[12294] in North American Network Operators' Group
Re: not rewriting next-hop, pointing default, ...
daemon@ATHENA.MIT.EDU (Sean M. Doran)
Thu Sep 11 21:09:58 1997
To: rja@corp.home.net (Ran Atkinson)
Cc: nanog@merit.edu
From: "Sean M. Doran" <smd@clock.org>
Date: 11 Sep 1997 21:06:00 -0400
In-Reply-To: rja@corp.home.net's message of "Thu, 11 Sep 1997 16:04:48 -0700"
rja@corp.home.net (Ran Atkinson) writes:
> Letting LSR in two hops lets outsiders trying to debug (e.g. some
> routing problem) at least perform first-level fault-isolation
> so they know whose NOC to call for further debugging assistance.
> So I view the 2-hop notion as an attempt at "reasonable
> compromise".
Cool, and I now view the 2-hop notion as the first
reasonable argument for encouraging people to totally
flatten their network into a full mesh.
Security policy should not under any circumstances prevent
the Internet as a whole from functioning reasonably well,
scaling decently, or make discovering and diagnosing
problems any harder than it already is.
Your opinion may vary with mine, but I am solidly in line
with Randy's suggestion that enabling LSRR on backbone
routers should be a requirement for peering. (This is not
surprising as I used to require it of a couple of peers in
a previous life, because in practice it is unfortunately
an irreplaceable diagnostic tool).
Sean.
