BGP MIBv2 Status; IETF 72

1
BGP MIBv2 Status IETF 72

• Jeffrey Haas jhaas_at_arbor.net
• Presented byDanny McPherson

2
MIB Doctor Review

• Many thanks to Joan Cucchiara for the early MIB
  Doctor review.
• MIB Doctor review has shaken out a number of
  minor issues.
• MIB is largely structurally stable.
• Were down to structural nits. These should be
  resolved in -08 once new I-D versions can be
  posted.

3
Open Issues - Configuration

• Do we really really absolutely positively for
  double sure not want to have configuration
  objects?
• Polled IDR and NANOG and all responses thus far
  have been against having configuration objects.
• Known issues that discouraged configuration
  objects were
• Base MIB couldnt sufficiently cover modern
  implementations without a number of extensions.
• Even with extensions, interactions with vendor
  configuration knobs a likely problem.
• BGP is all about policy and policy is highly
  vendor-specific. Standardizing abstract policy
  is outside the scope of this MIB.
• Current conclusion is we dont want configuration
  objects in MIB but extension mechanism doesnt
  preclude it in the future.

4
Open Issues - Abstract Indices

• Tables with many to one relationships need an
  abstract index to glue them together. E.g. NLRI
  has many prefixes sharing a common set of path
  attributes.
• While BGP implementers understand this
  relationship and can easily reflect this in the
  MIB, the MIB must be clear for users
• Better documentation as to how an abstract index
  relates one table to another.
• Better documentation and guidance for handling
  discontinuities - e.g. try not to re-use the same
  index too soon after BGP reboots.

5
Open Issues - Discontinuities

• The major discontinuity issue is in the previous
  slide - how to make sure people know what to do
  with the abstract indexes.
• Need to clarify a few remaining scenarios for
  discontinuities
• This is awkward because this edges the line
  between specification and implementation.
• An example is what happens when your BGP instance
  crashes. If your implementation keeps BGP state
  around for Graceful Restart, you may not need to
  reset your indices.

6
Open Issues - Structural Nits

• MIB Doctor has suggested we rename bgpExtensions
  to something more generic. Looking for a good
  name -- bgpIzel, perhaps?
• Also recommended that textual conventions be
  forked off into a related TC MIB
• Need IDR blessing to issue the related
  draft-ietf-idr-bgp4-mibv2-tc document.

7
Rough consensus, running code

• After -08 is issued, all open issues should be
  addressed.
• We should be ready for WG last call.
• Yakov has previously cited WG tradition of two
  interoperable implementations before we last call
  the drafts.
• Author (Jeff) intends to do a reference
  implementation prior to IETF 73 based on -08.
  Other vendors are encouraged to also implement.
• We should review status at IETF 73.

8
Questions?

• Email Jeff or idr_at_