yet more on distance vectors

Martha Steenstrup <msteenst@bbn.com> Thu, 21 May 1992 11:05 UTC

Received: from nri.nri.reston.va.us by ietf.NRI.Reston.VA.US id aa00486; 21 May 92 7:05 EDT
Received: from nri.reston.va.us by NRI.Reston.VA.US id aa10863; 21 May 92 7:11 EDT
Received: from PIZZA.BBN.COM by NRI.Reston.VA.US id aa10859; 21 May 92 7:11 EDT
Received: from pizza by PIZZA.BBN.COM id aa02493; 20 May 92 13:45 EDT
To: tli@cisco.com
cc: idpr-wg@bbn.com
Subject: yet more on distance vectors
Date: Wed, 20 May 92 14:43:19 -0400
From: Martha Steenstrup <msteenst@bbn.com>
Message-ID: <9205210711.aa10859@NRI.Reston.VA.US>

Hi Tony,

I want to make sure that I understand you.  You say that BGP handles
this case as you described in the previous message.  Your previous
message says that:

"When a route is advertised, it is implicitly a replacement of the
previous known route.  Thus, when Y advertises X ..  Y .. X .. Z back
to X, X is obligated to remove X .. Y .. Z."

My question is why?  I thought that BGP would reject X .. Y .. X .. Z
because it contained a known loop (i.e. X appears twice in the route).
But why would BGP also remove the previously acceptable route X .. Y
.. Z?  I don't understand.  I admit that I haven't looked at the BGP
specification in a long time, but I don't remember anything like this
in the specification.  Please point me to the reference.

Thanks,
m