Routing Area Open Meeting (rtgarea) IETF 97 (Seoul) =============================================================================== Area Directors: Alia Atlas (akatlas@juniper.net) Deborah Brungard (db3546@att.com) Alvaro Retana (aretana@cisco.com) Area Secretary: Jonathan Hardwick (jonathan.hardwick@metaswitch.com) Wiki: https://trac.tools.ietf.org/area/rtg/trac/wiki/WikiStart Scribe: Jonathan Hardwick (jonathan.hardwick@metaswitch.com) Location: Grand Ballroom 2, Conrad Seoul, KR Time: November 15, 2016, 1330-1530 (1:30pm-3:30pm) ------------------------------------------------------------------------------- 1. Administrivia ---------------- Alvaro: Congratulations to the two teams from the routing area that won prizes at the IETF Hackathon. Conversation regarding IAB statement on IPv6: https://www.iab.org/2016/11/07/iab-statement-on-ipv6/ David Sinicrope: Asking as BBF liaison. Did the IAB say how they would communicate this to the wider community? It will come as quite some news. Alia Atlas: It's released as a statement and there will be some press associated with it. Andy Malis: I took the action of distributing this statement to the ONF leadership. Stewart Bryant: There's a lot of stuff out there that will not support anything except IPv4. Some of this kit relates to national infrastructure. If the IETF does not support it until end of life, someone else will. The ITU may take up IPv4, for example. ------------------------------------------------------------------------------- 2. OPNFV Overview ----------------- Heather Kirksey - No comments or questions. ------------------------------------------------------------------------------- 3. Working with Standards Action Registries ------------------------------------------- Jeff Haas On a discussion of migrating from a temporary code point to an official one: Stewart Bryant: Hardware shippable with configuration is safer than code points burned into the silicon. Jeff Haas: Consider the blast radius of the code point. Stewart Bryant: It's easier to change software than to change and redeploy hardware. John Scudder: When reviewing a doc, if someone is squatting on a code point in the document, make them take it out. That should minimally happen at WG adoption time, but there's no need to wait for that. Discussing the different types of allocation policy that can be set on registries: Donald Eastlake: Standards action seems stronger than you need for most cases. Jeff Haas: Agree. Donald Eastlake: Occasionally things need to be stricter e.g. error codes. We use Expert Review. Extra is also a good solution. Michelle Cotton (IANA): Just to clarify - 5226bis has been mentioned, but the early allocation process is documented in RFC 7120. IANA has a desk near registration - please come to it and ask us any questions. Regarding First Come First Served registries, our goal is to do assignments in a few days. Email us at iana@iana.org in between meetings. Stewart Bryant: We have a 2 year drop dead date for early allocation. Is that still wise? Sometimes it takes more than 2 years to finish. Jeff encourages us to start early, but then we might hit the deadline. Michelle Cotton: RFC 7120 does not cover creating early registries. There's a chance RFC 7120 might be revved to cover that, and we can discuss the 2 year time limit at the same time. Stewart Bryant: It would be good to reconsider the limit. We hit the 2 year limit a few times in PWE3. Michelle Cotton: I don't think we have hit >2 years often, only once or twice. Ultimately, the IESG can leave it open a while longer. Alia Atlas: There was one case for the OSPF segment routing drafts, where an extensions was needed. Docs can take longer, but with a bit of motivation, things can happen quickly. ------------------------------------------------------------------------------- 4. ANRP Winner Talk: "Central Control Over Distributed Routing" Olivier Tilmans Hannes Gredler: Two existing solutions for ECMP path splitting; first is forwarding adjacencies (RSVP tunnel advertised in IGP); second is segment routing via a stack of node labels. Have you looked at these alternatives? Olivier Tilmans: With SR we disliked that you stil have to configure the ingress of the SR path. Doing this for 100 edge routers is tedious. Certainly though you could reuse this technology. Stewart Bryant: If you are talking about lots of updates then you need to run loop-free reconvergence otherwise you will glitch the traffic. However, SR will never cause a microloop. This will cause loops. Olivier Tilmans: You can compute the correct ordering of the updates. Stewart Bryant: But then you lose the benefit, because you can't do it in milliseconds. Acee Lindem: What router ID do you use for fake node A? How do you make the router IDs reachable? Olivier Tilmans: We maintain an adjacency with the fake router. Stewart Bryant: (1) How well does this complex approach compare to the old methods of downstream routes or LFAs? (2) Have you looked at this on some real topologies? We found during the FRR work that real topologies gave new insights. (3) Ordered-FIB is a way of controlling this so it doesn't give loops. Olivier Tilmans: The issue with LFAs was we did not understand how to explicitly control the chosen path. Stewart Bryant: It is hard to do multiple concatenated LFAs. But as soon as you get beyond 2 hops, SR is an easier way, and has no state on the new network. Olivier Tilmans: We have looked at a realistic topology. Stewart Bryant: Have you looked at non-planar topologies? (Cube or sphere.) Olivier Tilmans: Yes. Stewart Bryant: Your aims are to be able to make short-lived changes, but that is inconsistent with the need not to create microloops. SR avoids the risk of microloops. Robert Raszuk: If you have LDP network, how do you inject labels to fake node? Olivier Tilmans: We maintain a label to node mapping. ------------------------------------------------------------------------------- 5. Working Group and BoF Reports -------------------------------- Only discussion of points not written on the wiki is recorded in the minutes. See the wiki report for the detailed WG status reports. https://trac.ietf.org/trac/rtg/wiki/IETF97summary Stewart Bryant (PALS report): IPR took a long time on one of our drafts. There was a long list of authors, with historical authors too. Note to WGs - be careful how many authors you put on a draft, it can hold it up. Stewart Bryant: Should we poll contributors as well as authors? Alia Atlas: That is unnecessary. As it says in the note well, if you contribute something that has your IPR on it then you have to declare that IPR. Loa Andersson: In MPLS, we have the practice of calling for IPR on both contributors and authors. Thre is no evidence it takes more time. Stewart Bryant: The problem comes when people move jobs and their email addresses go dead. Alvaro Retana: This needs a longer discussion - please take your thoughts to the list. Alvaro Retana (for Sandy Murphy) (SIDR Report): SIDR is being closed as it has finished all its work; SIDROps is a new working group that is taking over. Sue Hares (TRILL report): TRILL has seen some deployment in data centers. ------------------------------------------------------------------------------- 6. Open Discussion / Any other business --------------------------------------- - None. -------------------------------------------------------------------------------