[36789] in North American Network Operators' Group
The ultimate routing loop?
daemon@ATHENA.MIT.EDU (Sean Donelan)
Fri Apr 20 02:35:25 2001
Date: 19 Apr 2001 23:28:26 -0700
Message-ID: <20010420062826.18663.cpmta@c004.sfo.cp.net>
Content-Type: text/plain
Content-Disposition: inline
Mime-Version: 1.0
To: nanog@merit.edu
From: Sean Donelan <sean@donelan.com>
Errors-To: owner-nanog-outgoing@merit.edu
I'm in washington this week, and as usual, some of the strangest
ideas come up.
Suppose I wanted the best chance for my packet to get through, no
matter what tier 1, 2, 3 network was still operating between points
A and B. I designate a "special" IP address block, and arrange for
the top 10 providers to accept announcements for the block both
directly, and transit through any of the other 10 providers. This
would allow transit via a third-party network to restore connectivity
across a partioned AS. For this network block, other than BGP loop
detection, it would bypass the normal peer/transit/peer filters.
If a AS was partioned, you might see a route cross two or three
primary backbones, and even the same backbone twice. But if the
packet went through, its worth it.
The questions is: Are more available paths really better? Or does
it just create more instability?
