Skip to main content

Minutes for ISIS at IETF-93
minutes-93-isis-4

Meeting Minutes IS-IS for IP Internets (isis) WG
Date and time 2015-07-20 13:20
Title Minutes for ISIS at IETF-93
State Active
Other versions plain text
Last updated 2015-08-07

minutes-93-isis-4
    IS-IS Meeting Minutes -   Monday, July 20th, 2015 1520-1720 CDT
    Chairs: Chris Hopps and Hannes Gredler
    Scribe: Acee Lindem (acee@cisco.com)

  - Intro, Adminastriva, Document Status
    Presenter: Chairs (Christian Hopps, Hannes Gredler)

  - IS-IS Path Computation and Reservation (IEEE 802.1Qca)
    Presenter: János Farkas
    Document: https://datatracker.ietf.org/doc/draft-farkas-isis-pcr/

    * See slides.
    Acee: Any implementations?
    Janos: Yes - Avaya has implementation and others have
           prototypes.
    Chris Hopps: No objections to WG Last Call.

 - IS-IS Prefix Attributes for Extended IP and IPv6 Reachability
   Presenter: Les Ginsberg
   Document: https://datatracker.ietf.org/doc/draft-ginsberg-isis-prefix-attributes/

   * See slides - New co-authors. WG last call requested.
   Chris: Any objections to WG last call? None
   Chris: Support WG last call: Significant
   Chris: Take to list but expect last call.
   Chris: Early allocation of code points done - was extremely easy.

   Les: we are requesting last call on two documents:

      draft-ietf-isis-prefix-attributes-01.txt
      draft-ietf-isis-route-preference-01.txt

   The latter document hasn't changed in the last two IETFs - but has
   a dependency on the prefix-attributes draft and has been on hold
   until the prefix-attributes draft was ready for last call.

 - Advertising L2 Bundle Member Link Attributes in IS-IS
   Presenter: Ahmed Bashandy
   Document: https://tools.ietf.org/html/draft-ginsberg-isis-l2bundles-00

   * See slides.
   Uma Chunduri: Prefer it would be existing Sub-TLV in TLV 22.
   Les Ginsberg: Information doesn't impact L3 topology. Would bloat
                 TLV 22 significantly. Isolate L3 topology changes
                 from L2 attributes.
   Ahmed: Agrees with Les.
   Stephane Litkowski: Why not just use L3 bundles?
   Ahmed: Would expose the L3 view to all the protocols running
          on the links - not just IS-IS.
   Jen Linkova: Loves the draft. Wants to keep L2 bundles rather than
                L3 bundles.
   Hannes: Why not run unnumbered on the parallel links?
   Ahmed: Unnumbered doesn't work on LANs while this does. Less bloating.
   Acee: Why not limited the identifier to a link identifiers. No IP
         addresses.
   Les: Wanted to leave the door open to define an IPv6 address per link.
   George Swallow: This is useful and we have this already in MPLS.
   Chris: (in resoponse to eariler assertion early allocation had been
          done) Early allocation not done yet.

 - IS-IS Extensions for Segment Routing
    Presenter: Stefano Previdi
    Document: https://datatracker.ietf.org/doc/draft-ietf-isis-segment-routing-extensions/

   * See slides.
   Chris Bowers: Does the algorithm 1 allow an LFA policy?
   Stefano: Algorithm 0 is whatever has been supported by
            IS-IS forever? LFA always is a different path.
   Chris: Text not clear.
   Acee: What do you mean by no local policy for algorithm 1?
   Stefano: It only applies within scope of IS-IS?
   Uma: Add clarifying text on changes.
   Stefano: Will add.
   Shraddha Hegde: Can you advertise a separate SID for different
                   algorithm?
   Stefano: Yes - each algorithm can advertise a separate SID.
   Bruno: What do you mean by right or wrong MS entry?
   Stefano: How can you choose when there are conflicts?
   Bruno Decraene: You can pick one without conflict. Needs to
                   be consistent.
   Peter Psenak: Why can't one use mapping server for prefixes from
                 prefix-aware routers as well?
   Stefano: Agress this a good idea.
   Hannes: Impossible to come up with tie breaking scheme?
   Stefano: Don't have consensus on tie breaker.
   Chris: Discuss this on list.
   Stefano: ALL routers in the domain must behave the same with the
            respect to SID conflicts.

 - IS-IS Autoconf
   Document: https://tools.ietf.org/html/draft-liu-isis-auto-conf-01
   Presenter: Bing Liu

   * See slides.
   Bing: Asks for WG adoption.
   Chris: Believes it is progressing. Anyone object to adoption.
   Chris: No objections and some support for adoption.
   Chris: Do we even want to do authentication?
   Acee: We have a simple key configuration option for OSPFv3
         authentication. This could be reused in this draft.

 - IS-IS Extensions for Flow Specification
    Presenter: Jianjie You
    Document: https://tools.ietf.org/html/draft-you-isis-flowspec-extensions-01

    * See slides:
    Acee: IS-IS is not defined as an IETF PE-CE protocol.
    Jianjie: Can remove use case.
    Hannes: Knows of at least one implementation of IS-IS as a
            PE-CE protocol.
    Jianjie: Working group adoption?
    Uma: Remove section on PE-CE protocol.
    Jianjie: Will remove.
    Jeff Haas: How do you limit these flow spec to rules in the
               routing domain?
    Jianjie: Will need to define.
    Chris: Take discussion to mailing list.

 - Advertising Encapsulation Capability Using IS-IS
    Presenter: Xiaohu Xu
    Document: https://tools.ietf.org/html/draft-xu-isis-encapsulation-cap/

    * See slides:
      Jianjie: Requests WG Adoption?

    Hannes: Objections for WG adoption?
      None.
    Chris: Support for WG adoption?
      Considerable amount.

 - IS-IS LSP lifetime corruption - Problem Statement
    Presenter: Bruno Decraene
    Document: https://tools.ietf.org/html/draft-decraene-isis-lsp-lifetime-problem-statement-00

    * See slides.
    Uma: This problem is discussed in the IS-IS KARP analysis. Is there a
         proposed solution?
    Bruno: We can talk of solutions.
    Chris: You must accepted the packet so the neighbor must have been
           authenticated.
    Bruno: Could be man-in-the-middle attack where lifetime is modified in
           authenticated packets since it is not protected.
    Chris: Someone would need a box physically placed in the network.
    Les: Security loophole in IGP begs question of how the attacker
           got in the network in the first place.
    George Swallow: Corruption can occur without attacker. For example,
           via gamma rays.
    Les: Believes the focus is on attacks rather than corruption.
    Bruno: Corruption is possible.
    Les: Cypto authentication will catch corruption.
    Bruno: Can discuss solutions.
    Chris: Discussion will continue on the list.

 - IS-IS Point-to-Multipoint Operation
    Presenter: David Lamparter
    Document: https://tools.ietf.org/html/draft-lamparter-isis-p2mp-00

    * See slides.
    Acee: Did you consider a hybrid approach with hellos still
          multicast?
    Chris: Multicast is VERY unreliable. Want to assure hellos are
           reliable once adjacency is established.
    Donald Eastlake: Why not represented as a mesh of P2P links?
    David: Link is seen as a broadcast network.
    Donald: IEEE 802 is defining P2P mesh for broadcast networks.
    David: Need to solve the problem for existing technologies and
           hardware.
    Donald: Why not use association?
    David: Described in draft.
    Les: Why do you need a new PDU?
    David: LAN Hellos used for discovery and P2P Hellos for
           adjacency maintenance.
    Les: You can use existing P2P hellos, instead of LAN hellos.
    Chris: Why not?
    David: Several use cases where we need to use LAN hellos.
    Les: Worst thing that could happen is that a router is stuck
         in INIT state. Only need to change destination addresses.
    Juliusz Chroboczek: How much do multicast hellos cost? Agress that
                        multicast hellos for discovery and for validation
                        of link layer.
    Les: LSPs and CSPs cannot be multicast due to the possibility of
         the lack of transitive connectivity.
    Hannes: In favor of anything that gets rid of psuedo-nodes.
    Acee: Even though this is a layer 2 problem, IS-IS P2MP is a good
          alternative in the immediate time frame.

 - IS-IS over IPv6
    Presenter: Christian Franke
    Document: https://tools.ietf.org/html/draft-franke-isis-over-ipv6-00

    * See slides.
    Les: There was a draft 15 years ago. One of the strengths of IS-IS is
         that it runs over layer 2.  Problems with requiring IPv6 up prior
         to running IS-IS. 25 years of history says we don't need this.
    Mikael Abrahamsson: There are devices on which the Layer 2 packet I/O
         APIs are not accessible or are hard to use. We need this to
         promote IS-IS implementation.
    Chris: I believe the pressure for this is coming from a few folks
           in homenet. Do we technically need this, or is it just to
           make some folks happy?
    Hannes: Have you considered interoperability?
    Christian Franke: Is is configured per circuit. Possible to handle both
            encapsulations on same link but there are problems.
    Hannes: What about MTU compatibility?
    Christian: Could do IPv6 fragmentation.
    Jen: IS-IS allows blackholing since it doesn't run over IPv6.
    Chris: Problem addressed with BFD.

 - IPv6 Source/Destination Routing using IS-IS
    Presenter: David Lamparter
    Document: https://tools.ietf.org/html/draft-baker-ipv6-isis-dst-src-routing-03

    * See slides.
    Chris: How many read?
    David: Now would be the time to read?
    Chris: Take discussion to the list.