Re: which protocols does MTUD help?

William Westfield <BILLW@MATHOM.CISCO.COM> Fri, 01 December 1989 13:42 UTC

Received: from decwrl.dec.com by acetes.pa.dec.com (5.54.5/4.7.34) id AA07606; Fri, 1 Dec 89 05:42:17 PST
Received: by decwrl.dec.com; id AA19964; Fri, 1 Dec 89 05:42:11 -0800
Date: Fri 1 Dec 89 05:41:57-PST
From: William Westfield <BILLW@MATHOM.CISCO.COM>
Subject: Re: which protocols does MTUD help?
To: mogul
Cc: craig@NNSC.NSF.NET, mtudwg
In-Reply-To: <8912010049.AA02017@acetes.pa.dec.com>
Message-Id: <12546623614.8.BILLW@MATHOM.CISCO.COM>

In a bridged environment between FDDI and ETHERNet, the systems will
normally assume that the MTU of the connected interface.  The FDDI
system would try to send 4k datagrams (or whatever), and these would
break at the bridge (it not knowing how to fragment).  MTU discovery
might be able to let both systems know more about the world.  (Mind you,
Im not sure how, since the bridge is supposed to be invisible.)

I think the usual case of this is handled by the way the MSS option
works NOW (most systems advertise their interface MTU, and refuse to
send a datagram larger than that even if the other host claims that it
is willing to accept it).  Something new might be needed to handle the
case where the FDDI-ETHER bridge is on some network at an intermediate
point in the path.

Bill W
-------