From van@lbl-csam.arpa  Wed Feb  3 09:28:03 1988
Posted-Date: Wed, 03 Feb 88 09:26:35 PST
Received-Date: Wed, 3 Feb 88 09:28:03 PST
Received: from LBL-CSAM.ARPA by venera.isi.edu (5.54/5.51)
	id AA11858; Wed, 3 Feb 88 09:28:03 PST
Received: by lbl-csam.arpa (5.58/1.18)
	id AA10955; Wed, 3 Feb 88 09:26:36 PST
Message-Id: <8802031726.AA10955@lbl-csam.arpa>
To: Jon Crowcroft <jon@cs.ucl.ac.uk>
Cc: end2end-interest@venera.isi.edu
Subject: Re: TCP Problem 
In-Reply-To: Your message of Wed, 03 Feb 88 15:23:42 GMT.
Date: Wed, 03 Feb 88 09:26:35 PST
From: Van Jacobson <van@lbl-csam.arpa>
Status: R

> A strange and interesting problem while doing a xTCP test UCL to
> NTA.
> 
> I had a telnet cx from ego to skage, then to Jotun (skage/jotun
> at nta), and was running a quick test, ttcp ego to jotun, which
> broke, with an ICMP net unreachable message at the jotun end (i
> guess from NTA's butterfly). My telnet cx stayed up.

Jon,

The xtcp telnet connection staying up isn't a bug, it's a
feature.  The icmp unreachable gets forwarded up to tcp but, at
least in the US, net unreachable is a transient condition so tcp
just calls "in_losing" (saying to ip `this path isn't working too
well, try a different route if there is one').  But the connection
won't go away until a) there is outbound data to send and b) the
retransmit timer maxes out trying to send it (which takes 12
retries & 15-20 minutes, usually). 

So, if the telnet connection was idle, I could see it staying
up while the active ttcp connection broke.  I could also see
the ttcp connection breaking quickly if one end was running
an older (4.2bsd based) tcp/ip.  Those treated the unreachable
as a fatal error & reset the local connection (then the other,
xtcp, end would get sent a reset as soon as the next packet
arrived).

 - Van