Skip to main content

Minutes IETF100: ospf
minutes-100-ospf-00

Meeting Minutes Open Shortest Path First IGP (ospf) WG
Date and time 2017-11-16 01:30
Title Minutes IETF100: ospf
State Active
Other versions plain text
Last updated 2017-11-16

minutes-100-ospf-00
   IS-IS/OSPF WG Agenda IETF-100

Time Slot (150m): Thursday, November 16, 2017 09:30-12:00 (GMT+8)
Scribe: Ayan Banerjee (ayabaner) <ayabaner@cisco.com>

  * Intro, Adminastriva, Document Status
      o Presenter: IS-IS Chairs <mailto:isis-chairs@ietf.org> and OSPF
        Chairs <mailto:ospf-chairs@ietf.org> (Acee Lindem
        <mailto:acee@cisco.com>, Abhay Roy <mailto:akr@cisco.com>,
        Christian Hopps <mailto:chopps@chopps.org>, Hannes Gredler
        <mailto:hannes@gredler.at>)

 Discussion of merging WGs:
   Chris Hopps: Let's discuss merging the two WGs.
   Acee: Initially against merger but now doesn't feel it is that bad due to
   advances in OSPFv3 TLV encoding and OSPFv2 prefix/link attributes. Chris
   Bowers: Thinks having IS-IS and OSPF together will have long-term benefits.

Document Status:

   Chris Bowers: June changes removed ERO and binding SID. Binding SID is
   needed. Jeff Tantsura: What about OSPF? Alia Atlas: ERO was not specified
   and didn't relate to SPRING working documents. Behavior and archiecture were
   not specified. Chris Hopps: Are we aligned? Jeff Tantsura: What is Chris
   worried about? Chris Hopps: The SR Binding TLV is needed to support context
   mirroring which is implmented. Jeff Tantsura: Binding SID could be published
   with use cases. Ketan Talaulikar: Whatever usage of ERO is removed so
   binding SID could also be removed. Shraddha Hedge: The binding SID TLVs are
   not implemented right now. There is a use case that could be added in a
   separate draft. Acee: Usage of Binding SID for context mirroring is
   sub-optimal. If required, we will define an OSPF Sub-TLV and SID for this
   purpose.

   Alia Atlas: OSPFv2 Segment Routing will be sent to the IESG with other
   SPRING documents.

   Chris Hopps: Is IS-IS YANG model in sync with OSPF?
   Acee: Some updates needed - will discuss with co-authors.

  * OSPF Graceful Restart Enhancements
      o Presenter: Acee Lindem <mailto:acee@cisco.com>
      o Document:
        https://datatracker.ietf.org/doc/draft-basavaraj-ospf-graceful-restart-enhancements/

    Shraddha Hedge: If you generalize the first, you need to handle the case
    where the peer router is slow forming an adjacency. Acee: Could defer
    convergence if any neighbors are in Exchange or greater state.

  * OSPF LLS Extensions for Local Interface ID Advertisement
      o Presenter: Ketan Jivan Talaulikar <mailto:ketant@cisco.com>
      o Document:
        https://datatracker.ietf.org/doc/draft-ietf-ospf-lls-interface-id/

    Acee Lindem: Does anybody have any reservations to requesting early IANA
    allocation for this draft?

                 No objections in the room.

  * ISIS Segment Routing Flexible Algorithm
      o Presenter: Shraddha Hedge <mailto:shraddha@juniper.net>
      o Document:
        https://datatracker.ietf.org/doc/draft-hegdeppsenak-isis-sr-flex-algo/

    Chris Hopps: Why are user defined algorithms needed?
    Shraddha: Constraints cannot be standardized.
    Dhruv Dhody: PCE has object types for metric types. Maybe metric types in
    PCE could be used. Ville Hallivuori: One algorithm conflict disable an
    algorith in the domain? Shraddha: If there is a chance of looping, you must
    revert to default behavior. Chris Hopps: This a misconfiguration and we
    don't need to optimize. Bruno: In theory, you need topology ID as well.
    With SR, we have two algorithms and these cannot be redefined. Shraddha: We
    don't use standard algorithms for user defined. Metric types are also
    standardized. Bruno: Standardized algorithm with user defined metric?
    Shraddha: If you have an algorithm that needs to be standarized, can go to
    IANA. Ketan: It is possible to define strict SPF with TE metric type.
    Stephane: This is not an algorithm, it is contraints for SPF. Shraddha:
    There are Sub-TLVs with contraints. Stephane: Constraints can be
    independent of algorithms. Shraddha: We will look at decoupling the
    algorthm and constraint specifications. Chris: MRT algorithms could be
    defined as standard algorithms. Constraints can be used with standard
    algorthims. Bruno: Could add standard algorithm field to existing TLVs.
    Stephane: Maybe could be made flexible by reusing TE objects. Acee: Need to
    handle case where multi-homed prefix is advertised with different
    algorithms by different routers. Acee: IPR exists on draft. Dhruv: Please
    push IPR declaration to the Working Group.

  * IS-IS TE Attributes per application
      o Presenter: Acee Lindem for Les Ginsberg <mailto:ginsberg@cisco.com>
      o Document: https://datatracker.ietf.org/doc/draft-ietf-isis-te-app/

  Chris Bowers: Flex Algorithm should not be referenced since it is not a WG
  document. Chris Bowers: There are differences between explicit and implicit
  enablement.

  * Enablement Discussion

   Chris Hopps goes through history of draft and his perspective on the
   adoption of the enablement draft.

   Chris Hopps: Will discuss enablement in next topic.
   Chris Hopps: Do we need explicit enablement decoupled from attribute
   advertisement? Acee: It is dependent on algorithm. For all but TE, it is
   clear that it is not necessary. Chris Bowers: With both documents WG
   documents, it is a difficult question. I see need for enablement draft.
   Solves immediate problem. Alia Atlas: How does enablement related to TEAS
   and GMPLS? Ketan: Application specific values is a problem that we need to
   solve. Shraddha: There are sometimes different solutions to the same problem
   in IETF. Chris Bowers: WG adopted IS-IS TE attributes draft didn't address
   enablement. Could adopt both drafts. Bruno: Application specific information
   is theoretical. Stephane: Application specific attributes could be complex.
   Chris Hopps: It is not that complex that it can't be implmented. Chris
   Bowers: Could be deployment complexities. Chris Hopps: We are not going to
   decide this today. Acee: There are much bigger issues in OSPF than IS-IS
   with the number of LSAs required for non-TE applications. In the case of
   OSPF, this is more important than how TE enablement is handled.

 * More Merging Workgroups

   Chris Hopps:
   Alia Atlas: Polling of who would only attend OSPF or IS-IS but not both>
       None
   Name for New WG:
          LSR - Link State Routing
          IGP-LS - IGP - Link State
          LaSeR - Link State Routing