Subject: Re: illegal network routes and a ponderance
From: David Laight <firstname.lastname@example.org>
Date: 02/20/2003 08:15:35
> I'm running 18.104.22.168/28 and 10.0.2.0/24 on the same "cable"
> (actually, 10baseT hub) right now. It works fine - because each
> machine involved has two addresses on the relevant interface, one on
> each network.
Doing that caused us much grief a few years ago. Consider the following:
A LAN segment with 3 subnets on it, say x.x.1.0 x.x.2.0 and x.x.3.0,
a 'private' subnet x.x.4.0 routed to by a 486 box with two ethernet
cards. The router was given addresses in all the main subnets (say)
x.x.1.4 x.x.2.x and x.x.3.4 in order to allow PC systems to route
via it (without the packets going to the LANs default gateway and back).
Someone then ran routed to advertise these routes, the system advertised
that it was a router between the 4 subnets on each subnet. This
caused other systems to start routing traffic via this system instead
of using their configured 'metric 0' route. We suddenly wondered
why packets were being dropped.
Co-located subnets are not expected by the IP spec, One way to deal
with them is not to worry that all the packets appear on the LAN twice.
The other is to allow a route to be defined that says that the
specified subnet is local and that addresses in it sould be arped for.
However is still doesn't make any sense to point a default route
at an ethernet interface, and the system doing the routing from a subnet
should have an address in that subnet.
David Laight: email@example.com