.40.0 - 204.94.47.0

   Coordinator:
      Klimenko, Igor  (IK5)  [email protected]
      (206) 343-7828

   Record last updated on 25-Aug-94.

The InterNIC Registration Services Host ONLY contains Internet Information
(Networks, ASN's, Domains, and POC's).
Please use the whois server at nic.ddn.mil for MILNET Information.
bosnia:~>

--Dave

-----------[000203][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 10:17:15 GMT
From:      [email protected] (Stephane Bortzmeyer)
To:        comp.protocols.tcp-ip
Subject:   Re: Know of a ftpable IPng spec?
In article <[email protected]>, [email protected] writes:

>I am looking for the IPng spec.... anyone know where I could find it?

Those who read French can have a look at the Web page:

http://www.cnam.fr/Network/IPng/

Those who don't should remember that IPng is far from being finished and 
that all documents are drafts. To understand IPng would require a lot
of digging in various archives since know one, to my knowledge, tried
to draw a synthesis yet.

IMHO (not a specialist or an IETF member, just a plain network 
administrator), you can begin with the documents on SIPP, IPng being 
based on it. My personal choice to start with:

- SIPP 16 bytes: 

ftp://ftp.doc.ic.ac.uk/computing/internet/internet-drafts/draft-ietf-sipp-spec-1.txt.Z

- SIPP addressing and routing: 

ftp://ftp.doc.ic.ac.uk/computing/internet/internet-drafts/draft-ietf-sipp-routing-addr-02.txt.Z

- transition from the present IPv4:

ftp://ftp.doc.ic.ac.uk/computing/internet/internet-drafts/draft-ietf-sipp-sst-overview-00.txt.Z

There is also a (not really comprehensive) page on the IETF Web server: 

http://www.ietf.cnri.reston.va.us/ipng/ipng.html

Stephane Bortzmeyer           Conservatoire National des Arts et Metiers	
[email protected]            Laboratoire d'Informatique
                              292, rue Saint-Martin			
tel: +33 (1) 40 27 27 31      75141 Paris Cedex 03
fax: +33 (1) 40 27 27 72      France	

"C'est la nuit qu'il est beau de croire a la lumiere." E. Rostand

http://web.cnam.fr/personnes/bortzmeyer/home_page.dom

					
	

-----------[000204][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 12:59:49 GMT
From:      [email protected] (Smirnov Dmitry)
To:        comp.protocols.tcp-ip
Subject:   OOB data.
Hello All!
In my program I want to use out-of-band data. But in some 
systems only one byte can be transfered. And second 
out-of-band packet overwrite first. Ugh .. My English
don't allow me to show all problem. May be anybody
have a detail description about using this mechanism.

--
           *******System programmer and administrator*******
              ******** Smirnov Dmitry Ivanovich *********
                  ********* BCEX K CTEHKE **********
                        ******Just a Unix******

-----------[000205][next][prev][last][first]----------------------------------------------------
Date:      Sat, 10 Sep 1994 17:50:19 GMT
From:      [email protected] (Don Dunstan)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
Tony Trehan ([email protected]) wrote:
: I need to provide TCP/IP (FTP, TELNET etc) in embedded
: industrial controllers running under real-time operating systems.....

Check out U S Software.  Information is available by anonymous
ftp to ftp.netcom.com (pub/ussw).

-----------[000206][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 18:23:37 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Making secondary DNS sites work
For some reason, it all works fine since I did a manual named-xfer.

Now, when I update the serial numbers in the db.* files, they're
automatically updated to the secondary.

Something must have gone out of whack that doing the named-xfer
repaired.

Thanks for everyone's advice and assistance.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000207][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 19:20:10 GMT
From:      [email protected] (Mike Redford)
To:        comp.protocols.tcp-ip
Subject:   Re: Any vendors of TCP (without IP)?
YOUR CONTRIBUTIONS ARE HIGHLY APPRECIATED.
PLEASE MAIL REPLY DIRECTLY TO ME.

City of Philadelphia  is think of Network
system that will support 70 clients. We need
to pick a server. I will like UNIWARE?
1. What type of server do I need?

2. Is the number of user dependent on the capacity of the server?

3. How do I choose a HUB?

4. With Uniware do I need network cards or modem for remote site?

5. Is it necessary to have 9600 buad rate or will 2400 baud rate do?

6.  Using the modem the remote sites, can dail in to connect to the server.

7.  To connect print to the 70 sites, the printers will 
    be controled by the server

    MIKE REDFORD



-----------[000208][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 19:44:15 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Registering reverse lookups (in-addr.arpa)
I submitted the template correctly filled out to InterNIC. They said,
"Sprintlink handles the set of Class C addresses in which you fall"
(204.94.44 is my address) "and so we've forwarded this information to
them." Unfortunately, no contact info was provided to me.

It's been about two weeks. Do I need to follow-up on this or is
Sprintlink so backlogged, it might take longer?

The reason it's annoying is clear to all: can't get to ftp.uu.net and
some other services without having the reverse nameservers registered.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000209][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 21:06:45 GMT
From:      [email protected] (Russo and Hale)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.protocols.ppp
Subject:   [Q] Low end router?
[ Article crossposted from comp.sys.ibm.pc.hardware.networking ]
[ Author was Russo and Hale ]
[ Posted on 10 Sep 1994 03:41:53 GMT ]

Does anyone know of a good router that is on the low cost end?  I want to 
connect a Class C ethernet (winsock) to a PPP dialup provider (Portal 
Communications).

Let me know if there is a FAQ or another newsgroup, and I will look there 
as well

Thanks in advance.

-----------[000210][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 23:29:32 GMT
From:      [email protected] (Ruediger Volk)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Registering reverse lookups (in-addr.arpa)
In article <[email protected]>, [email protected] (Glenn Fleishman) writes:
  > I submitted the template correctly filled out to InterNIC. They said,
  > "Sprintlink handles the set of Class C addresses in which you fall"
  > (204.94.44 is my address) "and so we've forwarded this information to
  > them." Unfortunately, no contact info was provided to me.
well, I think Dave Barr's response seemed to point at someone
at  Object Software Development  who is listed as contact for
the 8*C number block including your number.  This is very unlikely
to be person who can actually update the DNS for you - though you
probably have good reason to channel your request through him.

To find a mail address for the person(s) who should finally fo the registration
you can use the DNS: those persons are the zone contacts of the parent zone
who need to delegate your reverse mapping domain.
You should find their e-mail address in the SOA of the appropriate
parent zone;  the first to query in your case is 94.204.in-addr.arpa
(if that returns a "does not exist" or empty response
you would have to regress to 204.in-addr.arpa, etc)

(liking the host utility and being lazy to spell out the reverse domain name
I check such questions this way:)

[email protected]<271> host -Ci 204.94
94.204.in-addr.arpa     NS      ICM1.ICP.NET
icm1.icp.net    dns-admin.icm1.icp.net  (94090701 43200 3600 2149200 2149200)
94.204.in-addr.arpa     NS      NS1.SPRINTLINK.NET
icm1.icp.net    dns-admin.icm1.icp.net  (94090701 43200 3600 2149200 2149200)
94.204.in-addr.arpa     NS      NS2.SPRINTLINK.NET
Nameserver NS2.SPRINTLINK.NET not responding
94.204.in-addr.arpa SOA record not found, try again
94.204.in-addr.arpa     NS      NS3.SPRINTLINK.NET
Nameserver NS3.SPRINTLINK.NET not responding
94.204.in-addr.arpa SOA record not found, try again


So you might want to send a mail to:

To: "zone contact according to SOA 94.204.in-addr.arpa" <[email protected]>


You also should consider asking the party that actually assigned the
network number to you;  I guess that's Object Software Development.

--------
Ruediger Volk
Universitaet Dortmund, Informatik IRB
D-44221 Dortmund, Germany

E-Mail: [email protected]
Phone:  +49 231 755 4760                 Fax:  +49 231 755 2386

-----------[000211][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 23:44:24 GMT
From:      [email protected] (Ruediger Volk)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]>, [email protected] (don provan) writes:
  > Just to set the record straight, the NetWare TCP/IP implementation
  > has never had any such problem. In a subnetted class B address, the
  > subnet bits in the last byte can be anything at all so long as the
  > third byte of the address is neither 0 nor ff hex.
really?  do you impose any restrictions on the subnet masks?
Any mask from  18:14 to 30:2 bits allowed (I'm already restricting to
masks that conform to both RFC 950 *and* CIDR)?

Why would you want to forbid e.g. subnet numbers xx.xx.0.64/128/192
and xx.xx.255.64/128/192 under mask 255.255.255.192?

--------
Ruediger Volk
Universitaet Dortmund, Informatik IRB
D-44221 Dortmund, Germany

E-Mail: [email protected]
Phone:  +49 231 755 4760                 Fax:  +49 231 755 2386

-----------[000212][next][prev][last][first]----------------------------------------------------
Date:      10 Sep 1994 23:56:22 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip.domains,comp.protocols.tcp-ip
Subject:   Re: Registering reverse lookups (in-addr.arpa)
Thanks for advice. The folks at Objective Software Development aren't
responsible for the top-level information for reverse-nameserving.
Thanks for info on how to find this out in the future!

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000213][next][prev][last][first]----------------------------------------------------
Date:      Sun, 11 Sep 1994 06:39:29 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: OOB data.
In article <[email protected]> [email protected] (Smirnov Dmitry) writes:
>In my program I want to use out-of-band data. But in some 
>systems only one byte can be transfered. And second 
>out-of-band packet overwrite first. Ugh .. My English
>don't allow me to show all problem. May be anybody
>have a detail description about using this mechanism.

TCP doesn't really have out-of-band data.  It has a single "urgent
pointer", which tells the receiving system that all of the data up to a
particular byte should be expedited and the application should be notified.
The socket implementors chose to interpret this as meaning that the last
urgent byte should be delivered out of bound, since this matches how urgent
data is used in some protocols.

There's no pointer to the beginning of urgent data.  And since there's just
one urgent pointer, you can't have multiple out-of-band data.

The best way to get real out-of-band data is to use two connections, one
for the regular data and the other for the out-of-band data.
-- 
Barry Margolin                                                [email protected]


-----------[000214][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 1994 13:25:21 -0400
From:      [email protected] (Susan Ferebee Guion)
To:        comp.protocols.tcp-ip
Subject:   TCP-IP Packages for PC's (Again)
I'm once again looking for the best in tcp-ip packages for my PC's.

I'm had it woth FTP's PCTCP and the ~70K of low memory it requires.  Now
I'm looking for a replacement.  I'll need all the standards (ftp,telnet,
ping,...) but I need to preserve low memory.  I might even decide to go
with an X-terminal package if the proce war right.

I'm interested in experiences you might have had with any of them.
thanks in advance...

-- 
Susan Guion - Georgia Tech
Internet: [email protected]
uucp:	  ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!gt5582a

-----------[000215][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 02:17:34 -0700
From:      [email protected] (Samuel Lam)
To:        comp.protocols.tcp-ip
Subject:   Exotic ARP Protocol Questions
Two pretty odd ARP questions:

 1) Is sending outgoing ARP replies via MAC-level broadcasts
    instead of unicasts allowed?

 2) Is discarding incoming ARP replies arriving via MAC-level
    broadcasts allowed?

You guessed it!  We have two vendors' routers refusing to talk
to each other because they each implements one of the above
"features".

Pointers to references which I could cite to these vendors
would be great.

Thanks in advance.

...Sam
-- 
<[email protected]> -- Connectivity Technology Inc.

-----------[000216][next][prev][last][first]----------------------------------------------------
Date:      Sun, 11 Sep 1994 15:59:58 GMT
From:      [email protected] (Steven Weller)
To:        comp.protocols.tcp-ip
Subject:   Re: Embedded Systems TCP/IP
In article <[email protected]>, [email protected] (Don Dunstan) wrote:

> Tony Trehan ([email protected]) wrote:
> : I need to provide TCP/IP (FTP, TELNET etc) in embedded
> : industrial controllers running under real-time operating systems.....
> 
> Check out U S Software.  Information is available by anonymous
> ftp to ftp.netcom.com (pub/ussw).

Try Microware (OS-9 vendors) at 515 224 1929.

--
Steven Weller at Windsor Systems                 +1 502 425 9560 (voice)
2407 Lime Kiln Court, Louisville, KY 40222, USA  +1 502 426 3944 (fax)
<OS-9 Consultancy and Software>    [email protected] or [email protected]
Windsor Home Page: http://iglou.com/~stevenw/windsorhome.html

-----------[000217][next][prev][last][first]----------------------------------------------------
Date:      Sun, 11 Sep 1994 21:01:01
From:      [email protected] (Barry Trent)
To:        comp.dcom.lans.ethernet,comp.protocols.snmp,comp.protocols.tcp-ip
Subject:   Re: rmon(-alike) software for dos machine
In article <[email protected]> [email protected] (Wim.Holemans) writes:
>From: [email protected] (Wim.Holemans)
>Subject: rmon(-alike) software for dos machine
>Date: Wed, 7 Sep 1994 16:27:33 GMT
 
>Hi,
 
>i'm looking for software that works like a network monitor and enables me
>to get the information by snmp requests. I'm only interested in the following
>parameters :
>- number of packets/bytes seen on the net
>- number of errors seen on the net, if possible with a subdivision for types
>  of errors
 
>the info should be available for snmp requests and the software should run on
>a 386-DX with no harddisk and a SMC ethernet card.
 
>I got part of this functionality with the fergie(beholder) software, which
>offers me packets/bytes counts, but no error info.

My firm recently introduced a DOS-based RMON agent, called RMONster.  It 
doesn't do the "frame capture" group of RMON, but it does all the 
statistics groups (history, host/matrix, hostTopN, Alarm/Event, etc)  This 
appears to be what you are asking for.  Operation from a floppy and SMC 
ethernet support should not be a problem...

BTW, this is a commercial product, not share/freeware -- a license for the 
Ethernet software version is $595 US retail.

Be glad to snail mail or fax more info.
+-----------------------------------------------------------+
| Barry A. Trent            Of course I'm the center of     |
| Triticom                  the universe -- aren't you?     |
| Box 444180                                                |
| Eden Prairie, MN 55344   (The usual disclaimers apply)    |
| 612-937-0772                                              |
+-----------------------------------------------------------+

-----------[000218][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 04:42:06 -0400
From:      [email protected] (Chien Kwok)
To:        comp.protocols.tcp-ip
Subject:   Protocol comparisons
Hi all,

I hope this is not a stupid question... are there any FAQs out there that
compares the various commonly used protocols like TCP/IP, IPX/SPX, Appletalk,
X.25 etc. Suggestions for books also welcomed.

thanks in advance.

chien


-----------[000219][next][prev][last][first]----------------------------------------------------
Date:      Sun, 11 Sep 1994 22:07:48 GMT
From:      [email protected] (John Levine)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Can I have multiple distinguishable addresses for a single host?
I have three domains, call them a.com, b.com, and c.com.  But I only have one 
(maybe two, if I scrounge) server machine.  Is there some way that I could 
have three names with three IP addresses for the single machine, so I can 
distinguish FTP and, particularly, WWW connections for the three domains?  
That is, I'd like http://a.com/ and http://b.com/ to show different pages, 
even though they're served from the same host.

I'll be running BSDI's BSD/360 with full source, and I have plenty of free 
addresses in my subnet.  If it would make the software easier, I could plug 
in three separate Ethernet cards, since cards are cheap and I have enough 
slots.  Any wisdom would be appreciated.


Regards,
John Levine, [email protected]
Primary perpetrator of "The Internet for Dummies"


-----------[000220][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 1994 23:17:36 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single, host?
John, 

I'm sorry they made you name your book that.

Here's what we've done (someone can tell us if this is kosher).

We have a site called popco.com. We've registered several domain names
and created separate db.* files for them under DNS/BIND.

So we have a named.boot file that says

popco.com     204.94.44.1     db.popco
zap.com       204.94.44.1     db.zap
etc.

The db.* files contain the necessary information for these domains. The
sendmail files contain a "Cm" item that contains all of the names of
the other domains for which mail is received at popco.com.

To have things like www.popco.com, gopher.popco.com, we currently have
entries in our db.popco file:

gopher    IN CNAME  204.94.44.4
www       IN CNAME  204.94.44.4
etc.

When we bring a separate WWW machine online some day, we'll change that
record to read

www   IN A 204.94.44.178
(or whatever)
We'll also modify the TTL values during the week or so before we add
those records so that we can make sure no one is caching old values at
other sites for too long before the next flush is done.
----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000221][next][prev][last][first]----------------------------------------------------
Date:      11 Sep 1994 23:19:24 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single, host?
I realized I didn't answer part of your question:

What you're asking is: can I have three IP addresses in order to
distinguish what domain is being requested for service?

You can use the CERN http daemon to determine what the incoming request
is (so if someone does http://dummy.com/etc. vs
http://smartguy.com/etc., you can determine what they've entered and
route it to different directories).

This doesn't help you with FTP or Gopher, I don't think, unless there's
a great FTP daemon out there that can do the same kind of thing.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000222][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 00:17:54 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   online ATM sources?

Are there any on-line sources detailing asynchronous transfer mode?

daveriesz
[email protected]


-----------[000223][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 00:37:04 GMT
From:      [email protected] (William C. Fenner)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single, host?
In article <[email protected]>,
Glenn Fleishman <[email protected]> wrote:
>You can use the CERN http daemon to determine what the incoming request
>is (so if someone does http://dummy.com/etc. vs
>http://smartguy.com/etc., you can determine what they've entered and
>route it to different directories).

How?  The HTTP/1.0 protocol only passes the stuff after the hostname
to the HTTP daemon.  Is the CERN http daemon psychic?

  Bill
-- 
Bill Fenner                  [email protected]

-----------[000224][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 00:49:21 GMT
From:      [email protected] (Mark Andrews)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single  host?
In article <[email protected]> [email protected] (Glenn Fleishman) writes:
>John, 
>
>I'm sorry they made you name your book that.
>
>Here's what we've done (someone can tell us if this is kosher).
>
>We have a site called popco.com. We've registered several domain names
>and created separate db.* files for them under DNS/BIND.
>
>So we have a named.boot file that says
>
>popco.com     204.94.44.1     db.popco
>zap.com       204.94.44.1     db.zap
>etc.

	If your named.boot has lines identical to them IT WILL NOT
	WORK.

	They should be

	secondary popco.com     204.94.44.1     db.popco
	secondary zap.com       204.94.44.1     db.zap

	or

	primary popco.com db.popco
	primary zap.com db.zap


>
>The db.* files contain the necessary information for these domains. The
>sendmail files contain a "Cm" item that contains all of the names of
>the other domains for which mail is received at popco.com.
>
>To have things like www.popco.com, gopher.popco.com, we currently have
>entries in our db.popco file:
>

	CNAME's point to domains not addresses.

	try 

	gopher    IN CNAME  maigret.popco.com.
	www       IN CNAME  maigret.popco.com.

>gopher    IN CNAME  204.94.44.4
>www       IN CNAME  204.94.44.4
>etc.
>
 
>When we bring a separate WWW machine online some day, we'll change that
>record to read
>
>www   IN A 204.94.44.178
>(or whatever)
>We'll also modify the TTL values during the week or so before we add
>those records so that we can make sure no one is caching old values at
>other sites for too long before the next flush is done.
>----
>Glenn Fleishman, Point of Presence Company, Seattle, Washington
>        Send to <[email protected]> for information on our services
>        Moderator, Internet Marketing discussion list; send the message
>        INFO INET-MARKETING to the address [email protected] for info

	Mark

-----------[000225][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 01:52:01 GMT
From:      [email protected] (Vadim Antonov)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
John Levine ([email protected]) wrote:

: I'll be running BSDI's BSD/360 with full source...

Pray tell me where you got one!  I happened to have an ancient
IBM/360 and it would be way COOL to run DNS on it.

:-)

--vadim

-----------[000226][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 02:19:21 GMT
From:      [email protected] (S.T. Wong)
To:        comp.sys.hp.hpux,comp.protocols.tcp-ip
Subject:   Re: Help: How to close FIN_WAIT_2 connections ?
Hello,

Thanks to all who responded to the question:

[email protected] (Bruce Barnett)
[email protected] (Brian Hackley)
[email protected] (Scott Simpson)

We've tried to set the socket options NOLINGER and KEEPALIVE but doesn't work.
The script from Mr. Bruce Barnett works in this way (extracted from the in-line
comment of the scripts) :

"# Works by loading the 2 * msl timer (addr+16) in the Connection Control Block
# with a 1, which means it will time out in 1/2 second from now, and
# enter the CLOSE state, and the the CCB will be freed (so you will not
# see it with netstat -a)."

Have talked to local HP support staff, they told me that there will be 'per-
connection' time-out control in HP-UX 10.0.  However, no more information 
is available now.

Thanks again.

Regards,
ST Wong

=============================================================================
Bruce Barnett:

this is a bug in the operating system.
10 years ago I saw a program to fix this..

Don't have it anymore.

I sent you some programs that might help.

#! /bin/sh
# This is a shell archive.  Remove anything before this line, then unpack
# it by saving it into a file and typing "sh file".  To overwrite existing
# files, type "sh file -c".  You can also feed this as standard input via
# unshar, or by typing "sh <file", e.g..  If this archive is complete, you
# will see the following message at the end:
#		"End of shell archive."
# Contents:  3076.umcp-cs 914.rlgvax fixtcp fixtcp.mk get_tcp.c
# Wrapped by [email protected] on Tue Sep  6 09:10:18 1994
PATH=/bin:/usr/bin:/usr/ucb ; export PATH
if test -f 3076.umcp-cs -a "${1}" != "-c" ; then 
  echo shar: Will not over-write existing file \"3076.umcp-cs\"
else
echo shar: Extracting \"3076.umcp-cs\" \(1824 characters\)
sed "s/^X//" >3076.umcp-cs <<'END_OF_3076.umcp-cs'
XFrom: [email protected] (Steve D. Miller)
XNewsgroups: net.sources
XSubject: Re: Unhang TCP connections stuck in FIN_WAIT_2 state
XMessage-ID: <[email protected]>
XDate: Thu, 6-Feb-86 09:28:24 EST
X
X
X   Here's a better fix for the 4.2 FIN_WAIT_2 problem.  I don't remember
Xwhere I got it, but it works; the basic problem is that the code to
Xdrop the connection when nothing cares about it is there in vanilla
X4.2BSD, but is in the wrong place.  This fix moves it to the right
Xspot and mungs a conditional a little bit...
X
X   The fix is to netinet/tcp_input.c; I think this is for vanilla
X4.2, but your line numbers may vary.
X
X
X*** Vanilla (??) 4.2 tcp_input.c	Fri Jan 24 12:24:00 1986
X--- Fixed 4.2 tcp_input.c	Fri Jan 24 12:24:03 1986
X***************
X*** 358,363 ****
X--- 358,372 ----
X  			goto dropafterack;
X  		if (ti->ti_len > 0) {
X  			m_adj(m, ti->ti_len);
X+   			/*
X+  			 * If data is received on a connection after the
X+  			 * user processes are gone, then RST the other end.
X+  			 */
X+  			if ((so->so_state & SS_NOFDREF) 
X+  			    && tp->t_state > TCPS_CLOSE_WAIT) {
X+  				tp = tcp_close(tp);
X+  				goto dropwithreset;
X+  			}
X  			ti->ti_len = 0;
X  			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
X  		}
X***************
X*** 404,419 ****
X  			ti->ti_len -= todrop;
X  			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
X  		}
X- 	}
X- 
X- 	/*
X- 	 * If data is received on a connection after the
X- 	 * user processes are gone, then RST the other end.
X- 	 */
X- 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
X- 	    ti->ti_len) {
X- 		tp = tcp_close(tp);
X- 		goto dropwithreset;
X  	}
X  
X  	/*
X--- 413,418 ----
X
X
X-- 
XSpoken: Steve Miller 	ARPA:	[email protected]	Phone: +1-301-454-4251
XCSNet:	[email protected] 	UUCP:	{seismo,allegra}!umcp-cs!steve
XUSPS: Computer Science Dept., University of Maryland, College Park, MD 20742
END_OF_3076.umcp-cs
if test 1824 -ne `wc -c <3076.umcp-cs`; then
    echo shar: \"3076.umcp-cs\" unpacked with wrong size!
fi
# end of overwriting check
fi
if test -f 914.rlgvax -a "${1}" != "-c" ; then 
  echo shar: Will not over-write existing file \"914.rlgvax\"
else
echo shar: Extracting \"914.rlgvax\" \(7670 characters\)
sed "s/^X//" >914.rlgvax <<'END_OF_914.rlgvax'
X#--------------- CUT HERE ---------------
X#! /bin/sh
X# This is a shell archive, meaning:
X# 1. Remove everything above the #! /bin/sh line.
X# 2. Save the resulting text in a file.
X# 3. Execute the file with /bin/sh (not csh) to create the files:
X#	_get_tcp_.c
X#	fixtcp
X#	fixtcp.mk
X# This archive created: Wed Feb  5 15:16:26 EST 1986
X#
Xif test -f _get_tcp_.c
Xthen
Xecho shar: will not over-write existing file '_get_tcp_.c'
Xelse
Xecho x - _get_tcp_.c
X# ............    F  I   L   E      B  E  G  .......... _get_tcp_.c
Xcat << '\SHAR_EOF' > _get_tcp_.c
X/*
X * [email protected]
X * prints offsets of fields in TCP connection control block.
X * called by fixtcp sh script
X */
X#include <stdio.h>
X#include <sys/types.h>		/* u_char */
X#include <netinet/tcp.h>	/* tcp_seq typedef */
X#include <netinet/tcp_timer.h>	/* tcp timers */
X#include <netinet/tcp_var.h>	/* tcp connection control block */
X#include <netinet/tcp_fsm.h>	/* defines for tcp states */
X
X/* use S3/S5 strrchr(), but on 4.x systems, remap to Berkeley rindex */
X#ifdef BSD4
X#	define strrchr	rindex
X#endif
X
X#define STR_SAME !strcmp
X#define STR_DIFF strcmp
X
X/* fw non-int functions */
Xchar *basename();
X
X/* external non-int functions */
Xextern	char	*strrchr();
X
Xmain(argc, argv)
X	int	argc;
X	char	**argv;
X{
X	char	*cmd;
X	struct	tcpcb	*p = 0;
X
X	cmd = basename(argv[0]);
X
X	if (argc != 2)
X		{
Xusage:
X		fprintf(stderr, "usage: %s state|2msl|FIN_WAIT2|TIME_CLOSE\n", cmd);
X		exit(1);
X		}
X
X	if (STR_SAME(argv[1], "state"))
X		printf("0x%x\n", &p->t_state);	/* state offset */
X	else if (STR_SAME(argv[1], "2msl"))
X		printf("0x%x\n", &p->t_timer[TCPT_2MSL]);	/* timer offset */
X	else if (STR_SAME(argv[1], "FIN_WAIT2"))
X		printf("0x%x\n", TCPS_FIN_WAIT_2);	/* state value */
X	else if (STR_SAME(argv[1], "TIME_CLOSE"))
X		printf("0x%x\n", TCPS_TIME_WAIT);	/* state value */
X	else
X		goto usage;
X}
X
X/*
X * return basename of full path name
X */
Xchar *
Xbasename(path)
X	char	*path;
X{
X	char	*cp;		/* general char pointer */
X
X	if ((cp = strrchr(path, '/')) == NULL)	/* no rightmost slash */
X		return path;
X	else
X		return cp;
X}
X\SHAR_EOF
X# ............    F  I   L   E      E  N  D  .......... _get_tcp_.c
Xfi # end of overwriting check
Xif test -f fixtcp
Xthen
Xecho shar: will not over-write existing file 'fixtcp'
Xelse
Xecho x - fixtcp
X# ............    F  I   L   E      B  E  G  .......... fixtcp
Xcat << '\SHAR_EOF' > fixtcp
X# fixtcp
X# dennis bednar jan 24 86	[email protected]
X#
X# Unhang tcp connections which are stuck in the FIN_WAIT2 state
X# These connections can be seen by doing a 4.2 netstat -a command.
X#
X# Usage:
X# invoke as "fixtcp" to display kernel stuff for connections.
X# "fixtcp" by itself is HIGHLY RECOMMENDED for the first time!
X#
X# invoke as "fixtcp fix" to patch kernel memory - you must be root.
X# Then do a netstat -a command, and it should have gone away.
X#
X# CCI only symptom:
X# A symptom of this problem is that "startoftp" goes wild restarting
X# the receive daemon, and you see a lot of rcvlog.pid files being
X# created in the oftp spool directory.
X#
X# Symptom for everybody else:
X# In general, a symptom of this problem is that a tcpopen passive
X# will fail with the errno UNIX reason being "Address Already In Use".
X#
X#
X# To correct OFTP problem (CCI only):
X# su root
X# killoftp; fixtcp fix; startoftp
X#
X# Internals of how this script works:
X# Works by loading the 2 * msl timer (addr+16) in the Connection Control Block
X# with a 1, which means it will time out in 1/2 second from now, and
X# enter the CLOSE state, and the the CCB will be freed (so you will not
X# see it with netstat -a).
X# The proper offset for the 2 * msl timer can be seen by examining
X# /usr/include/netinet/tcp_var.h include file, plus other tcp*.h files
X# in the same directory.
X#
X# relies on
X#	_get_tcp_	a.out file that returns the offset of various
X#			fields in a connecton control block.
X#			There is a _get_tcp_.c file to create this.
X#			This was created to avoid problems of offsets
X#			being site-dependent, if your OS uses different
X#			offsets.
X#
X#
X
X# don't print full path name of command in error messages
Xcmd=`basename $0`
X
X# name of state to look for in the netstat command
X# state=ESTABLISHED	# debugging
Xstate=FIN_WAIT_2	# really
X
X# get the values of the offsets of the fields the the structure for adb
Xstateoff=`_get_tcp_ state`	# probably 0x8
Xtimer2msloff=`_get_tcp_ 2msl`	# probably 0x10
XFIN_WAIT2=`_get_tcp_ FIN_WAIT2`	# probably 9
XFIN_CLOSE=`_get_tcp_ TIME_CLOSE`	# probably 10
X
X
X
X# remove temp file if SIGHUP, SIGINT, SIGTERM
Xtrap "echo $cmd: interrupted; rm /tmp/fixtcp.$$; exit 1" 1 2 15
X
X# get kernel address of TCP CCB's in FIN_WAIT2 and save in a temporary file
Xnetstat -A | grep $state | sed '1,$s/ .*//p' >/tmp/fixtcp.$$
X
X# check if we got any addresses
Xif [ ! -s /tmp/fixtcp.$$ ]
Xthen
X#	file doesn't exist or is zero in length, therefore no addresses
X	echo "$cmd: Sorry, no tcp connections stuck in $state state."
X	rm /tmp/fixtcp.$$
X	exit 0
Xfi
X
Xecho "Before: only connections in state $state"
Xnetstat -a | grep $state
X
X
X# cat /tmp/fixtcp.$$	# debug
X
X# see if we want to patch kernel memory or just display it
Xif [ "$1" = "fix" ]
Xthen
X#	patch by writing
X	for addr in `cat /tmp/fixtcp.$$`
X	do
Xadb -w /vmunix /dev/kmem <<EOF
X0x$addr+$timer2msloff/w 1
X\$q
XEOF
X	done
X
X	sleep 2			# wait for connection to clear
X
X#	make sure it really got unstuck
X	netstat -a | grep $state >/tmp/fixtcp.$$
X	if [ -s /tmp/fixtcp.$$ ]	# file exists and size > 0
X	then
X		echo "$cmd: Sorry, TCP connections still hung!!"
X		rm /tmp/fixtcp.$$
X		exit 1
X	else
X		echo "$cmd: TCP connections in state $state have been unstuck."
X	fi
Xelse
X#	just display the current state flag and current 2 * msl timer
X	for addr in `cat /tmp/fixtcp.$$`
X	do
X		echo "The next two numbers displayed by adb should be $FIN_WAIT2 and 0."
X		echo "The state flag value of $FIN_WAIT2 represents the FIN_WAIT_2 state."
X		echo "The decimal 0 means the 2 * msl timer is off."
Xadb /vmunix /dev/kmem <<EOF
X0x$addr+8/d
X0x$addr+0x10/d
X\$q
XEOF
X	done
Xfi
X
X# cleanup intermediate file
Xrm /tmp/fixtcp.$$
X
Xecho "After: only connections in state $state"
Xnetstat -a | grep $state
Xexit 0
X\SHAR_EOF
X# ............    F  I   L   E      E  N  D  .......... fixtcp
Xfi # end of overwriting check
Xif test -f fixtcp.mk
Xthen
Xecho shar: will not over-write existing file 'fixtcp.mk'
Xelse
Xecho x - fixtcp.mk
X# ............    F  I   L   E      B  E  G  .......... fixtcp.mk
Xcat << '\SHAR_EOF' > fixtcp.mk
X#
X# [email protected] 2/4/86
X#
X# fixtcp.mk	Makefile, this file
X# fixtcp	shell script
X# _get_tcp_.c	C program
X# _get_tcp_	a.out program called by fixtcp
X# .fixtcp.mail	header for mail
X#
X# directions, type
X#	make -f fixtcp.mk	# to make necessary files
X#	edit fixtcp.mk and change INSTALLDIR
X#	make -f fixtcp.mk install
X#	cd $INSTALLDIR		# directory where you really installed it
X#	fixtcp			# to display tcp connections hung in finwait2
X#
X#				# don't do this if you have none to unstick
X#	su root			# required for adb write mode
X#	fixtcp fix		# to actually unstuck tcp connections
X
X# change this at your site
XINSTALLDIR = .
X
Xall: _get_tcp_
X
Xclean:
X	rm -f _get_tcp_
X
Xinstall: _get_tcp_
X	-cp _get_tcp_ $(INSTALLDIR)
X	-cp fixtcp $(INSTALLDIR)
X
X# distribute the latest version to the world, private for [email protected]
Xdist:
X	rm -rf /tmp/dpb
X	mkdir /tmp/dpb
X	cp fixtcp.mk /tmp/dpb
X	cp _get_tcp_.c /tmp/dpb
X	cp ../cmd/fixtcp /tmp/dpb
X	cp .fixtcp.mail	/tmp/dpb
X	(cd /tmp/dpb; makeshar * >>.fixtcp.mail)
X
X# please note that .fixtcp.mail was chosen so that makeshar *
X# doesn't try to append to itself.
X\SHAR_EOF
X# ............    F  I   L   E      E  N  D  .......... fixtcp.mk
Xfi # end of overwriting check
X# end of shell archive
Xexit 0
X-- 
X-Dennis Bednar
X{decvax,ihnp4,harpo,allegra}!seismo!rlgvax!dennis	UUCP
END_OF_914.rlgvax
if test 7670 -ne `wc -c <914.rlgvax`; then
    echo shar: \"914.rlgvax\" unpacked with wrong size!
fi
# end of overwriting check
fi
if test -f fixtcp -a "${1}" != "-c" ; then 
  echo shar: Will not over-write existing file \"fixtcp\"
else
echo shar: Extracting \"fixtcp\" \(3715 characters\)
sed "s/^X//" >fixtcp <<'END_OF_fixtcp'
X#!/bin/sh
X# fixtcp
X# dennis bednar jan 24 86	[email protected]
X#
X# Unhang tcp connections which are stuck in the FIN_WAIT2 state
X# These connections can be seen by doing a 4.2 netstat -a command.
X#
X# Usage:
X# invoke as "fixtcp" to display kernel stuff for connections.
X# "fixtcp" by itself is HIGHLY RECOMMENDED for the first time!
X#
X# invoke as "fixtcp fix" to patch kernel memory - you must be root.
X# Then do a netstat -a command, and it should have gone away.
X#
X# CCI only symptom:
X# A symptom of this problem is that "startoftp" goes wild restarting
X# the receive daemon, and you see a lot of rcvlog.pid files being
X# created in the oftp spool directory.
X#
X# Symptom for everybody else:
X# In general, a symptom of this problem is that a tcpopen passive
X# will fail with the errno UNIX reason being "Address Already In Use".
X#
X#
X# To correct OFTP problem (CCI only):
X# su root
X# killoftp; fixtcp fix; startoftp
X#
X# Internals of how this script works:
X# Works by loading the 2 * msl timer (addr+16) in the Connection Control Block
X# with a 1, which means it will time out in 1/2 second from now, and
X# enter the CLOSE state, and the the CCB will be freed (so you will not
X# see it with netstat -a).
X# The proper offset for the 2 * msl timer can be seen by examining
X# /usr/include/netinet/tcp_var.h include file, plus other tcp*.h files
X# in the same directory.
X#
X# relies on
X#	get_tcp	a.out file that returns the offset of various
X#			fields in a connecton control block.
X#			There is a get_tcp.c file to create this.
X#			This was created to avoid problems of offsets
X#			being site-dependent, if your OS uses different
X#			offsets.
X#
X#
X
X# don't print full path name of command in error messages
Xcmd=`basename $0`
X
X# name of state to look for in the netstat command
X# state=ESTABLISHED	# debugging
Xstate=FIN_WAIT_2	# really
X
X# get the values of the offsets of the fields the the structure for adb
Xstateoff=`get_tcp state`	# probably 0x8
Xtimer2msloff=`get_tcp 2msl`	# probably 0x10
XFIN_WAIT2=`get_tcp FIN_WAIT2`	# probably 9
XFIN_CLOSE=`get_tcp TIME_CLOSE`	# probably 10
X
X
X
X# remove temp file if SIGHUP, SIGINT, SIGTERM
Xtrap "echo $cmd: interrupted; rm /tmp/fixtcp.$$; exit 1" 1 2 15
X
X# get kernel address of TCP CCB's in FIN_WAIT2 and save in a temporary file
Xnetstat -A | grep $state | sed '1,$s/ .*//p' >/tmp/fixtcp.$$
X
X# check if we got any addresses
Xif [ ! -s /tmp/fixtcp.$$ ]
Xthen
X#	file doesn't exist or is zero in length, therefore no addresses
X	echo "$cmd: Sorry, no tcp connections stuck in $state state."
X	rm /tmp/fixtcp.$$
X	exit 0
Xfi
X
Xecho "Before: only connections in state $state"
Xnetstat -a | grep $state
X
X
X# cat /tmp/fixtcp.$$	# debug
X
X# see if we want to patch kernel memory or just display it
Xif [ "$1" = "fix" ]
Xthen
X#	patch by writing
X	for addr in `cat /tmp/fixtcp.$$`
X	do
Xadb -w /vmunix /dev/kmem <<EOF
X0x$addr+$timer2msloff/w 1
X\$q
XEOF
X	done
X
X	sleep 2			# wait for connection to clear
X
X#	make sure it really got unstuck
X	netstat -a | grep $state >/tmp/fixtcp.$$
X	if [ -s /tmp/fixtcp.$$ ]	# file exists and size > 0
X	then
X		echo "$cmd: Sorry, TCP connections still hung!!"
X		rm /tmp/fixtcp.$$
X		exit 1
X	else
X		echo "$cmd: TCP connections in state $state have been unstuck."
X	fi
Xelse
X#	just display the current state flag and current 2 * msl timer
X	for addr in `cat /tmp/fixtcp.$$`
X	do
X		echo "The next two numbers displayed by adb should be $FIN_WAIT2 and 0."
X		echo "The state flag value of $FIN_WAIT2 represents the FIN_WAIT_2 state."
X		echo "The decimal 0 means the 2 * msl timer is off."
Xadb /vmunix /dev/kmem <<EOF
X0x$addr+8/d
X0x$addr+0x10/d
X\$q
XEOF
X	done
Xfi
X
X# cleanup intermediate file
Xrm /tmp/fixtcp.$$
X
Xecho "After: only connections in state $state"
Xnetstat -a | grep $state
Xexit 0
END_OF_fixtcp
if test 3715 -ne `wc -c <fixtcp`; then
    echo shar: \"fixtcp\" unpacked with wrong size!
fi
chmod +x fixtcp
# end of overwriting check
fi
if test -f fixtcp.mk -a "${1}" != "-c" ; then 
  echo shar: Will not over-write existing file \"fixtcp.mk\"
else
echo shar: Extracting \"fixtcp.mk\" \(1079 characters\)
sed "s/^X//" >fixtcp.mk <<'END_OF_fixtcp.mk'
X#
X# [email protected] 2/4/86
X#
X# fixtcp.mk	Makefile, this file
X# fixtcp	shell script
X# get_tcp.c	C program
X# get_tcp	a.out program called by fixtcp
X# .fixtcp.mail	header for mail
X#
X# directions, type
X#	make -f fixtcp.mk	# to make necessary files
X#	edit fixtcp.mk and change INSTALLDIR
X#	make -f fixtcp.mk install
X#	cd $INSTALLDIR		# directory where you really installed it
X#	fixtcp			# to display tcp connections hung in finwait2
X#
X#				# don't do this if you have none to unstick
X#	su root			# required for adb write mode
X#	fixtcp fix		# to actually unstuck tcp connections
X
X# change this at your site
XINSTALLDIR = /sysbin
X
Xall: get_tcp
X
Xclean:
X	rm -f get_tcp
X
Xinstall: get_tcp
X	-cp get_tcp $(INSTALLDIR)
X	-cp fixtcp $(INSTALLDIR)
X
X# distribute the latest version to the world, private for [email protected]
Xdist:
X	rm -rf /tmp/dpb
X	mkdir /tmp/dpb
X	cp fixtcp.mk /tmp/dpb
X	cp get_tcp.c /tmp/dpb
X	cp ../cmd/fixtcp /tmp/dpb
X	cp .fixtcp.mail	/tmp/dpb
X	(cd /tmp/dpb; makeshar * >>.fixtcp.mail)
X
X# please note that .fixtcp.mail was chosen so that makeshar *
X# doesn't try to append to itself.
END_OF_fixtcp.mk
if test 1079 -ne `wc -c <fixtcp.mk`; then
    echo shar: \"fixtcp.mk\" unpacked with wrong size!
fi
# end of overwriting check
fi
if test -f get_tcp.c -a "${1}" != "-c" ; then 
  echo shar: Will not over-write existing file \"get_tcp.c\"
else
echo shar: Extracting \"get_tcp.c\" \(1503 characters\)
sed "s/^X//" >get_tcp.c <<'END_OF_get_tcp.c'
X/*
X * [email protected]
X * prints offsets of fields in TCP connection control block.
X * called by fixtcp sh script
X */
X#include <stdio.h>
X#include <sys/types.h>		/* u_char */
X#include <netinet/tcp.h>	/* tcp_seq typedef */
X#include <netinet/tcp_timer.h>	/* tcp timers */
X#include <netinet/tcp_var.h>	/* tcp connection control block */
X#include <netinet/tcp_fsm.h>	/* defines for tcp states */
X
X/* use S3/S5 strrchr(), but on 4.x systems, remap to Berkeley rindex */
X#ifdef BSD4
X#	define strrchr	rindex
X#endif
X
X#define STR_SAME !strcmp
X#define STR_DIFF strcmp
X
X/* fw non-int functions */
Xchar *basename();
X
X/* external non-int functions */
Xextern	char	*strrchr();
X
Xmain(argc, argv)
X	int	argc;
X	char	**argv;
X{
X	char	*cmd;
X	struct	tcpcb	*p = 0;
X
X	cmd = basename(argv[0]);
X
X	if (argc != 2)
X		{
Xusage:
X		fprintf(stderr, "usage: %s state|2msl|FIN_WAIT2|TIME_CLOSE\n", cmd);
X		exit(1);
X		}
X
X	if (STR_SAME(argv[1], "state"))
X		printf("0x%x\n", &p->t_state);	/* state offset */
X	else if (STR_SAME(argv[1], "2msl"))
X		printf("0x%x\n", &p->t_timer[TCPT_2MSL]);	/* timer offset */
X	else if (STR_SAME(argv[1], "FIN_WAIT2"))
X		printf("0x%x\n", TCPS_FIN_WAIT_2);	/* state value */
X	else if (STR_SAME(argv[1], "TIME_CLOSE"))
X		printf("0x%x\n", TCPS_TIME_WAIT);	/* state value */
X	else
X		goto usage;
X}
X
X/*
X * return basename of full path name
X */
Xchar *
Xbasename(path)
X	char	*path;
X{
X	char	*cp;		/* general char pointer */
X
X	if ((cp = strrchr(path, '/')) == NULL)	/* no rightmost slash */
X		return path;
X	else
X		return cp;
X}
END_OF_get_tcp.c
if test 1503 -ne `wc -c <get_tcp.c`; then
    echo shar: \"get_tcp.c\" unpacked with wrong size!
fi
# end of overwriting check
fi
echo shar: End of shell archive.
exit 0

=============================================================================
Scott Simpson:

Set the socket NOLINGER option using setsockopt. See Richard Stevens
book "Unix Network Programming".

=============================================================================
Brian Hackley:

Hello,

Sounds like you have applications with have those sockets
open.  If there is a server for those applications still running
on the HP, it SHOULD close those connections when it shuts down.
Check with ps -ef for any server running that is owned by init
instead of inetd (if the process is started by inetd).  A friendly
kill should permit the application to shut down those sockets.

If the socket application used the SO_KEEPALIVE option, those
connections would have been cleaned up by the OS.

Hope this helps,

-> Brian 

-- 
Brian Hackley         Network Support       [email protected]
Hewlett-Packard Chelmsford Response Center     Mailstop CHD-01-RC 
300 Apollo Drive  Chelmsford MA 01824               508-436-4478/
#include <std_disclaimer.h>   Telecommuting for now!   1-436-4478

=============================================================================

--
S.T. Wong                                | BITNET: [email protected]
Computer Services Centre                 | Internet: [email protected]
The Chinese University of Hong Kong      | Tel. No: (852) 609 8825
Shatin, N.T., Hong Kong                  | Fax  No: (852) 603 5001

-----------[000227][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 03:25:07 GMT
From:      [email protected] (Larry Drebes)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
Glenn Fleishman ([email protected]) wrote:
: You can use the CERN http daemon to determine what the incoming request
: is (so if someone does http://dummy.com/etc. vs
: http://smartguy.com/etc., you can determine what they've entered and
: route it to different directories).

My brief study into this subject is that the client doesn't
send the full url.  Does the cern daemon do something to request it?

While on the subject of multiple httpd's on a single machine
(with out using differnt ports), it should be possible to use
a loopback pseudo-device that can be ifconfig'ed to a unique 
ip address.  With a little /usr/etc/route magic it should work.

The NCSA httpd binds to INADDR_ANY, but that should be easy 
chang to a specific (configurable) address.  Having only thought
about this, and not tried it, their might be something obvious
that I'm missing.
larry-

-----------[000228][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 03:28:24 GMT
From:      [email protected] (Jeffrey Horn)
To:        comp.protocols.tcp-ip
Subject:   DNS: How to Advertise New Domain
I am setting up a Unix host for a company that recently was connected to
the internet.  I can access nearly everywhere in the outside world from
withing the company (mail, ftp, telnet, etc.), but can only access
addresses within the company by using actual numbers rather than machine
names and domain names.

How can I tell the DNS servers on the internet about a new domain?  The
connection is being done through netcom if that is of any importance.

-- Jeff Horn
-- 
Jeffrey Horn             |BELZER,BERNHARDT,BOETTCHER,BRANDENBURG,DRAVIS,
INTERNET:[email protected]|FETTER,GAPINSKI,GAUGER,HARMS,HIRSCHINGER,HORNE,
PHONE:(608) 274-5109     |JUECH,KLAJBOR,KROIS,KRONING,LEMKE,RUNGE,STOCK,
FAX:(608) 221-5008       |TAUBERT,TRESKE,WILLMERT,ZILLMER

-----------[000229][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 04:23:46 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
In article <[email protected]>
[email protected] (Larry Drebes) writes:

> My brief study into this subject is that the client doesn't
> send the full url.  Does the cern daemon do something to request it?

I was looking at the code in action.....I'll have to double-check. You
can set up .conf files that control access and route material depending
on what pages or URLs have been specified.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000230][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 04:24:37 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip
Subject:   Re: DNS: How to Advertise New Domain
Run, don't walk, to the nearest bookstore and buy DNS and BIND from
O'Reilley and Associates.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000231][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 94 10:41:00
From:      [email protected] (Paul A Vixie)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
This hasn't anything to do with domain name service but...

>The NCSA httpd binds to INADDR_ANY, but that should be easy 
>chang to a specific (configurable) address.  Having only thought
>about this, and not tried it, their might be something obvious
>that I'm missing.

Even better, leave it bound to INADDR_ANY and use getsockname() on each
incoming connection.  I've been meaning to teach this trick to inetd for
years.
--
Paul Vixie
Redwood City, CA
decwrl!vixie!paul
<[email protected]>

-----------[000232][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 07:34:55 +0000
From:      [email protected] (Phil Cramp)
To:        comp.misc,comp.protocols.tcp-ip,comp.sys.hp.hpux,comp.sys.hp.misc,comp.unix.misc,comp.unix.questions
Subject:   [Q] PC-NFS, pcnfsd, inetd and security
Internet gurus can you help me?
As a site we run a number of UNIX (mainly HP-UX) servers
on behalf of a growing number of customers, (FM if you will..)
One of our customers has expressed an interest in using SunSelect's
PC-NFS package as their means of communicating with us of a bridged
& routed internetwork. (Our requirements for them to connect to us
was TCP/IP and the ability for us to use a remote print server on
their site to deliver their printed output.) 

In order to print remotely to PC-NFS the PC-NFS "lpd" software
must be running on the remote PC. To run the "lpd" the customer
machine must be "logged in" on the network, requiring us to run
the pcnfsd (v2) on our HP-UX box. The HP-UX manuals indicate
that:
a) pcnfsd can be run directly from the netnfsrc script at boot time
   or from the inetd, using entries in inetd.conf (I have tried both
   successfully)
b) there is a pcnfsd.conf file which can be used to determine the 
   "validation" done by pcnfsd when requests are made of it.

My questions
1. What other means exist of validating incoming requests for pcnfsd?
   Is it possible to use entries in inetd.conf (if starting pcnfsd
   from here) to further "secure" pcnfsd requests.
2. Are there any other issues I should be aware of before deciding
   to go down this route with the customer.
3. Does anyone else have experience of this type of software in a 
   production environment, and are there any pitfalls to be aware of?

One other point - this customer will not be the only user of the HP-UX
server on which we will be running pcnfsd. How vulnerable will the machine
be once we have started this daemon?

Email or post, I'll summarise replies in all groups I have posted to.
Thanks in advance
-- 
Phil Cramp                          [email protected]
AT&T Istel Ltd., Redditch, UK
#include <std/disclaimer>
Plain .sig, not an original thought in my head...........

-----------[000233][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 07:44:54 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (Ruediger Volk) writes:
>In article <[email protected]>, [email protected] (don provan) writes:
>  > Just to set the record straight, the NetWare TCP/IP implementation
>  > has never had any such problem. In a subnetted class B address, the
>  > subnet bits in the last byte can be anything at all so long as the
>  > third byte of the address is neither 0 nor ff hex.
>
>really?  do you impose any restrictions on the subnet masks?
>Any mask from  18:14 to 30:2 bits allowed (I'm already restricting to
>masks that conform to both RFC 950 *and* CIDR)?

NetWare accepts any mask that allows at least two bits for the subnet
number and at least two bits for the host number. Well, technically
*any* division of the 16 bits is allowed, but the ones with a single
bit portion will always result in an address not allowed under
RFC-1122, so the binding will be rejected on those grounds.

>Why would you want to forbid e.g. subnet numbers xx.xx.0.64/128/192
>and xx.xx.255.64/128/192 under mask 255.255.255.192?

I'm sorry, I must not have been clear. Such addresses are fine and
NetWare accepts them (except I assume you meant to include xx.xx.255.0
instead of xx.xx.255.192 under the 255.255.255.192 mask, since that
latter causes subnet broadcast ambiguity). What I was saying was that
when the third byte is not 0 or 255, you have nothing to worry about
in the last byte. When the third byte *is* 0 or 255, then, of course,
you have to look at the third and fourth bytes together.

						don provan
						[email protected]

-----------[000234][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 08:03:42 GMT
From:      [email protected] (Martyn Spink)
To:        uk.announce,comp.protocols.tcp-ip
Subject:   TCP-IP Course
*****************************************************************************
*****************************************************************************
*****************************************************************************

There are a few places left on the following course. Please contact Helen
Sheehy a.s.a.p., if you are intrested in a place.

thanks

*****************************************************************************
*****************************************************************************
*****************************************************************************

		TCP/IP Protocols: A Practical Introduction
			   [Code:  TCP/IP-1]

			   Dr. Andy Carpenter

			  21st & 22nd September 1994
	
			     Cost : #280.00
                    [discount available for academics]

                                PEVE Unit
                       Department of Computer Science
                         University of Manchester 

*****************************************************************************
*****************************************************************************
*****************************************************************************

For Further Details:

Please contact Helen Sheehy via:

Phone:  061-275-6172

Email:  [email protected]

Fax:    061-275-6200

Post:   PEVE Unit
	Department of Computer Science 
        University of Manchester
	Oxford Road 
        Manchester M13 9PL

*****************************************************************************

Aim:

This course is intended to give a comprehensive introduction to the
TCP/IP protocols that are used on many Local Area and Wide Area
networks, and to impart an understanding of the problems in
establishing, managing, and resolving problems in such networks.

*****************************************************************************

Course Outline:

The first day starts by reviewing the Internet and the sort of
services that it has to provide.  Various terms and concepts are
introduced, in particular, the idea of protocol stacks and
addresses.  The rest of the day looks at the telnet and TCP
protocols, and RFCs.

The second day starts by examining other protocols central to the
operation of a TCP/IP network; e.g. UDP, IP and ICMP.  Various
aspects related to the operation of the protocols on particular
physical networks are then considered; for example, the need for ARP
on broadcast networks to discover the network address associated with
a particular IP address.  The course concludes with sections on
routing and advanced services.

The telnet protocol discussion covers the definition of the Network
Virtual Terminal (NVT) and the use of option negotiation to alter the
characteristics of an end-to-end connection.  Various problems of
remote terminal emulation are examined; for example, the perceived
delayed effect of control-C operations due to data buffering within a
network.  This section concludes with a paper exercise to decode a
telnet session, including option negotiation.

The section on the TCP protocol starts by examining the way in which
a reliable byte stream is obtained from a potentially unreliable
network.  Other aspects of the protocol covered are: establishing a
connection, sequence numbers, data transfers, flow control and the
format of the TCP header.  This section contains an exercise to
extract the data from a series of TCP packets and predict the
acknowledgements that are sent.

The IP protocol discussion primarily covers the format of the IP
header, the fragmentation and reassembly of packets that cross
networks, and the use of IP options.  In contrast, the section on UDP
and ICMP protocols concentrates on the uses of these protocols.  As
part of this section there are two paper exercises.  One of these is
on IP packet fragmentation; while the other requires the decoding of
the data passed in a telnet session from a dump of the packets
passing across a physical network.

Routing starts by examining the various IP address classes that are
defined, i.e. A, B and C, and their use.  It then goes on to look at
the routing gateway protocols that are available.  An example of the
operation of a dynamic gateway protocol is used to illustrate this.
The section concludes with a section on subnetting.

The advanced services section covers global naming and network
management.  The naming section covers the need for machine
registration, including the mechanisms involved, and name and domain
name servers.  The network management section introduces the
protocols used, e.g. SNMP and CMOT, and the need for object
identification.  Object identification leads onto network databases;
e.g. MIB.

*****************************************************************************

Course Delegates:

The course is suitable for network specialists who intend to become
involved in setting up TCP/IP-based computer networks, in maintaining
and trouble-shooting these networks, or in writing protocol code.

*****************************************************************************



--

-----------------------------------------------------------------
Martyn Spink, "Acting" Director, 
PEVE-IT Unit, Department of Computer Science, University of Manchester, 
              Oxford Road, Manchester, M13 9PL, U.K.
Tel: (+44) 61-275 6157         FAX: (+44) 61-275-6200
ARPA: [email protected]   
JANET: [email protected]    UUCP: ..!mcsun!uknet!mucs!mspink
-----------------------------------------------------------------

-----------[000235][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 94 08:23:57 GMT
From:      [email protected] (Hakan Nohre)
To:        comp.protocols.tcp-ip
Subject:   TCP ack-loop -- HELP
I would appreciate some help with the following problem :

I am running an application using a TCP connection between two nodes. 

Between the two nodes I have an X.25 network, capable of alternative routing.

Now sometimes when an X.25 connection goes down, TCP starts "acking like mad" :

  TCP Client			TCP Server

     ------ ACK Seq = x1, Ack=y1 ------>
     <---------ACK Seq=x2, Ack=y2-------

     ------ ACK Seq = x1, Ack=y1 ------>
     <---------ACK Seq=x2, Ack=y2-------

     ------ ACK Seq = x1, Ack=y1 ------>
     <---------ACK Seq=x2, Ack=y2-------

This goes on for a couple of minutes. Note that there is no "wait" so this traffic
is really clogging the network.

Note : These are empty acks, i.e. no user data.

Note :  y1 > x2 and y2 > x1, this means that both sides are acknowledging something
that the other side "has not sent yet".

Note : Since an X.25 connection has gone down, it is LIKELY that data in transit
       have been lost, and POSSIBLE that a TCP/IP datagram has been received 
       twice (if an ack was lost).

Note : The TCP code is based on the Berkeley code (1988).


QUESTION : WHY ? Is this a known problem ? Any ideas of what could be wrong ??


Regards Håkan




-----------[000236][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 94 08:33:10 GMT
From:      [email protected] (Hakan Nohre)
To:        comp.protocols.tcp-ip
Subject:   [Q] : TCP and NETWORK UNREACHABLE
Hello,

What would be the expected behaviour by TCP (in ESTABLISHED state) when an
NETWORK UNREACHABLE is received ??

Should it give up the connection or retransmit ?

The implementation of TCP that I am using returns an error from send/recv and sets the
error to ENETUNREACH. (Contrary to the UNIX implementation, according to the man-pages).

Whether I am supposed to close the connection here or not I do not know.
(I would rather not close the connection, and let TCP sort it all out).


Regards Håkan






-----------[000237][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 17:21:25 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Solaris 2.3 SO_KEEPALIVE - HELP
[email protected] (Tony DeFeo) writes:

>I am trying to detect a physically broken socket connection between two
>Sun workstations running Solaris 2.3. I have set the SO_KEEPALIVE option 
>on the socket using seteockopt() which sounds like it should do the trick
>according to the man pages, but doesn't seem to be working.

That's what it keepalive should do.  It's obviously misnamed.  A better
name might be "killmeif".


>What happens is, after I establish the socket connection between the two
>workstations, I physically disconnect one of them from the network. 
>I wait several (up to 30) minutes without sending any data, and neither

There might be some timer to set, too.  But I would think it to be
unreasonable to have it at or above 30 minutes.


>side detects that the other has gone away (I don't see any TX/RX activity
>on my transceiver either, I would expect to see activity from the SO_KEEPALIVE
>messages if it was working as advertised). Next I wrote some data on one side
>and waited again.  After about 3 minutes, the side that I wrote the data on
>detected the broken connection and shut down it's end of the socket properly.

But apparently not because of SO_KEEPALIVE.


>broken sockets? Optimally, what I would like is for both ends to time out
>and close the socket even though I am writing no data in either direction.

I would have just written some data.  But if you have to conform to an
existing protocol, this is probably not an option.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000238][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 17:20:06 -0400
From:      [email protected] (JameZ1)
To:        comp.protocols.tcp-ip
Subject:   Special Week of Network Seminars -PART I
UCSC Extension in Santa Clara
presents
A Week-Long Set of Seminars on
Emerging Technologies in Computer Networking
Monday-Friday, October 17-21
 PART II

JOINT SESSION
Friday, October 21

9-10:30 a.m.
Optically Transparent Switching and Its Use in Networks
J. W. Goodman, Chairman, Department of Electrical 
Engineering, Stanford University

10:45 a.m.-12 noon
New Frame Relay Directions
Cheryl VanderGriff, Product Manager, Sync Research and 
the Frame Relay Forum

1-2:15 p.m.
Putting Multimedia Technologies in Real Networks
Fouad Tobagi, Professor of Electrical Engineering and 
Professor of Computer Science, Stanford University.

2:30-3:30 p.m. 
Integration of Telecom and Networking Infrastructures 
within the Corporate Network
Joyce Axtell, Telecommunications Manager, Technology 
Funding

Closing Remarks
3:45 p.m.
Victoria Marney-Petix, Program Chair, and Coordinator, 
Certificate Program in Network Management

*******************
Fee: $975, 35 hours of Continuing Education credits. 
(No grade issued for this course). Enrollment limited.

Santa Clara (EDP 942D05)
5 meetings: Monday-Friday, October 17-21, 9 a.m.-5 p.m.  
UC Extension, 3120 De la Cruz Blvd. (Trimble at 101).
*******************
General Information
To Enroll
You may enroll by telephone or by FAX. When paying by 
VISA, MasterCard, or American Express, call 1(800) 660-
UNEX (8639); outside California; inside California, 
call (408) 427-6600 Monday- Friday, 8 a.m.-5 p.m. or 
send your enrollment form by FAX (408) 427-6608. Or 
mail the enrollment form included in this publication.

For Further Information
Please call or write:
(408)748-7375
University of California Extension
Science & Technology
3120 De la Cruz Blvd.,
Santa Clara CA 95054-2429

Extension's Santa Clara Location
The location for University Extension's Santa Clara 
Valley classrooms is 3120 De la Cruz Boulevard, Santa 
Clara,in the Oxford Business Park, at the intersection of De la Cruz
Boulevard and Trimble Road.





-----------[000239][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 11:31:33 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
In article <[email protected]> [email protected] (Ran Atkinson) writes:
>RFCs.  Smart folks look first at the relevant RFCs, including Host
>Requirements, and then at the BSD networking code.

I'd strongly suggest people treat the BSD networking code with caution -
especially on the matter of URGent data and ICMP error handling. RFC1122 and
the BSD code are very different.

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000240][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 12:52:49 GMT
From:      [email protected] (Don Rolph)
To:        comp.protocols.tcp-ip
Subject:   Re: PPP driver for TCP/IP32 from Microsoft
In article <[email protected]> [email protected] (Todd A Wallace) writes:
>From: [email protected] (Todd A Wallace)
>Subject: PPP driver for TCP/IP32 from Microsoft
>Date: 8 Sep 1994 14:30:07 GMT
 
>Does anyone know of a PPP driver that works with a dialup PPP account on
>the one hand, and the TCP/IP32 package that works with  Windows for
>Workgroups that you can get from Microsoft?
 
>----------------------------------------------------------------
>|    Todd Wallace               |  "A pessimist is surprised   |
>|    [email protected]    |   as often as an optimist,   |
>|-------------------------------|   but always pleasantly."    |
>| Expatriate Midwesterner (tm)  |            - Robert Heinlein |
>----------------------------------------------------------------

No but there is a slip driver nbr11 which is reputed to work.  I finally found 
a copy, but have not tried it yet.

Regards.
 
Don Rolph [email protected] WD3 MS10-13 (508)-699-1263

-----------[000241][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 13:38:33 GMT
From:      [email protected] (Felix Gaehtgens)
To:        comp.protocols.tcp-ip,de.comm.internet
Subject:   HELP! Complex routing with a Cisco 2503

Hi,

I have a question regarding the setup of a Cisco 2503
router for a complicating routing situation, which
involves two default routes which have to be used
according to the source IP address.

Here's a small overview of the network layout. The
IP addresses are not the "real" ones, I've just used
them for the example.

     Network A      Network B
         |              |    
         V              V
   194.100.101.3   194.100.102.5              
         |          |     <-- 2x 64 kbit Interfaces
      +-------------------+           
      |                   |           
      | Cisco 2503 Router |           
      |                   |           
      +--------+----------+           
               |   <-- Ethernet Interface (193.101.102.1)
  Network C -> |                                         
               |                                  
               |  +-------------+ 193.1.1.1       
193.101.102.2  +--+ UNIX Box    +-----+           
               |  |             |     | <------- Network D
               |  +-------------+     |                 
               |                      |  +-------------+
               |            193.1.1.2 +--+ UNIX Box    +
193.101.102.3  +                      |  |             |
               |                      |  +-------------+
            +--+---------+            |                 
            | UNIX Box   |            |  +-------------+
            |            |  193.1.1.3 +--+ UNIX Box    +
            +------------+            |  |             |
                                      |  +-------------+
                                      |                 
                                      |  +-------------+
                            193.1.1.4 +--+ UNIX Box    +
                                      |  |             |
                                      |  +-------------+
                                      |                 
                                         
                                      ...


Now, all traffic coming from the network C should
have the default route via network A. Traffic from
network D should have the default route via network
D. Traffic from network C should NOT be routed via
network B and network D should NOT be routed via
network A.

I understand the problem is that two default routes
should be used on the Cisco. I can set up ACLs
(access control lists) and tell the Cisco not to
route any traffic from network D to network A and
so on. but is it also possible to tell it explicitly
which interface to route the traffic over according
to the source IP address of the packets/connections?

There seems to be a mechanism for setting up routing
policies, but I'm not sure whether this can be used
for my situation.

I'd be very grateful for any hints on how to do this,
or, should this be impossible, for telling me. Please
reply via email and I'll make a summary upon request.

Thanks,
	Felix

-----------[000242][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 15:49:29 GMT
From:      [email protected] (Ran Atkinson)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
In article <[email protected]> [email protected] (Ran Atkinson) writes:
>>RFCs.  Smart folks look first at the relevant RFCs, including Host
>>Requirements, and then at the BSD networking code.

In article <[email protected]> [email protected] (Alan Cox) 
writes:
>I'd strongly suggest people treat the BSD networking code with caution -
>especially on the matter of URGent data and ICMP error handling. RFC1122 and
>the BSD code are very different.

Yes.  However, most of the installed base follows BSD conventions and
violates RFC-1122.  One has to pay attention to the BSD code if one
wants interoperability.

Ran

-----------[000243][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 17:53:21 GMT
From:      [email protected] (I.E.Kho)
To:        comp.protocols.tcp-ip
Subject:   Modified TCP Congestion Avoidance Algorithm
Can anyone tell me, where I can find the above article written by V. Jacobson, 1990?

Thanks,
[email protected]

-----------[000244][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 18:36:13 GMT
From:      [email protected] (Stefan Sharkansky)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.protocols.ppp
Subject:   Re: [Q] Low end router?
In article <[email protected]> [email protected] (Russo and Hale) writes:
>Does anyone know of a good router that is on the low cost end?  I want to 
>connect a Class C ethernet (winsock) to a PPP dialup provider (Portal 
>Communications).
>

Check out the following companies:

Morningstar Technologies 	http://www.morningstar.com/
Networks Northwest		 http://www.networksnw.com/

also, Rockwell has a similar thing that was reviewed in PC Magazine Network
Edition earlier this summer.

--

Stefan Sharkansky
Prospero Systems Research, Inc.
USMAIL	520 Frederick St. Box 19, San Francisco, CA 94117
VOICE	(415) 731-8114		FAX  (415) 731-3395
E-MAIL	[email protected]


-----------[000245][next][prev][last][first]----------------------------------------------------
Date:      12 Sep 1994 19:06:08 GMT
From:      [email protected] (Russo and Hale)
To:        comp.protocols.tcp-ip,comp.protocols.ppp,comp.protocols.tcp-ip.ibmpc
Subject:   [Q] FreeBSD Router Floppy
Does anyone know where I can find this?  I have been told that it routes 
nicely from PPP on a 386.  Does it use gated at all?

-----------[000246][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 19:38:10 GMT
From:      [email protected] (Tony DeFeo)
To:        comp.protocols.tcp-ip
Subject:   Solaris 2.3 SO_KEEPALIVE - HELP
I am trying to detect a physically broken socket connection between two
Sun workstations running Solaris 2.3. I have set the SO_KEEPALIVE option 
on the socket using seteockopt() which sounds like it should do the trick
according to the man pages, but doesn't seem to be working.

What happens is, after I establish the socket connection between the two
workstations, I physically disconnect one of them from the network. 
I wait several (up to 30) minutes without sending any data, and neither
side detects that the other has gone away (I don't see any TX/RX activity
on my transceiver either, I would expect to see activity from the SO_KEEPALIVE
messages if it was working as advertised). Next I wrote some data on one side
and waited again.  After about 3 minutes, the side that I wrote the data on
detected the broken connection and shut down it's end of the socket properly.
The other side however, just sat there.

Am I doing something wrong, or have I misunderstood what SO_KEEPALIVE is supposed
to do? If I have misunderstood SO_KEEPALIVE, is there some other way to detect
broken sockets? Optimally, what I would like is for both ends to time out
and close the socket even though I am writing no data in either direction.
Any help would be appreciated.

Thanks,

Tony DeFeo



-----------[000247][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 19:56:18 +0000
From:      [email protected] (steve johnson)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP Packages for PC's (Again)
In article <[email protected]>, [email protected] (Susan
Ferebee Guion) wrote:

> I'm once again looking for the best in tcp-ip packages for my PC's.
> 
> I'm had it woth FTP's PCTCP and the ~70K of low memory it requires.  Now
> I'm looking for a replacement.  I'll need all the standards (ftp,telnet,
> ping,...) but I need to preserve low memory.  I might even decide to go
> with an X-terminal package if the proce war right.

x terminal emulators take much more time and memory.  time because they've
got to constantly accept and redraw the entire x window as opposed to just
accepting the command, and memory to do it all.  a rooted window may take
2-3 meg, and rootless about 5.  i like them 'cuz i can sit at a mac or pc
and do *real* work, but they can be slow.  it also depends on how much ram
the x terminal has and how many people are using it.  and network traffic.

i've got windows for workgroups (windows 3.1.1) which i've heard is great
but haven't had the time to install it.  i hear it comes with tcp, telnet,
ping, etc.  one of my network buddies loves it.  it's all 32 bit.

whatever you do, make sure it (telnet, ping, x emulator...) is 32 bit. 
speeds things up a lot.

> I'm interested in experiences you might have had with any of them.
> thanks in advance...
> 
> -- 
> Susan Guion - Georgia Tech
> Internet: [email protected]
> uucp:     ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!gt5582a

-----------[000248][next][prev][last][first]----------------------------------------------------
Date:      Mon, 12 Sep 1994 22:15:21 GMT
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
In article <[email protected]> [email protected] (Glenn Fleishman) writes:
>In article <[email protected]>
>[email protected] (Larry Drebes) writes:
>> My brief study into this subject is that the client doesn't
>> send the full url.  Does the cern daemon do something to request it?
>I was looking at the code in action.....I'll have to double-check. 


I think you're thinking of the CERN daemon's "proxy HTTP" support, which is
triggered when it receives a full URL.  This requires the browser to be
configured to use proxy mode.  It doesn't happen automatically.
-- 
Barry Margolin                                                [email protected]


-----------[000249][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 10:03:40 -0700
From:      [email protected] (Lon Stowell)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Rob Tanner) writes:
>
>Direct from Rumor Control Central:  I have heard that IPX is a real
>bandwidth hog, and that if you don't want to see your TCP/IP severely
>degrade, don't let IPX run over the same segments that run TCP/IP.

   I see IPX and TCP/IP running on the same segment all the time.  
   Works fine, tastes great, etc.  Each protocol stack has
   peculiarities and benefits.   

   Some folks seem to be concerned about the chatty broadcasts of
   the Novell Service Advertisement Protocol.  If you use a Sniffer
   to note the actual percent of bandwidth utilization of Netware, it
   is really insignificant...it just LOOKS like lotsa frames.
   These ARE sent to the MAC layer broadcast address...I would believe
   some truly horrid implementations of the Unix LAN drivers would be
   too braindead to just discard them.  Never really seen a station
   get so overloaded with the SAP broadcast bursts that it lost
   tcp frames in close proximity...but again I guess it COULD happen
   with a horrid implementation. 

>
>I know zip about IPX, but it looks like we are going to be installing
>a Novell server anyway.  The Novell, of course, is for the PCs.  What
>concerns me as that we also have a number of Unix machines as well as
>an even larger number of X-stations that don't run IPX and don't
>concern themselves with the Novell server.

   How much IPX traffic the PC's will generate depends quite a bit on
   the LAN drivers and cards in the PC as well as sheer configuration.
   Most PC's just can't crunch sufficient bytes to bring LAN
   utilization up very much.   If you set your PC clients to always
   check LOCAL drives before the network drives you can cut a bit of
   traffic off the LAN, but again the real percent utilization is
   pretty low. 

   The only "quaint" thing I've observed is that the standard Windows
   multi-protocol LAN driver seems to service the individual protocol
   stacks in bursts.  First you'll see a buncha Novell frames coming
   out of the PC, then a buncha TCP frames...it doesn't seem to ever
   mix the two except in bursts.  This does make mixed protocol
   applications such as tcp client, ipx client run a bit slower than
   they need be....but if you aren't running dual stacks wouldn't
   worry about it. 


>
>Will my unix boxes feel the impact of IPX, or is the rumor unfounded?

   Lan traffic is lan traffic.  At a given utilization you will begin
   to note some tcp retransmissions in netstat.  Shortly after the tcp
   retransmissions begin to go above a coupla per cent, you may notice
   a slowdown in response time or bulk thruput.  In other words, if
   your LAN is already overloaded or near the edge, the Netware
   traffic will be a problem, but only because it is MORE traffic, not
   simply because it is Netware. 

   Note that I am assuming that none of your Unix stations are of the
   900+ Kbyte variety with continuous batch file xfers going on...in
   which case your IPX stations aren't gonna much like that LAN.  

   If your current Unix boxes are running the most common frame
   format, the EtherType or DIX format, set the Novell stations to
   that.  The only real problems I've ever seen was with "real" OSI
   stations on the same LAN with the 'novell hacked' quasi-802.3
   frame format in the IPX systems.  I believe Novell has discontinued
   this non-OSI compliant framing format.  If you use 802.3/SNAP
   format on all stations that also works fine.  

>I could isolate the IPX traffic, but to do that requires purchasing
>hardware and that means money that has to be justified.  Does anybody
>have any real performance figures that they could share.

I have some Sniffer traces rightchere on my desk.  

One set is from a PC-Novell type LAN.  The PC's pull in Oracle
Forms and such from a Novell Server rather than local drives.
The Oracle SQL*Net from the PC is then over tcp/ip to a Unix
Oracle Server.   There are a few handfuls of Sun's on that same
segment crunching away doing software development over X11 with the
Unix superserver that is also the Oracle server.

    The Novell stuff is taking less than 1% of the LAN.  

    There is about 1% Apppletalk.

    The rest is TCP/IP.

    Total frames per second averages about 300 which is roughly
    400Kbytes/second aggregate measured.  There is some tcp
    retransmission, but can't really tell if it is due to a slow
    station or what. 

The other trace is from a humongous largely Netware type LAN with
several of the big NetFrame servers on it.   The Unix stations are
in the minority, a few Suns, RS/6000's, X-stations and a big
superserver.  

    Novell is taking about 77% of the LAN.  

    TCP is getting about 9.9%

    Appletalk [those things are EVERYWHERE  >:-)  is the rest]

    The frame/second rate is rarely under 200 frames/second and
    regularly goes over 600+ for peaks of a few seconds.

    Both the TCP and Novell retransmission rates are a bit higher, in
    the several percent rate, but still overall response time is
    within targets for the users.  


Bottom line, I wouldn't believe too many rumors about mixed protocol
LAN's, particularly of the type you note....IF you don't exceed the
LAN's capacity you should have no problems.  If you do exceed the
LAN's capacity, it wouldn't matter if you were running Novell or even
Commodore/64 networking protocols on it. 


-----------[000250][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 06:17:41 -0400
From:      [email protected] (Eric V. Smith)
To:        comp.protocols.tcp-ip
Subject:   Re: BOOTP server answering DHCP client
In article <[email protected]>,
Alex Vickers <[email protected]> wrote:
 [ info about DHCP requests and replies deleted ]
>
>The bootpd sees this, modifies the BOOTREQUEST to a BOOTREPLY (as
>bootp normally does), assigns an ip-address to the client ip-address
>field and then, well, I initially made it just send that back but it
>was ignored, then after I'd dumped the vendor/options part of the
>packet I tried making it change the DHCPDISCOVER to a DHCPOFFER and
>send that but it still ignored it. The PC tries sending the same
>packet about three times then gives up and hangs. I've read RFCs 1533
>and 1534 which helped a little but don't seem to contain the answer to
>my problem. So my next strategy was to discover this newsgroup and ask
>here.
>
>If anyone can make any suggestions or point me at some useful RFCs
>then that would be very much appreciated.
>

Well, I'm about 2/3 finished with a DHCP server, so I'm pretty 
familiar with this.

Look at page 25 of RFC 1541.  It contains which fields in the BOOTP
message must be set, as well as which options must (or must not) be
present.  In partucular, note that the IP address lease time MUST be
present.

Also, be sure you look at the MSB of the flags field.  This tells you
whether you can unicast the reply to the client, or whether it must be
broadcast.  Daytona clients seem to request a broadcast when they are
first configured, but subsequent DHCPDISCOVER's can be unicast.

Eric.


-----------[000251][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 94 09:20:06 PDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   TCP/IP over OSI stack

We are look into implementing TCP/IP in the lower three layers of the OSI 
stack. In other words implement RFC 1006. I have two questions based on this 
subject:

	1.  Does anyone know of a FTP site that I can go get the source for RC 
	    1006 implemenetation OR know of a product that I can buy that 
	    implements RFC-1006 ?

	2.  If you have any experience with RFC-1006 implementation, please 
	    share it with me in term of performance, resource management, 
	    bandwidth requirements or any other topics that I have not 
	    mentioned.  

Thanks
Neil Butani
[email protected]


-----------[000252][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 07:47:47
From:      [email protected] (Clark Bremer)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]> [email protected] (Michael Salmon) writes:
>From: [email protected] (Michael Salmon)
>Subject: Re: TN5250/TN3270 Standards?
>Date: 7 Sep 1994 14:06:47 GMT
 
>In article <[email protected]>
>[email protected] (Clark Bremer) writes:
>|> In article <[email protected]>
>|> [email protected] (Michael Salmon) writes:
>|> >In article <[email protected]>
>|> >[email protected] (Rich Oldroyd) writes:
>|> >|> 
>|> >|> We're looking for the  specifications/RFCs covering the
>|> >|> TN3270 and/or TN5250 products. We're aware that we
>|> >|> can buy some off the shelf products supporting these
>|> >|> standards but would like the specs themselves. 
>|> >|> Any pointers?
 
>|> >To the best of my knowledge these are proprietry IBM protocols and you
>|> >have to by the specs. from them.
>|> 
>|> I'm not sure where its written down, but for TN3270, I'm sure of the answer -
>|> its simply to negotiate for "binary mode" on and "send end of record" on. 
>CB.
 
>Sorry but that's how you enter 3270 mode in Telnet. Telnet then become
>the carrier for the 3270 protocol.

Which is all that TN3270 is (3270 over telnet) , which is what the original 
question was about.  There are no RFC's for 3270 - That's all True Blue, who 
"don't need no steenking RFC's" to define protocols for their own equipment.  
CB.

===========================================================================
          _  _               Clark Bremer     [email protected]
         /  /_)              Software Engineer, NetStar Inc.
         \_/__)              10250 Valley View Road  MPLS, MN 55344

-----------[000253][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 03:22:55 GMT
From:      [email protected] (Alex Vickers)
To:        comp.protocols.tcp-ip
Subject:   BOOTP server answering DHCP client
Hello,
I am trying to modify our bootpd (2.4.0) so that it will dynamically
assign IP addresses to requesting PC's. My problem is not with getting
an IP address but with making the PC take notice of the packet that
bootpd sends. Our modified bootp used to work for bootp requests but
now we have a PC wanting to use DHCP and it ignores the BOOTP packets
with assigned IPs. I'll describe what I know happens and what I've
tried thus far:


The bootpd runs on an Ultrix 4.2 Decstation and is started by inetd.
The PC is running MS-Windows for WorkGroups 3.11 and loads the MS
tcpip stack. This sends a DHCP BOOTREQUEST packet with the following
DHCP options (ascii codes in hex):

63.82.53.63                         - DHCP magic cookie

35.01.01                            - DHCPDISCOVER (Message Type)

3d.07.01.<ethernet address of PC>   - Client Identifier

0c.09.<name of the PC>              - Host Name

ff                                  - End


The bootpd sees this, modifies the BOOTREQUEST to a BOOTREPLY (as
bootp normally does), assigns an ip-address to the client ip-address
field and then, well, I initially made it just send that back but it
was ignored, then after I'd dumped the vendor/options part of the
packet I tried making it change the DHCPDISCOVER to a DHCPOFFER and
send that but it still ignored it. The PC tries sending the same
packet about three times then gives up and hangs. I've read RFCs 1533
and 1534 which helped a little but don't seem to contain the answer to
my problem. So my next strategy was to discover this newsgroup and ask
here.

If anyone can make any suggestions or point me at some useful RFCs
then that would be very much appreciated.


Thanks In Advance,
  Alex Vickers    ([email protected])

Department of Survey and Land Information (New Zealand)
--

Bye,
Alex Vickers ([email protected]).

"Under the most rigorously controlled conditions of pressure,
 temperature, volume, humidity, and other variables, the organism will
 do as it damn well pleases."                          - Harvard's Law


-----------[000254][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 11:45:25 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: DNS: How to Advertise New Domain
[email protected] (Glenn Fleishman) writes:

| Run, don't walk, to the nearest bookstore and buy DNS and BIND from
| O'Reilley and Associates.

For those who can't read, it's the one with the grasshopper on the cover
spreading it's wings.  There is also a smaller grasshopper on the lowest
blue line.

If your bookstore sorts by author, look under Albitz/Liu.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000255][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 11:55:45 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
[email protected] (Samuel Lam) writes:

| You guessed it!  We have two vendors' routers refusing to talk
| to each other because they each implements one of the above
| "features".

Sounds to me like they have too many features in their products.  :-)


| Pointers to references which I could cite to these vendors
| would be great.

I'd start with the RFCs.  I don't remember which one has the right
information so I just fired up a grep and will mail you the results.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000256][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 03:42:03 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
In article <[email protected]>
[email protected] (Barry Margolin) writes:

> I think you're thinking of the CERN daemon's "proxy HTTP" support, which is
> triggered when it receives a full URL.  This requires the browser to be
> configured to use proxy mode.  It doesn't happen automatically.

Right, but it's not too bad to configure. If you use this, then you can
redirect the user's typed URL wherever you want.

So we can maintain
www.foo-bar-1.com
www.foo-bar-2.com
www.foo-bar-3.com
and use DNS records to have them resolve to the same address. The CERN
proxy support lets the HTTPd determine which of those paths was entered
and put them in the appropriate place on a single host. Voila!
----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000257][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 03:44:24 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single, host?
In article <[email protected]>
[email protected] (Mark Andrews) writes:
> >popco.com     204.94.44.1     db.popco
> >zap.com       204.94.44.1     db.zap
> >etc.
> 
>         They should be
> 
>         secondary popco.com     204.94.44.1     db.popco
>         secondary zap.com       204.94.44.1     db.zap
> 
>         or
> 
>         primary popco.com db.popco
>         primary zap.com db.zap

Absolutely. My apologies for quoting that incorrectly.

We have files that say

primary zap.com db.zap
primary popco.com db.popco

on our primary. Sorry!
----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000258][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 12:22:33 -0500
From:      [email protected] (John Schnizlein)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]>, [email protected] (Clark
Bremer) wrote:
...
> >|> >[email protected] (Rich Oldroyd) writes:
> >|> >|> 
> >|> >|> We're looking for the  specifications/RFCs covering the
> >|> >|> TN3270 and/or TN5250 products. We're aware that we
> >|> >|> can buy some off the shelf products supporting these
> >|> >|> standards but would like the specs themselves. 
> >|> >|> Any pointers?
 ...
> Which is all that TN3270 is (3270 over telnet) , which is what the original 
> question was about.  There are no RFC's for 3270 - That's all True Blue, who 
> "don't need no steenking RFC's" to define protocols for their own equipment.  
> CB.

I do not claim to know the contents, but tn3270 IS covered by RFCs: rfc1647
"TN3270 Enhancements" 7/15/94, and rfc1576 "TN3270 Current Practices"
1/20/94.

While it is true that IBM does not need to publish their technical
specifications through the RFC process, what most people need from their
3270 protocol is now well known (just not by me:-).

I am happy to note another "High Sierra" fan, I think.
-- 
disclaimer? we don't need no stinking disclaimer! | John M. Schnizlein
everybody knows nobody can represent the views of | [email protected]
435 elected policy makers.                        | router jockey

-----------[000259][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 94 13:15:03 -0500
From:      Richard N. Lussier <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Re: WANTED: ETHERNET SERVER with RS232 OUTPUT
ST Olektron Systems has an LI-1000 LAN INterface which accepts TCP/IP
ethernet packets and routes these packets to an rs-232 output. This
design can be reversed in your case. There are several options
available.
They can be reached at:
	ST Olektron Systems
	37 Sutton Rd.
	Webster, MA. 01570
	1(508)943-7440
	1(508)949-1804 fax

-----------[000260][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 11:59:23 -0400
From:      [email protected] (Jon Zeeff)
To:        comp.protocols.tcp-ip
Subject:   async 56k CSU/DSU with compression
I have a 115k bps async connection (with flow control) that I would like
to attach to a 56k leased line.  Does anyone make an async CSU/DSU
that has compression built in?

-- 
   Commercial Internet Advertising, Marketing and Consulting
Jon Zeeff	Branch Information Services	[email protected]
(313) 741-4442 	http://branch.com:1080/		gopher branch.com

-----------[000261][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 94 10:39:08 EST
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   TCP-IP Packages for PC's (Again)

IG>I'm had it woth FTP's PCTCP and the ~70K of low memory it requires.  Now
IG>I'm looking for a replacement.  I'll need all the standards (ftp,telnet,
IG>ping,...) but I need to preserve low memory.  I might even decide to go
IG>with an X-terminal package if the proce war right.

Try Ipswitch's Piper/IP.  It can take as little as 6K real-mode memory.
(617) 246-1150.

-- Bob Stein

--
==============================================================================
| ... The Galacticomm Demo System - 305/583-7808 - Home of The Major BBS ... |
==============================================================================


-----------[000262][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 06:12:28 GMT
From:      [email protected] (Epsilon 1993 Pty Ltd)
To:        comp.protocols.tcp-ip
Subject:   UPS Monitoring Software
We are trying to contact companies which supply UPS
Monitoring Software, Relay Contact Splitters and products
for monitoring the UPS via SNMP.

We Know of "Systems Enhancement Corporation" in the US
but would like to get in contact with other suppliers

Thank you in advance for any responses

                Edd

+---------------------+--------------------------------+
| Edward Hall         | Fax:   +61-2-674-3897          |
| Technical Support   | Email: [email protected] |
+---------------------+--------------------------------+


-----------[000263][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 13:25:40 -0400
From:      [email protected] (Tom Limoncelli)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In <[email protected]> [email protected] (Rob Tanner) writes:

>I know zip about IPX, but it looks like we are going to be installing
>a Novell server anyway.  The Novell, of course, is for the PCs.  What
>concerns me as that we also have a number of Unix machines as well as
>an even larger number of X-stations that don't run IPX and don't
>concern themselves with the Novell server.

IPX is a chatty protocol.  I don't have a lot of experience with it,
but I do have experience with AppleTalk, which is equally chatty.

Since you have two very disjoint sets of machines (i.e. two distinct
groups of machines that don't talk to each other much), why put them on
the same subnet?  If your wiring plant permits it (i.e. a 10Base-T star
configuration) why not put the Novell and PC machines on one subnet,
the Unix boxes and X Terminals on a different subnet, then put a router
between the two networks.

--tal
-- 
        Tom Limoncelli -- [email protected] (work) -- [email protected] (play)
  "Finally some good news!  The electrical fires caused by the cable cuts that
    precipitated the total power outage have been extinguished by the flood!"

-----------[000264][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 08:48:18 GMT
From:      [email protected] (Michael Salmon)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <[email protected]>
[email protected] (Eric Scouten) writes:
|> In article <[email protected]>
|> [email protected] wrote:
|> 
|> > Hello all!  I am doing some programming in TCP/UDP, and I want to confirm,
|> > or dispell something that was told to me.  From my question, you will
|> > see that I am a new comer to tcp/ip programming... :)
|> > 
|> > I was told that when using UDP (DGRAM), it doesn't make sure that the
|> > information arrives in order, or that it will arrive at all, but if it
|> > does arrive, IT IS THE CORRECT DATA THAT WAS SENT.
|> > 
|> > I was also told that when using TCP (STREAM), it makes sure that the data 
|> > arrives in the right order, and if it doesn't arrive, you will be notified,
|> > etc...  but IT DOES ***NOT*** MEAN THAT THE DATA THAT WAS SENT IS THE DATA
|> > YOU GET!
|> 
|> I'm not sure where you got that idea. TCP enforces data validity as well
|> as data sequence on the receiving end (i.e., the data you receive will be
|> the same data in the same order as it was sent). This means delays may
|> occur while the TCP daemons request retransmission. Notifications occur
|> only when the network completely fails to deliver data.
|> 
|> 
|> I think both TCP and UDP contain checksum-type information to enforce data
|> validity.

My understanding is that both can have checksums, in TCP it is
mandatory to use them but not in UDP (I believe that the default is not
to use them). So in fact it is the opposite to the original posters
idea, TCP more nearly ensures correct data than UDP.

-- 

Michael Salmon

#include	<standard.disclaimer>
#include	<witty.saying>
#include	<fancy.pseudo.graphics>

Ericsson Telecom AB
Stockholm

-----------[000265][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 94 14:13:31
From:      [email protected] (William )
To:        comp.protocols.tcp-ip
Subject:   LPD implementations
The RFC describing the LPD protocol (RFC1179) implies that it is preferred
to send the control file before sending the data files.  Does anyone know
which unix vendors ship with an lpd support that actually does this?  The
systems we have here all seem to send the data file first...

Thanks
Bill W

-----------[000266][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 18:15:16 -0400
From:      [email protected] (Jay Ashworth)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
[email protected] (Vadim Antonov) writes:

>John Levine ([email protected]) wrote:
 
>: I'll be running BSDI's BSD/360 with full source...
 
>Pray tell me where you got one!  I happened to have an ancient
>IBM/360 and it would be way COOL to run DNS on it.
 
>:-)

ROTFL!

I'll answer your question here, John, since it looks like none of the
remaining 2 articles will...

If I correctly understand all the pieces (based on my reading of about
6 of the Nutshell books), in order to answer to more than one IP
address with one MAC layer address, you'd need to ARP all of them...
and I don't believe ARP provides for this.  Now, if your primary link
is over a point to point link, it's may be a different story, but this
would still require you to be able to ifconfig the interface with
multiple simultaneous IP addresses... and I haven't seen anything like
this mentioned anywhere--and I read a _lot_.  :-)

Cheers,
-- jra
-- 
Jay R. Ashworth                                                       Ashworth
Designer                                                          & Associates
ka1fjx/4              High Technology Systems Consulting
[email protected]                                                +1 813 790 7592

-----------[000267][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 08:39:21 +0200
From:      [email protected] (Willy Janssen)
To:        comp.protocols.tcp-ip
Subject:   Looking for automated DNS update tool.
As we have to do a relevant number of DNS updates, I was wondering if there is
an automated tool around to include the update-info in DNS.

In first instance I'm thinking of automated generation of the reverse map, serial
numbers etc. At second there should be a syntax check on the input files to detect
all kind of typo's, and other errors.

Please let me now.

--

Willy Janssen - University of Nijmegen - Holland

Email: [email protected]


-----------[000268][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 12:18:10 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Re: Cross Platform RPC
In <[email protected]>, [email protected] writes:
>
>I would appreciate any comments/thoughts on the current state of RPC on
>the various *ix platforms (HP/SUN/IBM).  I spent significant time 3 years 
>ago looking at RPC and portability issues.  In the end the company that I
>worked with at the time went with Netwise RPC.  The Netwise package was
>fairly good.  I now need to rethink the cross platform communications issue
>for a new product and am leaning towards using the standard RPC libraries
>that come on the various platforms.  Some questions I have are: 
>
>Do most or all vendor provided RPC packages use XDR?  If not what are the
>various vendors using for the underlying data portability and how do they
>deal with a multi-platform environment?
>
>Are the vendor provided RPC packages fairly bug free.  When I evaluated 
>the Sun RPC several years ago it had quite a few bugs.
>
>Finally, what are the standards bodies doing with RPC, is there a universally
>accepted RPC standard in place or on the horizon.
>
>Thanks in advance for any comments you may have.
>
>Fred Covely.
>
>
Fred,

A couple of things to look at or think about.   The first is DCE based RPC's
this is the standard I believe you are looking for.   Problem here may be
availability on the variety of OS's that you want to provide the function.
A second issue is cost.   DCE is an OSF based solution so it requires
the DCE software on all the machines.

The SUN/ONC RPC is stable.   An issue here is do you want to go off
of UNIX to other platforms( Specifically Windows)  From a Windows 
perspective each TCP/IP vendor provides a different RPC API
that is unique to Windows.   There is a possible solution for this
contact Noblenet at 508-460-8222.   The provide a WINRPC that is
compatable with the UNIX RPC and it sits on WINSOCK.   They have
a compiler "similar" to RPCGEN that makes portability easier.
They have a developer license cost with no runtime fees.
Disclaimer: Not a product endorsement.


If you need further info drop me a note.

Jim Sliwa

--- When the snake and the mongoose meet the mongoose prevails
---     For it is the jaws with which we speak 
---     not the venom with which we say it
---     that determines our strength!




-----------[000269][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 12:49:19 GMT
From:      [email protected] (Brian Alan Weaver)
To:        comp.protocols.tcp-ip
Subject:   Re: Solaris 2.3 SO_KEEPALIVE - HELP

In article <[email protected]>, [email protected] (Phil Howard) writes:
>
>[email protected] (Tony DeFeo) writes:
>
>>I am trying to detect a physically broken socket connection between two
>>Sun workstations running Solaris 2.3. I have set the SO_KEEPALIVE option 
>>on the socket using seteockopt() which sounds like it should do the trick
>>according to the man pages, but doesn't seem to be working.
>
>That's what it keepalive should do.  It's obviously misnamed.  A better
>name might be "killmeif".
>
>
>>What happens is, after I establish the socket connection between the two
>>workstations, I physically disconnect one of them from the network. 
>>I wait several (up to 30) minutes without sending any data, and neither
>
>There might be some timer to set, too.  But I would think it to be
>unreasonable to have it at or above 30 minutes.

I ran into this problem when I was helping to write an EUI interface which
connected to an information server at IBM.  We tried to use the SO_KEEPALIVE
option but after lots of research (books and test), discovered that it has a 
timeout of about 2 hours before it will discover the broken socket. I think
I read this in a book by Stevens (can't remeber much more).  Anyway our
approach to this was to have a "handshake" protocol such that at regular
intervals data was sent by the server/client to determine the status of
the connection.

>
>
>>side detects that the other has gone away (I don't see any TX/RX activity
>>on my transceiver either, I would expect to see activity from the SO_KEEPALIVE
>>messages if it was working as advertised). Next I wrote some data on one side
>>and waited again.  After about 3 minutes, the side that I wrote the data on
>>detected the broken connection and shut down it's end of the socket properly.
>
>But apparently not because of SO_KEEPALIVE.
>
>
>>broken sockets? Optimally, what I would like is for both ends to time out
>>and close the socket even though I am writing no data in either direction.
>
>I would have just written some data.  But if you have to conform to an
>existing protocol, this is probably not an option.
>-- 
-- 
 _______________________________________________________________________
|				    					|
|	[email protected]						|
|									|
|	    _	/|	#include <Bill.The.Cat> 			|
|	    \'o.O'	#define  ACK   Pbbbbbbbttttt............	|
|	    =(___)=							|
|	       U		                           		|
|_______________________________________________________________________|

-----------[000270][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 12:51:29 GMT
From:      [email protected] (W. Richard Stevens)
To:        comp.protocols.tcp-ip
Subject:   Re: Modified TCP Congestion Avoidance Algorithm
> Can anyone tell me, where I can find the above article written by
> V. Jacobson, 1990?

The archives of the end2end-interest mailing list were recently made
available at ftp://ftp.isi.edu/end2end.  The 1990 archive is there,
and the referenced article is in that archive.

	Rich Stevens

-----------[000271][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 14:17:15 GMT
From:      [email protected] (Chris van der Merwe)
To:        comp.protocols.tcp-ip
Subject:   UDP port 520???
Hi!

Every time I do a name resolve request to out local name server I get a UDP 
datagram for port 520.

Can anybody tell me what this is??  As far as i can see, its something about 
a local routing thing.

THX!

-----------[000272][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 14:32:23 GMT
From:      [email protected] (Rob Tanner)
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Running IPX and TCP/IP over same ethernet
Direct from Rumor Control Central:  I have heard that IPX is a real
bandwidth hog, and that if you don't want to see your TCP/IP severely
degrade, don't let IPX run over the same segments that run TCP/IP.

I know zip about IPX, but it looks like we are going to be installing
a Novell server anyway.  The Novell, of course, is for the PCs.  What
concerns me as that we also have a number of Unix machines as well as
an even larger number of X-stations that don't run IPX and don't
concern themselves with the Novell server.

Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
I could isolate the IPX traffic, but to do that requires purchasing
hardware and that means money that has to be justified.  Does anybody
have any real performance figures that they could share.

Thanks,
Rob

      _ _ _ _           _    _ _ _ _ _  
     /\_\_\_\_\        /\_\ /\_\_\_\_\_\  
    /\/_/_/_/_/       /\/_/ \/_/_/_/_/_/  Robert J. Tanner
   /\/_/__\/_/ __    /\/_/    /\/_/       Ames Research Center
  /\/_/_/_/_/ /\_\  /\/_/    /\/_/        (415) 604-3451 (SETI)
 /\/_/ \/_/  /\/_/_/\/_/    /\/_/         (415) 604-5347 (Kuiper)
 \/_/  \/_/  \/_/_/_/_/     \/_/          [email protected]
 ____________________________________________________________________

-----------[000273][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 16:15:39 GMT
From:      [email protected] (Marcus J Ranum)
To:        comp.protocols.tcp-ip
Subject:   Re: How to get pass fire well?
>Most firewalls are rather useless -- they are based on the premise that
>somebody controls _all_ interconnection points to "outside". Large
>organizations that want firewalls almost never can guarantee this.

	Nope, but the problem is that large organizations even more
rarely can guarantee or maintain host security.

	Firewalls are not a panacea, and they're certainly not a
replacement for good or even mediocre host security. The problem is
that very very few people have good or even mediocre host security,
yet they connect their networks to untrusted networks and are
*surprised* when they have problems.

>Simply put, there is no substitute for securing information on secure
>machines, trying to isolate large internal networks often just gives a
>false sense of security.

	Right.

	But there's very few solutions that are more expensive than
securing a large network with adequate host based security. It just
doesn't work, unless you have the funding to have N systems admins
run and fix N^2 copies of sendmail on X different vendor's platforms
every 3 months or so.

	A firewall's not a solution, but it may help get a handle
on the problem. In practical terms, the two should be combined.
Often they are not, and the results can be ugly.

>  in particular because if one keeps sensitive information on unsecure
>  isolated hosts, if the isolation is breached, and it is as a rule
>  breached *behind* the firewall, catastrophe happens, as lots and lots
>  of hosts become easy to pick...
>
>"Firewall" is nearly just a cute moniker, it's no substitute for a
>serious information security setup.

	A firewall is a *PART* of a "serious information security setup"
as far as many of us are concerned.

mjr.

-----------[000274][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 18:14:06 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Triangular Network
Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]


               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]


-----------[000275][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 18:17:47 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Triangular Network
Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]





               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]


-----------[000276][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 18:44:42 GMT
From:      [email protected] (Jim Mercer)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]>, Shawn Curtis <[email protected]> wrote:
>There are no RFCs specifically for TN5250.

a simple "grep 5250 rfc-index" produces:
1205  Chmielewski, P.  5250 Telnet interface.  1991 February; 12 p. (Format:

RFC 1205 says:
   This RFC is being distributed in order to document the interface to
   the IBM 5250 Telnet implementation.  This information is being
   provided for hosts on the Internet that want to support the 5250 work
   station data stream within the Telnet protocol.  This memo provides
   information for the Internet community.  It does not specify any
   standard.  Distribution of this memo is unlimited.

as well as:
   This RFC describes the interface to the IBM 5250 Telnet
   implementation.  The purpose of this memo is to describe the details
   of the interface so that a person wanting to implement a client
   Telnet which emulates an IBM 5250 work station would be able to do
   so.  This memo does not describe all of the 5250 commands, aid codes,
   and other information specific to the 5250 data stream.  That
   information is contained in the IBM 5250 Information Display System,
   Functions Reference Manual, IBM publication number SA21-9247.
   Corrections and additions to this manual are documented in this RFC
   in section 5.

-- 
[ Jim Mercer   Reptilian Research  [email protected]  +1 416 506-0654 ]
[      "'Tis better to clutch a glowing squid than curse the darkness."       ]
[                                                     Archie McPhee Catalog   ]

-----------[000277][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 20:19:19 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Re: Triangular Network
Thanapan Ittisakulchai - SCCS - 3670506 ([email protected]) wrote:
: Hi all..
 
:    Now I am studying about feasibility to implement Triangular Network
: (Senior Project).


: Triangular Network Model.
: -------------------------
 
:    [A]  = Represent Network A
:    ---- = Represent Link Between [A]---[B]





:                [A]                         [A] ------------ [E]
:              /     \      modify to      /     \           /   \
:            /         \    --------->   /         \       /       \
:          /             \             /             \   /           \
:        [B]-------------[C]         [B]--------------[C]------------[F]
:                                      \             /
:                                        \         /
:                                          \     /
:                                            [D]
 
:       Figure 1.                         Figure 2.
:       ---------                         ---------
:       Triangular Network Model          Triangular Network Model
:       consist of 3 networks.            when add node[D],[E],[F]
 
: Scope of Project.
: -----------------
:       1. This system support fault tolerant.
:          (Route for new path and switch when old path is fail.)
 
:       2. Support for data real time query and update.
:          (Such as [A] can query and update database on Triangular Network.)
 
:       3. Develop the system on UNIX based machine.
:          (My advisor selected SCO-UNIX.)
 
: Please suggested me :
: ---------------------
:       1. About feasibility in Technically.
 
:       2. Suitable ? (If base on SCO-UNIX.)
 
:       3. What site can I get information?
 
:                                     Thank you for your advance,
 
:                                     Thanapan Ittisakuclahi
:                                     [email protected]


-----------[000278][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 20:25:08 GMT
From:      [email protected] (Quantum Books)
To:        comp.protocols.tcp-ip
Subject:   Stallings Talk
Quantum Books will host a luncheon talk on PGP by internationally
acclaimed author Bill Stallings at 12:30 p.m.  Thursday, September 22.
The talk PGP: A Peek Under the Hood will last approximately 30 minutes
and provide an overview on the internals of PGP, the e-mail privacy and
digital signature application for the masses. Bill Stallings, author of
the recently published Network and Internetwork Security and of a
forthcoming guide on PGP will also be available to sign copies of his
book. The talk is free and a light lunch will be provided. Seating is
limited so a reservation is required. RSVP [email protected]
-- 
Quantum Books		    | A Technical and Professional Bookstore   
----------------------------+------------------------------------------
Cambridge: 617-494-5042     | E-Mail:		 [email protected]
Philadelphia: 215-222-0611  | Mailing List:      [email protected]

-----------[000279][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 20:28:57 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Re: Triangular Network
Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]






               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]


-----------[000280][next][prev][last][first]----------------------------------------------------
Date:      13 Sep 1994 20:32:33 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Re: Triangular Network
From mars.mahidol.ac.th!mucc!usctit Wed Sep 14 03:28:15 1994
Path: mars.mahidol.ac.th!mucc!usctit
From: [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Triangular Network
Date: 13 Sep 1994 20:28:57 GMT
Organization: Mahidol University, Thailand.
Lines: 56
Message-ID: <[email protected]>
NNTP-Posting-Host: 202.14.162.1
X-Newsreader: TIN [version 1.2 PL2]

Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]








               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]




-----------[000281][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 21:47:07 GMT
From:      [email protected] (Kasey Hohenbrink)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Global list of TCPIP Vendors w/contact numbers

We are looking for a list of TCPIP vendors with
support/sales telephone/internet addresses.

Our software can be run from Xserver packages
using a wide variety of TCPIP.  Our tech support
needs a list of vendor names and numbers so that
we can interface with the vendors for our cust.

I would like anyone with such a list to email
me directly with it, if it exists.  If any of
you are TCPIP vendors, please email me with
any info you can....do not post to News, this
request needs to be filled quickly, and I would
not like to jam the News with postings.


Thank you in advance..

[email protected]

-----------[000282][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 22:15:22 GMT
From:      [email protected] (Charles Yamasaki)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
In <[email protected]> [email protected] (John Levine) writes:

>I have three domains, call them a.com, b.com, and c.com.  But I only have one 
>(maybe two, if I scrounge) server machine.  Is there some way that I could 
>have three names with three IP addresses for the single machine, so I can 
>distinguish FTP and, particularly, WWW connections for the three domains?  
>That is, I'd like http://a.com/ and http://b.com/ to show different pages, 
>even though they're served from the same host.
 
>I'll be running BSDI's BSD/360 with full source, and I have plenty of free 
>addresses in my subnet.  If it would make the software easier, I could plug 
>in three separate Ethernet cards, since cards are cheap and I have enough 
>slots.  Any wisdom would be appreciated.

I don't know of any way to have multiple addresses on the same
interface, but you could certainly plug in multiple NIC's.  The problem
here is that you need multiple router ports (maybe).

Come to think of it, you could actually have something like:

                             Your server
    +----------+           +------------------+
    | Internet | <------>  | NIC1         NIC2| 
    |  Router  |           |              NIC3|
    +----------+           +------------------+

And have the addresses for NIC2 and NIC3 be b.com and c.com.  NIC2 and
NIC3 would not even need to connect to anything at all since they don't
need to talk with anything on their networks, which don't even really
exist.  They do all their routing through NIC1 to the router.  In fact,
it should even work with a dummy interface driver is there is one.

As far as distinguishing Web connections, nope.  You're going to have
one server on this machine listening to port 80 on all three interfaces.
I don't know of any way around that.  Now you could start a different
server on another port, but that's an ugly solution.

Actually, the approach I would recommend would be to have www.b.com be a
canonical name for your a.com server and do the same with www.c.com. 
Then I'd tell users to connect to:

http://www.b.com/b/HomePage.html

        or

http://www.c.com/c/HomePage.html

and if you do have an http://www.a.com/index.html make it a reference to
any of the three home pages.

Hope that helps.
-- 
-----------------------+---------------------------------------------------
Charles "Chip" Yamasaki| The opinions expressed here are my own and are not
[email protected]     | supported or even generally accepted by OSHA. :-)
-----------------------+---------------------------------------------------

-----------[000283][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 22:37:22 GMT
From:      [email protected] (Ron Nadel)
To:        comp.protocols.tcp-ip
Subject:   Re: DNS: How to Advertise New Domain
In article <[email protected]> [email protected] (Jeffrey Horn) writes:

>I am setting up a Unix host for a company that recently was connected to
>the internet.  I can access nearly everywhere in the outside world from
>withing the company (mail, ftp, telnet, etc.), but can only access
>addresses within the company by using actual numbers rather than machine
>names and domain names.
 
>How can I tell the DNS servers on the internet about a new domain?  The
>connection is being done through netcom if that is of any importance.

You will need to have your provider coordinate their nameservice with yours so 
that you will appear on the "other side" - that is the Internet.  The root 
servers need to know you exist.

Someone else mentioned that you should  get the DNS/BIND book, but that is no 
answer to your question.  Those books will tell you how to configure your own 
DNS.

Ron

-----------[000284][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 23:12:46 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <[email protected]> [email protected] (Michael Salmon) writes:

> ...
>My understanding is that both can have checksums, in TCP it is
>mandatory to use them but not in UDP (I believe that the default is not
>to use them). So in fact it is the opposite to the original posters
>idea, TCP more nearly ensures correct data than UDP.

Your understanding is partly wrong.

    -TCP checksums are required by the TCP standard, but no one will
	come to take you away if you do not use them.  A few people
	(misguided in my view) do suppress TCP checksums.

    -UDP checksums are optional, but all reasonable systems now use them.

    -it does not make sense to talk about the "the default" with respect
	to checksums in the UDP protocol, although you can talk about
	the default use of UDP checksums in various implementations.

    -at one time, many implementations did not use UDP checksums.

    -that time was long ago.

    -of course, there are still many old systems out there with UDP checksums
	turned off.


Vernon Schryver    [email protected]

-----------[000285][next][prev][last][first]----------------------------------------------------
Date:      Tue, 13 Sep 1994 23:41:54 GMT
From:      [email protected] (Tom Fitzgerald)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
[email protected] (Samuel Lam) writes:

>  1) Is sending outgoing ARP replies via MAC-level broadcasts
>     instead of unicasts allowed?
 
>  2) Is discarding incoming ARP replies arriving via MAC-level
>     broadcasts allowed?
 
> You guessed it!  We have two vendors' routers refusing to talk
> to each other because they each implements one of the above
> "features".

I think they're both wrong.  RFC 826 gives the algorithm for ARP
processing, and by intent it has two properties that vendor #2 is
disobeying:

- The system receiving the packet must enter the IP:MAC mapping in its ARP
   table before even checking to see whether it's a request or reply.  In
   fact, the *only* distinction between the request and reply is that a
   reply doesn't require another reply to be sent in response.  Both the
   request and reply have the same side-effects on the local system.
- Nowhere is the receiver told to examine the destination MAC address.

If vendor #2 correctly handles ARP requests broadcast, then it must handle
ARP responses broadcast as well.

But vendor #1 is disobeying the last part of the protocol, the generation
of the response packet:

| 	Swap hardware and protocol fields, putting the local
| 	    hardware and protocol addresses in the sender fields.
| 	Set the ar$op field to ares_op$REPLY
| 	Send the packet to the (new) target hardware address on
| 	    the same hardware on which the request was received.

"New target hardware address" does not mean broadcast address.

There's a later note that further incriminates vendor #2:

| When a monitor receives an Address Resolution packet, it always
| enters the <protocol type, sender protocol address, sender
| hardware address> in a table.

Not only is the MAC destination address ignored, but the opcode (request or
reply) isn't used in this decision either.

-- 
Tom Fitzgerald      1-508-967-5278
Wang Labs           [email protected]         (account no longer active)
Lowell MA, USA      [email protected]       (use this address instead)

-----------[000286][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 07:02:32 -0400
From:      [email protected] (Sam Martin)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Robert J. Wilson) writes:


>> Direct from Rumor Control Central:  I have heard that IPX is a real
>> bandwidth hog, and that if you don't want to see your TCP/IP severely
>> degrade, don't let IPX run over the same segments that run TCP/IP.

We find that ipx and tcpip get along well. Cinfig is 3 Sun Sparc10, 1 NW_311,
lot's of traffic to a fibre backbone.
 
>Otherwise, IPX is fairly efficient. The problem is the absence of
>local-file- cache on the PCs so they won't keep downloading the same
>programs to the PCs time-and-time again. There is a "key-server" technique
>that lets you distribute local copies of the code to each PC and gives a
>lot of relief.

Key Server Technique: Sounds interesting, what does it mean?
Regards, Sam

-----------[000287][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 00:31:34 GMT
From:      [email protected] (Dan Thurman)
To:        comp.protocols.tcp-ip
Subject:   Need sources to "proxy-arp"...
	Hi Folks...

	Does anyone know where I can find sources for "proxy-arp"?  I heard
	someone at AT&T had written one up and is in the RFC's somewhere...

	Please send me e-mail, if possible.

	Thanks in advance.

---
---------------------------_----------------------------------------------
Dan Thurman            __(o o)__    EMAIL1: [email protected] (work)
Atlas Telecom          |w| U |w|    EMAIL2: [email protected]
4640 SW Macadam Ave.   | | | | |    VOICE: [USA] 1+503.228.1400 x251
Portland, OR 97201     | | | | |    FAX:   [USA] 1+503.228.0368
--------------------------------------------------------------------------



-----------[000288][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 00:55:18 GMT
From:      [email protected] (Glenn Fleishman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Proxies and multiple addresses
I checked this out again.

The CERN httpd will let you run -- without a proxy -- redirects to
multiple domains addressed as the same location.

So if you have domain1.com, domain2.com, domain3.com CNAME'd or IN A'd
to the same actual IP address, using the CERN httpd redirect commands
in the all.conf file, you can have a user specifying
http://www.domain1.com go to a direct home page than
http://www.domain2.com.

If you want more info on this, read the docs on the daemon, or send
mail to my business partner [email protected] (Todd Haedrich) who is
setting this up and can provide some additional details to interested
parties.

----
Glenn Fleishman, Point of Presence Company, Seattle, Washington
        Send to <[email protected]> for information on our services
        Moderator, Internet Marketing discussion list; send the message
        INFO INET-MARKETING to the address [email protected] for info

-----------[000289][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 01:00:09 GMT
From:      [email protected] (Robertino Benis)
To:        comp.protocols.tcp-ip
Subject:   ACER 5220A driver?
Hi!     
Has anyone met acer 5220A & packet driver for this eth. card?
Tnx.
robertino


-----------[000290][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 10:01:26 -0500
From:      [email protected] (Phil Howard)
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
[email protected] (Rob Tanner) writes:

>Direct from Rumor Control Central:  I have heard that IPX is a real
>bandwidth hog, and that if you don't want to see your TCP/IP severely
>degrade, don't let IPX run over the same segments that run TCP/IP.

Maybe.  IPX waits for acks for every packet, for instance.


>Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
>I could isolate the IPX traffic, but to do that requires purchasing
>hardware and that means money that has to be justified.  Does anybody
>have any real performance figures that they could share.

What I do know from seeing the effect from our own Novell server, watching
it from Unix, is that periodically the Novell box trashes the network for
about 2 to 4 seconds with zillions of broadcasts.  Also, Novell's TCP/IP
behaves very badly (it might also be non-conformant, but basic communications
with it from Unix does work and it usually works out as a router if you
don't stress it with too many packets over 1472 bytes).
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000291][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 01:51:36 GMT
From:      [email protected] (Christopher Davis)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
JA> == Jay Ashworth <[email protected]>

 JA> in order to answer to more than one IP address with one MAC layer
 JA> address, you'd need to ARP all of them...  and I don't believe ARP
 JA> provides for this.

Sure it does.  Proxy ARP, for example.  The Telebit Netblazer can be
configured to associate specific IP addresses with specific ports, so
that, say, "telnet blazer-port20" would get you a connection to whatever's
on serial port 20, such as a soda machine readout or whatever.

 JA> [...] this would still require you to be able to ifconfig the
 JA> interface with multiple simultaneous IP addresses... and I haven't
 JA> seen anything like this mentioned anywhere--and I read a _lot_.  :-)

I believe I've seen mention that it was possible under BSDI BSD/386 1.1
(which the original poster was running) using the "alias" parameter to
ifconfig.  I haven't yet tested this, as our BSDI box is busy doing more
important things that I really don't want to break...
-- 
Christopher Davis * <[email protected]> * (was <[email protected]>) * MIME * PGP * [CKD1]
  "It's 106 ms to Chicago, we've got a full disk of GIFs, half a meg of
     hypertext, it's dark, and we're wearing sunglasses."  "Click it."
      - Looking for: _The Big U_, by Neal Stephenson (out of print) -

-----------[000292][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 01:52:34 GMT
From:      [email protected] (Travis L Priest)
To:        comp.protocols.tcp-ip
Subject:   Re: LPD implementations
In article <[email protected]> [email protected] (William ) writes:

   From: [email protected] (William )
   Newsgroups: comp.protocols.tcp-ip

   The RFC describing the LPD protocol (RFC1179) implies that it is
   preferred to send the control file before sending the data files.
   Does anyone know which unix vendors ship with an lpd support that
   actually does this?  The systems we have here all seem to send the
   data file first...

I would be surprised if any vendors shipped lpds that sent the cf file
first (offhand, SGI, Sun (OS4.x), and Cray do not).  I suspect that
most vendors picked up the UCB source at some point and 'ported' it to
their systems without (m)any changes to the queueing procdure.  Since
the UCB lpd does not begin to process a job until the data connection
with the remote client is closed, it shouldn't matter if the chicken
or the egg comes first.

Travis

-----------[000293][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 00:58:56 +0100
From:      Simon <[email protected]>
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In <[email protected]> [email protected] (Rob Tanner) writes:

>Direct from Rumor Control Central:  I have heard that IPX is a real
>bandwidth hog, and that if you don't want to see your TCP/IP severely
>degrade, don't let IPX run over the same segments that run TCP/IP.
 
>Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
>I could isolate the IPX traffic, but to do that requires purchasing
>hardware and that means money that has to be justified.  Does anybody
>have any real performance figures that they could share.

Well, I found that I could bring our IPX network at work to a halt (ie wait
for 5 or so minutes to list a directory) by installing Solaris on a remote
machine. Okay, so this is kind of an extreme example.
I would say that when it comes down to it, an ethernet packet is an ethernet
packet (big surprise there :). It *really* depends on the apps you are
running. For example, the uni here banned people playing Doom on their LANs
as it had a tendancy to bring the network servers down.
But you are also right, IPX apps tend to fire a large number of packets down
the ethernet, which increases the network load. But if you don't have many
machines running IPX, it shouldn't be *much* of a problem.

--
Simon

-----------[000294][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 03:55:08 GMT
From:      [email protected] (James VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
[email protected] (Samuel Lam) writes:

> 1) Is sending outgoing ARP replies via MAC-level broadcasts
>    instead of unicasts allowed?

I agree with the other post analyzing RFC 826, but let's consider the
effects: Vendor 1's router distracts every host on the net with every
ARP reply, which may in the end reduce overall load on the router
while making many hosts' ARP caches incrementally larger.

> 2) Is discarding incoming ARP replies arriving via MAC-level
>    broadcasts allowed?

Vendor 2 put its MAC broadcast check further down the food chain than
they ought.  Again, the router looks a little faster, but it isn't
"liberal in what it accepts from others".
--
James B. VanBokkelen					Far Acres Farm
[email protected]{vax.ftp.com, asylum.sf.ca.us}			South Hampton, NH

-----------[000295][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 04:09:02 GMT
From:      [email protected] (James VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] TCP/IP MIL-STD
[email protected] (Ran Atkinson) writes:

>In article <[email protected]> [email protected] (Ran Atkinson) writes:
>>>RFCs.  Smart folks look first at the relevant RFCs, including Host
>>>Requirements, and then at the BSD networking code.
 
>In article <[email protected]> [email protected] (Alan Cox) 
>writes:
>>I'd strongly suggest people treat the BSD networking code with caution -
>>especially on the matter of URGent data and ICMP error handling. RFC1122 and
>>the BSD code are very different.
 
>Yes.  However, most of the installed base follows BSD conventions and
>violates RFC-1122.  One has to pay attention to the BSD code if one
>wants interoperability.

But as the individual posting re: "OOB data" has discovered, what 4bsd
actually implemented isn't as useful as a sockets programming manual
might lead you to hope.  While it is possible to design an API to TCP
that lets you use either RFC 1122 or bsd Urgent, you have to chose
which you'll use when you design the application protocol.
Furthermore, if any instances of your application are actually using
pure 4bsd sockets, you're still open to being done in by TCP
retransmission timing.  You can only avoid this by making the
application layer protocol strict lockstep...
--
James B. VanBokkelen					Far Acres Farm
[email protected]{vax.ftp.com, asylum.sf.ca.us}			South Hampton, NH

-----------[000296][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 04:13:42 GMT
From:      [email protected] (James VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP? That is the question.
[email protected] (Vernon Schryver) writes:

>    -TCP checksums are required by the TCP standard, but no one will
>	come to take you away if you do not use them.  A few people
>	(misguided in my view) do suppress TCP checksums.

A clarification: TCP as written up in the standards doesn't have
anything that corresponds to UDP's "don't bother to checksum this
packet" magic cookie.  Unless you build into your TCP some proprietary
mechanism of recognizing hosts or protocols that don't bother to
checksum incoming packets, unchecksummed packets will not interoperate.
--
James B. VanBokkelen					Far Acres Farm
[email protected]{vax.ftp.com, asylum.sf.ca.us}			South Hampton, NH

-----------[000297][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 04:26:42 GMT
From:      [email protected] (Russo and Hale)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Scripts and FreeBSD route
I have configured my FreeBSD router floppy (thanks to those who helped), 
but I can't get a tip session or script for the dialup.  My carrier 
requires a login and password.  How do I set this up?

Thanks in advance, again.
Michael Risch.

-----------[000298][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 94 06:16:03 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Cheap IP ethernet-to-asynch techniques?
Hi there.  Can anyone out there suggest really cheap ways that might be
used to connect two disjoint IP networks via Asynch modems?

We have two ethernets in two separate facilities that need are IP-connected
via a fractional T1 circuit.  In each location there is a CSU/DSU unit
along with a CISCO router.

What we would like to do is create a backup for this connection using 
as little money as possible, but which is completely independent of all the
T1 equipment.  The best solution would be something involving regular
asynch modems over voice lines.  The question then is: what is the cheapest
way to route IP from an ethernet out to a modem?

            |--- Ethernet             |--- Voice-grade phone line
            v                         v
(IP net #1)++++(black box)----(modem)====(modem)----(black box)++++(IP net #2)
                           ^
                Asynch-----|

Using this picture, then, the question becomes: what cheap solutions are
there for the black boxes?  Obviously commercial routers can do it for
around $2000 and up, but I was hoping for something we could put together
with chewing gum and chicken wire for considerably less.

Any suggestions?

-- John W. Manly  <[email protected]> (System Manager -- Amherst College)


-----------[000299][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 06:50:45 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Re: Triangular Network
From mars.mahidol.ac.th!mucc!usctit Wed Sep 14 03:28:15 1994
Path: mars.mahidol.ac.th!mucc!usctit
From: [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Triangular Network
Date: 13 Sep 1994 20:28:57 GMT
Organization: Mahidol University, Thailand.
Lines: 56
Message-ID: <[email protected]>
NNTP-Posting-Host: 202.14.162.1
X-Newsreader: TIN [version 1.2 PL2]

Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]


	                   TRIANGULAR NETWORK MODEL
                 	   ------------------------


               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]




-----------[000300][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 13:39:15 -0400
From:      [email protected] (Dave Cornejo)
To:        news.announce.newgroups,news.groups,comp.dcom.cell-relay,comp.dcom.isdn,comp.dcom.lans.ethernet,comp.dcom.lans.fddi,comp.dcom.lans.misc,comp.protocols.tcp-ip,bit.listserv.novell
Subject:   2nd CFV: comp.dcom.frame-relay
                     LAST CALL FOR VOTES (of 2)
               unmoderated group comp.dcom.frame-relay

Newsgroups line:
comp.dcom.frame-relay   Technology and issues regarding Frame Relay

Votes must be received by 23:59:59 UTC, 22 September 1994.

This vote is being conducted by a neutral third party. For voting
questions only contact Dave Cornejo <[email protected]>. For questions
about the proposed group please contact Allen Robel <[email protected]>

This CFV is also being sent to the cicso discussion mailing list
[email protected]


CHARTER

To provide a mechanism for the discussion of issues related to the
technology and application of frame relay in communications networks.

Topics expected to be discussed include (but are not limited to) the
following examples:

- frame relay standards and proposed standards
- applications of frame relay
- user experiences with frame relay
- relationship and comparison of frame relay to other technologies,
  including asynchronous transfer mode (ATM), cell relay, SMDS,
  X.25 packet switching, IP routing, ISDN, Broadband ISDN, etc.
- announcements of conferences, seminars, etc., related to frame relay
- news of the availability of frame relay products or services
- testing and interoperability of frame relay
- information related to performance and implementation of frame relay
- research findings
- activities of The Frame Relay Forum and other similar organizations
  with activities related to frame relay
- sources of additional information regarding frame relay

This newsgroup is not intended to be a forum in which standards are
developed, although it may be expected to relate news pertaining to
standards and to the progress of recognized standards bodies.

This list will be gatewayed to the Frame Relay Discussion list
<[email protected]>.


RATIONALE: 
 
Given that Frame Relay is beginning to emerge as a mainstream WAN service,
and given that discussions of Frame Relay technology are appearing with
increasing frequency on other newsgroups such as comp.dcom.cell-relay, it
makes sense at this time to create a newsgroup to provide a home for
discussion of this technology. 


HOW TO VOTE

Send MAIL to:   [email protected]
Just Replying should work if you are not reading this on a mailing
list.

The subject in your mail message must contain the name of the group to
allow your vote to be correctly handled by an automated vote-counter,
if you do not do this you will receive a message asking you to
resubmit your vote with the name of the group that you are voting for
in the subject line.

Please do not quote the entire CFV in your vote, it will confuse the
automatic vote counter and will earn you a duplicate vote message.

Your mail message should contain one of the following statements:
      I vote YES on comp.dcom.frame-relay
      I vote NO on comp.dcom.frame-relay

You may also ABSTAIN in place of YES/NO - this will not affect the
outcome.  Anything else may be rejected by the automatic vote counting
program.  The votetaker will respond to your received ballots with a
personal acknowledgement by mail - if you do not receive one within
several days, try again.  It's your responsibility to make sure your
vote is registered correctly.

Only one vote per person, no more than one vote per account.  No votes
will be accepted from system accounts (i.e. root, news, uucp) or
anonymous accounts.  Addresses and votes of all voters will be
published in the final voting results list.

BOUNCED ACKNOWLEDGEMENTS:

  [email protected]
  [email protected]
  boemia.pix.com.br!nelson
 	
-- 
Dave Cornejo                                There is nothing so subtle
Dogwood Media                                           as the obvious
Fremont, California

-----------[000301][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 07:06:21 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Triangular Network
Subject: Re: Triangular Network
Newsgroups: comp.protocols.tcp-ip
Summary: 
Keywords: 

Hi all..

   Now I am studying about feasibility to implement Triangular Network
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]









               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]



-----------[000302][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 14:31:02 -0400
From:      [email protected] (John W. Manly)
To:        comp.protocols.tcp-ip
Subject:   ARP debugging utilities?
Hi there.  Does anyone out there have, for ANY platform, some tools for
doing ARP-level debugging?  That is, I want to be able to manually 
issue ARP requests and record the responses that come back.  VMS, UNIX,
DOS, MAC -- anything would be good.

-- John W. Manly  <[email protected]> (System Manager -- Amherst College)


-----------[000303][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 14:44:42 -0400
From:      [email protected] (John W. Manly)
To:        comp.protocols.tcp-ip
Subject:   Re: ARP debugging utilities?
John W. Manly ([email protected]) wrote:
> Hi there.  Does anyone out there have, for ANY platform, some tools for
> doing ARP-level debugging?  That is, I want to be able to manually 
> issue ARP requests and record the responses that come back.  VMS, UNIX,
> DOS, MAC -- anything would be good.

While I'm at it, I'm also interested in similar tools for doing BOOTP
debugging.  I would like to be able to issue manual BOOTP queries for
arbitrary ethernet addresses.  

Of course, it occurs to me that this might not be possible if the BOOTP
protocol doesn't make a distinction between the ethernet address the
packet is being sent from and the ethernet address for which the BOOTP
server is to return the information.  On the other hand, maybe BOOTP
is smarter than this and does make a separation in the fields, and it 
just so happens that MOST of the time the source ethernet address and
the ethernet address being looked up are the same.

Finally, is there a TFTP client available for any platforms?  Again, while
trying to debug BOOTP/system-loading kinds of things, I would like to
be able to manually duplicate the TFTP operations that normally take
place automatically.

> -- John W. Manly  <[email protected]> (System Manager -- Amherst College)


-----------[000304][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 08:25:50 GMT
From:      Bill Melotti <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Mac/DEC SLIP problems - Help !
I am trying to establish a SLIP connection from my Mac to a SLIP server
which is actually a DEC terminal server. Normally I have no problems with
internet access on the Mac through Ethernet.

I am using Versaterm SLIP as the MacTCP LAP, and I can connect to the DEC
terminal server no problems and can use telnet via the Versaterm terminal
window, so no modem problems (I think). I then start SLIP on the DEC, and
then on the Mac.

I can then ping hosts for a while before the whole lot just seems to
stop......

Usually I can ping about 30 or so times before it falls over. This is
quicker if I try to ping something outside my subnet, sometimes to the
point where it won't even receive one reply (ie it fell over straight
away)

I asked the Sys Admin guy if he had any ideas. He set up a packet monitor
program, which showed the ICMP going out from the DEC server, but no
reply coming back, yet we can use the ping service of the DEC server no
problem. 

Any ideas anyone ?

Many thanks in advance

******
If you asked Rutherford Lab about this they'd probably deny my existence, 
let alone side with my opinions....

Bill Melotti    [email protected]
** Smaller sigs. save space and bandwidth, and have the added benefit of
  being kind to those people with dismally sad user interfaces **

-----------[000305][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 08:47:13 +0000
From:      [email protected] (Andy Leigh)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected]
(Rob Tanner) writes:

> Direct from Rumor Control Central:  I have heard that IPX is a real
> bandwidth hog, and that if you don't want to see your TCP/IP severely
> degrade, don't let IPX run over the same segments that run TCP/IP.

IPX, in its native form is ineficient because it doesn't do windowing
(it ping-pongs acks - one ack per data packet), and it, historically,
only uses binary sizes for maximimum packet size. e.g 512, 1024, 2048
etc. When passed through a router it defaults to 512 packet sizes.
However, all these things can be fixed by using VLMs (if you're brave)
and 3.12 and 4.01 (etc.) Netware.

BUT you ask about IPX being a bandwidth hog. It *isn't*, and neither is
IP. NCP (Netware Core Protocol) over IPX can certainly be a bandwidth
hog, as can NFS over IP. THe reason for this is that you are effectively
mounting your hard-drive over the wire (simplification alert for
purists). Watch the hard-disk light on a PC and then imagine that
happening over the ethernet, and you get a feel for NCP. In practice, if
all you have is a netware server on collision domain, with all the PC's
being of similar performance, you can achieve truly astonishing
efficiency figures for Ethernet. This is because the Disk seeks on the
Novell server are served in disk-cylinder order (called elevator
seeking). A workstation only gets answered by the server, when the
server head is over the correct cylinder to answer the request. The
result is that the major queue is in the servers disk/memory interface.

Watch, however low-speed WAN links. If you use a routed network, get
some SAP-filter software. Netware servers are always showing off with
broadcasts of their wares. With one server you don't tend to worry. With
300 odd, things can become interesting.

> I know zip about IPX, but it looks like we are going to be installing
> a Novell server anyway.  The Novell, of course, is for the PCs.  What
> concerns me as that we also have a number of Unix machines as well as
> an even larger number of X-stations that don't run IPX and don't
> concern themselves with the Novell server.

IPX and IP cohabit nicely, except that, because NCP is all about disk
mounting, large files are always being transfered, not unlike FTP. This
can lead to latency problems with telnet sessions and with X-Windows
sessions. In other words, a high speed PC talking to a fast server may
be able to use the Ethernet Capture Effect to chain together a large
number of big packets (with acks from the server) and deny a telnet
machine access to the wire for a few hundred milliseconds. The users
will experience a type ahead effect. BUT they would get eaxctly the same
effect if two UNIX hosts were doing an FTP, or NFS session in the same
domain, so this effect is not unique to IPX.

> Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
> I could isolate the IPX traffic, but to do that requires purchasing
> hardware and that means money that has to be justified.  Does anybody
> have any real performance figures that they could share.

On the LAN I administer, 95% of the traffic is Novell-based with, on
average 1000 users on a bridged (with bridged FDDI Backbone) network.
Some collision domains wobble between the 50% to 65% Ethernet usage
marks without customers complaining (to me at least :-) of slowness. We
have about 100 users in that domain who perform occasional Telnet
actions from the same PCs, and only very rarely do we see a type-ahead
effect.

There seems to be a 20-80 rule here. You can either cope with 80% normal
IP and 20% normal IPX usage, or the other way round. If you've already
got lots of IP traffic on a domain and you look like you're going to
require similar ammounts of IPX traffic, be prepared to do some cunning
re-design.

> Thanks,
> Rob

Hop this helps.


--

Andy Leigh               [email protected]  -  Home (preferred)
                         [email protected]    -  Work (RFC822)

-----------[000306][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 08:50:03 GMT
From:      [email protected] (George Ross)
To:        comp.protocols.tcp-ip
Subject:   Re: Solaris 2.3 SO_KEEPALIVE - HELP
In article <[email protected]>, [email protected] (Brian Alan Weaver) writes:
> ... that it has a 
> timeout of about 2 hours before it will discover the broken socket. ...

This is required by RFC 1122 section 4.2.3.6.  You can use "ndd" to tune it.
-- 
George D M Ross, Department of Computer Science, University of Edinburgh
     Kings Buildings, Mayfield Road, Edinburgh, Scotland, EH9 3JZ
Mail: [email protected]      Voice: 031-650 5147      Fax: 031-667 7209

-----------[000307][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 20:35:03 -0700
From:      [email protected] (Parik Rao)
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In <[email protected]> [email protected] (Phil Howard) writes:

>[email protected] (Rob Tanner) writes:
 
>>Direct from Rumor Control Central:  I have heard that IPX is a real
>>bandwidth hog, and that if you don't want to see your TCP/IP severely
>>degrade, don't let IPX run over the same segments that run TCP/IP.
 
>Maybe.  IPX waits for acks for every packet, for instance.

Enhancements by Novell to IPX such as burst mode (4-64 packets at a time,
without an ACK) and large packets make things a bit better.  

For Rob's case, with only one server, its not a big deal.  If you have
200 servers on a enterprise, bigger problem.  Every 60 seconds each
server will broadcast its routes (RIP) and services (SAP).  This can
be quite a number of packets.

Newer versions of IPX routing (NLSP v1.0) reduce this problem too, sending
out packets on a as-needed basis, instead of periodically.

Finally, you can nowadays run Netware over TCP/IP (using Flex/IP and
the VLM client kit) and not bother with IPX.  But then you miss out on
all the IPX fun... ("ok... who the hell misconfigured their servers
IPX network numbers").


-----------[000308][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 17:41:20 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: async 56k CSU/DSU with compression
[email protected] (Jon Zeeff) writes:

>I have a 115k bps async connection (with flow control) that I would like
>to attach to a 56k leased line.  Does anyone make an async CSU/DSU
>that has compression built in?

I just ran across a product update in Network Computing for one of these.
It is a combined unit that is also a bridge/router and includes compression.
It supports T1, frac T1, frame relay, and switched 56.  It does TCP/IP,
IPX/SPX, PPP, SLIP.  The price is shown as $1995.

Eastern Research
60 James Way
Southampton, PA 18966
tel 215-364-7955
fax 215-396-9822
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000309][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 09:08:06 MET
From:      [email protected] (Roger Hunen)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] : TCP and NETWORK UNREACHABLE
In article <[email protected]> [email protected] (Hakan Nohre) writes:
>What would be the expected behaviour by TCP (in ESTABLISHED state) when an
>NETWORK UNREACHABLE is received ??

NETWORK UNREACHABLE is generated by a router, not by an end node. Thus it 
may be that a link went down and routers along the path did not yet update 
their routing tables. After some time routers may start to reroute the packets.
As such NETWORK UNREACHABLE may be a temporary condition.

>Should it give up the connection or retransmit ?

Based on the above it must not give up.

Regards,
-Roger


-----------[000310][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 16:55:44 -0400
From:      [email protected] (Eric V. Smith)
To:        comp.protocols.tcp-ip
Subject:   Re: DHCP server and NT
In article <[email protected]>,
Alan Cox <[email protected]> wrote:
>In article <[email protected]> [email protected] (Eric V. Smith) writes:
>>I'm writing a DHCP server to run under Linux, with the specific 
>>purpose of serving NT Daytona clients (and later, Chicago).  I had 
 
>>by default.  Does anyone know why it doesn't?  Does it inspect the
>>class of the address I give it and assume something from that?
>
>I would hope it looks at the device involved and gets its netmask.
>

How would it do this before the device is configured, which is what
DHCP is all about?

Eric.


-----------[000311][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 11:31:17 GMT
From:      [email protected] (alain martineau)
To:        comp.protocols.tcp-ip
Subject:   Re: DNS: How to Advertise New Domain
In article <[email protected]>, [email protected] (Ron Nadel) writes:
>In article <[email protected]> [email protected] (Jeffrey Horn) writes:
>
>>I am setting up a Unix host for a company that recently was connected to
>>the internet.  I can access nearly everywhere in the outside world from
>>withing the company (mail, ftp, telnet, etc.), but can only access
>>addresses within the company by using actual numbers rather than machine
>>names and domain names.
 
>>How can I tell the DNS servers on the internet about a new domain?  The
>>connection is being done through netcom if that is of any importance.
>
>You will need to have your provider coordinate their nameservice with yours so 
>that you will appear on the "other side" - that is the Internet.  The root 
>servers need to know you exist.
>
>Someone else mentioned that you should  get the DNS/BIND book, but that is no 
>answer to your question.  Those books will tell you how to configure your own 
>DNS.
>
>Ron

No, the book tells how to be registered. So the answer is in the book.
Alain


-----------[000312][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 12:47:28 GMT
From:      [email protected] (Cliff Bedore)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
Rob Tanner ([email protected]) wrote:
: Direct from Rumor Control Central:  I have heard that IPX is a real
: bandwidth hog, and that if you don't want to see your TCP/IP severely
: degrade, don't let IPX run over the same segments that run TCP/IP.
 
: I know zip about IPX, but it looks like we are going to be installing
: a Novell server anyway.  The Novell, of course, is for the PCs.  What
: concerns me as that we also have a number of Unix machines as well as
: an even larger number of X-stations that don't run IPX and don't
: concern themselves with the Novell server.
 
: Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
: I could isolate the IPX traffic, but to do that requires purchasing
: hardware and that means money that has to be justified.  Does anybody
: have any real performance figures that they could share.

As with most things, the answer is, it depends.  If you are on a fairly
localized net, there should be little impact for normal operations.  If you
were using a dBASE application that was constantly doing lookups over the net,
you might see some degradation but otherwise probably not.

However, if you have a very large company with lots of Netware servers, and the
various company groups are connected by slower links, you can see some very
strange results.  I was asked by a company to look at why their software was
running poorly from a remote site to the main system.  The customer had T1
lines connecting the sites so they claimed it was not bandwidth.  (Interactive
terminal sessions are not net hogs.)  I saw a lot of netware traffic but the
customer pulled out a sniffer and showed that the peak bandwidth was 3Mbits/sec
on the ethernet which should not be a problem.  Unfortunately that 3Mbits/sec
was netware broadcasts which went down every router and T1 in the company.
Pushing 3Mbits/sec down a 1.5Mbit/sec T1 caused some serious delay in the
terminal sessions.  This was a large company with lots (>300?) of Netware 
servers that apparently all had to talk to each other.

So for normal use, should not be a problem but it will depend on how you use
it.



: Thanks,
: Rob
 
:       _ _ _ _           _    _ _ _ _ _  
:      /\_\_\_\_\        /\_\ /\_\_\_\_\_\  
:     /\/_/_/_/_/       /\/_/ \/_/_/_/_/_/  Robert J. Tanner
:    /\/_/__\/_/ __    /\/_/    /\/_/       Ames Research Center
:   /\/_/_/_/_/ /\_\  /\/_/    /\/_/        (415) 604-3451 (SETI)
:  /\/_/ \/_/  /\/_/_/\/_/    /\/_/         (415) 604-5347 (Kuiper)
:  \/_/  \/_/  \/_/_/_/_/     \/_/          [email protected]
:  ____________________________________________________________________


Cliff


-----------[000313][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 20:43:18 -0400 (EDT)
From:      [email protected] (Papa Smurf)
To:        comp.sys.mac.misc,comp.protocols.appletalk,comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Getting started w/Mosaic 'n' stuff
I am working as a lab assistant in a computer lab at my school, and I 
would to get a few of the computers here connected to our network with 
Mosaic. I really have no idea what I need to get this accomplished. First 
off, we have a Macintosh Quadra 605 with a built-in Ethernet card. I 
would like to know what software and hardware I'll be needing to connect 
this computer to the network through ethernet. If at all possible 
approximate prices for any extra materials would also be appreciated. I 
would also like to know if it's possible for Mac programs downloaded to a 
PC and copied to floppy by PC to work for a Mac. All extraneous 
information is also welcome, as I am trying to learn as much about this 
as possible. Thanx in advance. Please send responses directly to me 
because I don't have direct access to Usenet. 


-
Aaron Grant
[email protected]

-----------[000314][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 20:58:41 -0400
From:      [email protected] (The Bookpool)
To:        comp.protocols.tcp-ip
Subject:   Tech (TCP/IP) Book Discounts
We offer O'Reilly & Associates publications for a 25 to 35 percent discount.

"TCP/IP Network Administration", Craig Hunt
	Bookpool Price: $21.50    List Price: $29.95

We will stock any book that interests this newsgroup.  Please 
vote for the books that interest you, so that we can buy them in volume 
and offer them to you at a large discount.  We notify all voters when 
the book comes into stock.

For more information, mail a single line "send info" to [email protected]

-------------------------------------------------------------
 The Bookpool                       [email protected]
 233 Park Place, Suite 27           FAX (718) 638-8202
 Brooklyn, NY 11238

-----------[000315][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 14:31:15 GMT
From:      [email protected] (Thanapan Ittisakulchai - SCCS - 3670506)
To:        comp.protocols.tcp-ip
Subject:   Triangular Network
Subject: Re: Triangular Network
Newsgroups: comp.protocols.tcp-ip
Summary: 
Keywords: 

Hi all..

   Now I am learning about feasibility to implement Triangular Network 
(Senior Project).


Triangular Network Model.
-------------------------

   [A]  = Represent Network A
   ---- = Represent Link Between [A]---[B]






               [A]                         [A] ------------ [E]
             /     \      modify to      /     \           /   \
           /         \    --------->   /         \       /       \
         /             \             /             \   /           \
       [B]-------------[C]         [B]--------------[C]------------[F]
                                     \             /
                                       \         /
                                         \     /
                                           [D]

      Figure 1.                         Figure 2.
      ---------                         ---------
      Triangular Network Model          Triangular Network Model
      consist of 3 networks.            when add node[D],[E],[F]

Scope of Project.
-----------------
      1. This system support fault tolerant.
         (Route for new path and switch when old path is fail.)

      2. Support for data real time query and update.
         (Such as [A] can query and update database on Triangular Network.)

      3. Develop the system on UNIX based machine.
         (My advisor selected SCO-UNIX.)

Please suggested me :
---------------------
      1. About feasibility in Technically.

      2. Suitable ? (If base on SCO-UNIX.)

      3. What site can I get information?

                                    Thank you for your advance,

                                    Thanapan Ittisakuclahi
                                    [email protected]



-----------[000316][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 14:34:03 GMT
From:      [email protected] (John Skovron)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]>,
John Schnizlein <[email protected]> wrote:
>In article <[email protected]>, [email protected] (Clark
>Bremer) wrote:
>...
>> >|> >[email protected] (Rich Oldroyd) writes:
>> >|> >|> 
>> >|> >|> We're looking for the  specifications/RFCs covering the
>> >|> >|> TN3270 and/or TN5250 products. We're aware that we
>> >|> >|> can buy some off the shelf products supporting these
>> >|> >|> standards but would like the specs themselves. 
>> >|> >|> Any pointers?
 ...
>> Which is all that TN3270 is (3270 over telnet) , which is what the original 
>> question was about.  There are no RFC's for 3270 - That's all True Blue, who 
>> "don't need no steenking RFC's" to define protocols for their own equipment.  
>> CB.
>
  Snip...

>While it is true that IBM does not need to publish their technical
>specifications through the RFC process, what most people need from their
>3270 protocol is now well known (just not by me:-).
>
  Snip...

I'm pretty sure IBM does publish the technical specifications for 3270-style
devices, if anyone is _really_ interested.  I think the specification is in
a manual about 800 pages thick.  If you're in Florida or SoCal, you can still
use it for beach reading, but the rest of us are out of luck. :-O

Hold on...I _just_ checked with our local former-IBMer, and he's given me a
phone to call to get IBM technical specs: (800) 879-2755.  Just a warning:
specs may cost $$$.  You're looking for the 3270 Data Stream specification
or something like that.

-- 
-----------------------------------------------------------------------------
John Skovron
ILX Systems Inc.
[email protected]

-----------[000317][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 15:15:15 GMT
From:      [email protected] (Jyh-Shyang Wang)
To:        comp.protocols.tcp-ip
Subject:   How to know the size of packet's message?
Hi,
	When I get one packet, How can I know its size of it's message?
when I send one 'a' key, the packet's message show "a.." (. is not 
readable). but it's size is 1..... I know know how to get the size
in one packet. Now I can only get next packet to see its size...

thanx for any help...
--
      Internet Address:	[email protected]
Chinese Name: Wang2 Jyh4-Shyang2
¤¤ ¤å ©m ¦W :   ¤ý   §Ó  ²» 
English Name: Erik Wang    National Chiao-Tung University,Taiwan,R.O.C.


-----------[000318][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 16:13:27 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: DHCP server and NT
In article <[email protected]> [email protected] (Eric V. Smith) writes:
>I'm writing a DHCP server to run under Linux, with the specific 
>purpose of serving NT Daytona clients (and later, Chicago).  I had 
 
>by default.  Does anyone know why it doesn't?  Does it inspect the
>class of the address I give it and assume something from that?

I would hope it looks at the device involved and gets its netmask.

Alan

-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000319][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 16:43:44 GMT
From:      [email protected] (James Carlson)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]>, [email protected] (John Skovron) writes:
[...]
|> I'm pretty sure IBM does publish the technical specifications for 3270-style
|> devices, if anyone is _really_ interested.  I think the specification is in
|> a manual about 800 pages thick.  If you're in Florida or SoCal, you can still
|> use it for beach reading, but the rest of us are out of luck. :-O
|> 
|> Hold on...I _just_ checked with our local former-IBMer, and he's given me a
|> phone to call to get IBM technical specs: (800) 879-2755.  Just a warning:
|> specs may cost $$$.  You're looking for the 3270 Data Stream specification
|> or something like that.

Hmm.  That's odd.  I have the 3270 Information Display System Data
Stream Programmer's Reference (GA23-0059-07) right here on my desk.
It's about 430 pages (a *lot* thinner than our manuals) and cost about
$20.

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

-----------[000320][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 19:11:54 GMT
From:      [email protected] (Jim Shankland)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <Cw3CHA.FGx[email protected]> [email protected] (Vernon Schryver) writes:
>    -TCP checksums are required by the TCP standard, but no one will
>	come to take you away if you do not use them.  A few people
>	(misguided in my view) do suppress TCP checksums.

This statement certainly raised my eyebrows.  What does "suppress TCP
checksums" mean?  When such an implementation is preparing to send
out a TCP packet, what does it put in the checksum field?  Or does it
literally omit those bits from the sequence put on the wire?  Either
way, whom can such an implementation *talk* to?  (Not any of the TCP
implementations I'm familiar with!)

Or do these implementations simply ignore checksum errors in incoming
packets?  That's a little more credible, but horrible to contemplate.
Whose TCP does this?  Time to name names!

jas

-----------[000321][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 19:18:22 GMT
From:      [email protected] (Michel Paradis)
To:        comp.protocols.tcp-ip
Subject:   Suggestions...
We offer a fairly large business and leisure based BBS over the Internet and 
telephone lines. The system offers various types of visual databases and other
public and private features.

The problem is that all of the functions/programs are COM port based and so 
constantly require adding serial ports to allow the features.

Question...

Is there a way that we could have the serial port applications go to 
TCP/IP ports instead of the COM ports?

We've looked at COMt but it seems to be Windows based and cannot offer 
the applications under Windows. These are DOS based functions running 
under DV onto COM ports.
We've found nothing else other than telnetD type of utilities which only 
allow 1 user on a DOS machine, or others which cannot run the COM based 
utilities. There was a message posted once which talked about a fossil 
driver that took COM based applications, and routed the data to TCP/IP ports.
We've yet to ever find such a beast.

Ideas? Suggestions?
Thanks in advance.

--

Michel J Paradis - InaSec Inc. Box 70053, 160 Elgin St/Ott/Ont/Can/K2P-2M3
LiveWire Online: Telnet 198.53.239.3. (613)780-3569 - (819)682-6969/0610.
[email protected] or Mike within LiveWire Online.
        

                                                    

-----------[000322][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 19:38:51 GMT
From:      [email protected] (Kannan Thiruvengadam)
To:        comp.protocols.tcp-ip
Subject:   Application-Level BandWidth of Ethernet
Hi

For a hypothetical application that uses 10 Mbps Ethernet,
and communicates using UDP, doing negligible work apart from 
sending the packets themselves, what is the bandwidth available ?

Thanks.

--
Kannan Thiruvengadam
[email protected]
http://web.cs.ualberta.ca/~kannan

-----------[000323][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 20:41:19 GMT
From:      Gando
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: Global list of TCPIP Vendors w/contact numbers
In article <[email protected]> [email protected] (Kasey Hohenbrink) writes:
>
>We are looking for a list of TCPIP vendors with
>support/sales telephone/internet addresses.
>
>Our software can be run from Xserver packages
>using a wide variety of TCPIP.  Our tech support
>needs a list of vendor names and numbers so that
>we can interface with the vendors for our cust.
>
>I would like anyone with such a list to email
>me directly with it, if it exists.  If any of
>you are TCPIP vendors, please email me with
>any info you can....do not post to News, this
>request needs to be filled quickly, and I would
>not like to jam the News with postings.
>
>
>Thank you in advance..
>
>[email protected]

-----------[000324][next][prev][last][first]----------------------------------------------------
Date:      Wed, 14 Sep 1994 21:04:15 GMT
From:      Gando
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: Global list of TCPIP Vendors w/contact numbers
In article <[email protected]> [email protected] (Kasey Hohenbrink) writes:
>
>We are looking for a list of TCPIP vendors with
>support/sales telephone/internet addresses.
>
>Our software can be run from Xserver packages
>using a wide variety of TCPIP.  Our tech support
>needs a list of vendor names and numbers so that
>we can interface with the vendors for our cust.
>
>I would like anyone with such a list to email
>me directly with it, if it exists.  If any of
>you are TCPIP vendors, please email me with
>any info you can....do not post to News, this
>request needs to be filled quickly, and I would
>not like to jam the News with postings.
>
>
>Thank you in advance..
>
>[email protected]



    I have a found a TCP/IP that is fast
and absolutely a pleasure to use. It is 
TCP Pro from a company in California called 
Network TeleSystems and distributed by UB Networks. 

TCP Pro is both a Real mode and VxD so my memory
usage is much more efficient. The applications 
that come with TCP Pro (FTP, News, Telnet, etc.) 
are very easy to use and a plus even for a power 
user like myself.  I found that their tech support 
and customer service departments were easy to reach 
and very helpful. 

-----------[000325][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 22:23:14 GMT
From:      [email protected] (Aaron Leonard)
To:        comp.protocols.tcp-ip
Subject:   Re: ARP debugging utilities?

In article <[email protected]>, [email protected] (John W. Manly) writes:
|Hi there.  Does anyone out there have, for ANY platform, some tools for
|doing ARP-level debugging?  That is, I want to be able to manually 
|issue ARP requests and record the responses that come back.  VMS, UNIX,
|DOS, MAC -- anything would be good.

If you to force an ARP request, you can just clear out your ARP
cache (via the VMS command MU SET/ARP/FLUSH or its U*x equivalent)
and then attempt IP connectivity to the target node (via PING
or whatever), which will cause IP automatically to generate the
ARP request.

To see what's going on, you can run TCPDUMP (via the command
MU TCPD ETHER PROTO \ARP or its U*x equivalent.)

Aaron

Aaron Leonard (AL104), <[email protected]>
University of Arizona Network Operations, Tucson AZ 85721
  \ Don't lock yourself into open systems. /

-----------[000326][next][prev][last][first]----------------------------------------------------
Date:      14 Sep 1994 23:28:24 GMT
From:      [email protected] (Andrew B Consor)
To:        comp.protocols.tcp-ip
Subject:   Someone decipher this message
I am trying to figure out which type of service I need.  I want to 
connect to the internet but no one will tell ne what the different levels 
of access are.  Lets say I have a regular phone line and a 14.4k modem, 
now that should eliminate T1,leased line, and fractional T1 am I right.  
Then my choices are standard dialup whith unix shell access and something 
that I don't understand...slip access.  What makes slip access different 
if you are dialing up with the same phone line and modem.  What 
additional features/commands/speed enhancement do you get by chosing slip?
Thanks if you can help me?

New York: Re-elect Mario Cuomo We have'nt suffered enough !


-----------[000327][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 11:50:33 -0700
From:      [email protected] (Steve J. Hunt)
To:        comp.protocols.tcp-ip
Subject:   Re: Cheap IP ethernet-to-asynch techniques?
[email protected] writes:
> Hi there.  Can anyone out there suggest really cheap ways that might be
> used to connect two disjoint IP networks via Asynch modems?
> 


Just run SLIP or PPP on both sides.  No need for expensive black boxes.  
You didn't mention whether you have just PCs or also Unix boxes.  Its 
probably easier with Unix.  There are quite a few PD and commercial 
implementations of SLIP/PPP for PCs and for Unix.  Morningstar is the 
best commercial SLIP/PPP for Unix.

Steve Hunt
[email protected]


-----------[000328][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 10:26:55 -0600
From:      [email protected] (David Erin)
To:        comp.protocols.tcp-ip
Subject:   Re: online ATM sources?
[email protected] wrote:

: Are there any on-line sources detailing asynchronous transfer mode?
 
: daveriesz
: [email protected]

	Ditto.  (I'm going to check rtfm.mit.edu right now, but I'll have to
research the RFC list...does anyone know the number(s)?

Scott R. Bell

-----------[000329][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 13:22:28 -0700
From:      [email protected] (Lon Stowell)
To:        comp.protocols.tcp-ip
Subject:   Re: TN5250/TN3270 Standards?
In article <[email protected]> [email protected] writes:
>
>Hmm.  That's odd.  I have the 3270 Information Display System Data
>Stream Programmer's Reference (GA23-0059-07) right here on my desk.
>It's about 430 pages (a *lot* thinner than our manuals) and cost about
>$20.

  That manual along with either the 3174 or 3274 component description
  manual will cover just about everything you ever needed to know ab
  about 3270 devices and protocols...as well as the different flavors
  thereof and what the heck all those buzzwords mean.  If you must go
  for only a single manual, the 3274 manual has a section on
  programming along with descriptions of all the commands and orders,
  so although it isn't as comprehensive as the datastream ref manual,
  it will cover anything but the rarest and most exotic of situations.



-----------[000330][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 94 01:49:46 GMT
From:      [email protected] (Scott Miller)
To:        comp.org.sug,comp.protocols.snmp,comp.protocols.tcp-ip
Subject:   Summit?
I am looking for some information and I am hoping that someone will be able
to help me.  A couple of months ago, I saw some information about a
conference in November called the Enterprise Management Summit.  There was an
ad in Communications Week and also there was something posted on a mailing
list I subscribe to.  It sounded very interesting - very focused on
management issues and it included something about a plan to have vendors go
through a common set of tests with their products.

Unfortunately, I did not save any of the information about the conference. Is
it still going to happen?  Can anyone tell me more about it or how I can get
more information?  Thanks.

Scott Miller


P.S. Because this information is probably going to be hard to get, I am
posting this message to multiple news groups.


-----------[000331][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 03:39:37 GMT
From:      [email protected] (Chuck Smoko)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
Charles "Chip" Yamasaki - writes in resp. to this thread
|I don't know of any way to have multiple addresses on the same
|interface, but you could certainly plug in multiple NIC's.  The problem
|here is that you need multiple router ports (maybe).
|
|Come to think of it, you could actually have something like:
|
|                             Your server
|    +----------+           +------------------+
|    | Internet | <------>  | NIC1         NIC2| 
|    |  Router  |           |              NIC3|
|    +----------+           +------------------+
|
|And have the addresses for NIC2 and NIC3 be b.com and c.com.  NIC2 and
|NIC3 would not even need to connect to anything at all since they don't
|need to talk with anything on their networks, which don't even really
|exist.  They do all their routing through NIC1 to the router.  In fact,
|it should even work with a dummy interface driver is there is one.

[ Below is a idea that I found that workded for me on my Sunos 4.x
sun.  It is called vif and is written by a gent at columbia univ.  I
had to mod the code a tiny bit.  If you are intesreted in my sun mods,
I can send them via e-mail.  I would image that vif would work on just
about any unix that is derived from the UCB code.  I tried to mail to
the author, but never got a resp. - chuck smoko ]

> From ji Mon Feb 17 17:39:51 EST 1992
> From: [email protected] (John Ioannidis)
> Subject: Multiple IP addresses on a single Ethernet interface

This is a topic that comes up once in a while on comp.protocols.tcp-ip
and other newsgroups. The question is, how to get a machine with one
network interface to respond to more than one IP addresses. 

The other day, a colleague forwarded to me a request from the
namedroppers mailing list for exactly that. Here's my response:

> In article <[email protected]> [email protected] (Hitoaki Sakamoto) writes:
>    >In article <[email protected]> [email protected] (kuo-lin Hu) writes:
>    > >But it is just out of my curiousity that whether it is possible
>    > >I can assign dual IP addresses to a single controller?
 
>    >No,you can't.
> 
> Is this restriction common among UNIX TCP/IP implementations?
> Has anybody tried to modify this?
> 
> -- Kenji
> --
> Kenji Rikitake 
> [email protected] // [email protected] // ...!uunet!reseau!kenji 

I have a solution than might suit you.  For my doctoral work (there's
a paper about it in this year's ('91) SIGCOMM, also available for
anonymous FTP from cs.columbia.edu:/pub/ji/sigcomm*.ps.Z), I've
developed what I call the "Virtual Interface" (VIF). To the networking
code, it looks like an interface. It gets ifattach()ed when you open
the /dev/vif* device, and then you can ifconfig it as you like. It
does not have an if_input procedure; it only has an if_output. Packets
that it receives (from higher-level protocols) which have its
IP address, it simply loops back (like any well-behaved if driver).
Packets that it receives that are destined for some other address, it
encapsulates in an encapsulation protocol I call IPIP (IP-within-IP,
protocol number IPPROTO_IPIP == 94), and sends it to another machine
that groks that encapsulation protocol. This feature you won't need,
but here's how to have multiple IP addresses on a machine with a
single real interface:

Let's say your primary interface's IP address is 198.3.2.1, and you
also want it to respond to addresses 198.4.3.2 and 198.5.4.3 (note
that these are three distinct class C addresses in three distinct
class C nets). Here are the ifconfigs:

  ifconfig le0 198.3.2.1 up -trailers	# config primary interface

  ifconfig vif0 198.4.3.2 up 		# config first virtual interface
  route delete net 198.4.3 198.4.3.2	# delete spurious route 
  route add host 198.4.3.2 198.4.3.2 0	# add route for this i/f

  ifconfig vif1 198.5.4.3 up		# config second virtual interface
  route delete net 198.5.4 198.5.4.3	# delete spurious route 
  route add host 198.5.4.3 198.5.4.3 0	# add route for this i/f

The route deletes are needed because the ifconfig creates a default
route to the interface's network, which can cause problems; all that's
needed is the (host) route to the interface's address. 

Now, get le0's ethernet address (say, 8:0:20:3:2:1), and add the
following static ARP entries:

  arp -s 198.4.3.2 8:0:20:3:2:1 pub
  arp -s 198.5.4.3 8:0:20:3:2:1 pub

This will cause any ARP requests for the VIF addresses to be replied
with your machine's ethernet address. 

Now, make sure your default route is to your segment's gateway,
throught the real interface. FInally, make sure your routers and/or
hosts on the same segment as yours know that 198.4.3.2 and 198.5.4.3
are on that cable. 

Here's what you've accomplished.

ARP requests for any of your host's addresses will be replied to with
the host's ethernet address (the real one, because that's what it is,
the virtual ones because of the public static arp entries). Packets
reaching your host with any of these addresses will be accepted by the
ip_input routine because they match the address of one of the host's
interfaces. Packets leaving your host can have any of its addresses
(real and virtual). 

The code for vif follows. To use it, put the stuff in netinet/if_vif.c
and netinet/if_vif.h, configure your kernel with the number of
virtual interfaces you want using a line like:

pseudo-device	vif4		# Virtual IP interface

in your configuration file, and two lines like

netinet/if_vif.c	optional vif device-driver
netinet/if_vif.hc	optional vif device-driver

in the files file. Also, add the appropriate entries in conf.c, so
that you can access the if_attach() routine when you open the device:


------------------------in conf.c------------------------------------------

add this in the appropriate place in the headers of conf.c:

--------------------
#include "vif.h"
#if NVIF > 0
int	vifopen(), vifclose(), vifread(), vifwrite(), vifselect(), vifioctl();
#else
#define vifopen		nodev
#define vifclose	nodev
#define vifread		nodev
#define vifwrite	nodev
#define vifselect	nodev
#define vifioctl	nodev
#endif
--------------------

then, way down in the definition for cdevsw[]:

--------------------
	vifopen,	vifclose,	vifread,	vifwrite,	/*31*/
	vifioctl,	nodev,		nodev,		0,
	vifselect,	nodev,
--------------------

Make sure you remember the correct major device number!

------------------------in conf.c------------------------------------------

Finally, here's the code. It has the tunneling pieces removed (you
need more code to use that anyway), and it comes from a Mach 2.6
kernel; it should compile on any berkeley-derived unix with minor
changes (most likely only in the includes). 

---------------------netinet/if_vif.h--------------------------------------
typedef struct 
{
	struct ifnet	vif_if;
	struct ifnet	*vif_sif;	/* slave interface */
	int		vif_flags;
} vif_softc_t;

#define	VIFMTU	(1024+512)
---------------------------------------------------------------------------

and
---------------------netinet/if_vif.h--------------------------------------
/*
 * HISTORY
 * $Log:$
 */
 
/*
 * $Header:$
 *
 * Virtual IP interface module. 
 */

#include <multicast.h>

#include "param.h"
#include "systm.h"
#include "mbuf.h"
#include "socket.h"
#include "errno.h"
#include "ioctl.h"

#include "../net/if.h"
#include "../net/netisr.h"
#include "../net/route.h"

#ifndef MACH
#include "../machine/mtpr.h"
#endif

#ifdef	INET
#include "../netinet/in.h"
#include "../netinet/in_systm.h"
#include "../netinet/in_var.h"
#include "../netinet/ip.h"
#endif

#ifdef NS
#include "../netns/ns.h"
#include "../netns/ns_if.h"
#endif

#include "in_pcb.h"

#include "ip_var.h"
#include "ipip.h"
#include "ipip_var.h"
#include "micp.h"
#include "micp_var.h"

#include "if_vif.h"
#include "vif.h"

vif_softc_t vif_softc[NVIF];

int vifs_inited = 0;

int	vifoutput(), vififioctl();

vifattach()
{
	register int i;
	register struct ifnet *ifp;
	
	for (i=0; i<NVIF; i++)
	{
		ifp = &vif_softc[i].vif_if;
		ifp->if_name = "vif";
		ifp->if_unit = i;
		ifp->if_mtu = VIFMTU;
#if	MULTICAST
		ifp->if_flags = IFF_MULTICAST | IFF_NOARP;
#else	MULTICAST
		ifp->if_flags = IFF_NOARP;
#endif	MULTICAST
		ifp->if_ioctl = vififioctl;
		ifp->if_output = vifoutput;
		if_attach(ifp);
	}
}

vifopen(dev, flag)
int dev, flag;
{
	int unit;
	
	if (!vifs_inited)
	{
		vifattach();
		vifs_inited = 1;
		printf("vif initialized\n");
	}
	
	unit = minor(dev);
	if ((unit < 0) || (unit >= NVIF))
	{
		return ENXIO;
	}
	
	return 0;
}

vifclose(dev, flag)
int dev, flag;
{
	return 0;
}

vifread()
{
	return ENXIO;
}

vifwrite()
{
	return ENXIO;
}

vifselect()
{
	return ENXIO;
}

vifoutput(ifp, m0, dst)
	struct ifnet *ifp;
	register struct mbuf *m0;
	struct sockaddr *dst;
{
	int s;
	register struct ifqueue *ifq;
	struct mbuf *m;
	struct sockaddr_in *din;
	
	if (dst->sa_family != AF_INET)
	{
		printf("%s%d: can't handle af%d\n", 
		       ifp->if_name, ifp->if_unit,
		       dst->sa_family);
		m_freem(m0);
		return (EAFNOSUPPORT);
	}

	din = (struct sockaddr_in *)dst;
	
	if (din->sin_addr.s_addr == IA_SIN(ifp->if_addrlist)->sin_addr.s_addr)
	{
		printf("%s%d: looping\n", ifp->if_name, ifp->if_unit);
		
		/*
		 * Place interface pointer before the data
		 * for the receiving protocol.
		 */
		if (m0->m_off <= MMAXOFF &&
		    m0->m_off >= MMINOFF + sizeof(struct ifnet *)) {
			m0->m_off -= sizeof(struct ifnet *);
			m0->m_len += sizeof(struct ifnet *);
		} else {
			MGET(m, M_DONTWAIT, MT_HEADER);
			if (m == (struct mbuf *)0)
			  return (ENOBUFS);
			m->m_off = MMINOFF;
			m->m_len = sizeof(struct ifnet *);
			m->m_next = m0;
			m0 = m;
		}
		*(mtod(m0, struct ifnet **)) = ifp;
		s = splimp();
		ifp->if_opackets++;
		ifq = &ipintrq;
		if (IF_QFULL(ifq)) {
			IF_DROP(ifq);
			m_freem(m0);
			splx(s);
			return (ENOBUFS);
		}
		IF_ENQUEUE(ifq, m0);
		schednetisr(NETISR_IP);
		ifp->if_ipackets++;
		splx(s);
		return (0);
	}

	return EHOSTUNREACH;
}

/*
 * Process an ioctl request.
 */
/* ARGSUSED */
vififioctl(ifp, cmd, data)
	register struct ifnet *ifp;
	int cmd;
	caddr_t data;
{
	int error = 0;

	switch (cmd) {

	case SIOCSIFADDR:
		ifp->if_flags |= IFF_UP;
		/*
		 * Everything else is done at a higher level.
		 */
		break;

	default:
		error = EINVAL;
	}
 return (error);
}

vifioctl(dev, cmd, arg, mode)
dev_t dev;
int cmd;
caddr_t arg;
int mode;
{
	int unit;
	
	unit = minor(dev);
	if ((unit < 0) || (unit >= NVIF))
	  return ENXIO;
	
	return EINVAL;
}
---------------------------------------------------------------------------- 

To use it, compile your kernel, and reboot. Then create the vif
device:

# mknod /dev/vif c 31 0

(or whatever major number it ended up being), and echo something into
it:

# echo > /dev/vif

This will cause the device to be opened, which will if_attach the
interfaces. If you feel like playing with the code, you may want to
kmem_alloc() the vif_softc structrure at open time, and use the minor
number of the device to tell it how many interfaces to create. 

Now you can go ahead and ifconfig etc. 

I'll be happy to answer minor questions, and hear about success and
failure stories, but I cannot help you if you don't already know how
to hack kernels.

Good luck! 

/ji


-----------[000332][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 15:22:53 -0700
From:      [email protected] (Lon Stowell)
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Phil Howard) writes:

>Maybe.  IPX waits for acks for every packet, for instance.

  This would have a BENIGN effect as other stations could use the
  ethernet during this period of time....as you wouldn't see much
  back-back frames like you can easily generate with ip. 

>
>What I do know from seeing the effect from our own Novell server, watching
>it from Unix, is that periodically the Novell box trashes the network for
>about 2 to 4 seconds with zillions of broadcasts.  Also, Novell's TCP/IP
>behaves very badly (it might also be non-conformant, but basic communications
>with it from Unix does work and it usually works out as a router if you
>don't stress it with too many packets over 1472 bytes).

  Whether Novell's TCP/IP is conformant depends on which driver you
  have installed.  e.g. the standard token ring driver isn't--it just
  hacks in the LLC and SNAP header but doesn't respond to all the LLC
  mandatory frames.   Not picking on Novell, there are other
  miscreants out there.




-----------[000333][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 15:29:49 -0700
From:      [email protected] (Lon Stowell)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Tom Limoncelli) writes:
>
>IPX is a chatty protocol.  I don't have a lot of experience with it,
>but I do have experience with AppleTalk, which is equally chatty.

  True.  There are other 'chatty' protocols where servers and/or
  clients continually send out "Yo!  My name is Bob, and I can print"
  type messages.  

  If you compute the total amount of LAN bandwidth consumed by all
  those broadcasts for a modestly sized network, it is a nitpickingly
  small amount unless you have some truly misconfigured MAC layer
  routing loops configured with braindead forwarding.

  Yup, you see "Broadcast Storm" in bright red on the Sniffer's
  Expert summary.  Says something like "duration 5 seconds" or even
  '5 minutes' or something like that.  Then you look at the actual
  number of packets being passed and how much real bandwidth it
  consumes...and promptly go back to smacking the "n" key on netnews.

  

-----------[000334][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 13:03:40 -0500 (CDT)
From:      [email protected] ("Tom Clark (319)395-5045, COMNET 435-5045")
To:        comp.protocols.tcp-ip
Subject:   Variable Length Subnetting and RFC 1219
We are interested in talking to people who have adopted variable length
subnet masks to solve problems with address space limitations.  We are
preparing to implement VLSM in about a month and would like to amass as
much information as possible before then.

Also of interest, has anyone implemented the algorithm in RFC 1219 that 
they'd be willing to share?  We'd prefer a Perl implementation if possible
but anything would help.  Has anyone used this algorithm in the real world?
What were your experiences?  Are there any test suites available to ensure
completeness if we have to write our own?

Any help you may be will be greatly appreciated,

Tom

-----------[000335][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 05:14:17 GMT
From:      [email protected] (Tom Fitzgerald)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
[email protected] (James VanBokkelen) writes:

> [email protected] (Samuel Lam) writes:
 
> > 1) Is sending outgoing ARP replies via MAC-level broadcasts
> >    instead of unicasts allowed?
 
> I agree with the other post analyzing RFC 826, but let's consider the
> effects: Vendor 1's router distracts every host on the net with every
> ARP reply, which may in the end reduce overall load on the router
> while making many hosts' ARP caches incrementally larger.

I can see why vendor1 would be tempted to do this, but it's still a bad
idea (as well as violating the RFC).  Everybody's ARP table is
incrementally larger if there are only a few boxes from vendor1 on the LAN,
but if there are a LOT of vendor1's boxes, then everybody's ARP table is a
LOT larger.  Some systems have limits on the size of the ARP table, and
vendor1's actions could stress them beyond their limit.

I don't know where the "router" reference came from (do you recognize
vendor1's behavior?) but I can imagine a LAN with a huge number of routers
on it, so a router manufacturer can't exempt itself from the RFCs on the
grounds that most people won't be hurt; MAE-EAST (the 192.41.177 net) is
one example of a LAN full of routers.  (It's not huge yet, but it'll get
bigger.)

-- 
Tom Fitzgerald      1-508-967-5278
Wang Labs           [email protected]         (account no longer active)
Lowell MA, USA      [email protected]       (use this address instead)

-----------[000336][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 10:51:02
From:      [email protected] (Peter P. Morrissey)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Rob Tanner) writes:
>Path: newstand.syr.edu!travelers.mail.cornell.edu!news.kei.com!yeshua.marcam.com!usc!howland.reston.ans.net!agate!ames!george.arc.nasa.gov!tanner
>From: [email protected] (Rob Tanner)
>Newsgroups: comp.dcom.lans.ethernet,comp.protocols.tcp-ip
>Subject: Running IPX and TCP/IP over same ethernet
>Date: 13 Sep 1994 14:40:50 GMT
>Organization: NASA Ames Res. Ctr. Mtn Vw CA 94035
>Lines: 26
>Distribution: na
>Message-ID: <[email protected]>
>NNTP-Posting-Host: george.arc.nasa.gov
>Xref: newstand.syr.edu comp.dcom.lans.ethernet:11832 comp.protocols.tcp-ip:28133


>Direct from Rumor Control Central:  I have heard that IPX is a real
>bandwidth hog, and that if you don't want to see your TCP/IP severely
>degrade, don't let IPX run over the same segments that run TCP/IP.
 
>I know zip about IPX, but it looks like we are going to be installing
>a Novell server anyway.  The Novell, of course, is for the PCs.  What
>concerns me as that we also have a number of Unix machines as well as
>an even larger number of X-stations that don't run IPX and don't
>concern themselves with the Novell server.
 
>Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
>I could isolate the IPX traffic, but to do that requires purchasing
>hardware and that means money that has to be justified.  Does anybody
>have any real performance figures that they could share.

Don't waste your money building another network unless you need it. We run IPX 
andI P and they co-exist just fine. I have not seen any interactions between 
the two. As has been already aluded to, if there is an increase in utilization 
by the addition of more traffic, that is another issue, but it is no worse 
than adding more Unix/IP boxes that add more traffic.
Although I would qualify that a little by saying that in general with Novell 
you have a "File Server" which does exactly that, it serves a file. So, if you 
are searching through a Dbase or some other database, every single record that
you are searching through is sent to the client PC from the File server. One 
could argue that this process is more efficient on a Unix box which usually 
will be running the sorting application on the host and only send the results 
of the sort to the client. This is more of a design issue if you have a choice 
as to which platform you can run an application on. If you have a choice, it 
becomes even more critical with remote links, i.e. WAN or especially dial-up.

The other consideration is that you should use 3.12 or greater and VLM's. If 
you do this you will get more of the windowing that TCP gives you, thus 
overcoming another traditional weakness with Novell's protocol.

Bottom line though is that plenty of people have been doing all of the above 
for years without major problems. Take a look at your current utilization, do 
a baseline, then watch it as you add Novell servers and clients. Pay attention 
to the applications people are running and the effect it has on utilization. 
You should be able to intelligently predict when or if you need to re-design 
your network.

_Pete M.
Syracuse University

>Thanks,
>Rob
 
>      _ _ _ _           _    _ _ _ _ _  
>     /\_\_\_\_\        /\_\ /\_\_\_\_\_\  
>    /\/_/_/_/_/       /\/_/ \/_/_/_/_/_/  Robert J. Tanner
>   /\/_/__\/_/ __    /\/_/    /\/_/       Ames Research Center
>  /\/_/_/_/_/ /\_\  /\/_/    /\/_/        (415) 604-3451 (SETI)
> /\/_/ \/_/  /\/_/_/\/_/    /\/_/         (415) 604-5347 (Kuiper)
> \/_/  \/_/  \/_/_/_/_/     \/_/          [email protected]
> ____________________________________________________________________


-----------[000337][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 06:44:32 GMT
From:      [email protected] (Richard A. Bryan)
To:        comp.protocols.tcp-ip
Subject:   TERM Client for OS/2
Does there exist a TERM client for OS/2?  I know you can connect two UNIX
machines thru term to provide a pseudo connection, but is there client
software for OS/2 so I can run some kind of WWW browser or gopher client
without slip or PPP?



-- 
                           [email protected]

------------------------------------------------------------------------------ 
                         "Classify by ordnance."

STG2 Richard A. Bryan                                       OOO
FLEET ASW TRAINING CENTER, PACIFIC                        OO   OO 
Network Engineer                                     /   O       O   ////
Code 01J                                            <====================    
32444 Echo Lane, suite 100                           \   O       O   \\\\
San Diego, CA  92147-5199         			  O     O
------------------------------------------------    	 (_)   (_)
The opinions expressed above do not necessarily
reflect the views of my employer or the DOD.
Likewise, my employer or the DOD may not reflect
*MY* opinions.  So there.

-----------[000338][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 94 15:15:56 -0500
From:      [email protected] (Terry Kuny)
To:        comp.protocols.tcp-ip
Subject:   Query about ASCII limitation TCP/IP apps
Hi. I'm trying to find out a little about the reasons behind the inability of some Internet 
applications to use anything but US ASCII (obvious example, SMTP). Was it an historical 
accident? A choice to go with a lowest common denominator? What other applications are 
affected? Is the limitation only at the application level, or lower?

Any information anyone can send me (references to articles, books, net info, pointers to other 
sources, etc.) about the background/history of this issue would be appreciated.

Thanks,

Gary Cleveland
National Library of Canada
Ottawa
[email protected]


-----------[000339][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 94 14:41:59 PDT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Re: online ATM sources?

Re: looking for IP over ATM and related info.
Check RFC 1600 first--it contains a list of all the other RFCs, organized
by type, topic, status, etc. RFCs can be downloaded as follows:
RFCs may be obtained from the Internet host named DS.INTERNIC.NET via FTP, 
WAIS, and electronic mail. Using FTP, RFCs are stored as rfc/rfcnnnn.txt or 
rfc/rfcnnnn.ps where 'nnnn' is the RFC number, and '.txt' is the ASCII version, 
and '.ps' the PostScript version of the file.
To use FTP to get these documents, login to DS.INTERNET.NET as 'anonymous' and 
supply your e-mail address as the password (to keep track of who's downloading 
what). Using WAIS, use either your own local client or Telnet to 
DS.INTERNET.NET and login as 'wais' -- no password is required -- to access a 
remote WAIS client. Help information and a tutorial on using the program are 
available on-line. The database to search for the documents is named 'rfcs'.
AT&T Directory and Database Services also provides a mail server interface to 
the RFC documents. Send a mail message to [email protected], and include 
any of the commands in Table 3-6 to obtain the requested RFC file or files (a 
single message can include multiple requests).

Command	Explanation (remote command)
document-by-name rfcnnnn	'nnnn' is the RFC number.
                                 The text version is always sent.
file /ftp/rfc/rfcnnnn.yyy	'nnnn' is the RFC number. 
                                 If 'yyy' = '.txt' the text version is
                                 sent; if 'yyy' = '.ps' the PostScript version 
                                 is sent.
help	                         sends a text file on how to use the InterNIC 
                                 mail server
Table 3-6: Ordering RFCs for e-mail delivery

Hope this helps (stuff I cut from a manuscript I'm working on).


-----------[000340][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 17:12:07 -0400
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single
In article <[email protected]> [email protected] (Glenn Fleishman) writes:
>In article <[email protected]>
>[email protected] (Barry Margolin) writes:
>> I think you're thinking of the CERN daemon's "proxy HTTP" support
 
>Right, but it's not too bad to configure.

The problem is that it has to be configured by every browser user.  First
of all, not all browsers have proxy support.  And how do you tell every
potential user that they must access your server using proxy mode?
Presumably, the original poster's intent is to support people who access
his server by typing http://www.<company>.com/, for various values of
<company>; if they were accessing it from links in other documents, or even
from printed URLs, he could put the company names in the pathname portion
of the URLs.  Since he's trying to make it easy for users who don't have
detailed information about accessing these documents, a solution that
requires reconfiguring the browser is clearly out of the question.

Also, most browsers only support a single proxy server for each protocol.
So if you want people to access your HTTP server using proxy mode, they'll
have to specify that your server is their generic HTTP proxy gateway.  That
means that your server will be involved in *every* HTTP query they do.
-- 
Barry Margolin
BBN Internet Services Corp.
[email protected]

-----------[000341][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 17:36:44 -0400
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Proxies and multiple addresses
In article <[email protected]> [email protected] (Glenn Fleishman) writes:
>I checked this out again.
>
>The CERN httpd will let you run -- without a proxy -- redirects to
>multiple domains addressed as the same location.
>
>So if you have domain1.com, domain2.com, domain3.com CNAME'd or IN A'd
>to the same actual IP address, using the CERN httpd redirect commands
>in the all.conf file, you can have a user specifying
>http://www.domain1.com go to a direct home page than
>http://www.domain2.com.

The CERN httpd *is* the proxy when it gets requests that include the
protocol://domain prefix.

There are two ways that a client may send a request:

1) If it thinks it's talking to the real server, it sends:

	GET pathname HTTP/1.0

where pathname is the URL with the protocol://domain prefix stripped off.

2) If it thinks it's talking to a proxy gateway, it sends:

	GET url HTTP/1.0

By default, browsers talk directly to server, and use the first form.  It
requires explicit configuration of the browser or user environment to make
it use the proxy form.  And as I said in a previous post, most browsers
only allow you to configure a single proxy gateway for each protocol.
There's no way to specify "Always send full URLs"; nor would this be
desirable, since not all HTTP servers support the second form.
-- 
Barry Margolin
BBN Internet Services Corp.
[email protected]

-----------[000342][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 13:13:11 GMT
From:      [email protected] (John Alonzo Breen)
To:        comp.protocols.tcp-ip
Subject:   SLIP and local host name
I'm not sure this is the proper newsgroup for this, but I couldn't
think of a better one.  Feel free to redirect followups if
appropriate.

I connect my PC (via SLIP under Linux) to a SLIP server that
dynamically allocates IP addresses.  I was wondering what I should do
about my host's hostname (and domain, for that matter).  When it boots
up, my rc file assigns it a dummy hostname & domain.  While most
things seem to work fine (although mail is still a bit of a mystery),
it's annoying to have the machine think its name is something other
than what it really is.

Should I use the `hostname' command to set the hostname after I make
the SLIP connection?  Or should I just not worry about it, and let it
be happily ignorant as to its "real" name?

Thanks.
-- 
John A. Breen         | I teleported home one night with Ron & Sid & Meg.
[email protected]    | Ron stole Meggie's heart away, and I got Sidneys's leg.
                      |                            - Douglas Adams

-----------[000343][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 16:07:36 GMT
From:      [email protected] (Isidore Rigoutsos)
To:        comp.protocols.tcp-ip
Subject:   how to count used/available mbufs?

Hello netters.

We are dealing with the following unusual situation and would like to find a
way around it.

We have a distributed program that spawns 16 identical processes on 16 very
fast CPUs (running AIX 3.2.4 if it matters) and which are connected to one 
another through a very slow network (ethernet).

All 16 processes access a local disk and generate a lot of data which they then
send to a subset of the remaining 15 processes (typically we have around 10
distinct destinations to where the retrieved data will be sent).  The outgoing
messages are sent to the destinations through a series of rpc-based send()
operations.  There is *no* global synchronization -- intentionally;  the
processes finish as soon as they have issued the last send() operation to the
last destination.

Here is what appears to be happening:  all of the processes rush and fill the
local mbufs with outgoing messages.  And they will of course use them up very
quickly since they can beat the ethernet which delivers at a slow pace -- the
messages have to be largish so the exponential-kickback comes into play as well,
forcing things to move even slower.  Now, the outgoing buffers won't be emptied
until the messages have been delivered to their destinations.  But, all of the
destinations have very few mbufs left and even those are used not as fast as
they could because of the exponential kick-back (= a result of 16 processes
trying to send large messages to 16 destinations).  As a result, the outgoing
mbufs are drained at a very slow pace.

What we would like to do is police the sending processes so that they do not
issue any send() messages if, say, half of the mbufs are already in use. This
in essence will keep half of the available mbufs around for the incoming
messages.  If we artificially slow down the sending processes by injecting a
call to usleep() after each send() operation the network catches up and
performance improves dramatically.  The downside is that the argument to
usleep() is a function of the speed of the CPU's and of the network and we
would like to avoid something like this since the program is intended to run
under different configurations and there is not really a 'universal' value.
And of course, we want to avoid global communication to police the processes
because then the implementation would not scale up nicely.  Keeping track of
the used mbufs appears to be the only alternative.

So here is the question:

  how can I determine the number of available/used mbufs from inside a program?


Clearly, such information *is* available since 'netstat' reports it and
according to the man pages it is the memory handling routines that keep track
of those numbers -- unfortunately, a few hours of trying to find appropriately
named subroutine calls did not prove fruitful.

I do not read this newsgroup regularly so please email directly and I will
summarize if there is interest.


Thanks in advance for all and any help,

Isidore




--------------------------------------------------------------------------------
Isidore Rigoutsos, PhD                       | Internet:[email protected]
Computational Biology & Pattern Matching Grp | Bitnet:  [email protected]
IBM Thomas J. Watson Research Center         |----------------------------------
P.O. Box 704                                 | Phone:    (914) 784 7968
Yorktown Heights, NY 10598, USA.             | FAX:      (914) 784 7455
--------------------------------------------------------------------------------




-----------[000344][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 16:24:11 GMT
From:      [email protected] (Daryl Anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: UDP port 520???
In article <[email protected]> [email protected] (Chris van der Merwe) writes:
>Path: marlin.upe.ac.za!quagga.ru.ac.za!inet.up.ac.za!rkw-lan.cs.up.ac.za!s9448373
>From: [email protected] (Chris van der Merwe)
>Newsgroups: comp.protocols.tcp-ip
>Subject: UDP port 520???
>Date: Tue, 13 Sep 1994 14:17:15 GMT
>Organization: Computer Science Department, University of Pretoria
>Lines: 9
>Message-ID: <[email protected]>
>NNTP-Posting-Host: 137.215.164.45


>Hi!
 
>Every time I do a name resolve request to out local name server I get a UDP 
>datagram for port 520.
 
>Can anybody tell me what this is??  As far as i can see, its something about 
>a local routing thing.
 
>THX!
Helo,
This is RIP.
Check out your favorite Unix host and look at the file /etc/services
for popular ports that are used.

Regards
Daryl

-----------[000345][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 16:34:54 GMT
From:      [email protected] (Jon Zeeff)
To:        comp.protocols.tcp-ip
Subject:   Re: async 56k CSU/DSU with compression
>>I have a 115k bps async connection (with flow control) that I would like
>>to attach to a 56k leased line.  Does anyone make an async CSU/DSU
>>that has compression built in?

Having not gotten very far with this request, I'd also be interested in
ethernet bridges or routers that have compression and possibly a built-in
CSU/DSU for 56k.  Price is a factor.


-----------[000346][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994  17:15 gmt
From:      [email protected]  (Philip O' Toole)
To:        comp.protocols.tcp-ip
Subject:   SMTP from AS400 to Sun
Don't know if this is the right group to post this but.....
Does anyone know anything about using SMTP to transfer a datbase file
from a Sun server to an AS400??? We've been having some difficulty....
Any help would be appreciated....

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
|     Philip O'Toole. (My own thoughts......)     |
|     [Northern Telecom Ireland]                  |  
|Logic is the beginning of Wisdom, not the end....|
|_________________________________________________|

-----------[000347][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 94 17:21:34 GMT
From:      [email protected] (Jim Robinson)
To:        comp.protocols.tcp-ip
Subject:   Re: Help please--read() with sockets
Jim Shankland ([email protected]) wrote:
>In article <[email protected]> [email protected] (Paul Smith) writes:
 
>>No harm if errno where to be set to ENOCON or what ever in this scenario...
>>
>>And yes errno is 
>>indeterminate, but my style is to errno=0; prior to ever system call so as
>>to differentiate for sure who set errno to != 0.
 
>This, on the other hand, betrays some confusion.  errno is set as
>a result of a system call if and only if the system call fails (generally,
>returns -1).  Setting errno to 0 before the system call helps nothing.
>The correct test is to see if the system call has failed, and check
>errno only if it has.

This is true 99.99% of the time. The exception is dealing w/ sysconf() and
(f)pathconf(). It *is* necessary to set errno to 0 before the call in order
to differentiate between two separate cases upon failure (in one case errno
is unchanged). Am I the only one that considers this less than desirable?
-- 
Jim Robinson
[email protected]
{ubc-cs!van-bc,uunet}!mdivax1!robinson


-----------[000348][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 17:31:22 GMT
From:      [email protected] (Lawrence S. Brakmo)
To:        comp.protocols.tcp-ip
Subject:   snd_cwnd can wrap in NET-2 TCP (rare occurence)
It is possible to wrap the congestion window when receiving duplicate ACKs. 
This is a rare occurence since it requires TCP's send and receive  
buffers to be more than 32KB, and there must also be more than 32KB 
of data in transit when a loss is detected and the Fast Retransmit Mechanism
goes into action. 

The problem occurrs in tcp_input.c:

the following code snippets copyrighted:
/*
 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
 *      The Regents of the University of California.  All rights reserved.


                               } else if (tp->t_dupacks > tcprexmtthresh) {
                                        tp->snd_cwnd += tp->t_maxseg;
                                        (void) tcp_output(tp);
                                        goto drop;
                               }

tp->snd_cwnd can wrap since it is declared as u_short. This is not a problem
with the BSD-Lite TCP code since the windows are declared as u_long to handle
the big-windows option.

Lawrence S. Brakmo  ([email protected])
 



-----------[000349][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 18:30:30 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
In article <[email protected]> [email protected] (James VanBokkelen) writes:
>[email protected] (Samuel Lam) writes:
>> 1) Is sending outgoing ARP replies via MAC-level broadcasts
>>    instead of unicasts allowed?
>
>I agree with the other post analyzing RFC 826, but let's consider the
>effects: Vendor 1's router distracts every host on the net with every
>ARP reply, which may in the end reduce overall load on the router
>while making many hosts' ARP caches incrementally larger.

Why would any node's ARP cache be any larger? There's no reason to
remember an IP/MAC mapping that isn't interesting, regardless of
whether the mapping was delivered in a broadcast request or a
broadcast reply.

I don't mean to defend this behavior, but I believe the impact is
limited to the cost of processing the uninteresting broadcast traffic.
This implementation is being silly, but any implementation having
trouble with it is broken.

						don provan
						[email protected]

-----------[000350][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 19:16:45 GMT
From:      [email protected] (Bruce A. Mah)
To:        comp.protocols.tcp-ip
Subject:   Re: online ATM sources?
fsdpr  writes:
> Are there any on-line sources detailing asynchronous transfer mode?

Try the Cell Relay Gopher at:

gopher://cell-relay.indiana.edu/

Bruce.
--
Bruce A. Mah		   Graduate Student	          [email protected]
		Tenet Group, Computer Science Division
		 University of California at Berkeley

-----------[000351][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 04:24:10 -0500
From:      [email protected] (Stefan Grefen)
To:        comp.protocols.tcp-ip
Subject:   Re: how to count used/available mbufs?
In article <[email protected]>,
Isidore Rigoutsos <[email protected]> wrote:
>
>Hello netters.
>
>We are dealing with the following unusual situation and would like to find a
>way around it.
 [...]
>Here is what appears to be happening:  all of the processes rush and fill the
>local mbufs with outgoing messages.  And they will of course use them up very
>quickly since they can beat the ethernet which delivers at a slow pace -- the
>messages have to be largish so the exponential-kickback comes into play as well,

I assume here that you use TCP-based RPC. I would reduce the send and 
recv-buffer sizes to lessen the strain on the mbuf usage.  Of course this
would implicitly synchronize your applications a little bit more.
You get less data queued in the system which should prevent congesting
your network and reduce retransmissions.
If you want to be able to write one message without blocking than
increase only the sendbuf (and try to tune the amount of available mbuf-pages).
This will allow you to copy your whole message to the kernel but because
the recvbuffer-size determines the TCP-window size, you still have a 
reasonable flow control on the net.

>What we would like to do is police the sending processes so that they do not
>issue any send() messages if, say, half of the mbufs are already in use. This
>in essence will keep half of the available mbufs around for the incoming
>messages.  If we artificially slow down the sending processes by injecting a
>call to usleep() after each send() operation the network catches up and
>performance improves dramatically.  The downside is that the argument to
>usleep() is a function of the speed of the CPU's and of the network and we
>would like to avoid something like this since the program is intended to run
>under different configurations and there is not really a 'universal' value.
>And of course, we want to avoid global communication to police the processes
>because then the implementation would not scale up nicely.  Keeping track of
>the used mbufs appears to be the only alternative.

[...]

No, limit the mbuf usage through send/recv buffer sizes. In most systems
netstat reads kernel-memory to determine mbuf usage, in some systems there
is now a specific system call to get kernel information like this.

You get set send/recv buffer-sizes by using the setsockopt systemcall.

Stefan

>
>--------------------------------------------------------------------------------
>Isidore Rigoutsos, PhD                       | Internet:[email protected]
>Computational Biology & Pattern Matching Grp | Bitnet:  [email protected]
>IBM Thomas J. Watson Research Center         |----------------------------------
>P.O. Box 704                                 | Phone:    (914) 784 7968
>Yorktown Heights, NY 10598, USA.             | FAX:      (914) 784 7455
>--------------------------------------------------------------------------------
>
>
>


-- 
Stefan Grefen                          Convex Computer GmbH, Frankfurt, Germany
[email protected]		       Phone: +49-69-665270


-----------[000352][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 1994 20:24:08 +0000
From:      [email protected] (Niall Teasdale)
To:        comp.protocols.tcp-ip
Subject:   Re: IP bcast: is all 0's dead?
Bruce Wollen ([email protected]) wrote:

: I'm wondering if it's reasonable to even implement support for
: this old broken broadcast address in current networking software
: products.
 
: Any input is appreciated
: Bruce

Whether it's reasonable or not, it is required. Check RFC 1122, Host
Requirements - Communications Layers, section 3.3.6. It says that all
hosts should recognise and accept the all 0's broadcast address. It is
optional as to whether a host supports the sending of all 0's broadcasts,
but if it does it should default to all 1's (which Solaris 1 does not :-( ).

Niall.


-----------[000353][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 20:28:28 GMT
From:      [email protected] (Bob Baggerman)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc
Subject:   Re: Global list of TCPIP Vendors w/contact numbers
Kasey Hohenbrink writes:
>We are looking for a list of TCPIP vendors with
>support/sales telephone/internet addresses.

Take a look on the "http://www.bizweb.com/" Mosaic server under the 
"network.software" catagory.  There is a list of network software companies
with product information available via web servers.  

bob

-- 
Bob Baggerman                         !  [email protected]
Communications Laboratory             !  [email protected]
Georgia Tech Research Institute       !  [email protected]
Atlanta, GA  30332  USA               !  404-894-3525

-----------[000354][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 04:20:37 -0400
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: Query about ASCII limitation TCP/IP apps
In article <[email protected]> [email protected] writes:
>Hi. I'm trying to find out a little about the reasons behind the inability of some Internet 
>applications to use anything but US ASCII (obvious example, SMTP). Was it an historical 
>accident? A choice to go with a lowest common denominator? What other applications are 
>affected? Is the limitation only at the application level, or lower?

It's mostly a historical limitation.  Most of the old TCP/IP protocols are
directly translated from earlier Arpanet protocols.  These were developed
in the early 70's in the US -- at that time, I don't think there were any
8-bit character sets.  Even if there were international character sets, it
wasn't common for US programmers to worry about such things.

All these early protocols (TELNET, SMTP, FTP) make use of a common
character set, known as NETASCII.  This is an 8-bit character set that has
US-ASCII in its bottom half, and characters with the high-order bit set are
used as protocol control characters (e.g. TELNET's IAC).

Some of the protocols that are affected are TELNET, SMTP, FTP, FINGER, and
TIME.
-- 
Barry Margolin
BBN Internet Services Corp.
[email protected]

-----------[000355][next][prev][last][first]----------------------------------------------------
Date:      15 Sep 1994 22:47:19 GMT
From:      [email protected] (Dylan Greene)
To:        comp.protocols.tcp-ip
Subject:   Problem w/ NCSA + Sun
I've got a little net problem:

I'm networking a SUN SparcClassic on a net with a linux box, and
I'd like to setup a bunch of messy-dos machines with NCSA telnet
on them to use PC-NFS / TeX / VT-100 access to the sun and linux
resources..

But an interesting problem has arisin..  There was no problems
until I started putting the NCSA clients on the ethernet..

For some reason, the msdos's can only make a connection with the
linux, and not the sun.. Is there a conflict between NCSA
and SunOS I don't know of?  The cable lengths are allset, 
I'm really getting stuck here..

Thanks..
[email protected]


-----------[000356][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 94 23:00:57 GMT
From:      [email protected] (Sergey A. Elistratov)
To:        comp.protocols.tcp-ip
Subject:   Re: Help needed on KA9Q
In article <[email protected]>,
David Rudder <[email protected]> wrote:
>I have KA9Q running a router at a (somewhat) distant site.  I don't like 
>going over there, but have to make frequent changes to the setup.  All of 
>the stuff I do I could be doing over telnet.  I have setup telnet to run 
>on it, but I can't seem to create a password file.  As it stands, I can 
It's not difficult. In ka9q config directory ( c:\net by default) file
"ftpusers". Format: 
login password acsess_dir permissions

strat	1234	/strat	7
...
But i not find usefull for gateway staff commands.
Telnet session oriented for mailbox and telnet protocol
commands (send file etc.).
 
>telnet there, but can not login because it doesn't recognize any logins.  
>The docs don't explain this.
>
>			Thanks in Advance,
>			David Rudder
>			[email protected]
>
        Good luck,
	Sergey.


-----------[000357][next][prev][last][first]----------------------------------------------------
Date:      Thu, 15 Sep 94 23:10:21 GMT
From:      [email protected] (Sergey A. Elistratov)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and ARCHNET
In article <[email protected]>, Patrick Klos <[email protected]> wrote:
>In article <[email protected]>,  <[email protected]> wrote:
>>Some where I heard that TCP/IP can not run on top of ARCHNET.  
>>Is this true?  The only potential reason I can think of is the fact that
 ...
>
>This is NOT true.  ARCNET can support TCP/IP as well as any other network
>media.  ARCNET will be prone to fragmentation (either IP or ARCNET) since
>the maximum data size for any single packet is 508 bytes.

But where can i find drivers for ARCNET cards for UNIX ?
For FreeBSD etc. ?

>-- 
>============================================================================
>    Patrick Klos                           Internet: [email protected]
>    Klos Technologies, Inc.                Voice: (603) 424-8300
>    604 Daniel Webster Highway             FAX:   (603) 424-9300
>    Merrimack, New Hampshire  03054        BBS:   (603) 429-0032
>============================================================================

	Thanks for help,
          Sergey.


-----------[000358][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 01:53:47 GMT
From:      [email protected] (Stefan Sharkansky)
To:        comp.protocols.tcp-ip
Subject:   Re: Global list of TCPIP Vendors w/contact numbers
In article <[email protected]> Gando writes:
>    I have a found a TCP/IP that is fast
>and absolutely a pleasure to use. It is 
>TCP Pro from a company in California called 
>Network TeleSystems and distributed by UB Networks. 

Gee whiz, this posting comes from surprise, surprise, UB Networks.
(Although you had to do a little work to figure that out from Gando's
posting).  TCP Pro may or may not be a good product, but full
disclosure in posting would have been appropriate here.

>TCP Pro is both a Real mode and VxD so my memory
>usage is much more efficient. The applications 
>that come with TCP Pro (FTP, News, Telnet, etc.) 
>are very easy to use and a plus even for a power 
>user like myself.  I found that their tech support 
>and customer service departments were easy to reach 
>and very helpful. 

No comment.

-----------[000359][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 05:59:37 GMT
From:      [email protected] (Christoph Berendes)
To:        dc.general,dc.org.linux-users,comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.sys.novell,comp.sys.mac.comm
Subject:   Help get Hine JHS, in DC, on the Net.
Have you got Internet experience (gopher, Mosaic, listservs) you'd like to 
share with others?

Do you have experience with IP routers, Internet links, SMTP/gopher/
web-servers, Ethernet LANs, TCP/IP software for DOS/ Windows/Macintosh, or
Novell networks? 

If you answer "yes!" to either of these questions, here's a great
opportunity for you to use your skills in a worthy cause! 

Hine Junior High School in Washington, D.C., has gotten its own dedicated
Internet connection.  Although they have an existing LAN in their computer
lab, they are new to Internet technology and need assistance with
technology, training, and locating appropriate content areas on the Net. 

You can help them by advising them on

* how to use the Internet to support teaching and learning

* how to modify their existing Novell network to provide Internet access
via Telnet, FTP, email, etc.

* how to install and configure TCP/IP software and Ethernet adaptors for
Mac and PCs (DOS only and DOS/Windows); and

* how to keep an SMTP mailserver and Unix(FreeBSD) host up and running

Plus, if you've got modems/computers/Ethernet cards or hubs to donate, 
Hine could make very good use of them.

You can find more information about Hine and their Internet plans at

   URL   ftp://ftp.digex.net/pub/access/berendes/hine-background.txt

If you'd like to help out, please send email to Christoph Berendes at
[email protected]  

Thanks in advance.

Christoph Berendes
volunteer Internet coordinator
Hine JHS
7th St. and Penn. Ave, SE, oppo. Eastern Market metro
Washington, DC
[email protected]


-----------[000360][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 16:27:53 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Cheap IP ethernet-to-asynch techniques?
[email protected] writes:

>Using this picture, then, the question becomes: what cheap solutions are
>there for the black boxes?  Obviously commercial routers can do it for
>around $2000 and up, but I was hoping for something we could put together
>with chewing gum and chicken wire for considerably less.

Do you have any architecture specifications for chewing gum and chicken wire?

Actually it should be possible now days for someone to integrate an ethernet
and a small router with slip and ppp into the same box as a modem (V.FC/V.34
please) so that you have a BNC or switched ethernet twisted pair plus the usual
phone line.  If they can make this thing auto-dial on demand with a decent
scripting language they would sure have a hot product.

How about a 2-line load balancing version?  Or am I getting too greedy now?
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000361][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 08:20:47 GMT
From:      [email protected] (Chung-Chi CHUANG)
To:        comp.protocols.tcp-ip
Subject:   Re: online ATM sources?
[email protected] wrote:

: Are there any on-line sources detailing asynchronous transfer mode?
 
: daveriesz
: [email protected]

   1. rfc : rfc1695( ATM MIB), rfc1595( SONET MIB), rfc 1577 ( classical ip 
            and arp over ATM) , rfc1483( Multiprotocol Encasulation over
            AAL 5"

   2. ftp site :
        host :cell-relay.india
        directory : /pub/cell-relay/docs/current/CCITT
   3. Newsgroup :
        comp.dcom.cell-relay



                                     Chung-Chi Chuang
                                     Data Communications Institute
                                     Taipei, Taiwan
                                     [email protected]
 

-----------[000362][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 08:47:30 GMT
From:      [email protected] (Chung-Chi CHUANG)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP over OSI stack
[email protected] wrote:

: We are look into implementing TCP/IP in the lower three layers of the OSI 
: stack. In other words implement RFC 1006. I have two questions based on this 
: subject:
 
: 	1.  Does anyone know of a FTP site that I can go get the source for RC 
: 	    1006 implemenetation OR know of a product that I can buy that 
: 	    implements RFC-1006 ?
        
             You can try ISODE package which implemented rfc1006 , FTAM
             and other ISO stuff. The latest one is isode-8.0.

: 	2.  If you have any experience with RFC-1006 implementation, please 
: 	    share it with me in term of performance, resource management, 
: 	    bandwidth requirements or any other topics that I have not 
: 	    mentioned.  
 
: Thanks
: Neil Butani
: [email protected]


                                     Chung-Chi Chuang
                                     Data Communications Institute
                                     Taipei, Taiwan
                                     [email protected]


                            

-----------[000363][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 94 17:20:34 PST
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Radia Perlman's address?(Anybody knows BGP?)
Hi, my name is Ted.

I'm looking for the e-mail address of Radia Perlman, if anybody knows?
I'm doing research on the routing protocols (BGP, OSPF, etc.> and I don't
find a good resource which explains BGP well... If you knows about BGP,
I'd like to ask several questions (Note: I have read RFC1267 which explains
BGP's specification in detail).
What I need now is information on how BGP works in the Internet environment,
not just detail of its specification. Sincerely, Ted :)
Thanks in advance.

-----------[000364][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 94 16:42:47
From:      [email protected] (William )
To:        comp.protocols.tcp-ip
Subject:   Re: Cheap IP ethernet-to-asynch techniques?
    Hi there.  Can anyone out there suggest really cheap ways that might be
    used to connect two disjoint IP networks via Asynch modems?

    We have two ethernets in two separate facilities that need are IP-connected
    via a fractional T1 circuit.  In each location there is a CSU/DSU unit
    along with a CISCO router.

If you are interested in backing up THE LINK, you can use either SLIP or PPP
on the aux ports of the cisco routers, using "dial backup over dial-on-demand",
and get your backup connection "for free".

If you need to back up both the ciscos and the link, using slip or PPP on
any available unix systems is a fine idea, but you may have trouble figuring
out when the link should be "brought up" (of course, you could do it manually.)

BillW
cisco

-----------[000365][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 22:12:15 -0600
From:      [email protected] (bob alimi)
To:        comp.protocols.tcp-ip
Subject:   ospf Documentation: Looking for an overview
Can someone point me to some ospf documentation that is introductory
in nature.  I have only found (via archie) ietf docs that are very
technical.  I would really like to find something that would make sense
to a routing newbie (i.e., me).

Thanks much for any help,

Bob

-- 
 Bob Alimi                     | Information Resources      
 [email protected]          | Bureau of Land Management  
 [email protected]    | Sacramento, CA             

-----------[000366][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 13:04:51 GMT
From:      [email protected] (James VanBokkelen)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
[email protected] (don provan) writes:

>In article <[email protected]> [email protected] (James VanBokkelen) writes:
>>[email protected] (Samuel Lam) writes:
>>> 1) Is sending outgoing ARP replies via MAC-level broadcasts
>>> instead of unicasts allowed?
>>
>>... Vendor 1's router distracts every host on the net with every
>>ARP reply, which may in the end reduce overall load on the router
>>while making many hosts' ARP caches incrementally larger.
 
>Why would any node's ARP cache be any larger? There's no reason to
>remember an IP/MAC mapping that isn't interesting, regardless of
>whether the mapping was delivered in a broadcast request or a
>broadcast reply.

The canonical reason for caching whatever ARP IP/MAC mappings come
down the pike is the assumption that someone else might be about to
start talking to you, and it might be ICMP (no connection block to
check for the IP address).  I suppose you could discard uninteresting
Replies only, but code I'm familiar with doesn't, and I assume the
router vendor expects this, and is attempting to reduce the number of
incoming ARP Requests to the router.  Otherwise I can't come up with
even a selfish rationale for broadcasting Replies.
--
James B. VanBokkelen					Far Acres Farm
[email protected]{vax.ftp.com, asylum.sf.ca.us}			South Hampton, NH

-----------[000367][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 13:36:42 +0000
From:      [email protected] (Peter Gross)
To:        comp.protocols.tcp-ip
Subject:   Re: Getting duplicate IP addresses
I would suggest checking the /etc/services files on the DEC Ultrix machines.

We found that DEC appear to start all services by default and leave it to you
disable the ones you don't want. (We found this after a faulty terminal-server
kept requesting BOOTP from anyone and the Ultrix box repeatedly logging these
requests to a disk that filled-up :-().

PS: We don't use BOOTP here !!

This may be way of track, but I like to help if I can.

-- 
Peter Gross, UNIX Systems Manager, Brighton Health Care NHS Trust, UK

BTW, the ideas, suggestions, jokes, fantasies, complaints, comments, 
critiscms, cautions, insults, observations, diagrams, scripts, sarcastic 
remarks, stories, programs, and thoughts shown in the text above are mine 
and mine alone, and are NOT those of my employer or anyone else for that 
matter (unless I say otherwise ;-).

-----------[000368][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 13:45:05 GMT
From:      [email protected] (Ran Atkinson)
To:        comp.protocols.tcp-ip
Subject:   Re: Query about ASCII limitation TCP/IP apps
In article <[email protected]> [email protected] writes:
%  I'm trying to find out a little about the reasons behind the
% inability of some Internet applications to use anything but US ASCII
% (obvious example, SMTP). 

I fear you are not well informed about the abilities of Internet email.
If you want to transport 8-bit characters in the Internet, there is an
Internet standards-track protocol extension to do this.  The result is
something called Enhanced SMTP or ESMTP.  One can certainly use many
character sets other than US-ASCII in Email just by using the
Multipurpose Internet Mail Extensions (MIME).  MIME is available from
multiple vendors today.  Check the RFCs for more information on either
of these.

% Was it an historical accident? A choice to go with a lowest 
% common denominator? 

Hmmm.  Most of the applications in question existed before modern
computer character sets existed, so 7-bits was plenty wide enough.
There was also a certain amount of Euro-centric thinking (more than 7
bits is required by some Asian languages, but a single European
language can almost always be squeezed into a single ISO-646 variant).
MIME does fully support Asian languages (e.g. Chinese, Japanese, Korean)
and modern multi-language character sets (e.g. ISO-8859-x, ISO-10646).

% What other applications are affected? 
% Is the limitation only at the application level, or lower?

Read the specs for the applications you are interested in.  All the
RFCs are available online.  Use the online rfc-index and the most
recent version of IAB Official Standards to be sure you are locating
all of the relevant RFCs.


Ran
[email protected]



-----------[000369][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 14:21:35 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Well Known TCP Port 2049 (shilp)

Hello all,

	Does anyone know what well known port 2049 (shilp) is or
	what its function in life is ?  RFC 1340 does not state 
	a detailed description of the function of this well known
	port.  
	
	Any replies would be much appreciated. 

Thanks,
RAD

-----------[000370][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 14:23:32 GMT
From:      [email protected] (Daryl Anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: Problem w/ NCSA + Sun
In article <[email protected]> [email protected] (Dylan Greene) writes:
>Path: marlin.upe.ac.za!quagga.ru.ac.za!howland.reston.ans.net!math.ohio-state.edu!cs.utexas.edu!convex!news.duke.edu!MathWorks.Com!bigboote.WPI.EDU!tmok.res.wpi.edu!db70
>From: [email protected] (Dylan Greene)
>Newsgroups: comp.protocols.tcp-ip
>Subject: Problem w/ NCSA + Sun
>Date: 15 Sep 1994 22:47:19 GMT
>Organization: The Ministry of Knowledge BBS, Site 2
>Lines: 18
>Message-ID: <[email protected]>
>NNTP-Posting-Host: tmok.res.wpi.edu
>X-Newsreader: TIN [version 1.2 PL2]


>I've got a little net problem:
 
>I'm networking a SUN SparcClassic on a net with a linux box, and
>I'd like to setup a bunch of messy-dos machines with NCSA telnet
>on them to use PC-NFS / TeX / VT-100 access to the sun and linux
>resources..
 
>But an interesting problem has arisin..  There was no problems
>until I started putting the NCSA clients on the ethernet..
 
>For some reason, the msdos's can only make a connection with the
>linux, and not the sun.. Is there a conflict between NCSA
>and SunOS I don't know of?  The cable lengths are allset, 
>I'm really getting stuck here..
 
>Thanks..
>[email protected]
Helo,
You have come across one of the bugs in version 2.3.07
NCSA Telbin will not attach to a Solaris 2.2 or 2.3 that I know of.
The problem is the way in which NCSA handles a "do not fragment"
bit in the IP datagram header.

I fixed this and many other bugs.
You can get the latest on
  ftp.upe.ac.za:/pub/msdos/ncsa/tel23074.zip

All the new work I have done is enclosed with docs in this zip.
If you have problems email me, and we can make another plan.

Regards
Daryl

-----------[000371][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 15:29:11 GMT
From:      [email protected] (Art Berggreen)
To:        comp.protocols.tcp-ip
Subject:   Re: Query about ASCII limitation TCP/IP apps
In article <[email protected]> [email protected] writes:
>Hi. I'm trying to find out a little about the reasons behind the inability of some Internet 
>applications to use anything but US ASCII (obvious example, SMTP). Was it an historical 
>accident? A choice to go with a lowest common denominator? What other applications are 
>affected? Is the limitation only at the application level, or lower?

I think most of this dates back to the concept of the "Network Virtual Terminal" (NVT).
Remember that these protocols predate the general availability of PCs, individual
workstations and standard windowing display systems.  To deal with the variety of
terminal devices of the time, a standardized termninal model was defined (the NVT).
This virtual terminal has a printing character set based on 7 bit UASCII carried
in eight bit bytes.  The high order bit was turned on to escape to control functions.

>Any information anyone can send me (references to articles, books, net info, pointers to other 
>sources, etc.) about the background/history of this issue would be appreciated.

You might want to read the original RFC for Telnet over TCP/IP (RFC764).

Art


-----------[000372][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 15:35:08 GMT
From:      [email protected] (King)
To:        comp.protocols.tcp-ip
Subject:   Slip connection

I am trying to connect my PC which is running Slip, to a dialup
host that is running Slip.  I get a "Connect" message, and that's
it.  I would expect to get a login.  Has anyone run a dialup Slip?
If so what do you see after the Connect.

Jack
 

-----------[000373][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 15:46:53 GMT
From:      [email protected] (Russell Nelson)
To:        comp.protocols.tcp-ip
Subject:   Re: ACER 5220A driver?
In article <[email protected]> [email protected] (Robertino Benis) writes:

   Has anyone met acer 5220A & packet driver for this eth. card?

No, but it might be NE2000 compatible, so try NE2000.COM on it.

--
-russ <[email protected]>    http://www.crynwr.com/crynwr/nelson.html
Crynwr Software   | Crynwr Software sells packet driver support | ask4 PGP key
11 Grant St.      | +1 315 268 1925 (9201 FAX)  | What is thee doing about it?
Potsdam, NY 13676 | LPF member - ask me about the harm software patents do.

-----------[000374][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 17:12:49 GMT
From:      [email protected] (David D Kaas)
To:        comp.protocols.tcp-ip,vmsnet.networks.tcp-ip.misc
Subject:   VMS tcp/ip packages?
We have some Dec statio 3100s running VMS 5.x.  We would
like to install TCP/IP software.  What companies support
TCP/IP on VMS?  Are there any public domain or low cost
packages?

Thank you

Dave Kaas
| Dave Kaas                         | Internet: [email protected]               |
| Box 300 M/S A1-05                 |                                      |
| Boeing Computer Services Richland |           [email protected]      |
| Richland, Wa 99352                | Phone:    (509) 376-6386             |
-- 
| Dave Kaas                         | Internet: [email protected]               |
| Box 300 M/S A1-05                 |                                      |
| Boeing Computer Services Richland |           [email protected]      |
| Richland, Wa 99352                | Phone:    (509) 376-6386             |

-----------[000375][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 17:22:28 GMT
From:      [email protected] (don provan)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Peter P. Morrissey) writes:
>Although I would qualify that a little by saying that in general with Novell 
>you have a "File Server" which does exactly that, it serves a file. So, if you 
>are searching through a Dbase or some other database, every single record that
>you are searching through is sent to the client PC from the File server. One 
>could argue that this process is more efficient on a Unix box which usually 
>will be running the sorting application on the host and only send the results 
>of the sort to the client.

Just a minor clarification: several companies have database products
for NetWare which provide server based searching as well as other
advanced database functions that can limit the network traffic to the
data actually of interest to the client. Even out of the box, NetWare
provides a little of this, but I don't know if it would cover what
you're imagining.

I suspect you knew that, since you did say "in general", but readers
might have been misled. It's just as easy to eat up network bandwidth
accessing database files directly using NFS. I don't myself know
whether using NetWare to do it is more or less common.

						don provan
						[email protected]

-----------[000376][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 18:59:36 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Running IPX and TCP/IP over same ethernet

In article <[email protected]>, 
<[email protected]> writes:
> 
> In article <[email protected]> 
 [email protected] (Robert J. Wilson) writes:
> 
> 
> >> Direct from Rumor Control Central:  I have heard that IPX is a real
> >> bandwidth hog, and that if you don't want to see your TCP/IP severely
> >> degrade, don't let IPX run over the same segments that run TCP/IP.
> 
> We find that ipx and tcpip get along well. Cinfig is 3 Sun Sparc10, 1 NW_311,
> lot's of traffic to a fibre backbone.
>  

One thing to be careful of is SAP (Service Advertising Protocol).  Each
server on the IPX internet babbles its state every so often, and the routers
pass this on to all the other IPX networks.  On large networks, this can
consume large portions of your bandwidth.  We've seen networks where the
routers essentially stop ever 15-20 seconds while they process all the SAP
updates (of course, those are not Network Systems' routers ...).

The problem of SAP consuming bandwidth has been well understood for quite
a while.  Novell has a Netware Loadable Module that will filter SAP updates
to let you keep local updates from poluting your network.  Refer to the
excellent book from Novell Press, _Netware_LAN_Analysis_, by Laura Chappell.

Other than SAP (and a little RIP, but the issues are the same as IP RIP),
there doesn't seem to be much difference in efficiency between IP and IPX
in a packet-on-the-wire performance sense.  IPX (well, SPX) is typically
much less than TCP, since it uses a window size of 1, but that's a different
matter.

Sorry for the length of this posting, and for taking bandwidth from the
tcp-ip group.  It's probably time to take this to comp.dcom.novell, or
whatever.

- Ted

--------------------------------------------------------------------------
Ted Doty, Network Systems Corporation | phone:      +1 301 596-2270
8965 Guilford Road, Suite 250         | fax:        +1 410 381-3320
Columbia, MD, 21046 USA               | voice mail: (800) 233-1485
--------------------------------------------------------------------------
The opinion expressed in this message is fictitious.  Any resemblence to
real opinions, living or dead, is purely coincidental.

-----------[000377][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 19:03:50 GMT
From:      [email protected] (Tom Bodoh)
To:        comp.protocols.tcp-ip
Subject:   CMU TCP/IP
We used to run CMU TCP/IP on some of our VAXen.  We later upgraded to Multinet
but we now have a need for some TCP/IP capabilities on a 'temporary'
system.  Anyone know what the current status of CMU TCP/IP is?  Last I heard
it was being abandoned to the user community to support.  What was the
latest version available (6.6?), will it run on 5.52 VMS and where can it
be found?  I have located 6.6 on the net.  Thanks...

-- 
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ Tom Bodoh - Section Manager, Systems Engineering and Management, Hughes STX +
+ USGS/EROS Data Center, Sioux Falls, SD, USA	57198     (605) 594-6830      +
+ Internet; [email protected] (152.61.192.66)   Amateur radio call; N0YGT +
+	"Welcome back my friends to the show that never ends!" EL&P	      +
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

-----------[000378][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 20:15:02 GMT
From:      [email protected] (Marten Terpstra)
To:        comp.protocols.tcp-ip
Subject:   Re: Well Known TCP Port 2049 (shilp)
In <[email protected]> [email protected] writes:

>	Does anyone know what well known port 2049 (shilp) is or
>	what its function in life is ?  RFC 1340 does not state 
>	a detailed description of the function of this well known
>	port.  
>	
>	Any replies would be much appreciated. 

Most common use of port 2049 is NFS.

-Marten

% grep 2049 /etc/services
nfs		2049/udp			# usually but not always!
nfs		2049/tcp			# only on bsdi boxes

-----------[000379][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 20:23:25 GMT
From:      [email protected] (Rick Jones)
To:        comp.protocols.tcp-ip
Subject:   Re: Well Known TCP Port 2049 (shilp)
: 	Does anyone know what well known port 2049 (shilp) is or
: 	what its function in life is ?  RFC 1340 does not state 
: 	a detailed description of the function of this well known
: 	port.  

$ grep 2049 /etc/services
nfsd        2049/udp                 # NFS remote file system

More information can be found in the RFC for NFS.

rick jones

-----------[000380][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 21:50:04 GMT
From:      [email protected] (Byron Lunz)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.ibmpc,comp.protocols.ppp
Subject:   Re: [Q] Low end router?
In article <[email protected]>, [email protected] (Russo and Hale) says:
>
>[ Article crossposted from comp.sys.ibm.pc.hardware.networking ]
>[ Author was Russo and Hale ]
>[ Posted on 10 Sep 1994 03:41:53 GMT ]
>
>Does anyone know of a good router that is on the low cost end?  I want to 
>connect a Class C ethernet (winsock) to a PPP dialup provider (Portal 
>Communications).

Copied from:

                              W E B s t e r
                              -------------                          
Vol. 1: No. 1              The Cyberspace Surfer              Sept. 14, 1994 


Compatible Systems Pursues the Corporate Connection
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  A local area network to Internet router for under $1,000 was introduced by
Compatible Systems Corp. of Boulder, Colo. The company made the new product
announcement for the MicroRouter 900i, priced at $995, on Sept. 6. 

  "Companies are beginning to use Internet connectivity as a competitive
advantage," says Matt McConnell, president of Compatible Systems. "The
MicroRouter 900i allows full Internet access for all computers and
workstations on a corporate LAN. No other IP router even comes close to the
management and performance features you get with the MicroRouter 900i." 

  The router supports high-speed leased or switched line service through a
CSU/DSU. ISDN service is available with standard terminal adapters. The
router can also provide connects over standard low-speed telephone lines via
a modem. Dedicated or dial-on-demand connections are available to Internet
service providers. IP traffic is routed using  PPP protocol. The router comes
with Windows and Macintosh management software.

-----------[000381][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 23:09:37 GMT
From:      [email protected] (Herb Rosenberg)
To:        comp.protocols.tcp-ip
Subject:   Looking for Reomote Control Software via TCP/IP?
Subject: Looking for Reomote Control Software via TCP/IP?
Newsgroups: comp.dcom.lans.ethernet
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
Summary: 
Keywords: 

I am looking for some remote control software like a PC Anywhere, Close 
UP, Carbon Copy, Reach Out, etc, but one that I can use for remote 
control across a WAN using TCP/IP and not IPX/SPX.  All of these other 
products (to the best of my knowledge) require ipx/spx to work.

I am trying to minimize or eliminate IPX across my wan, with only ipx on 
local segments, and wan / backbone traffic to be only IP.  HOwever, I 
have a large number of using that need to do trure pc to pc remote 
control and I would like an IP based solution.

Any suggestions or comments would be greatly appreciated.

Thanks.



-- 
[email protected]

-----------[000382][next][prev][last][first]----------------------------------------------------
Date:      Fri, 16 Sep 1994 23:09:43 GMT
From:      [email protected] (Syed Zafar Taqvi)
To:        comp.protocols.tcp-ip
Subject:   FAQ
Is there a FAQ for this group?
If it exists, where can i access it?

thanks,
zafar

-----------[000383][next][prev][last][first]----------------------------------------------------
Date:      16 Sep 1994 23:31:47 GMT
From:      [email protected] (William C. Fenner)
To:        comp.protocols.tcp-ip
Subject:   Re: Cheap IP ethernet-to-asynch techniques?
In article <[email protected]>,
Phil Howard <[email protected]> wrote:
>Actually it should be possible now days for someone to integrate an ethernet
>and a small router with slip and ppp into the same box as a modem (V.FC/V.34
>please) so that you have a BNC or switched ethernet twisted pair plus the usual
>phone line.  If they can make this thing auto-dial on demand with a decent
>scripting language they would sure have a hot product.

Sounds something like the Telebit Netblazer PN1, and/or the Rockwell NetHopper.

  Bill
-- 
Bill Fenner                  [email protected]

-----------[000384][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 94 00:14:09 GMT
From:      arnaud
To:        comp.protocols.tcp-ip
Subject:   Could someone explaine me what's TCP/IP?


	Hi,

	I have been hearing this word for too long,
	Could someone explain me in details what is
	TCP/IP...

	Thanks,

	If I could get some answer by mail at :	
	[email protected]

-----------[000385][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 94 00:26:17 GMT
From:      [email protected] (Mitch Butler)
To:        comp.protocols.tcp-ip
Subject:   ARP question
What is supposed to happen when a multicast MAC address is returned in
an ARP reply?

Our workstations (Axil, running SunOS 4.1.3) don't care. If they ARP to
find the Ethernet MAC address for some IP address and get a multicast
MAC address in the reply, they will happily place it in the ARP cache
and send packets to the multicast address.

However, our router does care.  If it ARPs to find the MAC address for
some IP address and gets a multicast MAC address in the reply, it
throws it away and continues to ARP, looking for a "valid" MAC
address.

Our router vendor thought that there was an RFC which specified this
behavior. But they didn't know the RFC number, and I can't find
anything.  So, is there an RFC which deal swith this case?

P.S. I know all about "true" IP multicasting (RFC 1112) and how it
     maps multicast IP addresses to multicast MAC addresses.

     In this case, we want to map a "normal" IP address to a multicast
     (not broadcast) MAC address.

======================================================================
    [email protected]
======================================================================



-----------[000386][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 1994 01:04:38 GMT
From:      [email protected] (Tom Fitzgerald)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
> [email protected] (don provan) writes:
 
> > Why would any node's ARP cache be any larger? There's no reason to
> > remember an IP/MAC mapping that isn't interesting, regardless of
> > whether the mapping was delivered in a broadcast request or a
> > broadcast reply.

[email protected] (James VanBokkelen) writes:

> The canonical reason for caching whatever ARP IP/MAC mappings come
> down the pike is the assumption that someone else might be about to
> start talking to you, and it might be ICMP (no connection block to
> check for the IP address).  I suppose you could discard uninteresting
> Replies only, but code I'm familiar with doesn't

Hmmm.  Don is correct (and my post concerning ARP table overflows was dead
wrong).  By RFC 826, the receiver of an ARP packet should only act on it if
its IP address is the target-IP of the ARP, or if the sender's IP address
is already in the receiver's ARP cache (in which case the receiver should
update the MAC address).  No new ARP entry will be added by any system
except the one with the matching target-IP address.

> and I assume the
> router vendor expects this, and is attempting to reduce the number of
> incoming ARP Requests to the router.  Otherwise I can't come up with
> even a selfish rationale for broadcasting Replies.

It might not reduce the number of incoming ARPs, though.  Systems that
don't already have that ARP mapping cached will ignore the reply.  It might
reset the timeout counter in existing ARP cache entries, so maybe it will
avoid some ARP cache timeouts.....

Other than that, the only "positive" side effect of the broadcast reply is
that it updates the MAC address in ARP caches that already contain that
mapping.....  this might be useful in cases where a system changes its MAC
address, but a better solution is for the router to make a single
unsolicited ARP request broadcast following the change, containing the new
MAC address.

-- 
Tom Fitzgerald      1-508-967-5278
Wang Labs           [email protected]         (account no longer active)
Lowell MA, USA      [email protected]       (use this address instead)

-----------[000387][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 1994 03:14:11 GMT
From:      [email protected] (Richard Huddleston)
To:        comp.protocols.nfs,comp.protocols.tcp-ip,comp.unix.admin
Subject:   NFS Proxy Servers?
As is usual, when I saw information on this topic I didn't need it
--and now that I need it, I can't remember where I saw it.

Any pointers appreciated, and I will post a summary.

Richard

-- 
Richard Huddleston	: A clever saying escapes me at the moment. 
University of Maryland	:	
CMSC/ANTH		: 
ObVanity_URL	  	: http://www.wam.umd.edu/~reh/ 

-----------[000388][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 1994 06:49:28 GMT
From:      [email protected] (Frederick Scott)
To:        comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
[email protected] (Phil Howard) writes: 

>[email protected] (Rob Tanner) writes:
>
>>Direct from Rumor Control Central:  I have heard that IPX is a real
>>bandwidth hog, and that if you don't want to see your TCP/IP severely
>>degrade, don't let IPX run over the same segments that run TCP/IP.
>
>Maybe.  IPX waits for acks for every packet, for instance.

Did you mean SPX?  IPX is a datagram protocol.

Fred

-----------[000389][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 1994 10:48:24 GMT
From:      [email protected] (Pete A. Zaitcev)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and ARCHNET
In <[email protected]> [email protected] (Sergey A. Elistratov) writes:

>In article <[email protected]>, Patrick Klos <[email protected]> wrote:
>>
>>This is NOT true.  ARCNET can support TCP/IP as well as any other network
>>media.  ARCNET will be prone to fragmentation (either IP or ARCNET) since
>>the maximum data size for any single packet is 508 bytes.

This happens only if you use RFC-1051 framing. Current implementation
must follow the RFC-1212 which includes datalink level fragmentation.
RFC-1212 is good because it does not duplicate IP headers in fragments
and it allows retries when the hardware reports a packet loss.

>But where can i find drivers for ARCNET cards for UNIX ?
>For FreeBSD etc. ?

Write it yourself, the hardware is extremely simple. I did this once.
Other sources are Russ Nelson's packet driver (Sorry Russ, it was bad
design: lost interrupts, transmitter lockups) which is RFC-1051, driver
for Coherent (not sure was it Mark Williams :-), and exellent Igor
Chechik's driver for the ISC UNIX. Unfortunately, Chechik's pet is for
pseudo 802.2. But hardware management is correct. You do not know
where he lives? Excuse me, every russian hacker must know. And western
ones probably wann't to waste their time with ARCnet. :-)

Pete

-----------[000390][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 1994 11:56:35 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Trumpet Winsock slip problems
does anyone have the correct settings for running trumpet winsock under slip?



-----------[000391][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 94 21:20:23 -0500
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Ipswitch Inc's TCP/IP Stack?
	
Hello,
 
	I'm trying to find the email address of Ipswitch but haven't been
able to.  Is it possible to use this tcp/ip stack as a server?  Run it on
a machine and than have it act as a SLIP/PPP server?
	Is this possible with all of the tcp/ip stacks?  Or just particular
ones?
 
	Thanks & Take care.
 
	Vikram

-----------[000392][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 1994 21:09:44 -0400
From:      [email protected] (Craig Weeks)
To:        news.announce.newgroups,news.groups,comp.client-server,comp.os.misc,comp.os.ms-windows.networking.tcp-ip,comp.os.os2.networking.tcp-ip,comp.protocols.tcp-ip,comp.unix.osf.osf1
Subject:   RFD: comp.soft-sys.dce
                 REQUEST FOR DISCUSSION
                 ----------------------

         Group Name:  comp.soft-sys.dce
             Status:  unmoderated
       Distribution:  world wide
            Summary:  Discussion of DCE implementations and applications
                      on all platforms (UNIX, OS/2, Windows, etc.)
        Proposed by:  Craig Weeks
                      ([email protected])


This is a formal Request For Discussion (RFD) on the creation of an un-
moderated newsgroup comp.soft-sys.dce.

This RFD is being cross-posted to the following newsgroups:

     news.announce.newgroups
     news.groups
     comp.client-server
     comp.os.misc
     comp.os.ms-windows.networking.tcp-ip
     comp.os.os2.networking.tcp-ip
     comp.protocols.tcp-ip
     comp.unix.osf.osf1

The Call For Votes (CFV) will be posted 21-30 days after posting this RFD.
Vote counting will be conducted by a neutral party.

RATIONALE

There is currently no newsgroup dedicated to the Distributed Computing
Environment (DCE) developed by the Open Software Foundation (OSF).  DCE is
discussed occasionally in other groups today but the subject addressed by
those groups is too broad to make discussion of DCE difficult for those who
are not interested in the other subject areas these groups address.

CHARTER

The newsgroup comp.soft-sys.dce is intended for discussions of any and all
DCE implementations and applications across all hardware platforms and
operating system.  The topics to be discussed include (but are not limited to)
the following:

     - Porting DCE applications between platforms
     - Configuring DCE cells
     - Installation of DCE
     - DCE performance
     - Network transport questions
     - Suggestions for future enhancements
     - Success Stories
     - Application development tools

Craig Weeks
IBM Corp.
Austin, Texas

512-838-3041

These opinions are mine, not IBM's.

-----------[000393][next][prev][last][first]----------------------------------------------------
Date:      Sat, 17 Sep 1994 18:00:49 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   TCP/IP On NetWare Client
	Could someone please tell me if a NetWare client
can communicate with a NetWare v3.11 server using TCP/IP
instead of IPX?

Can this be done with LAN Workplace For DOS?

If so, how would the client be configured/installed?

			Thanks!

-----------[000394][next][prev][last][first]----------------------------------------------------
Date:      17 Sep 1994 22:06:55 GMT
From:      [email protected] (Andrew B Consor)
To:        comp.protocols.tcp-ip
Subject:   Is dial-up SLIP telnetable?
Thanks to you guys I now understand the difference between SLIP and 
regular dial-up.  But what I do not understand is why you would want 
SLIP other than the fact that you can use mosaic. Let's say you have a 
dial-up SLIP connection. Can you do things like telnet to yourself from
another point on the internet.  Can you use your entire hard drive as an
FTP site for people to get files from? What I am asking is basically if
you need to be hooked up to the internet via a phone line then what good
does an IP address do you if the answer of my previous questions is no?  
Also if you have only a 14.4k modem and a regular phone line..  Would'nt 
you want to use dial-up because then you would be using your internet 
service provider's connection which is probably T1.  If you do something 
like WWW using your connection it is going to be dirt slow, right.
I have probably confused half the people here but it is my nature.


((((((((((((((((())))))))))))))))
( This SIG is dedicated to      )
( Richard "Tricky Dick" Nixon   )
*********************************
| Conservative Republicans Rule!|
| Support NRA: fight the power  |        
| Right to Life                 |
| Ollie for Senate !            |
 *******************************


-----------[000395][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 00:43:15 GMT
From:      [email protected] (steve c. whalen)
To:        comp.protocols.tcp-ip
Subject:   WfWG TCP/IP
While browsing the Microsoft BBS under Windows for Workgroups I found two 
files concerning TCP/IP.  They are 1) wfwt32.exe described as "TCP/IP-32 
for WFW" and 2) wfwtcp.exe described as "MS TCP Protocol for WFWG 3.11".
Can these two files be used in conjunction with Mosaic, and if so how are 
they used?  Thanks


-----------[000396][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 03:35:33 GMT
From:      [email protected] (Matthew Gream)
To:        comp.protocols.tcp-ip
Subject:   Verification of a TCP implementation

What's the usual way people go about verifying a TCP/IP implementation?

Even if I can show both forwards and backwards traceability between
specs and implementation, is there something like a "test suite"
that'll exercise those bits of the protocol that usually come out wrong
or are problematic? 

The choice of a "bake off" with other implementations doesn't appeal to
me because it seems like it's only going to test general functionality
-- ie. only half a solution.

Ideas ?

Matthew.

--
Matthew Gream 
<[email protected]>
(02) 821-2043
(sw/hw engineer)

-----------[000397][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 94 12:27:39 -0500
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Re: Is dial-up SLIP telnetable?
Andrew B Consor <[email protected]> writes:
 
>Thanks to you guys I now understand the difference between SLIP and 
>regular dial-up.  But what I do not understand is why you would want 
>SLIP other than the fact that you can use mosaic. Let's say you have a 
>dial-up SLIP connection. Can you do things like telnet to yourself from
>another point on the internet.  Can you use your entire hard drive as an
 
 
 
	You can have bi-directional telnet, ftp, irc, muds, usenet, email
gopher sites, www server and anything you want with a SLIP account.  The
big limitation is that you are using a telephone line which is obviously
limited by bandwith.
	SLIP stands for "Serial Line Internet Protocol".  It is a way of
transferring tcp/ip packets over normal voice grade telephone lines.
Dedicated SLIP accounts are often employed by small buisnesses in order for
their office LANs to have a direct tcp/ip connection.
	However, if you wish to deal with IP over serial lines, you should
deal with PPP.  It is more robust, supports error cor{ection, assign IP
addresses automatically.  It is also the standard by the IETF.
	Feel free to correct any mistakes I made and have a nice day.
 
	

-----------[000398][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 94 04:50:17 GMT
From:      [email protected] (Tom Evans)
To:        comp.protocols.appletalk,comp.protocols.tcp-ip,comp.sys.mac.comm
Subject:   Re: How good is MacTCP?
In article <[email protected]>, [email protected] (Thomas Kofler) writes:
> Hi everyone.
> I am using MacTCP 2.0.2, 

Get the 2.0.2->2.0.4 patcher, it fixes some bugs. From the tidbits
site and ftp.apple.com and mirrors.

> and I got the impression (by watching
> TCP traffic using etherfind on a Sun) that the implementation
> is not optimal. In particular, I wonder wether MacTCP causes
> the problem of not acknowledging fragment IP packets (TCP packest)
> sometimes (used in conjunction with MacPPP 2.0.1).

TCP Acks and IP fragments SHOULD be happeneing at way different layers
of the IP stack. Test something simpler, like UDP or ICMP first.

If you use MacPing and send ICMP Echoes of different sizes to
another Macintosh, it goes haywire when you exceed the MTU -
especially more than 576 (odd) bytes over LocalTalk. Also bad over
ARA, implying timing problems over slow links (as yours is over
PPP). If you can possibly manage it, don't send or receive IP
fragments.

Test it by sending pings to your Mac from an IP host - try packet sizes
from 50 to 5000 and see if anything "interesting" happens.

My best effort was sending large ICMP pings mac-to-mac and having both
of them crash simultaneously and horridly. Of course this happened on
a customer's site...

========================
Tom Evans  [email protected]
Webster Computer Corp P/L, 11 Glenvale Crescent Mulgrave, Melbourne 3170
Victoria, Australia 61-3-561-9999  FAX ...560-0067  A.C.N. 004 818 455


-----------[000399][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 10:03:45 +0000
From:      [email protected] (Andy Leigh)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP On NetWare Client
In article <[email protected]>
        [email protected] writes:

>       Could someone please tell me if a NetWare client
> can communicate with a NetWare v3.11 server using TCP/IP
> instead of IPX?

Not at the moment. I have heard that Novell plans to run Netware Core
Protocol over IP, but until that is released, you are stuck with IPX as
your network layer protocol.

It might be wise to take this to comp.sys.novell

--

Andy Leigh               [email protected]  -  Home (preferred)
                         [email protected]    -  Work (RFC822)

-----------[000400][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 18:35:18 EDT
From:      [email protected] (The Old Bear)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Re: Question on "firewalls"
In article <[email protected]> [email protected] ( Ed Alcoff) writes:
>From: [email protected] ( Ed Alcoff)
>Subject: Question on "firewalls"
>Date: 19 Sep 1994 18:20:18 GMT
 
>Greetings:
 
>I apologize in advance if this is a FAQ:
 
>I understand that there are both commercial and p.d.
>software that allows one to set up a "firewall" to protect
>a TCP/IP network from external access.
 
>1.  How is this different than what can be done with
>a decent router?
 
>2. Where can I find such products?



For a good discussion of the issues, get hold of a copy of the following
book:
        Firewalls and Internet Security
        (Repelling the Wily Hacker)
        by
        William R. Cheswick and Steven M. Belleview

        Addison-Wesley Professional Computing Series   $26.75




-----------[000401][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 15:08:56 GMT
From:      [email protected] (Pete A. Zaitcev)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
In <[email protected]> [email protected] (Tom Fitzgerald) writes:

>Other than that, the only "positive" side effect of the broadcast reply is
>that it updates the MAC address in ARP caches that already contain that
>mapping.....

Detects the duplicate IP address which is equal to *your*.
No other way to catch this. Or only ARP asking system will
see duplicates. Helpful for debug typo errors in /etc/hosts?

Pete

-----------[000402][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 15:36:16 GMT
From:      [email protected] (Daryl Anderson)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP On NetWare Client
In article <[email protected]> [email protected] writes:
>Newsgroups: comp.protocols.tcp-ip
>Path: marlin.upe.ac.za!quagga.ru.ac.za!howland.reston.ans.net!torn!watserv2.uwaterloo.ca!undergrad.math.uwaterloo.ca!att-out!pacbell.com!ames!agate!darkstar.UCSC.EDU!news.hal.COM!olivea!charnel.ecst.csuchico.edu!yeshua.marcam.com!MathWorks.Com!zombie.ncs
>.mil!news.duke.edu!godot.cc.duq.edu!newsfeed.pitt.edu!dsinc!netnews.upenn.edu!newsserver.jvnc.net!adl33cc!SABLE.ADELPHI.EDU!FISCHMAN
>From: [email protected]
>Subject: TCP/IP On NetWare Client
>Message-ID: <[email protected]>
>Sender: [email protected] (USENET News System)
>Nntp-Posting-Host: sable.adelphi.edu
>Reply-To: [email protected]
>Organization: Adelphi University, Garden City, NY
>Date: Sat, 17 Sep 1994 18:00:49 GMT
>Lines: 9


>        Could someone please tell me if a NetWare client
>can communicate with a NetWare v3.11 server using TCP/IP
>instead of IPX?
 
>Can this be done with LAN Workplace For DOS?
 
>If so, how would the client be configured/installed?
 
>                        Thanks!
I think Netware/IP is the beast you looking for.
You load a TSR I believe is called NWIP.EXE on top of TCPIP.EXE.
This gives IPX over IP!
NetBIOS is apparently not fully supported, but look into the product
yourself.  I have only heard about this and do not use it myself.
Regards
Daryl

-----------[000403][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 16:17:46 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
In article <[email protected]> [email protected] (Charles Yamasaki) writes:
>I don't know of any way to have multiple addresses on the same
>interface, but you could certainly plug in multiple NIC's.  The problem
>here is that you need multiple router ports (maybe).

4.4BSD has multiple addresses per NIC. Linux has the 'dummy' virtual driver
for hanging addresses onto. Any any other host just stuff in some fake 
slip links.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000404][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 16:21:07 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: Triangular Network
In article <[email protected]> [email protected] (Thanapan Ittisakulchai - SCCS - 3670506) writes:
>   Now I am studying about feasibility to implement Triangular Network
>(Senior Project).

	Get some papers on Spanning Tree Algorithms.

>      1. About feasibility in Technically.

	Easy. It's done in the real world. Whats harder is spotting breaks
fast and handling them cleanly as opposed to recovering over timeouts (see
RIP)

>      2. Suitable ? (If base on SCO-UNIX.)

If you need to do no kernel side hacking. It's probably best done in user
mode for a project anyway.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000405][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 16:23:38 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] : TCP and NETWORK UNREACHABLE
In article <[email protected]> [email protected] writes:
>What would be the expected behaviour by TCP (in ESTABLISHED state) when an
>NETWORK UNREACHABLE is received ??

Depends if your stack is compliant or old BSD. A compliant stack _MUST_
report all ICMP messages to the user (RFC1122).
>Should it give up the connection or retransmit ?
Up to you. Most people leave the system to retry out as unreachables are
often transient.

Alan


-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000406][next][prev][last][first]----------------------------------------------------
Date:      Sun, 18 Sep 1994 16:25:50 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP over OSI stack
In article <[email protected]> [email protected] writes:
>	1.  Does anyone know of a FTP site that I can go get the source for RC 
>	    1006 implemenetation OR know of a product that I can buy that 
>	    implements RFC-1006 ?

ask archie for ISODE
>
>	2.  If you have any experience with RFC-1006 implementation, please 
>	    share it with me in term of performance, resource management, 
>	    bandwidth requirements or any other topics that I have not 
>	    mentioned.  

Read the ISODE manual

Commiserations in advance

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000407][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 16:38:26 GMT
From:      [email protected] (Pat Miller)
To:        comp.protocols.tcp-ip
Subject:   SLIP & MTU

 Does anyone know of a way to figure out the best MTU settings for 
Various speed

I am running a slip connection over a terminal server direct connection 
--(the quality of the line does not seem to be the best I seem to get better 
Throughput at 9800 baud but I can go up to 38400 with no problem when acting
as a terminal.  Next Iam telneting to a Linux machine supplying me with
a CSLIP connection.  



-----------[000408][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 16:48:03 GMT
From:      [email protected] (Sam Oscar Lantinga)
To:        comp.protocols.tcp-ip
Subject:   Extending the IP Protocol?

	I would like to implement a new option in the IP protocol:

IPOPT_RELAY

	This would be used to relay IP packets from one network
to another, providing a sort of distributed subnet in which not all
of the machines on the subnet are on the actual subnet wire.
The reason I would like to do this is so that my home machine can
dialup a SLIP server and appear as a machine on my network at work,
as well as a machine on the SLIP server's subnet.  

	My question is....  Will packets with an unrecognized
IP option like this be forwarded across routers?  I noticed that
in the Linux network code that if a packet has an unrecognized IP
option, the packet is discarded.  Is this standard behavior?  Or
will my custom packets make it safely across the network?

	On another note, it may be necessary to do full packet
encapsulation for this to work properly.  Is it possible to select
a new protocol number say "encap" that would correspond to a simple
encapsulation protocol? 

Thank your for any insights or suggestions. :)

	-Sam Lantinga		([email protected])



-----------[000409][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 19:14:17 +0200
From:      [email protected] (Casper H.S. Dik)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
[email protected] (Jay Ashworth) writes:

>If I correctly understand all the pieces (based on my reading of about
>6 of the Nutshell books), in order to answer to more than one IP
>address with one MAC layer address, you'd need to ARP all of them...
>and I don't believe ARP provides for this.  Now, if your primary link
>is over a point to point link, it's may be a different story, but this
>would still require you to be able to ifconfig the interface with
>multiple simultaneous IP addresses... and I haven't seen anything like
>this mentioned anywhere--and I read a _lot_.  :-)

It's really no problem to arp multiple addresses on on ethernet port.
ARP maps IP->ethernet and the way this mapping is used, a many to one
mapping is not illegal.  (Ethernet addresses never need to be
mapped back to the IP address, only the IP address needs to be
mapped to a MAC address).

In Solaris 2.x it's possible to do so:

	# Primary address:
	ifconfig le0 ....
	# More addresses:
	ifconfig le0:1 ....
	# Or even more
	ifconfig le0:100 ....


I don't think there's a limit.  BSD 4.4. does something similar, I think.
(Obviously, you don't read comp.unix.solaris :-)

Casper

-----------[000410][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 22:58:28 GMT
From:      [email protected] (William C. Fenner)
To:        comp.protocols.tcp-ip
Subject:   Re: Exotic ARP Protocol Questions
In article <[email protected]>, Pete A. Zaitcev <[email protected]> wrote:
>Detects the duplicate IP address which is equal to *your*.
>No other way to catch this.

That's not true; you can ARP for your own address if you're interested in
finding out if someone else is using it.

  Bill
-- 
Bill Fenner                  [email protected]

-----------[000411][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 1994 23:26:58 GMT
From:      [email protected] (Josh Cohen)
To:        comp.protocols.tcp-ip
Subject:   Congestion control (TCPIP)

I have been reading various sources on TCPIP and such, and I have a 
question on congestion control.  I manage tcpip nets over data packet
wan and routers.  My question is more about the standards than technical
detail.  

We frequently have problems when a HP station spools LPD data from a
token ring lan across the wan via an IBM 6611 router.  As expected there
is a congestion bottleneck at the router ( it connects the t/r to a 56k
link)  Certain works indicate that the sender should be halfing its
window in the presence of congestion... and so on.  Also I have read that
the router should be sending ICMP source quence msgs, but I have also
read that this is prohibited.

What gives? people constantly mention 'the new standard' and that it
recommends x y or z for congestion control ( slow start, multiplicative
decrease, etc on the TCP level and ICMP on the IP level )

Where can I find the REAL new standard?  
I have the RFC for TCP and IP, I have the jacobson paper, and other do dads
but I have found conflicting reports... 

I am trying to test products on our WAN for safety and protocol compliance.
It seems easier to see what the products are doing than to find what the 
'real' standard is ... ;-)

Could someone please point me in the right direction.

Thanks alot..
Josh Cohen
( if you reply with a post, please email it to me also.. )

-----------------------------------------------------------------------------
Josh Cohen					    UPSnet Technical Services
Telecommunications Analyst			           Network Management 
United Parcel Service		ATLAS 283-8803	   UNIX Systems Administrator
[email protected] 		(201)-828-8803		 Scott Mcnealy is GOD
-----------------------------------------------------------------------------
"The opinions expressed here are my own and do not represent the views
or opinions of United Parcel Service."
-----BEGIN PGP PUBLIC KEY BLOCK----------------------------------------------
mQCNAi5wpK8AAAEEAPbwPQUT3259SrGDPsgZx+YS5z+7iw1b56Hg4Md/q+8aRakW
CgX9OSeOh7TPwYmp/Iws3A+MDjleRHd1K1+H1MZuVopZYbQRhHMdN7Ofzh58lRcS
OqU1G9fT6iDIIjnVs8/i10/GPrulI0iJsafh6P3TCz2Vtj5xiQWsUmXn5cnLAAUT
tC5Kb3NoIFIuIENvaGVuICh0ZWwyanJjKSA8am9zaEB0ZWxlY29tLnVwcy5jb20+
tCRKb3NoIFIuIENvaGVuIDxqb3NoQHRlbGVjb20udXBzLmNvbT4=
=f4Bu
-----END PGP PUBLIC KEY BLOCK------------------------------------------------
--
-----------------------------------------------------------------------------
Josh Cohen					    UPSnet Technical Services
Telecommunications Analyst			           Network Management 
United Parcel Service		ATLAS 283-8803	   UNIX Systems Administrator
[email protected] 		(201)-828-8803		 Scott Mcnealy is GOD
-----------------------------------------------------------------------------
"The opinions expressed here are my own and do not represent the views
or opinions of United Parcel Service."
-----BEGIN PGP PUBLIC KEY BLOCK----------------------------------------------
mQCNAi5wpK8AAAEEAPbwPQUT3259SrGDPsgZx+YS5z+7iw1b56Hg4Md/q+8aRakW
CgX9OSeOh7TPwYmp/Iws3A+MDjleRHd1K1+H1MZuVopZYbQRhHMdN7Ofzh58lRcS
OqU1G9fT6iDIIjnVs8/i10/GPrulI0iJsafh6P3TCz2Vtj5xiQWsUmXn5cnLAAUT
tC5Kb3NoIFIuIENvaGVuICh0ZWwyanJjKSA8am9zaEB0ZWxlY29tLnVwcy5jb20+
tCRKb3NoIFIuIENvaGVuIDxqb3NoQHRlbGVjb20udXBzLmNvbT4=
=f4Bu
-----END PGP PUBLIC KEY BLOCK------------------------------------------------

-----------[000412][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 94 13:00:17 -0800
From:      [email protected] (Troi Eisler)
To:        comp.protocols.tcp-ip
Subject:   Re: SMTP from AS400 to Sun
If you are trying to transfer a file to/from AS400 and Unix machine you
should be using FTP not SMTP.  SMTP is for Simple Mail (Transport Protocol)
while FTP is File Transport Protocol.  FTP works very well for transfering
files between these two machines.  I am using it at my site.

-----------[000413][next][prev][last][first]----------------------------------------------------
Date:      18 Sep 94 23:53:46 GMT
From:      [email protected] (Dan Lanciani)
To:        comp.protocols.tcp-ip
Subject:   Re: LPD implementations
In article <[email protected]>, [email protected] (Travis L Priest) writes:
| In article <[email protected]> [email protected] (William ) writes:
| 
|    From: [email protected] (William )
|    Newsgroups: comp.protocols.tcp-ip
| 
|    The RFC describing the LPD protocol (RFC1179) implies that it is
|    preferred to send the control file before sending the data files.
|    Does anyone know which unix vendors ship with an lpd support that
|    actually does this?  The systems we have here all seem to send the
|    data file first...
| 
| I would be surprised if any vendors shipped lpds that sent the cf file
| first (offhand, SGI, Sun (OS4.x), and Cray do not).  I suspect that
| most vendors picked up the UCB source at some point and 'ported' it to
| their systems without (m)any changes to the queueing procdure.  Since
| the UCB lpd does not begin to process a job until the data connection
| with the remote client is closed, it shouldn't matter if the chicken
| or the egg comes first.

Actually, I believe there are systems which will fail if you try to
send them the control file first.  (Or, to be more exact, they will
fail if you do not send the control file last.  If I remember correctly,
something gets kicked when the control file is completely received
and that something will get upset if the expected data files aren't
there.)

The important thing to remember about RFC1179 is that it is a description
of the way the author would like an lpd protocol to operate.  It is by
no means a description of the lpd protocol currently in use (or ever
used) and it is a terrible guide from which to implement lpd-compatible
code.

				Dan Lanciani
				[email protected]*

-----------[000414][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 02:56:18 UTC
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   [Q] RFC854 includes telnet server

If a specification states that a product must provide 'telnet' in accordance
with RFCs 854, 855, 856, 857, 858, 859, 860, and 861 that is your basic
'telnet' product that allows you to communicate with other machines running
the TCP/IP suite.  Does that also mean that the product must also provide a 
'telnet server' or is that discussed in some other RFC ?

-------------------------------------------------------------------------
To find out more about the anon service, send mail to [email protected]
Due to the double-blind, any mail replies to this message will be anonymized,
and an anonymous id will be allocated automatically. You have been warned.
Please report any problems, inappropriate use etc. to [email protected]

-----------[000415][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 05:09:06 GMT
From:      [email protected] (Sam Oscar Lantinga)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
Rob Janssen ([email protected]) wrote:
: >	I would like to implement a new option in the IP protocol:
 
: >IPOPT_RELAY
 
: Please explain how this option is going to accomplish anything...

	Sure.  A modified version of tcpdump snarfs packets 
destined for the non-existent host.  It appends the destination
address to the packet using the IPOPT_RELAY option, and rewrites
the destination address as the SLIP-connected host.  The packet
is then sent back out on the network with a new destination, leaving
the source address unchanged (using raw sockets and a simple kernel
hack)  When the packet is received by the destination host, the 
IPOPT_RELAY option is detected, the destination address is re-written
to be the original destination
_and_sent_out_a_second_loopback_interface ifconfig'd to have the
proper destination address to "exist" on the network at work.

	Complex, I admit, but I'm not sure of any other way
to acomplish having my machine on the SLIP connection also appearing
on the network at work.

: IP-in-IP encapsulation is already being used to tunnel IP packets
: between nonstandard networks over another network.  In fact, *two*
: protocol numbers have already been allocated to this purpose, probably
: by oversight...  (numbers 4 and 94)

	Where can I find out more about this?

: Can't it be done by plainly adding some static routes on the routers?
: Or by using proxy ARP?

	How is this done?  I have tried adding host routes to the
Linux machine at work (not a router), but I can't detect any packets
getting to my machine on the SLIP connection (and yes, IP forwarding
IS enabled).  The SLIP router will not (as far as I can tell) forward
any packets to the SLIP connected host that are not destined for it.
I have tried adding the "ghost" IP address to my system's arp cache
using 'arp -s IP_address HW_address', but when an arp request goes
out for the arped address, the Linux box does not reply.
Is this proper proxy-arp method?  I've seen alot of discussion of
what it does, but none on how to set it up. 

Any help would be appreciated. :)

Thanks!

	-Sam


-----------[000416][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 12:38:36 -0400
From:      [email protected] (Barry Margolin)
To:        comp.protocols.tcp-ip
Subject:   Re: network breakdown
In article <[email protected]> [email protected] (Alex Neutjens) writes:
>If the network of a Unix site goes down,
>then the time between the physical network-breakdown and the detection for an 
>application that there is no network-communication available, is exactly 20 
>minutes.
>This is a standard in tcp-ip. In case of a "semi" real-time application, this 
>time-out is unacceptable.
 
>Does tcp-ip provide a better way to solve this problem ? 

No.  TCP/IP is designed to survive across temporary network problems.
Remember its initial goal: to serve as the protocol for the military, and
allow connections to stay up even when the enemy has knocked out portions
of the network.

For instance, some of the common routing protocols take several minutes to
stabilize when a router goes down and a backup must take over.  On media
like thick and thin coax ethernet, when a new segment is being added to a
daisy-chained segment, that segment will be down while cables are being
unplugged and reconnected.
-- 

Barry Margolin
BBN Internet Services Corp.
[email protected]

-----------[000417][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 06:02:55 GMT
From:      [email protected] (Jim Raykowski)
To:        comp.protocols.tcp-ip
Subject:   Re: Slip connection
In article <[email protected]> [email protected] (King) writes:
>From: [email protected] (King)
>Subject: Slip connection                                    
>Date: Fri, 16 Sep 1994 15:35:08 GMT


>I am trying to connect my PC which is running Slip, to a dialup
>host that is running Slip.  I get a "Connect" message, and that's
>it.  I would expect to get a login.  Has anyone run a dialup Slip?
>If so what do you see after the Connect.
 
>Jack
> 
Jack,
  You should get some indication that the slip dirver is up and running.  On 
my system I get "packet mode enabled".  That tells me that all is ok.

TTFN,

Jim Raykowski          internet:  [email protected]
San Diego, CA          OS/2 BBS:  WZ02186
FAX: (619) 435-6814

-----------[000418][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 17:49:56 -0700
From:      [email protected] (Guy Harris)
To:        comp.protocols.tcp-ip
Subject:   Re: [Q] : TCP and NETWORK UNREACHABLE
In article <[email protected]>,
Alan Cox <[email protected]> wrote:
>In article <[email protected]> [email protected] writes:
>>What would be the expected behaviour by TCP (in ESTABLISHED state) when an
>>NETWORK UNREACHABLE is received ??
>
>Depends if your stack is compliant or old BSD. A compliant stack _MUST_
>report all ICMP messages to the user (RFC1122).

Where does it say something that amounts to "it MUST report all ICMP
messages to the user"?  I see:

         4.2.3.9  ICMP Messages

            TCP MUST act on an ICMP error message passed up from the IP
            layer, directing it to the connection that created the
            error.  The necessary demultiplexing information can be
            found in the IP header contained within the ICMP message.

			...

            o    Destination Unreachable -- codes 0, 1, 5

                 Since these Unreachable messages indicate soft error
                 conditions, TCP MUST NOT abort the connection, and it
                 SHOULD make the information available to the
                 application.

                 DISCUSSION:
                      TCP could report the soft error condition directly
                      to the application layer with an upcall to the
                      ERROR_REPORT routine, or it could merely note the
                      message and report it to the application only when
                      and if the TCP connection times out.

			...


      4.2.4  TCP/APPLICATION LAYER INTERFACE

         4.2.4.1  Asynchronous Reports

            There MUST be a mechanism for reporting soft TCP error
            conditions to the application.  Generically, we assume this
            takes the form of an application-supplied ERROR_REPORT
            routine that may be upcalled [INTRO:7] asynchronously from
            the transport layer:

               ERROR_REPORT(local connection name, reason, subreason)

            The precise encoding of the reason and subreason parameters
            is not specified here.  However, the conditions that are
            reported asynchronously to the application MUST include:

            *    ICMP error message arrived (see 4.2.3.9)

            *    Excessive retransmissions (see 4.2.3.5)

            *    Urgent pointer advance (see 4.2.2.4).

            However, an application program that does not want to
            receive such ERROR_REPORT calls SHOULD be able to
            effectively disable these calls.

            DISCUSSION:
                 These error reports generally reflect soft errors that
                 can be ignored without harm by many applications.  It
                 has been suggested that these error report calls should
                 default to "disabled," but this is not required.

"Network unreachable" is code 0, so, as I read RFC1122, it says only
that an application must be able to request that it be notified of
"network unreachable" errors, not that *all* applications *must* be
notified of them - which, as far as I'm concerned, is as it should be,
given that "network unreachable" often just means "some router crashed,
it's rebooting, honest, don't worry, be happy, all will be well soon".

-----------[000419][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 11:52:59
From:      [email protected] (Hamish Whittal)
To:        comp.protocols.tcp-ip
Subject:   Subnetting
I am still a little cofused by subnetting and subnet masking. I have read 
quite a couple of explainations, but still can't seem to fully understand it. 
Is there someone out there that can explain it to me, of possibly some place I 
could read about it.

any help appreciated

regards

hamish

-----------[000420][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 08:50:59 GMT
From:      [email protected] (Gregory Bond)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
In article <[email protected]> [email protected] (Charles Yamasaki) writes:


   As far as distinguishing Web connections, nope.  You're going to have
   one server on this machine listening to port 80 on all three interfaces.
   I don't know of any way around that.  Now you could start a different
   server on another port, but that's an ugly solution.

Well, you could use the TCP wrappers to listen on port 80 and then
decide what program to run (or what homepage to use, or whatever)
depending on that the local address was....

Greg.
--
Gregory Bond <[email protected]> Burdett Buckeridge & Young Ltd Melbourne Australia

Atilla The Hun's Maxim: If you're going to rape, pillage and burn, be sure to 
do things in that order.  -- P. J. Plauger, Programming On Purpose, p147


-----------[000421][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 17:12:41 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   is exponential backoff really right?
Or more specifically, is an exponential backoff where each try is 2 times
as long as the previous really right?

Consider:

You are linked via a SLIP or PPP dialup line.  The line connection drops
and is redialed, either automatically, or manually.  It takes about 45
seconds minimum to restablish the phone connection.  This is the best case.
This does not consider the delays involved in detecting the condition.

During this 45 seconds, the exponential backoff has been working against
you.  Retries are now up in the 30 to 60 second range.  If you didn't get
reconnected really quick, it can be worse.

Given a period of outage, the delay in reestablishing traffic flow could be
as long as the outage.  For reasons I am not sure of, I've seen many where
the delay (lag) is a lot longer.

This is particularly annoying with interactive session (telnet, rlogin).
Your session is frozen for quite some time.

Another condition I have seen is where the backoff is different on each
end.  After getting the dialup reconnected, I'm getting the output from
my session a LOT sooner than my input is being accepted.

I would like to see an exponential backoff that is not as harsh as this.
Instead of doubling the number of seconds, how about a proportion that is
less than 2.

instead of      how about
     1               1
     2               1
     4               2
     8               2
    16               3
    32               4
    64               5
                     6
                     8
                    10
                    12
                    16
                    20
                    25
                    32
                    40
                    50
                    64

In this case the proportion is approximately 1.25 and using millisecond
or microsecond resolution could make the numbers easier to calculate
instead of using a table lookup (although I think a table lookup is just
fine).

If I were to experiment with such an exponential backoff, what kind of
impact might there be on the existing network?
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000422][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 09:17:23 GMT
From:      [email protected] (Stef Van Dessel)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
Sam Oscar Lantinga ([email protected]) wrote:

: I have tried adding the "ghost" IP address to my system's arp cache
: using 'arp -s IP_address HW_address', but when an arp request goes
: out for the arped address, the Linux box does not reply.
: Is this proper proxy-arp method?  I've seen alot of discussion of
: what it does, but none on how to set it up. 

It's "arp -s <ip_address> <hw_address> pub". I've succesfully connected
two networks using the same network number - but different ip's for the
hosts ;-).

Say you have NET1-ethernet- GATE1 <point-to-point-link> GATE2 -ethernet-NET2

Then GATE1 needs to proxy arp for each host on NET2 and for GATE2. GATE1
also needs host routes to these machines via the point-to-point link with
GATE2. Do the same for the opposite direction. It's trivial to write a little
script that will snarf the IP's from a file, and do the arp publishing and
route adding. 

Note that you can control what hosts can get out: if there is a host on
NET1 you don't want to access GATE2 or NET2, then do not do the arp stuff
or the route add on GATE2.


stef


-----------[000423][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 09:39:02 GMT
From:      Chris Christoff <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Window and segment size

Can anybody tell me how to find out the tcp-ip window size and segment
size on a Sun 
Sparc Server 10 running Solaris ?

Cheers,

Chris
[email protected]

-----------[000424][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 10:20:40 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: IP/TCP/UDP checksum implementation in C?
In article <[email protected]> [email protected] (Jonathan Walker) writes:
>Does anyone have some source, or know where I can get some source, for 
>a C impementation of the checksum that is present in the various IP
>headers? 

Unless you have a miracle quality compiler or a very risc and compiler
oriented machine don't do it. Use assembler. Going from assembler to a very
optimised assembler routine on a 386 with Linux added 40K/second to some
peoples ftp performance. Needless to say going from C to good assembler
on a 386 would be far more drastic.

KA9Q has some stuff for C implementations of checksums.

Alan

-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000425][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 94 15:29:11
From:      [email protected] (William )
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
Isn't there an IP mobility working group of the IETF working on things
like this?

BillW
cisco

-----------[000426][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 09:58:51 UNDEFINED
From:      [email protected] (Alex Neutjens)
To:        comp.protocols.tcp-ip
Subject:   network breakdown

If the network of a Unix site goes down,
then the time between the physical network-breakdown and the detection for an 
application that there is no network-communication available, is exactly 20 
minutes.
This is a standard in tcp-ip. In case of a "semi" real-time application, this 
time-out is unacceptable.
The only solution I see at the moment, is to setup an internal timer to look 
at the response time of the proces that has send a request over the network.

Does tcp-ip provide a better way to solve this problem ? 


-----------[000427][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 13:58:27 GMT
From:      [email protected] (Patrick Klos)
To:        comp.protocols.tcp-ip
Subject:   Re: Ipswitch Inc's TCP/IP Stack?
In article <[email protected]delphi.com>,  <[email protected]> wrote:
>	
>Hello,
> 
>	I'm trying to find the email address of Ipswitch but haven't been
>able to.  Is it possible to use this tcp/ip stack as a server?  Run it on
>a machine and than have it act as a SLIP/PPP server?
>	Is this possible with all of the tcp/ip stacks?  Or just particular
>ones?
> 
>	Thanks & Take care.
> 
>	Vikram

You should be able to get to thier tech support at "[email protected]".

-- 
============================================================================
    Patrick Klos                           Internet: [email protected]
    Klos Technologies, Inc.                Voice: (603) 424-8300
    604 Daniel Webster Highway             FAX:   (603) 424-9300
    Merrimack, New Hampshire  03054        BBS:   (603) 429-0032
============================================================================

-----------[000428][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 94 14:48:07 GMT
From:      [email protected] (Keith Eagling)
To:        comp.protocols.tcp-ip
Subject:   socket write/send problem
I have been having problems writing some data over a AF_INET stream
socket. If I call a write or send asking for 8 bytes to be sent
they will return 8 but nothing is actually sent over the ethernet
(I been using a sniffer). Does Unix buffer till some threshold
of buffer data is reached before the TCP packet is sent (I know
telnet can send individual characters in a TCP packet) or is it
something else. Incidently the write/send will cause a TCP 
packet to be transmited if I sleep before calling either of them.

Any ideas

thanks 

Keith Eagling.

-----------[000429][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 19:58:13
From:      [email protected] (A. Grant)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (don provan) writes:
>NetWare accepts any mask that allows at least two bits for the subnet
>number and at least two bits for the host number.

Yes, but does it accept different masks on different interfaces,
and does it route correctly in that case?

-----------[000430][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 20:08:22
From:      [email protected] (A. Grant)
To:        comp.dcom.lans.etherne,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
In article <[email protected]> [email protected] (Phil Howard) writes:
>Maybe.  IPX waits for acks for every packet, for instance.

Not true.  IPX is a datagram protocol like IP.  It is NCP (the RPC
protocol that does file sharing etc.) that, in its original
implementation, waits for an ack for every request.  This is true
whether NCP is running over IP or IPX.  Whatever plans Novell have
to replace IPX with IP as a carrier for NCP, they are not going 
to replace NCP itself, but just make it slightly more efficient with 
burst requests/replies etc.

IPX is a quite nice protocol, very similar to XNS.  I am amazed that
the "next generation" IP was not based on IPX, especially as IPX is
already the world's most widely used protocol. 

-----------[000431][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 15:21:47 GMT
From:      [email protected] (Mario Klebsch)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
Hello!
[email protected] (Casper H.S. Dik) writes:

>It's really no problem to arp multiple addresses on on ethernet port.
>ARP maps IP->ethernet and the way this mapping is used, a many to one
>mapping is not illegal.  (Ethernet addresses never need to be
>mapped back to the IP address, only the IP address needs to be
>mapped to a MAC address).


>In Solaris 2.x it's possible to do so:
 
>	# Primary address:
>	ifconfig le0 ....
>	# More addresses:
>	ifconfig le0:1 ....
>	# Or even more
>	ifconfig le0:100 ....

This sounds very interesting. I tried it, as a regular user I typed:

	$ ifconfig ie0:1

But afterwards, when I type

	$ ifconfig -a
	lo0: flags=849<UP,LOOPBACK,RUNNING,MULTICAST> mtu 8232
        	inet 127.0.0.1 netmask ff000000 
	ie0: flags=863<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST> mtu 1500
        	inet 134.169.36.10 netmask ffff0000 broadcast 134.169.255.255
	        ether 8:0:20:a:5b:4e 
	ie0:1: flags=842<BROADCAST,RUNNING,MULTICAST> mtu 1500
        	inet 0.0.0.0 netmask 0 

Is there to get rid of these entries? And why can I create them
without any privileges?

Thanks in advence,

	Mario
--
Mario Klebsch, DG1AM, [email protected]		+49 531 / 391 - 7457
Institut fuer Robotik und Prozessinformatik der TU Braunschweig
Hamburger Strasse 267, 38114 Braunschweig, Germany

-----------[000432][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 15:27:25 GMT
From:      [email protected] (anil.verma)
To:        comp.misc,comp.protocols.tcp-ip,comp.unix.misc
Subject:   Domain Name Server
I am trying to find out more about Domain Name Servers.  Is this 
discussed in any newsgroups?  Are there any archieves that have info
on DNS?  Please respond to [email protected]

TIA

Anil Verma

-----------[000433][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 16:12:34 GMT
From:      [email protected] (Jarek Lis)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
Tom Limoncelli ([email protected]) wrote:
: In <[email protected]> [email protected] (Rob Tanner) writes:
 
: >I know zip about IPX, but it looks like we are going to be installing
: >a Novell server anyway.  The Novell, of course, is for the PCs.  What
: >concerns me as that we also have a number of Unix machines as well as
: >an even larger number of X-stations that don't run IPX and don't
: >concern themselves with the Novell server.
 
: IPX is a chatty protocol.  I don't have a lot of experience with it,
: but I do have experience with AppleTalk, which is equally chatty.

I must disagree. IPX is nice, plug&play protocol. It just works, works well
and you have nothing to do with it (which is advantage , or disadvantage).

: Since you have two very disjoint sets of machines (i.e. two distinct
: groups of machines that don't talk to each other much), why put them on
: the same subnet? 
 Why not? Big advantage is that you have both TCP/IP and IPX avalaible
for all machines.

:              If your wiring plant permits it (i.e. a 10Base-T star
: configuration) why not put the Novell and PC machines on one subnet,
: the Unix boxes and X Terminals on a different subnet, then put a router
: between the two networks.

Novell file server is very effective, IPX and IP router itself. And cost you
only second network card.

If you want to have something really efective, resonable cheap and universal,
consider using EtherSwitch in place of hub.


Jarek.


-----------[000434][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 16:15:23 GMT
From:      [email protected] (Vernon Schryver)
To:        comp.protocols.tcp-ip
Subject:   Re: IP/TCP/UDP checksum implementation in C?
In article <[email protected]> [email protected] (Alan Cox) writes:
>Unless you have a miracle quality compiler or a very risc and compiler
>oriented machine don't do it. Use assembler. Going from assembler to a very
>optimised assembler routine on a 386 with Linux added 40K/second to some
>peoples ftp performance. Needless to say going from C to good assembler
>on a 386 would be far more drastic.
> ...


The best code to do the TCP checksum on an R3000 or R4000 in 32-bit mode
that I know of is the same that is produced by the standard compiler
from careful C.  Because of the nature of the 386, I bet the same applies
to most 386 C compilers.  Of course, you have keep your head together,
and not write simple, byte-at-a-time C code such as that which I've seen
in some 386 UNIX implementations.  I think you can convince most C
compilers to produce a sequence of addc's, a single add to the index,
and a single comparison of the index.

Of course, in many cases you don't care about speed and so the simplest,
byte-at-time code is the right solution.  At the other extreme, many
people in-line some checksum computations, such as the IP, TCP, and UDP
headers.


Vernon Schryver    [email protected]

-----------[000435][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 16:36:39 GMT
From:      [email protected] (Jarek Lis)
To:        comp.protocols.tcp-ip,comp.dcom.lans.ethernet
Subject:   Re: Running IPX and TCP/IP over same ethernet
[email protected] wrote:
: > >> Direct from Rumor Control Central:  I have heard that IPX is a real
: > >> bandwidth hog, and that if you don't want to see your TCP/IP severely
: > >> degrade, don't let IPX run over the same segments that run TCP/IP.
: > 
: > We find that ipx and tcpip get along well. Cinfig is 3 Sun Sparc10, 1 NW_311,
: One thing to be careful of is SAP (Service Advertising Protocol).  Each
: server on the IPX internet babbles its state every so often, and the routers
: pass this on to all the other IPX networks.  On large networks, this can
: consume large portions of your bandwidth.  We've seen networks where the
: routers essentially stop ever 15-20 seconds while they process all the SAP
: updates (of course, those are not Network Systems' routers ...).

Even thosands bytes send every 20 second (I think it is exacty 30 seconds)
by several tens of servers isn't problem for Ethernet. Problem starts, when
you have networks connected via routers and phone lines. On our campus,
about 70 Novell servers, SAP eats about 1200 bps (bits/sec).


-----------[000436][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 94 16:59:12 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   RE: Congestion control (TCPIP)
In Article <[email protected]>
[email protected] (Josh Cohen) writes:

>What gives? people constantly mention 'the new standard' and that it
>recommends x y or z for congestion control ( slow start, multiplicative
>decrease, etc on the TCP level and ICMP on the IP level )
>
>Where can I find the REAL new standard?  
>I have the RFC for TCP and IP, I have the jacobson paper, and other do dads
>but I have found conflicting reports... 

Get a copy of RFC1122 and RFC1123 from your favorite NIC. These update many
older RFCs and are critical to building a functional implementation.

In the area of congestion management, see RFC1122, 4.2.2 and 4,2,3. There have
been substantial improvements to the sliding window algorithm, mostly by Van
Jacobson of LBL, since the original TCP RFC came out and 1122 made many of the
REQUIRED.
 
R. Kevin Oberman			
Energy Sciences Network (ESnet)
National Energy Research Supercomputer Center (NERSC)
Lawrence Livermore National Laboratory (LLNL)
Internet: [email protected]		+1 510-422-6955

-----------[000437][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 94 17:11:03 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   RE: Window and segment size
In Article <[email protected]>
Chris Christoff <[email protected]> writes:

>Can anybody tell me how to find out the tcp-ip window size and segment
>size on a Sun 
>Sparc Server 10 running Solaris ?

MSS is dependent on whether the connection is on or off net. If on-net, it is
the MTU for the interface used less overhead (1460 for Ethernet). For off-net,
it is 576 less overhead (536). The definition is usually that the same subnet
is "on-net", but his may be configurable. Solaris implements MTU discovery, I
believe, and should be able to increase segment size even for off-net addresses
in many cases.

Not sure about the default window. A quick check of a SunOS 4.1.3 system show
it starts out with a 4096 byte window for a telnet connection, but there is no
reason to believe Solaris is similar.

The best way to find this is to use tcpdump(8). It displays window and MSS
negotiation by default.

R. Kevin Oberman			
Energy Sciences Network (ESnet)
National Energy Research Supercomputer Center (NERSC)
Lawrence Livermore National Laboratory (LLNL)
Internet: [email protected]		+1 510-422-6955

-----------[000438][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 18:02:05 GMT
From:      [email protected] (Sam Oscar Lantinga)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
Stef Van Dessel ([email protected]) wrote:
: Say you have NET1-ethernet- GATE1 <point-to-point-link> GATE2 -ethernet-NET2
 
: Then GATE1 needs to proxy arp for each host on NET2 and for GATE2. GATE1
: also needs host routes to these machines via the point-to-point link with
: GATE2. Do the same for the opposite direction. It's trivial to write a little
: script that will snarf the IP's from a file, and do the arp publishing and
: route adding. 

I have:

	NET1_work-ethernet- Linux_box - NET1_work-ethernet- ROUTER1 -
INTERNET_router_router_router_INTERNET - SLIP_server -serial-Home_Linux_box

	I want my Linux box on the network at work to act as a proxy
for my Home_Linux_box, making it appear as though my home Linux box is
on the same network as my machine at work.  I think the proxy arp will
work, but the issue is routing subnet packets all the way past the
SLIP server to my home linux box.

	... I've gotten proxy arp working, but routing is still a
problem, hence the need for the IPOPT_RELAY extension... ?
Is there any way to force non-specific IP packets along a certain
route?  Or to broadcast routing information telling the routers that
one particular machine on the subnet can be reached through an
entirely different network?

Thanks for all the help. :)

	-Sam Lantinga	([email protected])


-----------[000439][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 18:20:18 GMT
From:      [email protected] ( Ed Alcoff)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Question on "firewalls"
Greetings:

I apologize in advance if this is a FAQ:

I understand that there are both commercial and p.d.
software that allows one to set up a "firewall" to protect
a TCP/IP network from external access.

1.  How is this different than what can be done with
a decent router?

2. Where can I find such products?

Thx, in advance,

Ed Alcoff

-----------[000440][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 18:20:40 GMT
From:      [email protected] (Steve Crumley)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
Gregory Bond ([email protected]) wrote:
> In article <[email protected]> [email protected] (Charles Yamasaki) writes:


>    As far as distinguishing Web connections, nope.  You're going to have
>    one server on this machine listening to port 80 on all three interfaces.
>    I don't know of any way around that.  Now you could start a different
>    server on another port, but that's an ugly solution.
 
> Well, you could use the TCP wrappers to listen on port 80 and then
> decide what program to run (or what homepage to use, or whatever)
> depending on that the local address was....
 
> Greg.
> --
This sounds like a good solution.  Does anyone have any more details
on this approach?

> Gregory Bond <[email protected]> Burdett Buckeridge & Young Ltd Melbourne Australia
 
> Atilla The Hun's Maxim: If you're going to rape, pillage and burn, be sure to 
> do things in that order.  -- P. J. Plauger, Programming On Purpose, p147
 

-----------[000441][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 19:19:34 GMT
From:      [email protected] (Rob Janssen)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
In <[email protected]> [email protected] (Sam Oscar Lantinga) writes:

>Rob Janssen ([email protected]) wrote:
>: >	I would like to implement a new option in the IP protocol:
 
>: >IPOPT_RELAY
 
>: Please explain how this option is going to accomplish anything...
 
>	Sure.  A modified version of tcpdump snarfs packets 
>destined for the non-existent host.  It appends the destination
>address to the packet using the IPOPT_RELAY option, and rewrites
>the destination address as the SLIP-connected host.  The packet
>is then sent back out on the network with a new destination, leaving
>the source address unchanged (using raw sockets and a simple kernel
>hack)  When the packet is received by the destination host, the 
>IPOPT_RELAY option is detected, the destination address is re-written
>to be the original destination
>_and_sent_out_a_second_loopback_interface ifconfig'd to have the
>proper destination address to "exist" on the network at work.
 
>	Complex, I admit, but I'm not sure of any other way
>to acomplish having my machine on the SLIP connection also appearing
>on the network at work.

This is much too complex.  Proxy ARP will do what you want in a
much simpler and more efficient way.

>: IP-in-IP encapsulation is already being used to tunnel IP packets
>: between nonstandard networks over another network.  In fact, *two*
>: protocol numbers have already been allocated to this purpose, probably
>: by oversight...  (numbers 4 and 94)
 
>	Where can I find out more about this?

This kind of numbers are listed in the most current "Assigned numbers"
RFC, which also usually points to a reference.
There is also an RFC about encapsulation (RFC 1241)

>: Can't it be done by plainly adding some static routes on the routers?
>: Or by using proxy ARP?
 
>	How is this done?  I have tried adding host routes to the
>Linux machine at work (not a router), but I can't detect any packets
>getting to my machine on the SLIP connection (and yes, IP forwarding
>IS enabled).  The SLIP router will not (as far as I can tell) forward
>any packets to the SLIP connected host that are not destined for it.
>I have tried adding the "ghost" IP address to my system's arp cache
>using 'arp -s IP_address HW_address', but when an arp request goes
>out for the arped address, the Linux box does not reply.
>Is this proper proxy-arp method?  I've seen alot of discussion of
>what it does, but none on how to set it up. 

"Proxy ARP" means that a system which knows the route to a certain
destination (e.g. it has a SLIP link to it) replies to ARP requests
for that destination on another network.  It replies with its own
hardware address.

E.g.: one system is on an ethernet and has address 192.0.0.1
A second system is linked to it using SLIP and has address 192.0.0.2
Now, when some system on the ethernet asks for the hardware address of
192.0.0.2, the first system replies with its hardware address.  All
packets for 192.0.0.2 will then be sent to the hardware address of
system 192.0.0.1, and it will route them to the proper SLIP link without
further modifications.  So, there is no need to check every packet on
the ethernet, only the ARP requests (which are all received anyway).

I don't know if Linux supports this in some way.  If not, it can be
added.

Rob
-- 
-------------------------------------------------------------------------
| Rob Janssen                | AMPRnet:   [email protected]           |
| e-mail: [email protected]     | AX.25 BBS: [email protected]#UTR.NLD.EU     |
-------------------------------------------------------------------------

-----------[000442][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 19:24:56 GMT
From:      [email protected] (Bernt Habermeier)
To:        comp.protocols.tcp-ip
Subject:   simulators wanted


I'm looking for a free software package that is well suited for the
simulation of end to end network protocols.  I want to measure the
performance of a protocol and measure how the protocol contributes to
net congestion.

I want to be able to throw new ideas into the simulation rather
quickly and get a feel of what happens in the real world.  That means,
I don't want to write low level code to tell the simulation how the
protocol works--in that case I might as well just implement the real
thing.  The whole reason for this is to decrease the
implementation-reimplementation cycle.

It would be good if there was a mechanism in which I can simulate at
least one intermediate node of the communication process.

Variables I want to be able to control (other than the protocol
itself): node to node throughput, probability of errors on each
node, intermediate node-packet droppage (source quench doesn't really
need to be included here)..

I could imagine that there are some general purpose simulators that can
do the above, lemme know what they and where I could find them.

Thanks,

Bernt.




-----------[000443][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 02:06:48 -0400
From:      [email protected] (Jay Ashworth)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
[email protected] (Christopher Davis) writes:
>JA> == Jay Ashworth <[email protected]> [ me... ]
> JA> in order to answer to more than one IP address with one MAC layer
> JA> address, you'd need to ARP all of them...  and I don't believe ARP
> JA> provides for this.
>Sure it does.  Proxy ARP, for example.  The Telebit Netblazer can be
>configured to associate specific IP addresses with specific ports, so
>that, say, "telnet blazer-port20" would get you a connection to whatever's
>on serial port 20, such as a soda machine readout or whatever.

Oh.  Ok...  I get it.  One MAC address, multiple IP's... right.  But that
was really a secondary argument anyway... However...

> JA> [...] this would still require you to be able to ifconfig the
> JA> interface with multiple simultaneous IP addresses... and I haven't
> JA> seen anything like this mentioned anywhere--and I read a _lot_.  :-)
>I believe I've seen mention that it was possible under BSDI BSD/386 1.1
>(which the original poster was running) using the "alias" parameter to
>ifconfig.  I haven't yet tested this, as our BSDI box is busy doing more
>important things that I really don't want to break...

Possibly... but someone else made a fairly good point, to wit: configure a
couple of loopback interfaces with your other IP addresses, (b and c, in
Vix's example), and hang servers off of them.  Correct me if I'm wrong,
but don't you tell the socket calls which IP address to live on?  It
should be possible to hang one WWW server on each address, while only
using 1 box and 1 interface card, if you got the routing right.

I haven't the resources to test this, but the theory sounds tight.

And theory has been very, very good, to me...

Cheers,
-- jr 'So, does it _work_, Paul?' a
-- 
Jay R. Ashworth                                                       Ashworth
Designer                                                          & Associates
ka1fjx/4              High Technology Systems Consulting
[email protected]                                                +1 813 790 7592

-----------[000444][next][prev][last][first]----------------------------------------------------
Date:      19 Sep 1994 22:15:21 GMT
From:      [email protected] (Kenneth Martig)
To:        comp.protocols.tcp-ip
Subject:   Using multiple Class C's on a sigle wire
I know this question has been asked before, but I couldn't find it in the
faq and I haven't seen it in about a month.  Sorry.

How do you get multiple class C's to be seen on a single wire.  We are in
an etherswitched, low traffic environment.  With printers and
concentrators and coke machines :) needing IP addresses, the number of IP
addresses has become an issue before routing has become an issue.

What do I need to do to get a machine to see a different class C address
without going through a router.  I have tried all that I could think of
to no avail.  I would appreciate any help that you all can give me.

Thanks bunches,


Ken Martig
ZymoGenetics, Inc.
[email protected]

-----------[000445][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 1994 22:59:17 GMT
From:      [email protected] (Randy Davis)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for automated DNS update tool.
In article <[email protected]>,
Willy Janssen <[email protected]> wrote:
|As we have to do a relevant number of DNS updates, I was wondering if there is
|an automated tool around to include the update-info in DNS.

  Yes, there is.  Although not specifically designed as such, I think it would
do the job fairly easily.  The "h2n" perl script, described by the O'Reilly
_DNS_And_Bind_ book (which I highly recommend), is designed to convert an IP
address lookup file (in the form of the Sun NIS /etc/hosts file) to the DNS
format, including the forward and reverse resolution files, and can handle
multiple domains, MX records, etc...

  All you would do is create a /etc/hosts file, in the form of:

IP_Address <tab or space> hostname

  Then run the script to create all the needed named files except the boot
file (which doesn't need to be recreated, anyway).

|In first instance I'm thinking of automated generation of the reverse map, serial
|numbers etc. At second there should be a syntax check on the input files to detect
|all kind of typo's, and other errors.
|
|Please let me now.

  The h2n script is available from:

ftp.uu.net:/published/oreilly/nutshell/dnsbind/dns.tar.Z

Randy Davis                                   Email: [email protected]
Corporate Network and System Administrator
Megatek Corporation, San Diego, California


-----------[000446][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 00:23:00 GMT
From:      [email protected] (Ron Nadel)
To:        comp.protocols.tcp-ip,vmsnet.networks.tcp-ip.misc
Subject:   Re: VMS tcp/ip packages?
In article <[email protected]> [email protected] (David D Kaas) writes:

>We have some Dec statio 3100s running VMS 5.x.  We would
>like to install TCP/IP software.  What companies support
>TCP/IP on VMS?  Are there any public domain or low cost
>packages?

Dave, there are several packages that will run under VMS.  Probably the most 
robust is Multinet from TGV (Two Guys and a VAX).  It is a full function, easy 
to use package, and their tech support is very good.  Carnegie-Mellon provides 
a very inexpensive package called CMUTECH.  It costs almost nothing (outside 
of media/shipping).  It is not as complete as Multinet, and of course they do 
not have the tech support of a commercial vendor.  But it does work.  The last 
number I had for them was 412-268-5896.

Wollongong makes a VMS IP product called WIN/TCP but it I would recommend the 
others above well before this, and DEC does have their UCX product, but 
consensus is that it's not a great product.  We run a Silo archiving product 
on one VAX here that REQUIRES UCX, but Multinet has UCX hooks in it so that it 
responds and handles all UCX calls - so we do not run UCX.

All in all, Multinet is the superior product for VMS, CMUTECH is the cheapest.

Email if you have questions.

Ron

-----------[000447][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 00:58:39 GMT
From:      [email protected] (Arnt Gulbrandsen)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
In article <[email protected]>,
Steve Crumley <[email protected]> wrote:
>This sounds like a good solution.  Does anyone have any more details
>on this approach?

No.  But I do something similar another way, http://www.nvg.unit.no/
is available in two languages and the server guesses which language
is appropriate for the client.  index.html looks something like

<!--#exec cmd="/local/www/cgi-bin/homepage.cgi"-->

homepage.cgi is a small program that prints one of two pages
depending on the client's address (it uses getpeername()), but it
could easily use getsockname() and select page according to its own
address.

--Arnt


-----------[000448][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 01:17:04 GMT
From:      [email protected] (Elena Leong)
To:        comp.protocols.tcp-ip
Subject:   Follow up: Dupe. IP Addrs
Hi again.

Thanks to all the suggestions I got about why I was getting "Duplicate IP!" 
messages on my system.  A few people suggested that it might be because of
proxy Arps happening on the LAN (the LAN I am connected to is bridged using
Ciscos with proxy Arps), and the Suns seeing Arps requested by a host for 
itself.  So I booted my machine to ignore arps on this interface (I have 2
cards - le0 runs TCP/IP, le1 runs OSI), by putting this line in /etc/rc.boot:
	ifconfig le1 up -arp
and havent seen any messages since.  Running etherfind also helped me under-
stand the problem better.

However, the network guy rings me today, and tells me that he has found 2
hosts that have the same IP addresses as 2 other hosts!  So there you go.
(BTW he has 3 other culprits to go, because I have seen 5 ethernet addresses
being complained about... I am glad I am the only one responsible for handing
out IP addresses on my own little network here!)

Hope that helps someone else.

Elena.

-----------[000449][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 01:44:19 GMT
From:      [email protected] (C. M. Rahman jr)
To:        comp.protocols.tcp-ip
Subject:   my subdomain can't access internet
Hi,

I been trying to add a subdomain to my system but it was able to communicate
with my domain but can't access internet via my network.

here is the picture
my domain is ccsi.com
my subdomain is spur.com

  spur.com <-> ccsi.com <-> INTERNET (Alternet)

spur.com can access ccsi.com but can't access Internet
ccsi.com can access spur.com and internet. How come I can't get spur.com
to access Internet ?

I have added the spur.com to routing table, that is why I was able to 
access ccsi.com.  

Any help will be appreciated.. !! thanks

--
  #################################################################
  # C.M. Rahman jr.                |   Public Network Provider    #
  # Network/System Administrator   |       Austin, Texas          #
  # Commuter Communication Systems |"Pray for what you want, but  #
  # [email protected]        | work for the things you need"#
  # Tel/Fax :- (512) 467-6591      |  Modem:- (512) 467-9859      #
  #################################################################

-----------[000450][next][prev][last][first]----------------------------------------------------
Date:      Mon, 19 Sep 94 23:25:00 +0200
From:      [email protected]fi (Petri Kejonen)
To:        comp.protocols.tcp-ip
Subject:   Linux and SLIP/CSLIP
 Hi!

  I have a leased line in my home and the speed of my direct connection to
the internet is 19200bps. I have been testing my connection with SLIP
protocol and everything seems to run just fine (well, maybe not so fast, but
ok anyway ...). But when I try to use CSLIP (header compression)
I get no connection. My host is set up for the compression now and they can
actually ping "me", but I get nothing through when I try to ping "them".
  I'm using Linux (1.0.9) and I have set up the serial port with the
"slattach" -command to use the CSLIP protocol. It would be a pity
if I could not use the compression, since I haven't been able to persuade
my provider to use the PPP protocol.

  If you have any comments or advice, please reply ... And please excuse my
possibly bad english :)

--
[email protected]
[email protected]

-----------[000451][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 08:24:36
From:      [email protected] (Daryl D. Rester)
To:        comp.protocols.tcp-ip
Subject:   Subnetting Class "C"
Can anyone tell me if there are any utilities that assist in the subnetting of 
a class "c" network number, if so were might I find them?

Daryl D. Rester       [email protected]

-----------[000452][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:09:35 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
[email protected] (Sam Oscar Lantinga) writes:

>	This would be used to relay IP packets from one network
>to another, providing a sort of distributed subnet in which not all
>of the machines on the subnet are on the actual subnet wire.
>The reason I would like to do this is so that my home machine can
>dialup a SLIP server and appear as a machine on my network at work,
>as well as a machine on the SLIP server's subnet.  

My terminal server lets me do this.  But the SLIP server is on that subnet.
This may not be your case.


>	My question is....  Will packets with an unrecognized
>IP option like this be forwarded across routers?  I noticed that
>in the Linux network code that if a packet has an unrecognized IP
>option, the packet is discarded.  Is this standard behavior?  Or
>will my custom packets make it safely across the network?

Sounds like the IPOPT_DISCARDME feature :-)

Seriously it should discard it.


>	On another note, it may be necessary to do full packet
>encapsulation for this to work properly.  Is it possible to select
>a new protocol number say "encap" that would correspond to a simple
>encapsulation protocol? 

Sounds like you need some sort of tunneling.

I once tried the following:

I connected COM3 to COM4 on my Linux machine with a null modem.  I dialed
to my SLIP server from COM2.  I then used the SLIP server (but you can use
a machine that supports SLIP) with one port in dialout mode (telnet to it)
which was wired with a null modem to one of the regular SLIP ports.  I was
able to have a second SLIP connection on COM4 tunneled through TELNET.
There was no real benefit since it only got me back on the same wire.
But in theory I could put someone from anywhere in the Internet onto my
wire this way.  You might be able to rig up something like this.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000453][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:14:23 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: network breakdown
[email protected] (Barry Margolin) writes:

>No.  TCP/IP is designed to survive across temporary network problems.
>Remember its initial goal: to serve as the protocol for the military, and
>allow connections to stay up even when the enemy has knocked out portions
>of the network.

Ironically, we now use the feature to keep connections up even when portions
of the network are knocked out by bad equipment, poor lines, ditch diggers,
etc.


>For instance, some of the common routing protocols take several minutes to
>stabilize when a router goes down and a backup must take over.  On media
>like thick and thin coax ethernet, when a new segment is being added to a
>daisy-chained segment, that segment will be down while cables are being
>unplugged and reconnected.

Unfortunately we still have too many hosts that won't keep the connections
up long enough.  It's getting better.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000454][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:16:48 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: network breakdown
[email protected] (Alex Neutjens) writes:

>If the network of a Unix site goes down,
>then the time between the physical network-breakdown and the detection for an 
>application that there is no network-communication available, is exactly 20 
>minutes.

I've seen other time periods, including 2 minutes, 10 minutes, 15 minutes,
and 30 minutes.


>This is a standard in tcp-ip. In case of a "semi" real-time application, this 
>time-out is unacceptable.
>The only solution I see at the moment, is to setup an internal timer to look 
>at the response time of the proces that has send a request over the network.
 
>Does tcp-ip provide a better way to solve this problem ? 

Use UDP.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000455][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:39:36 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Using multiple Class C's on a sigle wire
[email protected] (Kenneth Martig) writes:

>I know this question has been asked before, but I couldn't find it in the
>faq and I haven't seen it in about a month.  Sorry.

Then the new FAQ should be out real soon now.


>How do you get multiple class C's to be seen on a single wire.  We are in
>an etherswitched, low traffic environment.  With printers and
>concentrators and coke machines :) needing IP addresses, the number of IP
>addresses has become an issue before routing has become an issue.

I will await an authorative answer on this, too.  In the mean time, if your
class C addresses are grouped, go ahead and try a supernet mask, e.g. one
with fewer than 24 bits of 1's.


>What do I need to do to get a machine to see a different class C address
>without going through a router.  I have tried all that I could think of
>to no avail.  I would appreciate any help that you all can give me.

I'm not sure what you mean by "see a different class C address".  If you
mean to make it behave like the address is on the same wire and in the same
subnet, see the supernet test above.

I have not tried any of this and could be totally wrong.  I'm even betting
that I'm totally wrong.  But until a real answer arrives, it's something to
try and see if it works.
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000456][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:45:51 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: my subdomain can't access internet
[email protected] (C. M. Rahman jr) writes:

>I been trying to add a subdomain to my system but it was able to communicate
>with my domain but can't access internet via my network.
>
>here is the picture
>my domain is ccsi.com
>my subdomain is spur.com

That's not a subdomain.  That's just another domain (both are peer subdomains
of the domain "com").


>  spur.com <-> ccsi.com <-> INTERNET (Alternet)


>spur.com can access ccsi.com but can't access Internet
>ccsi.com can access spur.com and internet. How come I can't get spur.com
>to access Internet ?

How about some addresses?  I have 198.6.201.1 for ccsi.com, but none for
spur.com.  Does it exist?  Is it known to your provider?


>I have added the spur.com to routing table, that is why I was able to 
>access ccsi.com.

Is your provider routing it your way for the returns?
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000457][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 04:26:48 GMT
From:      [email protected] (D. Hobbs)
To:        comp.protocols.tcp-ip
Subject:   Using DNS
First, Yes, I have read DNS and BIND from O'Reilly.

I'm trying to get our RS/6000 to resolve domain names.  Our
DEC machine is pointing to some SPRINTNET name server, so it
seems to work fine.  However, I'd like to point all of our
other Unix boxes (including the RS/6000) at the DEC machine.
When I do what the DNS book says, the RS/6000 seems to eventually
grind to a slow halt.  I'm guessing it has to do with the  domain
name cache size, but I'm only guessing.

In the /etc/resolv.conf, I've specified:
nameserver 5.5.5.5 (assuming that is the valid IP address)
domain my.domain.com

The RS/6000's IP address is 5.5.5.4, and the DEC machine is
5.5.5.5.  The domain name is not registered, but the IP addresses
are (these domain names and IP are not valid -- protecting the
innocent :).

Anyway, when I do the above, the RS/6000 eventually just slows down
to nothing (rlogin, telnet, etc. -- user processes are OK and the
system load doesn't seem to change).  Any clues?

Thanks,

D.
-- 
David Hobbs
[email protected], [email protected]
All typos are my own.  The spelling errors are someone else's.
"I am Fudd of Borg.  Pwepare to be assimiwated."

-----------[000458][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 12:51:00 -0500
From:      [email protected] (Phil Howard)
To:        comp.protocols.tcp-ip
Subject:   Re: Subnetting Class "C"
[email protected] (Daryl D. Rester) writes:

>Can anyone tell me if there are any utilities that assist in the subnetting of 
>a class "c" network number, if so were might I find them?

Such as doing what?  Figuring out how many hosts you can have in each of
how many subnets?

Pick one:

A.  2 hosts in 62 subnets (124)
B.  6 hosts in 30 subnets (180)
C.  14 hosts in 14 subnets (196)
D.  30 hosts in 6 subnets (180)
E.  62 hosts in 2 subnets (124)

You can probably do some mixing as well.  I doubt if any "tools" do that
for you.

If you want "tools" that are GUI interfaces to your network, contact your
network equipment/systems vendor(s).
-- 
/***** Phil Howard KA9WGN *********** How about universal JOBS? **************\
*      Unix/Internet/Sys Admin        Let's de-Foley-ate congress in 94       *
*      CLR/Fast-Tax                   Don't let Annie get your gun!           *
\***** [email protected] ************* Just say NO to CIX extortion ***********/

-----------[000459][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 16:02:39 -0400
From:      [email protected] (William A White III)
To:        comp.unix.aix,comp.unix.questions,comp.protocols.tcp-ip
Subject:   ASCII Terminals, Telnet, x.25 and Satellites
We are attempting to run applications on an HP/9000 from a remote RS/6000.	The pieces consist of an ascii terminal (a wyse370 with an Enhanced PC keyboard)connected to an RS/6000 which has tcp/ip running over x.25.  The RS/6000 is	connected to an AT+T Tridom Clearlink In Door Unit (our private satellite	WAN).  The other end of this another multiplexor connected by land lines to 	AT&T Tridom's Atlanta site (we are in Edison, New Jersey) which is connected 	to an RS/6000's x.25 port.  The Edison RS/6000 routes the tcp/ip traffic 	across a 10-Base-2 v2 ethernet segment to the HP/9000.  The HP/9000 is running	the Oracle Financials application.	

We tested the HP application on a wyse terminal on an RS/6000 connected to the 
'router' RS/6000 via token-ring.  When we moved the testing to an x.25/
satellite connected wyse terminal we found that the escape sequences 	
generated by the function keys and keypad keys were being sent in multiple
packets.  Pressing the F1 key produced a packet with an escape, another 
packet with an O and a third packet with a P.  We know that this is the 
nature of x.25 across satellites and that we need a method of blocking the
telnet sessions.

We looked around for a tn3270 type server to run on the HP or even the 
'router' RS/6000 and we cannot seem to find any available.  We also cannot
imagine that we are the first to come across this problem.  If someone could
point us in the right direction or tell us how they solved this problem we
would appreciate this assistance.

Thank You,
Bill
-- 
William White  [email protected]

-----------[000460][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 17:44:24 -0500
From:      [email protected] (Richard Schenke)
To:        comp.protocols.tcp-ip
Subject:   LPD printer rejects a quick reopen
Greetings all,

I am using LPR/LPD to print files from a Unisys 2200/400 mainframe to
a Hewlett Packard LaserJet 4siMX, using the DirectJet network interface 
in the printer.  The traffic is TCP/IP all the way from Unisys to printer.  

My problem occurs when there are multiple files in the Unisys print
queue.  After the first file is printed, Unisys closes the first
LPD connection and tries to open a new LPD connection, by sending 
SYN to port 515.  The HP printer sends RST in response.  The Unisys
comm software (CMS1100) reports this to the LPR application (Teksouth
ReMoat) as 'Network Unreachable'.  The LPR driver reacts by giving up
further attempts to write to the remote printer.

If there is a delay between closing the connection for file 1 and
opening it for file 2, everything is OK.  From this I suspect that
the printer has not yet performed a passive open to port 515 when
the next SYN comes in.  The printer sends RST because it is not in 
the correct state to react to the SYN.

I can also address the same printer through a Novell FlexIP NLM 
gateway, residing in a file/print server.  It provides the LPD function,
and sends it by IPX from server to printer.  If I repeat the same test,
the Sniffer shows two packets from the server to the Unisys after the close
of connection 1, sending RST on connection 1, and a window size of 0.
When Unisys sends a SYN for connection 2, FlexIP accepts it and opens the
connection for file 2.

Questions:
1) Why is the DirectJet rejecting the connection?
2) Is there a way for the printer to say, "Come back later, I'm busy now?"
(Instead of, "Go away, there is nobody here by that name!")
3) Is there a way to distinguish between reasons for getting RST?
4) What do the unsolicited RSTs from FlexIP accomplish?

As a workaround, the Teksouth developer will retry the connection 
in 15 seconds if it had ever opened successfully, and give up if it
fails on the first try.

advTHANKSance,
Richard Schenke                    e-mail: [email protected]  
Grumman Technical Services         Phone (713) 280-2611
Information Systems Contract, Johnson Space Center, NASA
16511 Space Center Blvd, Houston TX 77058

-----------[000461][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 94 10:33:53 GMT
From:      [email protected] (Andy Harper, KCL Systems Manager)
To:        comp.protocols.tcp-ip
Subject:   Re: CMU TCP/IP
In article <[email protected]>, [email protected] (Tom Bodoh) writes:
> We used to run CMU TCP/IP on some of our VAXen.  We later upgraded to Multinet
> but we now have a need for some TCP/IP capabilities on a 'temporary'
> system.  Anyone know what the current status of CMU TCP/IP is?  Last I heard
> it was being abandoned to the user community to support.  What was the
> latest version available (6.6?), will it run on 5.52 VMS and where can it
> be found?  I have located 6.6 on the net.  Thanks...

Read the CMU FAQ available from:

    ftp://ftp.kcl.ac.uk/cmu-tcpip/cmu.faq


Regards,

Andy Harper
Kings College London

-----------[000462][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 10:56:15 GMT
From:      [email protected] (Steve Thompson)
To:        comp.protocols.tcp-ip,vmsnet.networks.tcp-ip.misc
Subject:   Re: VMS tcp/ip packages?
In article <[email protected]>, [email protected] (Ron Nadel) writes:
>Dave, there are several packages that will run under VMS.  Probably the most 
>robust is Multinet from TGV (Two Guys and a VAX).
>[...]                        ^^^^^^^^^^^^^^^^^^^

They're probably Twenty Guys and a VAX now :-) Maybe a change is needed;
lessee, TPA ("Twenty Persons and an Alpha").

-steve

---------------------------------------------------------------------------
Steve Thompson, System Mangler   Internet: [email protected]
School of Chemical Engineering   Phone:    (607) 255 5573
Olin Hall, Cornell University    FAX:      (607) 255 9166
Ithaca NY 14853                  "Time is just one damn thing after another"
---------------------------------------------------------------------------

-----------[000463][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 17:33:51
From:      [email protected] (Talina)
To:        comp.protocols.tcp-ip
Subject:   Getting my own PPP IP Address
I'm not sure if I'm using the correct "terminology" - but I recently heard 
that  you can get your own static IP address for PPP acounts from Us.Domain or 
something to that effect.

Does anyone understand it, and if you get that far anyone have an idea of 
where I'd find docs/info on getting one. (I am currently using Dynamic 
connection - or at least attempting to.) *grin*




------------------------------------------------------------------------------

[email protected]


-----------[000464][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 94 14:15:01 GMT
From:      [email protected] (Design Service)
To:        comp.protocols.tcp-ip
Subject:   Re: Could someone explaine me what's TCP/IP?
In Article <[email protected]>, arnaud wrote:
>
>        Hi,
as in "Hi Jack"?

>        I have been hearing this word for too long,
so stop reading.

>        Could someone explain me in details what is
>        TCP/IP...
>
>        Thanks,
>
>        If I could get some answer by mail at : 
>        [email protected]
Don't do it, it's a trap!  No one would actually post such a question unless
they were trying to lure the unsuspecting sap who would help them. 

-----------------------------------------------------
Ken Lucius, Sys Admin  !  Yield to temptation...
Design Service         !  it may not pass your way again!
[email protected]      !             --Lazarus Long

-----------[000465][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 14:32:16 GMT
From:      [email protected] (Gert Doering)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
[email protected] (Sam Oscar Lantinga) writes:

>	Complex, I admit, but I'm not sure of any other way
>to acomplish having my machine on the SLIP connection also appearing
>on the network at work.

Why not simply using Proxy ARP on the SLIP server? Sounds a lot easier +
faster.

gert
-- 
Yield to temptation ... it may not pass your way again!  --  Lazarus Long
                                                            //www.muc.de/~gert
Gert Doering - Munich, Germany                             [email protected]
fax: +49-89-3243328                         [email protected]

-----------[000466][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 14:46:46 GMT
From:      [email protected] (Dan Flowers)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.

In article <[email protected]>, [email protected] writes:
|>I am sorry if this is posted in a FAQ somewhere, the FAQ's I found didn't
|>mention it that I could find...  (how do you find the FAQ's anyway)  is
|>there a FAQ news group???).
|>
|>Thanks, Jeff

I'm not sure about a newsgroup, but I have found a WWW page:

http://www.cis.ohio-state.edu/hypertext/faq/usenet/FAQ-List.html#L

(But I don't recall seeing this newsgroup's faq listed...)

Dan

-----------[000467][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 15:16:04 GMT
From:      [email protected] (Jim Brown)
To:        comp.protocols.tcp-ip
Subject:   Looking for shareware for ppp running under Solaris
Working on a project and require shareware software dealing with ppp running 
under Solaris.  If you could respond and give me some information to 
point me in the right direction I'd appreciate it.  Please send responses 
to my email address:  [email protected]

thank you


-----------[000468][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 11:57:18 +0200
From:      [email protected] (Casper H.S. Dik)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?
[email protected] (Mario Klebsch) writes:

>This sounds very interesting. I tried it, as a regular user I typed:
 
>	$ ifconfig ie0:1
 
>But afterwards, when I type
 
>	$ ifconfig -a
>	lo0: flags=849<UP,LOOPBACK,RUNNING,MULTICAST> mtu 8232
>        	inet 127.0.0.1 netmask ff000000 
>	ie0: flags=863<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST> mtu 1500
>        	inet 134.169.36.10 netmask ffff0000 broadcast 134.169.255.255
>	        ether 8:0:20:a:5b:4e 
>	ie0:1: flags=842<BROADCAST,RUNNING,MULTICAST> mtu 1500
>        	inet 0.0.0.0 netmask 0 
 
>Is there to get rid of these entries? And why can I create them
>without any privileges?

They don't hurt, but only the superuser can get rid
of them (ifconfig le0:xxx 0.0.0.0).


I just discovered there's a maximum of 255 addresses per interface.

Casper

-----------[000469][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 15:49:46 GMT
From:      [email protected] (Curtis Justus)
To:        comp.protocols.tcp-ip,biz.sco.general,comp.protocols.tcp-ip.domains
Subject:   DNS Primary Setup Questions
Hi,
   I am working with 2 SCO UNIX Boxes, an NT Box with microsoft mail, an 
AS400, and a UNISYS 2200.  What I want to do is get one of my SCO boxes 
to be the Primary DNS.  I looked through the setup guide for TCP/IP and I 
don't think I set it up correctly.  If I give a quick configuration, 
could one of the whizzes here please give me a hand?

Op Sys, etc.     Mach.  Name        Address
------------     -----------        -------
SCO Box 1:        savo51            51.0.0.251     Primary DNS
SCO Box 2:        prosserv          51.0.0.252     Secondary DNS

How would I get just these two to work together... I could wait with the 
NT box for another post...

Thanks,
cj

--
*****************************************************************************
*  Curtis J. Justus                     *  Phone:  (414)-457-4433           *
*  Schultz Sav-o Stores, Inc.           *  "We Picked a bad year to have a  *
*  2215 Union Avenue                    *   good year."  - Ken Griffey Jr.  *
*  Sheboygan, WI  53082                 *    on his and Frank Thomas' year  * 
*****************************************************************************

-----------[000470][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 17:59:38 GMT
From:      [email protected] (Craig Morgan)
To:        comp.protocols.tcp-ip
Subject:   Re: Looking for automated DNS update tool.
In article <[email protected]>
[email protected] (Willy Janssen) writes:

> As we have to do a relevant number of DNS updates, I was wondering if there is
> an automated tool around to include the update-info in DNS.
> 
> In first instance I'm thinking of automated generation of the reverse map, serial
> numbers etc. At second there should be a syntax check on the input files to detect
> all kind of typo's, and other errors.
> 
> Please let me now.
> 
> --
> 
> Willy Janssen - University of Nijmegen - Holland
> 
> Email: [email protected]
> 

Try checking the archie servers for a program called h2n, it comes with the
nutshell guide to DNS and Bind. It's available in the bind distribution also.

It's a PERL script which will generate the relevant info from a single hosts
file, including all reverse address files, notification of duplicates, etc.

Well worth using, mail me if you can't find a copy and I'll send you mine
(slightly modified!).


--
Craig

  Craig Morgan                       Lecturer, CS Group
  School of Computing                WWW, Gopher, Anon FTP Maintainer        
  
  Staffordshire University, UK       Email: [email protected] 

-----------[000471][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 01:14:39 -0400
From:      [email protected] (???)
To:        comp.protocols.tcp-ip
Subject:   Linux TCP/IP to Novell
I am trying to set-up my Linux server with Novell server.  Can someone tell me where to start or where I can go for information? Any help will be appretiated.  Thank you.

[email protected]

-- 
	-jOhN lEE -	rUtGeRs UnIvErSiTy	[email protected]

-----------[000472][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 18:41:37 GMT
From:      [email protected] (Bill Richter)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP-IP Packages for PC's (Again)
In article <[email protected]>,
[email protected] (steve johnson) wrote:

> In article <[email protected]>, [email protected] (Susan
> Ferebee Guion) wrote:
> 
> > I'm once again looking for the best in tcp-ip packages for my PC's.
> > 
> > I'm had it woth FTP's PCTCP and the ~70K of low memory it requires.  Now
> > I'm looking for a replacement.  I'll need all the standards (ftp,telnet,
> > ping,...) but I need to preserve low memory.  I might even decide to go
> > with an X-terminal package if the proce war right.
> 

If you can find a September 19, 1994 copy of INFOWORLD you can read the
Reviews/Product Comparison of TCP/IP products starting on page 68.  From
their executive summary they ranked products as follows:

    DOS/Windows
      WRQ RNS/TCP Connection Windows 4.0 Reflections 2VT3200 WIndows 4.11 -
7.5
      InterCon TCP/Connect II for Windows 4.0                             -
7.1
      Novel LAN WorkPlace for DOS 4.2 NFS Client for LAN WorkPlace 3.0    -
7.0
      NetMange ChameleonNFS TCP/IP for Windows 4.0.1                      -
6.9
      Frontier Super TCP/NFS for Windows 4.0.0                            -
6.9
      FTP Software PC/TCP OnNet 1.1 for Windows                           -
6.7
      SunSoft PC-NFS 5.1                                                  -
5.3
      Beame & Whiteside BW-Connect NFS for DOS & Windows 3.1              -
5.1
      Wollongong PathWay Access 3.0 Client NFS 2.0.1 Runtime 2.0          -
5.1

   Macintosh
      Wollongong PathWay Access 2.12 Client NFS 2.0                       -
6.2
      InterCon TCP/Connect II for Macintosh 1.2.1 NFS/Share 1.3.3         -
6.1

   OS/2
      IBM TCP/IP 2.0 for OS/2 Base Kit, NFS kit,DOS,Windoes Access kit    -
6.5
      FTP Software PC/TCP for OS/2 1.3                                    -
6.1

I tried to COPY this from their article the best I could.  Typically, each
issue has a correction to previous articles so sometime in the feature one
of the above ratings may change..but then again they might not.

As for my experience, when I was using an OS/2 machine we used the IBM
TCP/IP packages.  Overall it's pretty good.  Over a year ago when we needed
it, it was the fastest FTP around and we were doing large file transfers. 
Also, it was the only one at the time with a 5250 client.  We needed that
for connecting to AS/400's.  I did participate in an IBM phone interview of
users of their product and passed on some comments for improvement.  First,
the keyboard mapping could be done through a GUI tool or something rather
than having to edit files for 3270/5250/etc. sessions.  Second, 2.0 didn't
install over previuos (1.?) version.  I know, I tried it 4 - 5 times before
giving up.  Also, it took about 3 tech. support people at IBM to get around
that one.

Now that I'm on a Mac, I'm using some NCSA telnet programs and they work
ok.

> x terminal emulators take much more time and memory.  

Didn't work much with X terminal software on the IBM, but did a little
testing with the IBM product.  The early version was real slow, but when we
upgraded to the latest release with all of the CSD's, we were able to move
right along.

All testing was done and worked at 6M and 16M.  6M works, but is not
recommended.

> 
> > I'm interested in experiences you might have had with any of them.
> > thanks in advance...
> > 
> > -- 
> > Susan Guion - Georgia Tech
> > Internet: [email protected]
> > uucp:     ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!gt5582a
 
-- 
.------------------------------------------------------------------------.
| Bill Richter, IL02/SH-6       | Motorola E-mail:  cpgn32               |
| Motorola, Inc.                |        Internet:  [email protected] | 
|1301 E. Algonquin Road         |                                        |
|Schaumburg, IL  60196-1077 USA |                                        |
'------------------------------------------------------------------------'

-----------[000473][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 18:43:26 GMT
From:      [email protected] (Sam Oscar Lantinga)
To:        comp.os.linux.development,comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
Gert Doering ([email protected]) wrote:

: Why not simply using Proxy ARP on the SLIP server? Sounds a lot easier +
: faster.

	Two reasons.  

1)	The SLIP server is on a different subnet.
2)	I don't have access to the SLIP server.

Thanks,

	-Sam


-----------[000474][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 18:50:26 GMT
From:      [email protected] (Tom Bodoh)
To:        comp.protocols.tcp-ip,vmsnet.networks.tcp-ip.misc
Subject:   Re: VMS tcp/ip packages?
In article <[email protected]>, [email protected] (Ron Nadel) writes:
|> In article <[email protected]> [email protected] (David D Kaas) writes:
|> 
|> >We have some Dec statio 3100s running VMS 5.x.  We would
|> >like to install TCP/IP software.  What companies support
|> >TCP/IP on VMS?  Are there any public domain or low cost
|> >packages?
|> 
|> Dave, there are several packages that will run under VMS.  Probably the most 
|> robust is Multinet from TGV (Two Guys and a VAX).  It is a full function, easy 
|> to use package, and their tech support is very good.  Carnegie-Mellon provides 
|> a very inexpensive package called CMUTECH.  It costs almost nothing (outside 
|> of media/shipping).  It is not as complete as Multinet, and of course they do 
|> not have the tech support of a commercial vendor.  But it does work.  The last 
|> number I had for them was 412-268-5896.
|> 

Note that CMU abandoned CMUIP to the public domain a few years ago.  I just
recently found it on csus.edu and it installed and ran fine.  Well - I do
know the shortcomings from previous experience; crashes and slowness but hey
it's free.  We use Multinet on our other VAXen and it is great...

-- 
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ Tom Bodoh - Section Manager, Systems Engineering and Management, Hughes STX +
+ USGS/EROS Data Center, Sioux Falls, SD, USA	57198     (605) 594-6830      +
+ Internet; [email protected] (152.61.192.66)   Amateur radio call; N0YGT +
+	"Welcome back my friends to the show that never ends!" EL&P	      +
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

-----------[000475][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 19:01:39 GMT
From:      [email protected] (Sam Oscar Lantinga)
To:        comp.protocols.tcp-ip
Subject:   Re: Extending the IP Protocol?
: >	My question is....  Will packets with an unrecognized
: >IP option like this be forwarded across routers?  I noticed that
: >in the Linux network code that if a packet has an unrecognized IP
: >option, the packet is discarded.  Is this standard behavior?  Or
: >will my custom packets make it safely across the network?
 
: Sounds like the IPOPT_DISCARDME feature :-)
 
: Seriously it should discard it.


	Hmm... Sounds like IP tunneling is the way to go.
Not a problem.  Someone mentioned that this has been done before...
Can anyone tell me where I can get the specs on it?

: to my SLIP server from COM2.  I then used the SLIP server (but you can use
: a machine that supports SLIP) with one port in dialout mode (telnet to it)
: which was wired with a null modem to one of the regular SLIP ports.  I was
: able to have a second SLIP connection on COM4 tunneled through TELNET.

	*grin*  You read my mind.  My other secret plan was to figure
out how to set my Linux box up as a SLIP server, running sliplogin on
the telnet pty.  I dunno.....

Thanks!

	-Sam


-----------[000476][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 19:35:19 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: Novell TCP/IP routing problem
In article <[email protected]> [email protected] (A. Grant) writes:
>In article <[email protected]> [email protected] (don provan) writes:
>>NetWare accepts any mask that allows at least two bits for the subnet
>>number and at least two bits for the host number.
>
>Yes, but does it accept different masks on different interfaces,
>and does it route correctly in that case?

The currently shipping products support only RIP-I routing topologies,
so they do not support splitting a single network into subnets of
different sizes. The exception is proxy ARP support in the
MultiProtocol Router product, but I assume that's not the kind of
thing you're talking about.

(Actually, you can configure NetWare with variable subnet masks and,
so far as it's been tested, it routes fine.  But without the aid of a
routing protocol that handles variable sized subnets, the caveats to
such a configuration are too extensive to go into, so officially they
are not supported.)

					don provan
					[email protected]

-----------[000477][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 19:38:46 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   Unix and TCPIP etc
I am interesting in finding out what is needed to have a "host"
one where i can offer the email, telnet, ftp, etcc.
I assume i will need the TCPIP stuff.
DO i need unix? 
thanks for any replies(privately please)
Thanks,
Ed

-----------[000478][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 19:39:17 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP and ARCHNET
In article <[email protected]> [email protected] (Pete A. Zaitcev) writes:
>This happens only if you use RFC-1051 framing. Current implementation
>must follow the RFC-1212 which includes datalink level fragmentation.

The correct number for the RFC describing the modern IP encapsulation
on Arcnet is RFC-1201, not RFC-1212.

						don provan
						[email protected]

-----------[000479][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 19:45:40 GMT
From:      [email protected] (don provan)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP/IP On NetWare Client
In article <[email protected]> [email protected] writes:
>	Could someone please tell me if a NetWare client
>can communicate with a NetWare v3.11 server using TCP/IP
>instead of IPX?

Someone already mentioned the NetWare/IP product, which is the best
way to do what you're describing. Depending on how simple your needs
are, you may only need the trivial approach of IP tunnelling.

>Can this be done with LAN Workplace For DOS?

The DOS components for tunnelling were shipped with LAN WorkPlace
version 4.0 (or so) and later. The NetWare components are in NetWare
v3.11 and later. The documentation for setting up a tunnel is in the
NetWare manual on TCP/IP configuration.

						don provan
						[email protected]

-----------[000480][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 19:59:03 GMT
From:      [email protected] (Richard Kurtz (816)995-3204)
To:        comp.protocols.tcp-ip
Subject:   Dummy's Routing Question - Class B Network
I am trouble shooting a routing problem in a SunOS 4.1.X environment.  We use
Class B networks, subnetted, normally in this environment.  Recently a group
has moved a LAN segment such that it now is no longer attached to the original
path and Class B network.  They want to route to different subnets of a Class B
network on different interfaces of a Sun.

This does not appear to be a workable solution.

My question is, what is the RFC that I can point them to to convince them to
either readdress the LAN segment so it is on a contiguous Class B network or
move it back to it's original Class B network?

Thanks 
---
+=============================+
| Richard Kurtz               |
| AT&T                        |
| Room C500M                  |
| 2121 East 63rd Street       |
| Kansas City, MO 64130-3440  |
 +=============================+
| Email Richard [email protected] |
| Phone (816)995-3204         |
+=============================+



-----------[000481][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 20:15:06 GMT
From:      [email protected] (Dale Stilwell)
To:        comp.dcom.lans.ethernet,comp.protocols.tcp-ip
Subject:   Re: Running IPX and TCP/IP over same ethernet
Rob Tanner ([email protected]) wrote:
> Direct from Rumor Control Central:  I have heard that IPX is a real
> bandwidth hog, and that if you don't want to see your TCP/IP severely
> degrade, don't let IPX run over the same segments that run TCP/IP.
 
> I know zip about IPX, but it looks like we are going to be installing
> a Novell server anyway.  The Novell, of course, is for the PCs.  What
> concerns me as that we also have a number of Unix machines as well as
> an even larger number of X-stations that don't run IPX and don't
> concern themselves with the Novell server.
 
> Will my unix boxes feel the impact of IPX, or is the rumor unfounded?
> I could isolate the IPX traffic, but to do that requires purchasing
> hardware and that means money that has to be justified.  Does anybody
> have any real performance figures that they could share.
 
> Thanks,
> Rob

Rob,
	Most of those rumours come out of the religious "protocol" wars. 
The bandwidth hogging nature of IPX is greatly exagerated. IPX's "bad"
reputation comes from the fact that 1) it broadcasts to announce it's 
services. 2) it broadcasts to announce it's routing (Hey who doesn't) and
3) it's NCP layer is not very efficient. To put this in perspective, the
Service broadcasts are a very small number of packets. The only time I've
ever seen a link or segment swamped by SAPs and RIPs was a single WAN 
segment in a IPX environment made up of over 15000 services. In this case
a single 56K circuit ended up carrying nothing but SAPs and RIPs. In your 
case you may end up with maybe 20 services (including printing, faxing, 
backups, virus scanning ...) so you won't even notice them. Netware's RIP
protocol probably doesn't generate any more bytes than OSPF. And NCP is
not any less efficient than NFS.
 
I wouldn't worry about it.

Dale

P.S. I've got 30 or so segments at 12 different sites that run both IP and
IPX as well as OSI and SNA/LLC all at the same time. No problems.
--
======================================================================
  Dale Stilwell ([email protected])	  Kraft General Foods Canada
  Phone:(514) 340-2544 (Montreal)	  Fax:(514) 340-2297
  		      #include <standard/disclaimer.h>
  MS LanMan: It's got everything I hate about Netware and then some...
======================================================================

-----------[000482][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 21:12:14 GMT
From:      [email protected]
To:        comp.protocols.tcp-ip
Subject:   FTP on AS/400 (OS/400) And VAX(VMS)

  We need to transfer files from an AS/400 (OS/400) to a VAX (VMS) over a 
TCP/IP connection. Except for FTP, we have been unable to find a vendor who 
sells file transfer software for this combination of platforms over TCP/IP. 
Connect:Direct from Sterling Software does not work for this combination and 
MLINK from Legent does not support these two platforms at all.

   Does anyone know of any other vendors who has a product that will work? FTP 
does not allow a restart from where it left off if there is an interruption 
while the transfer is in progress. We need this capability because of the size 
of the files and time constraints for transferring these files.

My email address is: [email protected]


Edward Nash
System Administrator/Programmer/Software Evaluator
Republic Information And Communications Services Inc.
New York, NY


-----------[000483][next][prev][last][first]----------------------------------------------------
Date:      Tue, 20 Sep 1994 23:13:25 GMT
From:      [email protected] (Paul C. Brant)
To:        comp.protocols.tcp-ip
Subject:   HELP REQUIRED: Problem using connect and socket stuff !
Hi,

I am trying to write a program that will do the following:

connect to remote terminal server port and start capturing data to buffer.
check buffer against lookup up file for key works ( CRASH, MOUNT TAPE etc.)
distribute data to any other client that might telnet my program.

My problem is with the client side. IE connecting to the remote terminal server.

When ever my function issues a connect the program stops with a msg like:

tclient: Connect Error: Invalid Argument

My code for tclient is as follows:


#include "plog.h"

tclient(hostname, portnum)
{
	register int h;
	int hpid;
	struct sockaddr_in hostsock;
	struct hostent *hp;
	char hostbuf[80];

	if((hp = gethostbyname(hostname)) == 0) {
		perror("tclient: Bad host name");
		return(1);
	}

	if(h = socket(AF_INET, SOCK_STREAM, 0) == -1) {
                perror("tclient: Socket Error");
                return(1);
	}

	hostsock.sin_family = AF_INET;
	hostsock.sin_port = htons(portnum); 
	bcopy(hp->h_addr, &hostsock.sin_addr, hp->h_length);

	if(connect(h, &hostsock, sizeof(hostsock)) < 0) {
		perror("tclient: Connect Error");
		return(1);
	}

	if((hpid = fork()) == -1 ) {
		perror("tclient: fork error");
		return(1);
	}

	if(hpid == 0) {
		while(1) {
			if(read(h, hostinbuf, sizeof(hostinbuf)) == -1) {
				perror("tclient: Error recieving from host");
				return(1);
			}
			printf("Recievied the following from host: %s\n", hostinbuf);
			if(write(h, hostoutbuf, sizeof(hostoutbuf)) == -1) {
				perror("tclient: Error sending to host");
				return(1);
			}
 printf("Sending the following data to host: %s\n", hostoutbuf);
		}
	close(h);
	printf("tclient: closed connection !\n");
	return 0;
	}
}

All help / comments / suggestions welcome


-----------[000484][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 01:22:03 GMT
From:      [email protected] (Ftp Maintainer)
To:        comp.protocols.tcp-ip.ibmpc,comp.protocols.tcp-ip
Subject:   Where to find source of WATTCP FTP
Hi,
  We need to modified some command of WATTCP FTP. But I am fail to get
it's source from dorm.rutgers.edu:/pub/msdos/wattcp (Main archiver). :(
Is there any other site I can get it????

  Thanks for help. :)


--

<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
<>   CS Lee	(Lee Chee Siong)		<>  If you can give	<>
<>   System Manager of Computer Center          <>        nothing else, <>
<>   National Chung Hsing University,		<>     :-)    (-:       <>
<>   Taichung, Taiwan, Republic of China.	<>  at least give a	<>
<>   Internet Address: [email protected] <>        cheerful face <>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>

-----------[000485][next][prev][last][first]----------------------------------------------------
Date:      Wed, 21 Sep 1994 05:32:47 GMT
From:      Robert Minich <[email protected]>
To:        comp.protocols.tcp-ip
Subject:   Re: is exponential backoff really right?
Please pardon the extensive quoting.

Phil Howard, [email protected] writes:
>Consider:
>
>You are linked via a SLIP or PPP dialup line.  The line connection drops
>and is redialed, either automatically, or manually.  It takes about 45
>seconds minimum to restablish the phone connection.  This is the best
 case.
>This does not consider the delays involved in detecting the condition.
>
>During this 45 seconds, the exponential backoff has been working against
>you.  Retries are now up in the 30 to 60 second range.  If you didn't get
>reconnected really quick, it can be worse.
>
>Given a period of outage, the delay in reestablishing traffic flow could
 be
>as long as the outage.  For reasons I am not sure of, I've seen many
 where
>the delay (lag) is a lot longer.

If your outage is on the order of a minute or less, I don't think a
one minute delay in resuming sessions is a real problem. If the link
was down longer, the TCPs I've seen all put an upper bound on
timeouts at about one minute. I admit I've seen situations where I
could start a new TELNET session and caused output from the remote
side which kick started a TELNET session that was currently waiting
~60s to retransmit. Maybe we need an ICMP Reachable message? :-)

>Another condition I have seen is where the backoff is different on each
>end.  After getting the dialup reconnected, I'm getting the output from
>my session a LOT sooner than my input is being accepted.

As soon as one side sees a packet from the other, an ACK should sent
along with a segment of data. At that point (barring a very lossy
network) the only thing slowing you down is slow start. That's a
small price to pay for avoiding polluting an internet with
retransmissions that aren't likely to get through.

>I would like to see an exponential backoff that is not as harsh as this.
>Instead of doubling the number of seconds, how about a proportion that is
>less than 2.
 [table comparing x^2 and x^1.25 backoffs snipped]
>In this case the proportion is approximately 1.25 and using millisecond
>or microsecond resolution could make the numbers easier to calculate
>instead of using a table lookup (although I think a table lookup is just
>fine).

You've just increased the number of transmissions unlikely to get through
from 7 to 14 or 15 over two minutes.

If you _really_ want to attack this problem, let the data link level
(SLIP/PPP in your case) notify upper layers when: (1) the link
is found to be down nothing is going to get through so don't bother
sending it and (2) the link comes back up and the upper layers
should attempt to resume their activities (for example, send a TCP
segment now and get the show on the road.) This won't help when the
downed link is somewhere distant but between the connection endpoints.
In that case you might as well deal with a "harsh" exponential backoff.

>This is particularly annoying with interactive session (telnet, rlogin).
>Your session is frozen for quite some time.

In the more difficult general case I mentioned above, you might
consider (CAREFULLY) an option to let the user cause an immediate
timeout and and restransmission, perhaps by kicking the DLL layer with
a retry_now message of some sort that gets passed upwards.

>If I were to experiment with such an exponential backoff, what kind of
>impact might there be on the existing network?

Retransmissions due to your local SLIP/PPP line going down won't be a
real problem. However, when some other host out in the Internet
becomes unreachable for some reason, you will be sending a lot of
packets that are likely end up in the bit bucket. If everyone does
this, we all get stuck with more traffic than with an exponent of 2.

-----------[000486][next][prev][last][first]----------------------------------------------------
Date:      Wed, 21 Sep 1994 05:43:54 GMT
From:      [email protected] (Stanley Chan)
To:        comp.protocols.tcp-ip
Subject:   Re: Using DNS
D. Hobbs ([email protected]) wrote:
: I'm trying to get our RS/6000 to resolve domain names.  Our
: DEC machine is pointing to some SPRINTNET name server, so it
: seems to work fine.  However, I'd like to point all of our
: other Unix boxes (including the RS/6000) at the DEC machine.
: When I do what the DNS book says, the RS/6000 seems to eventually
: grind to a slow halt.  I'm guessing it has to do with the  domain
: name cache size, but I'm only guessing.
 
: In the /etc/resolv.conf, I've specified:
: nameserver 5.5.5.5 (assuming that is the valid IP address)
: domain my.domain.com
 
: The RS/6000's IP address is 5.5.5.4, and the DEC machine is
: 5.5.5.5.  The domain name is not registered, but the IP addresses
: are (these domain names and IP are not valid -- protecting the
: innocent :).
Is the DEC machine running named ?? If not I dont think it will work
because the resolv.conf must point to a host with DNS setup.
If you dont want to setup a local DNS, then your RS/6000 must point to where
the DEC is pointing to.

: Anyway, when I do the above, the RS/6000 eventually just slows down
: to nothing (rlogin, telnet, etc. -- user processes are OK and the
: system load doesn't seem to change).  Any clues?
Just because it tried to find a named server for the DNS request, if it
did not find one, it will keep on trying unitl timed out.
Ohter processes not using DNS are not affected.


: Thanks,
 
: D.
: -- 
: David Hobbs
: [email protected], [email protected]
: All typos are my own.  The spelling errors are someone else's.
: "I am Fudd of Borg.  Pwepare to be assimiwated."

--
Stanley Chan				
E-mail [email protected] 		(Ph 61-7-8771016 Fax 61-7-8771120)
Snail  Golden Casket Art Union Office
       Locked Bag 7, Coorparoo DC QLD Australia 4151



-----------[000487][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 06:31:10 GMT
From:      [email protected] (Jon Kay)
To:        comp.protocols.tcp-ip
Subject:   Re: TCP or UDP?  That is the question.
In article <[email protected]> [email protected] (Vernon Schryver) writes:
>    -TCP checksums are required by the TCP standard, but no one will
>	come to take you away if you do not use them.  A few people
>	(misguided in my view) do suppress TCP checksums.

Thanks for the introduction, Vernon.  I beg your pardon - I'm afraid
I'm not going to debate this as fiercely as on earlier occasions,
since I'm busy looking for a job (anybody need a TCP/IP performance
expert?) AND writing my Ph.D. thesis.

However, I do think it's time for the quarterly Checksum Redundancy
FAQ posting:


	Checksum Redundancy Avoidance Frequently Asked Questions

Discussion about checksum redundancy avoidance (CRA) seems to arise
about once a quarter starting from when we published a Usenix paper on
the subject.  Thus, I have decided to write a FAQ on the subject and
simply post it whenever the subject arises, perhaps with the
occasional update.

NOTE: This FAQ details a slightly more advanced version of CRA than I
have previously described.  CRA has strategies for detecting bridges
and bad adapters now.

> What is checksum redundancy avoidance?

CRA is a technique for considerably raising the throughput of TCP/IP
implementations by permitting them to avoid doing checksumming in many
circumstances.  The basic observation is that both TCP/IP and LAN
hardware implement a checksum.  To a certain extent, the two are
redundant.  Specifically, they are redundant for packets that are
being sent over a single LAN, which is most packets.  Our design does
not checksum such packets; it does checksum routed and bridged
packets, since CRCs don't cover routers, and they don't cover some
bridges.

> Where can I get more information than just this &^%&* FAQ??
> How can I run CRA?

A paper and implementations for Ultrix 4.2a and DEC OSF/1 1.3 are
available for anonymous FTP from the directory

ucsd.edu:pub/csl/fastnet

usenixw93.ps is the name of the paper; fetch the README to figure out
what you want in the way of implementations and for more information
about my work and other UCSD Computer Systems Lab work.

Incidentally, the CRA implemented described in the paper is an earlier
and less sophisticated version than is described in this FAQ.  We are
working on publishing the upgraded scheme more formally. 

> How is CRA different from turning off UDP checksums?

Just turning off UDP checksums means that all UDP transmissions, no
matter to where in the Internet, are unprotected.  Many WANs (notably
SLIP links) are not protected by CRCs.  So running NFS over a SLIP
link could spell disaster if you just turn off UDP checksums, whereas
CRA would detect this situation and automatically checksum packets.
Additionally, no router is protected by a CRC, and the chance of
corruption within a router goes up exponentially with the number of
hops a packet travels. 

Also, CRA applies to TCP just as much as to UDP.  Note that most data
is transferred across UDP (via NFS) these days.

> I'm looking at the output of netstat -s on my machine and it shows
> that checksum errors have been detected.  Doesn't that mean that CRA
> is a bad idea?

There's no way of telling whether those errors occurred on your local
LAN or on a WAN or in a router.  It's my belief (and my experience
tends to back this up) that almost all checksum errors crop up across
WANs.

> What about end-to-end ness?  Aren't you sacrificing reliability by
> no longer checking memory-to-memory but rather adapter-to-adapter?

An small amount of reliability within the network subsystem, yes.
Reliability within the system as a whole, no.  Nobody checksums disk
activity or memory copies or program correctness to memory either
except in the most perfunctory way (LFS checksums one word per data
block, less protection than the IP header checksum).  In a system that
is subject to errors from many influences, does it make sense to
demand complete reliability from the network subsystem, especially at
such a high cost as is imposed by data checksumming?

> What about whole bad designs of adapters like the DEQNA, that were a
> major factor in the decision to always checksum?

The new version will have statistical checksumming: checksumming every
tenth or twentieth packet.  The hope is that it will catch bad
adapters relatively quickly.  If too many errors are seen, the
implementation will print an error message and turn off CRA on that
interface.

This may result in higher reliabilities, since it would make bad
adapters easier to detect.

Remember that your SCSI interface can (and sometimes does) betray you
without any sign for years.  Yet you haven't yet raised a lynch mob to
hang the guy who wrote your filesystem.

> How do you tell that a host is on the same LAN?  What about bridges?

Whenever a host with this extension answers an ARP request, it also
transmits an IEEE spanning tree packet configured to look innocuous
(highest priority numbers, etc.).  Neither routers nor bridges will
forward such a packet.  Routers will simply ignore it (it isn't
IP/IPX/etc.), and bridges are compelled to interpret without
forwarding spanning tree packets.  If a host receives such a packet,
it can be sure that the sending host is on the same LAN segment.

This is different from the earlier version of the scheme and the
implementations, which both depend entirely upon IP routing
information.  The old scheme was usually good enough, but this new
scheme is safer. 

The new scheme still checks IP routing information so that
checksumming will be done in the case of multiple logical subnets on a
single piece of wire.  Sometimes packets travel through a router even
though they are on the same physical network.

> What about proxy ARP?  Can CRA detect that and "do the right thing"?

Yes.  It doesn't matter to CRA whether IP knows that a packet is
routed.  It only avoids checksumming if the host is actually on
the same piece of wire.

> What about repeaters?

The LAN CRC covers repeaters.  Repeaters don't recalculate the CRC, so
if any changes happen across the repeater, they show up as a CRC error
at the destination's adapter.

One problem with the earlier version of CRA is that multimedia bridges
do generally recalculate CRCs because of the change in header and lack
of an obvious associative property or subtraction operator in CRC
calculation.  

> What about bridges that don't support the spanning tree protocol?

There are some out there.  If it does follow the bridge spec enough to
forward packets with unmodified CRCs, or it's even vaguely reasonably
reliable, then it doesn't matter.  In other words, three totally
separate things have to be wrong with a bridge for CRA not to work.
The notion is that that's probably less likely than other sources of
system failure.

> In a world where you can buy interfaces that do Internet
> checksumming, why do we need this scheme?

CRA is basically free.  No extra hardware required.  I have yet to see
a CHEAP interface that implemented Internet checksumming.  Admittedly,
it's easier than it might be for FDDI because those kinds of networks
so far seem to require an onboard processor for SMT basics and the
like.  However, SGI's checksumming FDDI board, for example, requires
an AMD 32000 32-bit RISC processor while DEC's FDDI board is able to
make do with the 16-bit MC68000.

> What mechanisms do you use to negotiate nonuse of checksums?

UDP already has a mechanism for this - just have the sender put a zero
in the checksum field.  We have the sender of UDP packets decide
whether CRA should be used and use that protocol.

For TCP we use Zweig and Partridge's TCP Alternate Checksum Option
(RFC1146).  The IANA has kindly granted us alternate checksum 3 to
be no checksum at all.

> Whose idea was it?

A number of people seem to have had the idea over the course of last
decade or so.  It's not the world's subtlest idea.  However, as far as
we know, Joe Pasquale and I are the first to actually implement it and
write it up and pursue it and refine it.
-- 
WWW/Mosaic Home Page:  http://www-cse.ucsd.edu/users/jkay
Email:                 [email protected]

-----------[000488][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 07:59:25 GMT
From:      [email protected] (Brad Lemings)
To:        comp.protocols.tcp-ip
Subject:   FTP welcome and help messages...
I could find the answer to this question if I did enough research but, in
hopes of a quicker answer: how does an FTP server display welcome messages,
help messages, and other such messages automatically?  Some servers display
messages before you even log on, others after you log on, change directories,
send/receive files, etc., etc.

Eric Lemings

-----------[000489][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 08:23:56 GMT
From:      [email protected] (Frederic CHOLLET)
To:        comp.protocols.tcp-ip
Subject:   tcp/ip via isdn

hello, I'm a french guy who's got a big problem. At first, I'm sorry for
my poor english. 
Ok, let's now speek about it.
I'm a student in Telecommunication (Eurecom Institute, NICE) and I have to find
a solution to connect a PC or a Mac via an ISDN link to a Web Server 
(on Internet). The problem is that I don't know if it exists any gateway 
between ISDN and TCP/IP (a software or a hardware) and what sort of difficulties
this link presents.

I need all pieces of information you've got about it,
if you know some newsgroups more appropriate than this one, tell me their name.

	thank you very much, Fred

my e-mail : [email protected] 

-----------[000490][next][prev][last][first]----------------------------------------------------
Date:      Wed, 21 Sep 1994 14:02:05
From:      [email protected] (Danny Hughes)
To:        comp.protocols.tcp-ip
Subject:   CLASS C SUBNET?
I need help subnetting a Class C address on a Novell token ring file server.

I currently have 2 nic's in the file server and would like to use the same 
class c address for tcp/ip traffic.  Is there any way to subnet this address 
so that one card uses xxx.xxx.xx.1 through xxx.xxx.xx.127 and the other card 
uses xxx.xxx.xx.128 through xxx.xxx.xx.254?  Any help would be greatly 
appreciated.

Thanks in advance

Danny Hughes

-----------[000491][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 21:48:55 -0700
From:      [email protected] (Greg Earle)
To:        comp.protocols.appletalk,comp.protocols.tcp-ip
Subject:   Patches to "tcpdump" to directly grok AppleTalk (EtherTalk) packets?
I noticed that tcpdump 2.2.1 has support for de-mangling EtherTalk packets
(DDP and some subsets thereof, like NBP and ATP), but this is only invoked
if the programs discovers them hidden (i.e., encapsulated) inside a UDP
packet, as done by KIP.

We've got a Sun-4/370 here with "netatalk" installed, not to mention many,
many Macs on my subnet with Ethernet cards emitting EtherTalk packets, and
I'd love to have a version of "tcpdump" that can grok these packets directly.

I've spent some time looking at print-ether.c and gencode.c et al., and I'm
feeling like maybe it's just a SMOP, but I'm sure this is a wheel that's been
invented before somewhere out there.

Does anyone have patches to print-ether.c et al. to directly grok AppleTalk
(EtherTalk) packets and feed them to the ddp_print() et al. routines in
tcpdump's print-atalk.c?  If so, could you respond to me directly?

Thanks in advance,

-- 
- Greg Earle                    WWW: http://www-mipl.jpl.nasa.gov/~earle/
  Phone: (818) 353-8695                 FAX: (818) 353-1877 [Call # again if
  Internet: [email protected]			     you get !FAX tone]
  UUCP: [email protected] a.k.a. ..!{ames,usc}!elroy!isolar!earle

-----------[000492][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 10:52:17 GMT
From:      [email protected] (Robert Bannocks)
To:        comp.protocols.tcp-ip
Subject:   Re: Help needed on KA9Q
the number is more flexable you can gain access to the net> prompt if you
set the right bit.

try ftpusers with guest guest / 127 
you should be able to login as guest and type @ to get to the net prompt

Sergey A. Elistratov ([email protected]) wrote:
: In article <[email protected]>,
: David Rudder <[email protected]> wrote:
: >I have KA9Q running a router at a (somewhat) distant site.  I don't like 
: >going over there, but have to make frequent changes to the setup.  All of 
: >the stuff I do I could be doing over telnet.  I have setup telnet to run 
: >on it, but I can't seem to create a password file.  As it stands, I can 
: It's not difficult. In ka9q config directory ( c:\net by default) file
: "ftpusers". Format: 
: login password acsess_dir permissions
 
: strat	1234	/strat	7
: ...
: But i not find usefull for gateway staff commands.
: Telnet session oriented for mailbox and telnet protocol
: commands (send file etc.).
:  
: >telnet there, but can not login because it doesn't recognize any logins.  
: >The docs don't explain this.
: >
: >			Thanks in Advance,
: >			David Rudder
: >			[email protected]
: >
:         Good luck,
: 	Sergey.


--
------------------------------------------------------------------------------
YES, the loonies are running the asylum
------------------------------------------------------------------------------
If you have an american express card you are the problem
------------------------------------------------------------------------------
Robert Bannocks                              Email:  [email protected]
Systems Programmer, Kingston University, Penrhyn Road, Kingston-upon-Thames.
------------------------------------------------------------------------------

-----------[000493][next][prev][last][first]----------------------------------------------------
Date:      Wed, 21 Sep 1994 11:09:48 GMT
From:      [email protected] (Alan Cox)
To:        comp.protocols.tcp-ip
Subject:   Re: Verification of a TCP implementation
In article <[email protected]> [email protected] (Matthew Gream) writes:
>The choice of a "bake off" with other implementations doesn't appeal to
>me because it seems like it's only going to test general functionality
>-- ie. only half a solution.

A bake off tests the functionality normally used. Its a very good start.
It'll also show you stupid real world problems (bad ack handling on old
wintrumpets, invalid sequence numbers from a certain vendors old terminal
servers, bizarre systems that use window < mss (try feeding that through
the SWS avoidance algorithms).

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  [email protected]   //  [email protected]#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''

-----------[000494][next][prev][last][first]----------------------------------------------------
Date:      Wed, 21 Sep 1994 11:22:28 +0200
From:      [email protected] (Interpac's master)
To:        comp.protocols.tcp-ip
Subject:   sz through a telnet session ?
Yes, I know this question already has been debated ... I even know one
solution... but here is the question :

Our users are connecting to a terminal server servicing a modem rack.
Upon login, the terminal server establish a telnet session to our
SunOs 5.3 host. But the users aren't able to download files from
the unix box using the Zmodem protocol.

Now, the "classic" answer is :
  begin the session normally
  exit to the terminal server and issue something like
    "session telnet client profile binary"
  come back to the session

But we don't want, for security reasons, to allow people to issue commands
to the terminal server. And we think this kind of operation should be totally
transparent.

Don't tell me : issue the connection in binary mode. It doesn't work. Due to
CR-LF translation rules, the user can't even enter his password on the unix
box.

Deseperate ... need help ... thanks

--
 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
/Jean-Francois "Jef" Stenuit | Interpac NV/SA               |    // Amiga    \
\Internet administrator      | 350/358 Avenue Louise Box 11 | \\//  fanatic  /
/Phone (32)(2) 646-6000      | B-1050 Brussels              |  \/            \
\Fax   (32)(2) 640-3638      | Belgium                      | & SparcStation /
/Email [email protected]       |                              |   adict ...    \
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

-----------[000495][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 22:32:22 +1000
From:      [email protected] (David J. Hughes)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?

[email protected] (Ricard Wolf) writes:

>There is a group of class A B and C networks which are designed to
>be used internally in networks that will never be directly connected
>to the Internet (however, connection via a router is possible), which
>are defined in RFC 1597. Could this be what the vendor meant? 
>Don't know about the term "sider" though; never heard it before.

CIDR (Classless Inter-Domain Routing) is a mechanism that was developed to
extend the life of the IP address range.   From what I've read it was
developed by the likes of Peter Ford, Hans-Werner Braun, Yakov Rekhter and
others working with them.  The purpose of CIDR is to allow the C-class
address space to be utilised effectively without blowing out the core
routing tables.

CIDR works by using variable length netmasks and advertising a contiguous
block of C-class nets under the one prefix (i.e. one route entry).  If an
organisation has more than 254 allocated addresses, they can either (a) try
to get a B-class from the nic (good luck), (b) grab several C-class
addresses and advertise a route to each C-class to the rest of the net (not
friendly on the core routers), or (c) use a group of contiguous C-class nets
and advertise them under a single route using a CIDR address block.

For more info, there's a good paper in the proceedings of INET'93 called
"CIDR and the Evolution of the Internet Protocol" by Braun, Ford and
Rekhter.


   __                                  David J. Hughes  -  [email protected]
  /  \               /  /   /           http://Bond.edu.au/People/bambi.html
 /___/ __   _   ____/  /   / _     
/   \ /  \ / \ /   /  /   / / \  /   Senior Network Programmer, Bond University
\___/ \__//  / \__/   \__/ /  / /    Qld.   4229   AUSTRALIA    (+61 75 951450)

-----------[000496][next][prev][last][first]----------------------------------------------------
Date:      20 Sep 1994 22:44:45 +1000
From:      [email protected] (David J. Hughes)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?

[email protected] (Gregory Bond) writes:

>Well, you could use the TCP wrappers to listen on port 80 and then
>decide what program to run (or what homepage to use, or whatever)
>depending on that the local address was....

I've come in late on this as it's the question I came here to pose (not
a frequent reader of this group).  So, if it's been mentioned then
forgive me.

You could just add a line or two to the server that called getsockname()
after accept()ing the connection and have a look at the address. 
Mapping the address onto a new homepage is a trivial problem.  This way
you only get one copy of the server hanging around.

The only question remaining (and this was the question I was going to
ask) is how to assign multiple addr's to the same interface.  The 
SIOCSIFADDR ioctl only appears to allow a single address per protocol family
to be specified.  The fact that the interface data structure provides a
list of addresses would appear to handle this although if you can't get
in there to set things up it doesn't help much.  In particular I'm
interested in SunOS and Solaris.  Any pointers?



   __                                  David J. Hughes  -  [email protected]
  /  \               /  /   /           http://Bond.edu.au/People/bambi.html
 /___/ __   _   ____/  /   / _     
/   \ /  \ / \ /   /  /   / / \  /   Senior Network Programmer, Bond University
\___/ \__//  / \__/   \__/ /  / /    Qld.   4229   AUSTRALIA    (+61 75 951450)

-----------[000497][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 09:08:20 +1000
From:      [email protected] (David J. Hughes)
To:        comp.protocols.tcp-ip
Subject:   Re: What are "sider" IP numbers?
[email protected] (Ricard Wolf) writes:
 
>There is a group of class A B and C networks which are designed to
>be used internally in networks that will never be directly connected
>to the Internet (however, connection via a router is possible), which
>are defined in RFC 1597. Could this be what the vendor meant? 
>Don't know about the term "sider" though; never heard it before.
 
CIDR (Classless Inter-Domain Routing) is a mechanism that was developed to
extend the life of the IP address range.   From what I've read it was
developed by the likes of Peter Ford, Hans-Werner Braun, Yakov Rekhter and
others working with them.  The purpose of CIDR is to allow the C-class
address space to be utilised effectively without blowing out the core
routing tables.
 
CIDR works by using variable length netmasks and advertising a contiguous
block of C-class nets under the one prefix (i.e. one route entry).  If an
organisation has more than 254 allocated addresses, they can either (a) try
to get a B-class from the nic (good luck), (b) grab several C-class
addresses and advertise a route to each C-class to the rest of the net (not
friendly on the core routers), or (c) use a group of contiguous C-class nets
and advertise them under a single route using a CIDR address block.
 
For more info, there's a good paper in the proceedings of INET'93 called
"CIDR and the Evolution of the Internet Protocol" by Braun, Ford and
Rekhter.
 
 
   __                                  David J. Hughes  -  [email protected]
  /  \               /  /   /           http://Bond.edu.au/People/bambi.html
 /___/ __   _   ____/  /   / _     
/   \ /  \ / \ /   /  /   / / \  /   Senior Network Programmer, Bond University
\___/ \__//  / \__/   \__/ /  / /    Qld.   4229   AUSTRALIA    (+61 75 951450)
 


-----------[000498][next][prev][last][first]----------------------------------------------------
Date:      21 Sep 1994 09:09:08 +1000
From:      [email protected] (David J. Hughes)
To:        comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains
Subject:   Re: Can I have multiple distinguishable addresses for a single host?

[email protected] (Gregory Bond) writes:
 
>Well, you could use the TCP wrappers to listen on port 80 and then
>decide what program to run (or what homepage to use, or whatever)
>depending on that the local address was....
 
I've come in late on this as it's the question I came here to pose (not
a frequent reader of this group).  So, if it's been mentioned then
forgive me.
 
You could just add a line or two to the server that called getsockname()
after accept()ing the connection and have a look at the address. 
Mapping the address onto a new homepage is a trivial problem.  This way
you only get one copy of the server hanging around.
 
The only question remaining (and this was the question I was going to
ask) is how to assign multiple addr's to the same interface.  The 
SIOCSIFADDR ioctl only appears to allow a single address per protocol family
to be specified.  The fact that the interface data structure provides a
list of addresses would appear to handle this although if you can't get
in there to set things up it doesn't help much.  In particular I'm
interested in SunOS and Solaris.  Any pointers?
 
 
 
   __                                  David J. Hughes  -  [email protected]
  /  \               /  /   /           http://Bond.edu.au/People/bambi.html
 /___/ __   _   ____/  /   / _     
/   \ /  \ / \ /   /  /   / / \  /   Senior Network Programmer, Bond University
\___/ \__//  / \__/   \__/ /  / /    Qld.   4229   AUSTRALIA    (+61 75 951450)
 


-----------[000499][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1994 17:35:30 -0700
From:      [email protected] (Sean T. Lamont)
To:        comp.protocols.tcp-ip
Subject:   Frame Relay Info wanted
Looking for some frame relay information:

#1: Looking for information about FRAD-devices: vendors, specifics on
    usage, etc.

#2: Wondered if anybody has hooked up a cheap PC with specialized
    hardware to act as a TCP/IP router for frame relay, or even the
    general case of a cheap PC acating as a TCP/IP router with a
    standard serial connector (v.35, etc.)

(As you might have been able to guess, I'm trying to get high-speed
TCP/IP connectivity without shelling out wads of cash.)

Thanks.
-- 
Sean T. Lamont, Abstract Software       | Ask me about the WSI-Fonts
NEXTSTEP development, TCP/IP consulting | Professional collections for NEXTSTEP
[email protected]                 | http://www.abstractsoft.com

-----------[000500][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1994 13:49:57 -0400
From:      [email protected] (Jay Ashworth)
To:        comp.protocols.tcp-ip,biz.sco.general,comp.protocols.tcp-ip.domains
Subject:   Re: DNS Primary Setup Questions
[email protected] (Curtis Justus) writes:
>Op Sys, etc.     Mach.  Name        Address
>------------     -----------        -------
>SCO Box 1:        savo51            51.0.0.251     Primary DNS
>SCO Box 2:        prosserv          51.0.0.252     Secondary DNS

51.0.0.0?  Who's net is that?  

I think it's time for the "get your own assigned net number" speech,
here...

Cheers,
-- jra
-- 
Jay R. Ashworth                                                       Ashworth
Designer                                                          & Associates
ka1fjx/4              High Technology Systems Consulting
[email protected]                                                +1 813 790 7592

-----------[000501][next][prev][last][first]----------------------------------------------------
Date:      22 Sep 1994 15:47:49 -0500
From:      [email protected] (Joshua Harding)
To:        comp.protocols.tcp-ip,comp.protocols.ppp
Subject:   Configuri