Path: utzoo!utgpu!watserv1!watmath!att!ucbvax!TERMINUS.UMD.EDU!dzoey
From: dzoey@TERMINUS.UMD.EDU
Newsgroups: comp.protocols.tcp-ip.ibmpc
Subject: Re: Token Ring promiscuous software.
Message-ID: <9105241754.AA21363@terminus.umd.edu>
Date: 24 May 91 17:54:43 GMT
References: <9105241202.aa05099@louie.udel.edu>
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The Internet
Lines: 31


> From: Charlie Turner <CHARLIE%UMVMA.bitnet@umrvmb.umr.edu>

> On the other hand, I don't understand how a non promiscuous T/R
> adapter could work in a bridge but they do. So are they *really*
> non promiscuous?

The token ring bridge isn't really a bridge, but a brouter (yes, folks
it's more confusing terminology!  Question: What's a gateway?  Answer:
about 10 pounds.  But I digress.) The TR bridge doesn't forward all
packets between the two rings.

When a machine on one ring (HOSTA) wants to communicate with a machine
on another ring (HOSTJ), it sends out (for the sake of the IP world)
an ARP which is transmitted as a (possibly all-rings) broadcast.  When
the bridge receives a broadcast it records its (the bridge's) address
in the packet's source routing field and broadcasts it on the other
ring.  When HOSTJ receives the broadcast packet, it stores HOSTA's
address and the source routing information that was collected on the
way.  When HOSTJ replies to HOSTA, it puts the source routing
information it collected about HOSTA into the packet.  Each bridge the
packet passes reads the source route and gets the address for the next
hop.

Source routing is not part of the 802.5 standard (or it wasn't last
time I was involved in this stuff. It's been a while since I looked).
It's an IBM implementation standard.  The IBM token ring tech ref.
does a reasonable job of explaining source routing more detail.

			Joe Herman
			U. of Maryland