BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)
draft-ietf-idr-rfc2796bis-02

This comment replies to the document only citing RFC2385 in the Security
Considerations.  That section should now also cite the BGPbis spec, because
that has expanded the security considerations for BGP - although there's still
a lot of emphasis on 2385, it is able as well to point to draft-ietf-bgp-vuln.
Discussing the risks well is a big step forward.  If noone in the IESG objects to
the DS of RFC2796 making a textual departure (I don't), then RFC2796 should
directly cite draft-ietf-bgp-vuln.  Sometimes there's a formalistic requirement
to change very few words from PS to DS.

First, I share Sam's DISCUSS concern.
No need to hold 2 discusses on that.

Citation/Reference [7] (i.e. RFC2119) must be a
Normative Reference.

$ idnits draft-ietf-idr-rfc2796bis-01.txt
idnits 1.77 (21 Aug 2005)

draft-ietf-idr-rfc2796bis-01.txt:


  Checking nits according to http://www.ietf.org/ID-Checklist.html:
  * The document seems to lack an IANA Considerations section.
  * Looks like you're using RFC 2026 boilerplate. Better change to RFC
    3978/3979.

From Gen-ART review by Lakshimnath Dondeti:

...
2. ROUTER_ID is now referred to as BGP Identifier.  Both terms have been around for a long while now.  Perhaps the authors should explain what they have in mind in changing that term.
...
4. Editorial Nit:   Replace  "With the existing BGP model," in Page 3 with something like "In BGP-4"

  A Table of Contents would have helped me review this document.

I originally entered the following discuss.  There is question within
the IESG about whether this is actually the requirement or not.  I
don't want to block this document indefinitely while we answer general
questions.

All the implementations that sent in implementation report forms
claimed to test against Cisco and Juniper.  However as far as I can
see reading the document Cisco and Juniper did not actually fill out
the form.

I think you need to have two implementations who both participate in
the implementation report interoperate.


I'm willing to be convinced that this is not a requirement, but it
does significantly concern me that none of the implementations in the
implementation report tested against each other.