IS-IS WG Agenda IETF-96 Time Slot (90m): Wednesday, July 20th, 2016 1550-1720 CEST Chairs: Christian Hopps and Hannes Gredler Scribe: Acee Lindem (acee@cisco.com) - Intro, Adminastriva, Document Status (See slides) Presenter: Chris Hopps Les Ginsberg: Did we not move L2 bundles past WG call? Hannes: It is pending my shepherd's report. Acee: IS-IS YANG is waiting on Ops-state discussion conclusion. I think it is fully featured and ready from a content standpoint. Will discuss at the next IS-IS YANG model call. Chris Bowers: The MRT mechanism is complete but have to decouple a controlled convergence timer that we need to decouple from MRT in a separate RTG WG draft. - Update to IS-IS Multi-Instance (See slides) Presenter: Les Ginsberg Document: https://datatracker.ietf.org/doc/draft-ietf-isis-mi-bis Les: Ready for WG last call? Chris: One implementation originally implemented the behavior according to the draft. Chris: Who thinks it is ready? Room: A few Chris Who objects to it going to WG last call? Room: None - ISIS Auto-Configuration Presenter: Bing Liu (Leo) Document: https://datatracker.ietf.org/doc/draft-ietf-isis-auto-conf/ Ian Farber: The question I meant to ask is "How do you know when you've synchronized with all your neighbors?" The text doesn't address this. Leo: One minute should be reasonable for home network. But may not be right for larger network. Hannes: May want to measure LSP arrival rates to determine convergence. Les: Problem is not database synchronization. It is how long do you wait to discover all your neighbors? Chris: This is ONLY for duplication protection? Leo: It is not duplicate detection. It is for startup mode before network is available. Leo: Requests last call. Chris: I noticed in a TBD in the text. The WG Last Call will happen on the list after open questions resolved. Ian: TBD is removed in the latest version. - Update to RFC 5316 Presenter: Mach Chen Document: https://datatracker.ietf.org/doc/draft-chen-isis-rfc5316bis Mach: WG adoption? Chris: Is this the same change we made in draft-ietf-isis-rfc4971bis-01? Mach: Yes. Chris: Anyone object to adopting this draft? Room: No one. Chris: I think we should move quick on this one. We will take this the adoption to the list. - IS-IS Extensions for Segment Routing Presenter: Stefano Previdi Document: https://datatracker.ietf.org/doc/draft-ietf-isis-segment-routing-extensions/ Chris: So effectively TLV is really "SR Additional Algorithms"? Stefano: Yes - it is really the backward compatibility question? Must poll implementations for backward compatibility on change to text to unconditionally support algorithm 0? Alia: Perhaps Juniper may want to look at their implementation to see if it impacted by the backward compatibility issue. - IS-IS extensions for SPRING Multicast Presenter: David Allan Document: https://datatracker.ietf.org/doc/draft-allan-isis-spring-multicast Dave: Not looking for WG adoption. Contingent on PIM WG acceptance. Stafano: Computation is based on topology. Is it confined to area? Dave: Currently, it is but one could postulate a multi-area computation. It would be a new draft. Chris: Should IS-IS and OSPF meet together for items like this that are presented in both WGs. Alia: I think it would be an interesting experiment to meet together for proposal like this. Chris: Some overlap. Alia: PIM and MBONED have joint meetings. I don't the RTG WG is the right WG for everything that hits more than one WG. Hannes: No reference to multi-topology in the draft? Is it limited? Dave: You could do it for multi-topology? However, you'd drive up the computation linearly. Hannes: Not pushing for relaxation - just needs to be specified. Dave: Not sure it is a good idea but we left this out intentionally to make it agnostic. Hannes: Some networks have non-congruent topologies for IPv4 and IPv6. Ahmed: Draft is already referencing multi-topology TLVs so it is supported by definition. - Signaling MSD (Maximum SID Depth) using IS-IS Presenter: Jeff Tantsura Document: https://tools.ietf.org/html/draft-tantsura-isis-segment-routing-msd Chris: TLV is 32 bits but value only requires 8 bits. Jeff: WG adoption? Microsoft wants this? Les: Usage of MSD 0 is inconsistent with MSD draft in PCE. Jeff: MSD 0 means the router can't push any labels on the label stack. Ahmed Bashandy: What is the link MSD? Jeff: This is for the traffic being egressed from this node. Can only push this number of labels on the MPLS label stack. Chris: Does anyone object to WG adoption? Room: No one. Chris: Does anyone support it? Room: A few. Chris: Will take to the list. - Carrying Geo Coordinates Information In IS-IS Presenter: Naiming Shen Document: https://tools.ietf.org/html/draft-shen-isis-geo-coordinates Chris: Does one flag indicate centimeters vs meters? Naiming: Yes. Alia: Have you talked to anyone in ART about how to do privacy with GPS coordinates? Naiming: No. Alia: Send me an E-mail and I will send you a reference. Chris: Are you seeing routers with GPS HW so they can know the value? Naiming: If you jitter the result - all should be done in the same. Chris: I was thinking about dispatcher a support personnel. Hannes: Does this apply to prefixes or the node? Naiming: This applies to the node not the prefixes. In BGP, could apply to prefix. Julian Lucek: Centimeter granularity could cause flapping advertisement if GPS on router is periodically polled. Julian: Have you considered BGP-LS? Naiming: Have not considered BGP-LS yet. Stefano: Who is the consumer? Can you advertise it in BGP-LS only? Acee: We will add BGP-LS encodings in the BGP draft. Robert Raszuk: If you redistribute into BGP, how is this handled? Enke Chen: Could associate the informaiton with a loopback. Hannes: How would you handle multi-area? Naiming: In the interest of time, will take discussion to the list. - IS-IS Routing fo Spine-Leaf Topology Presenter: Naiming Shen Document: https://tools.ietf.org/html/draft-shen-isis-spine-leaf-ext Tony Przygienda: Very cool. Did you look at RPL and the dynamic ranking? With the defaults, you advertise the exceptions. Hannes: The problem of sparse protected nodes. One way to handle this is with multiple IS-IS L1 instances? Chris: I think you may get into trouble with leaf-to-leaf routing. Naiming: Please send this to the list. Julien Meuric: I think TRILL has a similar mechanism. You may want to look at it. Naiming: I will look at this. Chris: Okay, we are 5 minutes over. Let's take the discussion to the list.