Skip to main content

Minutes for OSPF at IETF-96
minutes-96-ospf-2

Meeting Minutes Open Shortest Path First IGP (ospf) WG
Date and time 2016-07-22 08:00
Title Minutes for OSPF at IETF-96
State Active
Other versions plain text
Last updated 2016-07-26

minutes-96-ospf-2
IETF 96: Open Shortest Path First (OSPF) WG Agenda
Friday, July 22, 2016 (CEST) 10:00 - 12:00 - Morning Session I
Location: Potsdam II
========================================================

Chairs: Acee Lindem
        Abhay Roy
Scribe: Les Ginsberg (ginsberg@cisco.com)

WG Status Web Page: http://tools.ietf.org/wg/ospf/

1) Administrivia
 - Blue sheets
 - Scribe (Les Ginsberg)/jabber
 - Jabber room: ospf@jabber.ietf.org

2) WG Status Update (See slides) - Abhay Roy
  Alia Atlas: TTZ - expert review found significant number of
    issues - more review from WG would be a good idea.
  Acee Lindem: Maybe not lack of review but ambitious goals.
  Alia Atlas: Probably correct
  Abhay Roy: MRT draft has normative reference to "OSPFv3 Extended
    LSAs" which is stuck at the moment.
  Abhay Roy: Operator defined TLVs - service Routing may be an
    alternative (will be presented today)

3) OSPF Segment Routing Update (See slides) - Peter Psenak
 - https://datatracker.ietf.org/doc/draft-ietf-ospf-segment-routing-extensions/
 - https://datatracker.ietf.org/doc/draft-ietf-ospf-ospfv3-segment-routing-extensions/
  Acee Lindem: Hope IS-IS will do the same on algorithm sub-TLV
  Les Ginsberg: Situation in IS-IS is different. Algorithm sub-TLV
    is optional.
  Chris Bowers: Was SR algorithm change discussed with WG?
  Alia Atlas: This is necessary to be done.
  Chris Bowers: Wants to be sure WG reviewed the change
  Alia Atlas: Was that done after WG last call? Concerned about
    transparency about spec development. Particularly concerned
    because early allocation of codepoints have expired. Do we need
    to go to IESG for extension of early allocation of codepoints?
    Will review draft in early August.
  Acee Lindem: Probably need IESG extension
  Acee Lindem: May be overdesigning SR conflict resolution - which
    is cause for some of the OSPF draft changes.
  Alia Atlas: Codepoints have expired
  Peter Psenak: Page says expire 10-22-2016
  Alia Atlas : Will need to have changes reviewed - which means we will
    need IESG extension of codepoints.

4) OSPF YANG Model Update (See slides) - Yingzhen Qu
  - https://datatracker.ietf.org/doc/draft-ietf-ospf-yang/
  Acee Lindem : YANG ip-address is a union of IP/IPv6 address
  Acee Lindem : Model has gotten quite large. Would like to freeze
    the feature set unless oper state is an issue.
    Most significant change is in regards to virtual links. Makes
    sense for virtual links to be associated with backbone area. Cisco
    implementation can map it in this way. Aligns config and oper state
    in this regard.One request for an ospfv2/ospfv3 interface but only
    3 attributes different which can be handled by "when" clause.
  Jeff Haas: There is a SPRING YANG model for OSPF.
  Yingzhen Qu: This is common model for all protocols - is imported by
    OSPF draft.
  Jeff Haas: Thinks there is still some overlap.
  Acee Lindem : Have been trying to import/extend.
  Jeff Haas: Wants to complete alignment across IGPs.
  Alia Atlas: Do we have implementation info?
  Yingzhen Qu: The last comments are primarilty from implementation
    experience.
  Alia: What is the timeline for finishing the draft? Want to
    complete YANG models "soon".
  Yingzhen Qu: Biggest dependency is on oper state. Then will need
    editorial changes.
  Acee Lindem: Big model intimidates reviewers (100+ pages).
    Have had many experienced reviewers (Derek, Stephane, Helen,
     etc.)
  Yingzhen: OSPF YANG model meetings are held every 2 weeks.
  Alia Atlas: Encourage prioritizing completion
  Chris Hopps: What are the oper state blocking issues?
  Acee Lindem: Do we need to prune based on applied state
  Chris Hopps: Thinks this should not be a blocking issue.
  Acee Lindem: Can leave things in applied and would not have to
    wait for implementations of applied state.
  Chris Hopps: Split between config ans state probably won't change
  Yingzhen Qu: Group for interface config used in oper state. Also
    counters and other interface state data.
  Chris Hopps: We don't like the split but don't want to block models.
  Acee Lindem: Is someone from Huawei looking at the OSPF
    model? I know we had participation for the key chain model.

5) OSPF extensions for Computed Multicast applied to
   MPLS based Segment Routing (See slides) - David Allan
  - https://datatracker.ietf.org/doc/draft-allan-ospf-spring-multicast/
  Peter Psenak: Why use extended prefix TLVs? What is relationship
    between prefix and Multicast Groups.
  Dave Allan: Have to go back and check
  Peter Psenak: Think we need a separate object.
  Jeff Zhang: Plan how to specify how edge joins are handled.
    Multi-area/multi-level will be addressed.
  Acee Lindem: Would PIM WG own the solution as to whether
    this progresses?
  Dave Allan: Yes - SPRING WG has pushed in this direction.
  Acee: As they killed M-OSPF this makes sense ;^)
  Chris Hopps: Where did use OUI come from?
  Dave Allan: Allow vendor specific extensions.
  Chris Hopps: Algorithms need to be agreed upon.
  Acee Lindem: Everybody in the domain needs to use the
   same algorithm and come up with the same multicast
   distribution trees (MDTs).
  Chris Hopps: Thinks IANA registry for algorithms is sufficient.
  Acee Lindem: Agrees with Chris Hopps.
  Chris Bowers: Use of OUI comes from TRILL/SPB.
  Dave Allan: Used those two documents as a template.


6) OSPF TE Attributes for non-TE Applications
   Alternatives Update (See slides) - Peter Psenak
 - https://datatracker.ietf.org/doc/draft-ppsenak-ospf-te-link-attr-reuse/
  Robert Raszuk: Any relation between affinity bits and
   application bits?
  Peter: Affinity bits are one of the application specific
   advertisements - application bits are a new thing.
  Jeff Haas: This is NOT link coloring.
  Peter Psenak: Trying to advertise something which is used
   by RSVP-TE.
  Acee Lindem: New draft revision will come soon.

7) OSPF Routing with Cross-Address Family
   MPLS Traffic Engineering Tunnels Revisited (See Slides) - Acee
 - https://datatracker.ietf.org/doc/draft-smirnov-ospf-xaf-te/
  Acee Lindem: Perhaps do AD sponsored
  Alia Atlas: Normally AD sponsored drafts go to RTGWG - not
   appropriate in this case.
  Chris Bowers: Seems odd for AD sponsosored drafts to be used
   in this case.
  Alia Atlas: Does WG believe this is a good idea?
  Les Ginsberg: Why aren't authors more enthusiastic?
  Chris Bowers: Find new authors?

8) Service Distribution using OSPF (See slides) - Padma Pillay-Esnault
 - https://datatracker.ietf.org/doc/draft-pillay-esnault-ospf-service-distribution/
  Acee Lindem: Transport instance was of interest to GMPLS
   optical - but did not get implementations - so did not
   take it forward.
  Padma Pillay-Esnault: Did not have a "killer app".
  Acee Lindem: SDR acronym clashes with "Secure Domain Router" ;^)
  Padma Pillay_Esnault: Name will change in next revision
  Acee Lindem: Reserve a range of identifiers for local use?
  Tony Przygienda: Scope of a service does not align with flooding
   in all cases. S-BFD use case is an example. Discipline required in
   allocating codepoints - is concerned about proliferation.
  Padma Pilay-Esnault: Agree on scope issues - needs more thought.
  Chris Hopps: What is definition of a service? Are you creating
    a directory.
  Padma Pillay-Esnault: Might want to have IDs using different metrics.
  Chris Hopps : Similar to dockers/containers.
  Padma Pillay-Esnault: Agreed
  Tony P: Don't combine service metric and reachability metric.
  Padma Pillay-Esnault: Some service might want to follow IGP metric.
  Acee: Presentation is different than draft - draft uses
    client/server semantics.
  Padma: How do we want to use the service database?
  Robert Raszuk: Why flood information of addresses of services
    rather than node queries?
  Padma Pillay-Esnault: DNS does not advertise services.
  Tony P: There is service record DNS
  Chris Hopps: Jabber does this - but rarely used otherwise.
  Tony: Services are layered in terms of scope. S-BFD example
    provides an access point - don't want to use DNS.
  Les Ginsberg: Should we correlate GENAPP and this draft?
  Padma Pillay-Esnault: Thinking this may be a good idea
  Alia Atlas: Suggest to send out a video presentation on
   use cases - may help progress the draft.
  Padma Pillay_Esnault: Wants to simplify the draft to remove
   use cases that are not appropriate.
  Acee: Please continue discussion actively - before next IETF.
  Tony P: Stay away from semantics of the application.
  Padma Pillay-Esnault: Agreed

9)OSPF Link Overload (See slides) - Shraddha Hegde
- https://datatracker.ietf.org/doc/draft-ietf-ospf-link-overload/
 Acee Lindem: Wants geo-location draft to also use link
   scope RI LSA.
 Peter Psenak: RI LSA seems like node info - should use
   Extended Link LSA.
 Tony P: If all routers do not understand the extension is
   there a problem?
 Shraddha Hegde: Only changes link metrc if understood.
 Chris Hopps: Why not take link down?
 Shraddha Hegde: Use as link of last resort - may be take out
  of service after traffic is diverted.
 Les Ginsberg: Separate operation for TE/IGP?
 Shraddha Hegde: No
 Acee Lindem: Stub router affects traffic to the node - but this
   affects traffic in one direction - which does not matter in stub.
   Need this in non-stub case since other links will have "normal"
   metrics and will not discourage use of the link.

9) OSPF Draft Review - Acee Lindem
  Acee Lindem:Need more eyes reviewing documents. Suggest:
   1)OSPF Review team
   2)Assign document shepherd early

  Alia Atlas: Agree on more reviews. Participation in reviews
   shows there is interest in a document. Forcing review team
   does not show interest by community.
  Adrian Farrell: Agree with Alia. I do not review documents
   that I think are poor as this may indicate support when that
    is not intended.
  Tony P: Sometimes in RTGDIR review authors do not provide
    responses. We have become more permissive - don't kill idea with
     little support quickly if at all.
  Jeff Haas: RTGWG does not shepherd all documents by ADs - so assign
     shepherd after WG adoption
  Acee Lindem: Seems like #2 above is the better idea.
  Alia Atlas: This is an important topic. Agree asking for document
     shepherd earlier is a good idea. If document does not progrees in
     a year - kill it. Interested parties are not participating.
  Lou Berger: Problem is not documents sitting around.
     "Do you believe document is ready?" is the question
     asked in the shepherding template. Two duties of shepherds:
       1)Improve document
       2)Decide documentis ready to pub
  Padma Pillay Esnault: Authors should recognize shepherd is
    trying to help.