Single subnet

Dave Katz <katz@merit.edu> Fri, 18 May 1990 15:34 UTC

Received: from merit.edu by NRI.NRI.Reston.VA.US id aa09213; 18 May 90 11:34 EDT
Received: Fri, 18 May 90 10:33:12 EST by merit.edu (5.59/1.6)
Date: Fri, 18 May 1990 10:33:12 -0500
From: Dave Katz <katz@merit.edu>
Message-Id: <9005181533.AA04837@merit.edu>
To: fddi@merit.edu
Subject: Single subnet
Status: O

I'm working on an article for SIGCOMM CCR on running IP and CLNP over FDDI
and came upon an interesting question:  Is there any reasonable way of
having single-MAC stations on both rings when using the single-subnet
approach?  The problem comes when a station on each ring needs to
talk to a station on the other ring (and both are single-MAC).  The
initiator has no knowledge of where the destination is, but knows it's
on the same subnet, and sends an ARP.  The real destination cannot
respond (since it won't hear the ARP), so the only logical choice is
a dual-MAC router.  However, that router will not have a priori knowledge
of where the destination is either, so it cannot blindly answer the ARP.
I suppose that the router could ARP out both rings for the destination
and if it finds it on the other ring could then send an ARP reply (but
the timing on this could be tricky).  Ambiguities due to the network
wrapping could presumably be eliminated by using a "different" ARP on
each ring (and each single-MAC station would ignore the other's ARP
in this case).
 
Comments?
 
By the way, CLNP does not suffer from this problem, since routers know
a priori where all the hosts are.