Routing and MTUs

Philippe Prindeville <philipp@gipsi.gipsi.fr> Wed, 06 December 1989 16:40 UTC

Received: from decwrl.dec.com by acetes.pa.dec.com (5.54.5/4.7.34) id AA02965; Wed, 6 Dec 89 08:40:09 PST
Received: by decwrl.dec.com; id AA22064; Wed, 6 Dec 89 08:40:00 -0800
Received: by mcsun.EU.net with SMTP; Wed, 6 Dec 89 17:39:46 +0100 (MET)
Received: from gipsi.gipsi.fr by inria.inria.fr (5.61+/89.0.8) via Fnet-EUnet id AA21569; Wed, 6 Dec 89 14:33:54 +0100 (MET)
Received: by gipsi.gipsi.fr; Wed, 6 Dec 89 14:34:18 -0100 (MET)
Date: Wed, 6 Dec 89 14:34:18 -0100
From: Philippe Prindeville <philipp@gipsi.gipsi.fr>
Message-Id: <8912061334.AA00404@gipsi.gipsi.fr>
Phone: +33 1 30 60 75 25 / +33 1 47 34 42 74
To: MTU Discovery <mtudwg>
Subject: Routing and MTUs

I know this is not a new idea, and no doubt there will be some groans
at the mention of it again, but isn't MTU one of the type-of-service
parameters that should be conveyed with routing information?  I know
that someone mentioned (sorry, mailing lists never stay in my already
voluminous mailbox for long) that it isn't reasonable to maintain a
subnet MTU for each subnet at MIT in the Stanford routing tables.

Everyone will (I hope) agree with this.  But as most of the last
few scenarios have pointed out, it is usually the transit links
between complete autonomous systems/domains that have the widely
differing MTUs.  And most FDDI sites will be using FDDI to hang
ethernet (etc) off, not as a direct connection.  Therefore, in
most cases, the limiting factor of an AS MTU will be a near-universal
average (ie. ethernet is everywhere) or a border-gateways MTU.

So maybe this idea is worth revisiting.

-Philip