IETF 89 - ISIS Minutes
      Scribe: Acee Lindem - acee.lindem@ericsson.com
      WG Chair: Chris Hopps, Hannes Gredler

 * Status - Chris Hopps (See Slides)
      - 1 new RFCs (RFC 7142)
      - 5 WG documents (3 Active, 1 WG Last Call, 1 Submitted for Publication


 * IS-IS Extended Sequence Number - Albert Tian (See slides)
      - Comments all addressed - many incorporated into 02 version.
      - Ready for WG Last Call

 * Advertising Per-Node Tags - Chris Bowers (See slides)
      - Node level rather than existing Link Level (RCF 5305)
      - Node Level rather than exiting Prefix Tag (RFC 5130)
      - Unbounded list of 32-bit tags in TLV

   Robin Li: What is the usage?
   Chris: LFA backup path policy for selection, Identify rLFA targets that
          support targeted LDP sessions.
   Robin: Automesh is also concept of color for who will join the mesh group.
   Chris: Are you suggesting auto-mesh as an application of this draft?
   Robin: No - Automesh has an existing way.
   Les Ginsberg: Are you using the same examples as OSPF?
   Chris: Yes.
   Les: Why don't you use common tag or bit for well-known applications.
   Chris: The SPs would rather leave it flexible even if they have to
       configure it.
   Les: So the tags will have separate meanings per node?
   Chris: rLFA needs to be able to exclude nodes with a certain tag.
   Les: You're not defining what paths are red paths?
   Hannes: For example, please try all LFAs for which there is no red node
        on the path.
   Chris: May not support low bandwidth on links of colored node.
   Les: Is the prefix tag name space separate?
   Chris: Yes
   Acee: Alternate is IANA registry for well-known applications.
   Eric Yu: 64 tags in OSPF while in ISIS it is unbounded?
   Chris: OSPF tags go into single instance RI LSA and have bounded length.
   Eric: How do I cancel a tag?
   Hannes: Send an LSP with a higher sequence number.
   Anton Smirnov: Difference is OSPF RI LSA is a single instance LSA.
       Applications are just examples which may or may not be be
       implemented and are not mandatory.

 * ISIS Autoconfiguration - Leo Liubing (See Slides)
    - See slides for changes from 01 version

   Chris Hopps: Was on honeymoon last meeting. Are operators now wanting to
      connect their IS-IS networks to the home?
   BT Guy: Not home network environment. It is about zero configuration.
   Ian Farrer: Very similar function in IEEE for SPB. There will be a BCP for
       SPB (Shortest Path Bridging).
   Les Ginsberg: Need to look at it in the framework of not being in the home.
       Two cases of system ID duplicate resolution. You need to also cover the
       case of a conflict with your neighbor prior to adjacency formation.
   Leo: MAC address duplication is a problem that cannot easily be solved.
   Les: You could have multiple interfaces and choose a different
        MAC for system ID than the interface on the link.
   Les: OSPF targets small networks. ISIS targets larger networks for
        autoconfiguration (e.g., TRILL).
   Leo: Homenet decided not to support multiple IGPs?
   Les: Strong support for single routing protocol.
   Chris: May not need in ISIS as well.
   Gregory Cauchie: ISIS was not chosen but there were many in Homenet who
        were undecided and the final Homenet protocol has not been
        decided.
   Ian Farrer: Need to get work in front of Homenet.
   Chris Hopps: Doesn't seem ready yet for work group adoption, need to sync up with
        other efforts and see if homenet even wants it.

 * ISIS Advertisement of Global Labels - Xiaohu Xu (See slides)
   Hannes: Conflict with authors of ISIS Segment Routing draft?
           Have you been able to resolve?
   Xiaohu: Uses different TLVs?
   Hannes: Difference is that this draft uses existing TLV with tweaked semantics, like the high metric.
   Xiaohu: Uses global labels rather than label indexes as well.
   Acee: Differences in SPRING Arch should be debated in SPRING as
         opposed to via IGP encodings in the IGP WGs.

 * ISIS Advertisement of Entropy Label Capability - Xiaohu Xu (See slides)
   Hannes: Curtis made comment that the ELC property is a link label property
           rather than a node property.
   Xiaohu: RFC 6790 says all interfaces should support entropy label processing
           rather than a mix of those supporting and not supporting.
   Stephane Litkowski: Why not use capability bit rather than new TLV?
   Xiaohu: You could.
   Chris: Should work out in SPRING.

 * ISIS Extensions for MPLS Multi-Topology - Robin Li (See slides)
   Acee: So the gist of the last slide is that global label is needed beyond
         segment routing?
   Robin: I think so.
   Alia: It would be Good to see the architecture differences presented in the
         SPRING WG. Will there be requirements for this MPLS Multi-Topology?


   Chris: Any comments on other drafts?
   Hannes: See you in Toronto.