Re: revised draft of BGP-4

"Michael F. Nittmann" <nittmann@wis.com> Thu, 15 June 1995 17:29 UTC

Received: from ietf.nri.reston.va.us by IETF.CNRI.Reston.VA.US id aa06698; 15 Jun 95 13:29 EDT
Received: from CNRI.Reston.VA.US by IETF.CNRI.Reston.VA.US id aa06694; 15 Jun 95 13:29 EDT
Received: from interlock.ans.net by CNRI.Reston.VA.US id aa11696; 15 Jun 95 13:29 EDT
Received: by interlock.ans.net id AA69867 (InterLock SMTP Gateway 3.0 for iwg-out@ans.net); Thu, 15 Jun 1995 13:21:42 -0400
Message-Id: <199506151721.AA69867@interlock.ans.net>
Received: by interlock.ans.net (Protected-side Proxy Mail Agent-2); Thu, 15 Jun 1995 13:21:42 -0400
Received: by interlock.ans.net (Protected-side Proxy Mail Agent-1); Thu, 15 Jun 1995 13:21:42 -0400
Date: Thu, 15 Jun 1995 12:20:15 -0500 (CDT)
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: "Michael F. Nittmann" <nittmann@wis.com>
To: Curtis Villamizar <curtis@ans.net>
Cc: Yakov Rekhter <yakov@cisco.com>, bgp@ans.net
Subject: Re: revised draft of BGP-4
In-Reply-To: <199506151640.MAA12188@laplink-lt.brookfield.ans.net>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Thu, 15 Jun 1995, Curtis Villamizar wrote:

> 
> In message <199506151425.AA43047@interlock.ans.net>et>, "Michael F. Nittmann" writ
> es:
> > 
> > Following case: assume a multihomed AS splits up due to internal 
> > connectivity loss. Both parts announce routes to systems with transit 
> > agreement. 
> > By accepting routes with the own AS path, the other part of the AS can be 
> > reached via the transit system.
> > this shouls be the default.
> 
> Partition repair is possible, but should not be the default.  Would
> you like to repair an AS690 partition (can't happen, just hypothetic :)
> 
> > The limitation in this paragraph is only important for single homed AS 
> > systems, which are not the majority case.
> > 
> > I would suggest to invert this, that the default behaviour will be to 
> > accept routes with the same AS path, but not to duplicate the same route 
> > at all, e.g.
> > 
> > 	1.0.0.0 is announced by two routers of AS 0 to AS 1 and AS 2 .
> > 	
> > 	AS 0 splits due to internal connectivity loss (split means, the 
> > 	fallback link is too slow to take all traffic)
> > 	
> > 	Router A announces 1.0.0.0, router B (of AS 0) lost the route due to 
> > 	the split and picks it up from AS 2, which has a transit agreement with
> > 	AS 0.
> > 
> > 	Router B can announce to the severed part of AS 0 that 1.0.0.0 is 
> > 	now not reachable best via the internal 'natural' link, but via AS 2.
> 
> Having an agreement to do partition repair may be a different from a
> transit agreement.  Some providers (us) don't announce you back to
> yourself so you can't prefer our backbone in favor of your own or
> intentionally partition your own backbone to overcome congestion

  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

You are right. It should not be the default. That's a scheme I did not 
think of.

We do not announce back either, in fact we don't announce any peers to no 
peer, only our dependents (if everything is configured right ;-) ).



> (unless you default at us, which we also prohibit in our agreement).
> 
             ^^^^^^^^^

A good network should be default free on the backbone anyways. Right?


> Partition repair (accepting your AS in the path exactly once) can be
> configured but should not be the default.  If you have an agreement
> with your provider to do partition repair, then configure it.
> 
> Curtis
> 

--------------------------------------------------------------------------------
Michael F. Nittmann                                             nittmann@wis.com
Network Architect						nittmann@b3.com 
B3 Corporation, Marshfield, WI (CIX Member)               (715) 387 1700 xt. 158
US Cyber (SM), Washington DC				  (715) 573 2448
							  (715) 831 7922
--------------------------------------------------------------------------------