Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!usc!ucsd!ucbvax!ulysses!ulysses.att.com!smb
From: smb@ulysses.att.com (Steven Bellovin)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Wither Ethernet?
Message-ID: <13019@ulysses.att.com>
Date: 25 May 90 20:53:09 GMT
References: <9005211233.AA15946@gateway.mitre.org>
Sender: netnews@ulysses.att.com
Lines: 35

In article <9005211233.AA15946@gateway.mitre.org>, hal@GATEWAY.MITRE.ORG (Hal Feinstein) writes:
> It has been suggested by some standards gurus that in the future
> all LANs will be "standardized" to ISO/IEEE802 and that ethernet is a dead
> (or possibly dying)  protocol.

I've never been a worshipper of ``standards'' that try to impose rather
than describe.  802.3 would seem to me to meet that criterion, I'm
afraid.

At least in the TCP/IP world, I don't see that much movement to
802.3 (and away from Ethernet) simply because there's no compelling
reason to.  The current technology is here, it works, and it's
compatible with (almost) everyone else.  I'm sure folks will correct
me if I'm wrong, but I thought that HP was the only major vendor
to try to use 802.3 encapsulation for IP over Ethernet, and that
a few others (i.e., Cisco) would accomdate them, mostly for
compatibility.  Furthermore, as a result, it was hard to talk
to some HP machines.

The 802.3 format -- and in particular the 802.2 goo you're supposed to
use -- makes sense if you're trying to do routing at that layer.
In ``our'' world, there's a better, more flexible way to do routing,
and that's at the IP layer.  So why switch?

The situation is different for other media.  Token ring is inhabited
by bluish PCs, and FDDI was decreed from the outset to use 802.2.
For those, a simpler (or at least different) encapsulation has never
had a chance to take hold.  Thus, 802.2 will rule, even for IP.

The higher levels? I'm *not* going to start another TCP/IP vs OSI flamefest
here today...

		--Steve Bellovin

P.S.  I don't have a very good track record as a prophet; beware...