Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!UDEL.EDU!Mills
From: Mills@UDEL.EDU
Newsgroups: comp.protocols.tcp-ip
Subject: Re:  routing - very strange stuff
Message-ID: <8909030944.aa00525@huey.udel.edu>
Date: 3 Sep 89 13:44:49 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 20

Alan,

You are seeing much the same thing as described in my Internet monthly
report for July. While I have no hard evidence and have lost direct
ARPANET connectivity formerly used to diagnose such zoo events as
EGP route churn, the characteristics of the problem appear very much
like multiple EGP routes with the same "metric," causing route changes
on every EGP update. We all know that EGP is being used as a routing
algorithm in many places in the Internet, although such was not the intent
in the design. However, considering that pragmatic fact, there are a number
of things that could be done to EGP in order to improve its "routing"
performance, some already done to at least some implementations of EGP.
I have in mind split-horizon, hold-down, hysterisis and other tricks of
the trade. The question in most minds, it seems, is whether we should go
to the trouble of doing these gruesome things as against waiting for
something better to come along. Unfortunately, this question has been in
our communal minds since 1983 and only recently has led to what appears to
be real progress in various working groups of the IETF.

Dave