Re: bgp4-17 5.1.3 NEXT_HOP attribute determination

Alex Zinin <azinin@nexsi.com> Fri, 11 January 2002 16:30 UTC

Received: from trapdoor.merit.edu (postfix@trapdoor.merit.edu [198.108.1.26]) by nic.merit.edu (8.9.3/8.9.1) with ESMTP id LAA12015 for <idr-archive@nic.merit.edu>; Fri, 11 Jan 2002 11:30:27 -0500 (EST)
Received: by trapdoor.merit.edu (Postfix) id 659AD9128F; Fri, 11 Jan 2002 11:29:59 -0500 (EST)
Delivered-To: idr-outgoing@trapdoor.merit.edu
Received: by trapdoor.merit.edu (Postfix, from userid 56) id 313AF91290; Fri, 11 Jan 2002 11:29:59 -0500 (EST)
Delivered-To: idr@trapdoor.merit.edu
Received: from segue.merit.edu (segue.merit.edu [198.108.1.41]) by trapdoor.merit.edu (Postfix) with ESMTP id 398619128F for <idr@trapdoor.merit.edu>; Fri, 11 Jan 2002 11:29:58 -0500 (EST)
Received: by segue.merit.edu (Postfix) id 118D35DF0D; Fri, 11 Jan 2002 11:29:58 -0500 (EST)
Delivered-To: idr@merit.edu
Received: from relay1.nexsi.com (relay1.nexsi.com [66.35.205.133]) by segue.merit.edu (Postfix) with ESMTP id D53315DF09 for <idr@merit.edu>; Fri, 11 Jan 2002 11:29:57 -0500 (EST)
Received: from mail.nexsi.com (ripper1.nexsi.com [66.35.212.35]) by relay1.nexsi.com (Postfix) with ESMTP id 10CF23F65; Fri, 11 Jan 2002 08:32:41 -0800 (PST)
Received: from khonsu.sw.nexsi.com (cscovpn4.nexsi.com [172.16.213.4]) by mail.nexsi.com (8.9.3/8.9.3) with ESMTP id IAA24475; Fri, 11 Jan 2002 08:33:44 -0800
Date: Fri, 11 Jan 2002 08:29:04 -0800
From: Alex Zinin <azinin@nexsi.com>
X-Mailer: The Bat! (v1.51) Personal
Reply-To: Alex Zinin <azinin@nexsi.com>
Organization: Nexsi Systems
X-Priority: 3 (Normal)
Message-ID: <140685068856.20020111082904@nexsi.com>
To: Tom Petch <nwnetworks@dial.pipex.com>
Cc: idr@merit.edu
Subject: Re: bgp4-17 5.1.3 NEXT_HOP attribute determination
In-Reply-To: <001e01c2b978$fa384140$c490bc3e@tom3>
References: <001e01c2b978$fa384140$c490bc3e@tom3>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Sender: owner-idr@merit.edu
Precedence: bulk

Tom,

> I think that there is a problem with logic and terminology
> in the determination of NEXT_HOP.  There are four cases
> listed,
> - a form of "third party" NEXT_HOP attribute
> - a second form of "third party" NEXT_HOP attribute
> - "first party" NEXT_HOP attribute
> - by default
> and respectively the BGP speaker
> - can use
> - can use
> - may use
> - should use
> which seems inconsistent; I would prefer SHOULD.

Since use of third party next hops is an optimization
by nature and does not *have* to be implemented,
and given that "should" is used for the default,
I think "may/can" is ok. If we change the optional
cases to "should" to increase the level on insistence,
we should do the same to the default case and make
it a must.
If you ask me, I think we're all right now.


> But logically, the first party (common subnet) overlaps the
> two third parties (common subnet with IBGP or locally
> originated and common subnet with EBGP) which obviates
> SHOULD.

We know how to deal with overlapping cases already---
the more specific one wins. Same here---third party
cases are more specific in their description.
I guess it might be clearer if we changed "-if -if"
to "-if -otherwise if".

> And what does locally originated actually mean?

It means that the route was not received from another
BGP peer and was created by the guy we're talking about.

> Presumably IGP and static except that the subsequent use of
> internal router does not quite match that.

> I wonder if it would be clearer if the focus was on the
> common subnet rather than where the route came from.  Can eg
> the third party cases be combined as

> 'if the external peer X shares a common subnet with the
> immediate next hop that the BGP speaker itself has installed
> in its routing table for this route, then the speaker SHOULD
> use this immediate next hop in the NEXT_HOP attribute'

When I changed this text to what it is now, I tried to preserve
the spirit of the old one, and keep some "educational" info.
I think both ways are fine and functionally equivalent, it is
more a question of preference.
So, I guess my vote is to stet it :)

Alex.