North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

[Idr] Configuration objects in BGP MIB v2: Call for consenus

  • From: Jeffrey Haas
  • Date: Sun Jul 06 21:57:25 2008

Please pardon this intrusion in the usual operational chatter.

I have been working on the successor to the BGP MIB within IETF over the
last several years.  As part of a review of the current draft of this
MIB (http://tools.ietf.org/html/draft-ietf-idr-bgp4-mibv2-07) I have
been requested to gather operational consensus regarding being able to
configure your BGP peering sessions from within the MIB.  Please see the
forward below for the details.

The BGP MIBv2 is attempting to address some operational holes with
regards to the current BGP MIB.  These issues include IPv6 support and
also better counters for your peering session.  In addition to feedback
on the configuration objects issue I'd appreciate general feedback on
the MIB on or off the list.

My goal is to resolve all remaining open issues with the MIB within the
next few months and get it into working group last call.  A reference
implementation in Quagga will likely follow.

-- Jeff

----- Forwarded message from Jeffrey Haas <[email protected]> -----

Date: Sun, 6 Jul 2008 21:50:00 -0400
From: Jeffrey Haas <[email protected]>
To: [email protected]
Subject: [Idr] Configuration objects in BGP MIB v2: Call for consenus

Working Group,

Back around 2005 I had a number of discussions with people who had
provided input for the BGP MIBv2 work.  These conversations were
specifically regarding the configuration objects in the MIB.
draft-ietf-idr-bgp4-mibv2-05 was the last version of the MIB that
contained the proposed configuration objects.

The results of those discussions were effectively that the configuration
mechanisms in that MIB were too complex and had some potential issues.
In particular:

- Modern BGP implementations tend to be more complex than the feature
  set covered by the proposed MIBs.  It was not possible to
  configure all session specific features from the MIB.  Since the base
  MIB is not intended to cover all possible current and future features
  this is problematic.
- Configuration of peering sessions are not sufficient to fully
  implement BGP in an operational network.  BGP fundamentally requires
  policy for the population of the Ribs.  Policy elements and algebra
  vary considerably among vendors.  

  Providing a general policy engine for BGP in the MIB is likely out of
  scope of this work.

Presume that the structural issues from draft-05 may be addressed.
Should they be addressable, do we wish to pursue including configuration
objects in the BGP MIB?

Given the operational impact of this issue, I would appreciate it if
this call for consensus was distributed within and without IETF.  I will
be forwarding this to NANOG as a first step.

-- Jeff