Minutes IETF98: isis
minutes-98-isis-01

IS-IS WG IETF-98
Time Slot (90m): Friday, March 31, 2017 11:50-13:20 CDT
Chairs (Christian Hopps, Hannes Gredler)
Scribe: Acee Lindem (acee@cisco.com)

- Intro, Adminastriva, Document Status - Chris Hopps (See slides)
     Alia Atlas: Some discussions surrounding using 
                 IS-IS for link-bundles. 
     Chris Hopps: Draft has been WG document for months and 
            months. Why are there questions now?
     Alia: Requires further discussion. Need to notify
           RTG WG and OSPF WG.                 
     Les: What is the progrssion of this document. Past
          WG last call since June. What is timeline?
     Alia: Wasn't completely aware of timeline and will
           do another AD review. Assuming no major concerns
           during IETF last call it could go to IESG telechat
           in May.
     Peter Lothberg: This is holding up required IS-IS protocol
           deployment.        
           
- YANG Data Model for IS-IS protocol - Stephane Litkowski (See slides)
   Document: draft-ietf-isis-yang-isis-cfg
   Document: draft-ietf-isis-sr-yang
   
   Rob Shakir: Expects to update the route policy model this week. 
   Chris Hopps: We should make a decision very soon on revised 
                datastore.
   Alia: Route policy and BGP models will be updated soon. In longer
         term, we're moving away from the dual tree but is worried
         about implementations and applications based on the dual 
         tree.
   Chris Hopps: Is the base SR model WG last called in SPRING?
   Stephane: It is close to last call but there is some cross
             checking.        
   Jeff Haas: Need to relook at what BFD has done and whether you
              want to schema mount it. Also, should there be BFD 
              configuration in the IGPs. 
   Acee: I think schema mount is the wrong hammer here.
   Alia: Need to go ahead with WG last call of models and put
         them in "Waiting for Implementation" state.
   Yingzhen Qu: Need YANG doctor reviews of models. 
     
- IS-IS Flooding Reduction in MSDC - Xiaohu Xu (See slides)
  Document: draft-xu-isis-flooding-reduction-in-msdc
  Chris Hopps: Why unicast hellos on LAN before DIS selection?
  Xiaohu: Avoid cost of multicast hellos being received by every 
          routers.
  Peter Lothberg: Elaborate on backup model. Assume that the control
          ethernet is broken and then your design will be good.
  Chris Hopps: Is management LAN used for traffic? 
  Donald Eastlake: There isn't much resource used here.  
  
- IS-IS TE Attributes per application - Les Ginsberg (See slides)
  Document: draft-ginsberg-isis-te-app
  Chris Bowers: Where would the new TLVs go? 
  Les: Set up a new IANA registery for TLV 22 sub-TLVs. 
  Chris Bowers: I'm not for this draft, but if we set up this 
         registry we would need to constrain the usage.
  Chris Bowers: Should not hard-wire the applictions. Should use 
         attribute applicaition ID. Two applications may want two
         separate attribute groups.
  Les: Could allocate 2 bits for two SR applications.
  Chris Bowers: Could have attribute group ID. 
  Les: This does not change the basic approach.
  Chris Bowers: We need to solidify the text before adopting. 
  Chris Hopps: Maybe you should incorporate this.
  Robert Razuk: Could have dynamic allocation of bits. 
  Les: Very open to discussing.
  Stephane: Believes something generic is desirable.  
  Uma Chunduri: Like backward compability option and agrees we
            need multiple values per application. 
  Alia Atlas: Why do you believe there are no backward compatibility
              issue?
  Les: We need to deal with the problems of backward compatibility
       and
  Chris Bowers: We need to define constraints on attributes.
  Chris Hopps: AI: So Les and Chris to merge ideas about generic group
               identification rather than hard coded app ids and we
               will then have a single solution. [Agreement from Chris
               and Les that both are open to this solution - Chris to
               post suggestions to list]
    
- IS-IS Extensions to Support Segment Routing over IPv6 Dataplane - Les
  Document: draft-bashandy-isis-srv6-extensions
  Les: Can we WG last call the base set of SR extensions for the MPLS
       data plane?   
  Chris Hopps: Polls room for having read and support WG last call? 
        Seems reasonable but will discuss on list.  
    
- IS-IS Routing for Spine-Leaf Topology - Naiming Shen
   Document: draft-shen-isis-spine-leaf-ext
   Juliusz Chroboczek: Is that not equalivant to put each leaf in
                       its own area?
   Les: Answer is "No". Leaf node only needs default in most cases and
        is not the same.
   Chris Hopps: This is the difference between IS-IS and OSPF with 
        respect to areas, an IS-IS instance can only be in 1 area so
        the spine nodes don't work.
   Nikos Triantafillis: Concerned with computational space wiht many 
        spines and many prefixes.        
   
      
- Overhead Reduction for IS-IS Enabled Spine-Leaf Networks - Zhe Chen
   Document: draft-chen-isis-sl-overheads-reduction
   Chris Hopps: Is this standard IS-IS? 
   Les: This is not standard IS-IS since there are not multiple 
        L1 areas.
   Chris Hopps: Does each leaf (could be many) in the diagram require
                another area?
   Zhe: Yes
   
- Avoiding Traffic Black-Holes for Route Aggregation in IS-IS - Zhe Chen
   Document: draft-chen-isis-black-hole-avoid
   Les: Does the spine have to segregate the unreachable prefix? 
   Zhe: For this specific space. 
   Chris Hopps: Link-Scoped LSPs 
   Les: Exactly the opposite of what is being proposed in the other
        draft.
   Chris Hopps: How does this relate? Same solution in a different way?
   Zhe: Different solution.
   Naiming: There are situations where you don't know what you are
            missing. That is whey negative routes don't work. 
   Zhe: Maybe we can use configurations to solve this problem. 
   Naiming: Route informatin from leaf can be dynamic. 
   Zhe: Will think about in next version of the draft. 
   
   
Alia: IEEE Liason for Jumbo LLCs - Allocated a new ethertype for 
      jumbo LLC frames. 
Les: This is huge. Everybody has implemented jumbo frames using the 
     old ethertype. 
Alia: Thank you for the perspective. 
Peter Lothberg: Changing the ethertype is a disaster. 
Alia: The IEEE can't find out who owns the ethertype. 
Donald Eastlake: If this is only used with multicast? 
Chris: No - unicast as well. 
Alia: Need help replying back to IEEE.