WG Status 09:00 / chairs, 20 min Adrian: about the documents in Revised I-D Needed state. Some documents stay a very long time in that state. Looks like the authors do not want to publish anymore. Loa: do not think it is the case. Adrian: should we then return it to the WG rather than dealing this between authors, AD, and shepherd? Loa: decision is on your side. no problem with it. we can learn from this. [state for smp-requirements is Revised ID Needed] draft-ietf-mpls-multipath-use-02 09:20 / 09:25 Curtis Villamizar, 10 min Lucy Y.: how is this draft related to the referenced composite link document? Curtis: the draft says you can use mpls as server layer to mpls-tp if you use and EL and vice versa and so it describes rules to do so. draft-ietf-mpls-forwarding-02 09:30 / 09:28 Curtis Villamizar, 36 min Eric O.: good suff, but nothing mpls specific. Curtis: there is some mpls specific stuff, and topics here are too often ignored The draft tells you what the problems are. Eric O.: there are a lot of SHOULD that made me nervous in the draft Curtis: we are discussing things which can impact LSRs George: I think is is very useful material Lucy: very usefull indeed. how do you see new stuff coming up e.g., global labels, fit in your study? Curtis: this is a snapshot in time, difficult to state with regards to future technology evolutions This is an informational document. George: in some case you might want to use lower cases for must/should Lucy: to the chairs, should we move forward with this document or wait for new stuff (e.g., global label) to stablize Wim: if we go forward, a piece is missing which is convergence Curtis: I beleieve we have something in the FRR section but please share this on list Wim: we are indeed talking about forwarding performance, but convergence should also be discussed Eric O.: I had two concerns: 1/ use of 2119 terms; 2/ not specific to mpls Curtis: tell me what you think should be removed Robin: I appreciate your work very much. you could add some performance analysis on the depth of the label stack Curtis: some vendors can deal with deep label stack, up to 6, 8 labels Nick D.: really good document. [missed] Sharam: very good document. maybe you could add guidelines (e.g., use EL for balancing, ...) Curtis: giving numbers is between vendors and customers, and I get the feedback that there is already too much recommendations. draft-akiya-mpls-lsp-ping-reply-mode-simple-00 10:10 / 10:14 Nobo Akiya, 10 min Greg: very useful to extend lsp ping/tracert. What happens if rules for formating are not respected i.e., duplicate entries ? Nobo: good point. Sam A.: Why can't you use a control channel? Nobo: there might not always be a control channel back Sam A.: how do you do in mixed environments where not all implementations support this? Nobo: valid concern Curtis: if data plane and control plane are out of sync, the next hop could be on a control channel but it will be the wrong one. So the only reliable mode is using v4/v6 Nobo: you are right but some operator prefer using the reverse LSP if it exists Curtis: but then there might be the error condition I mentionned. You thus might want to change the preference order Nobo: I would like to hear from operators on this. Nick: we would like the reverse path to the reverse LSP but how do you determine the reverse LSP Nobo: FEC based Lucy: [missed] draft-akiya-mpls-entropy-lsp-ping-00 10:20 / 10:28 George Swallow, 10 min Lucy: what is "IP based LB, imposing entropy"? George: a strange corner case Curtis: "imposing" is a bad work. You should rather say, "adding the EL". George: yes Curtis: also, there are more cases that these 4. It is not "mpls or ip". It Might be both. George: this is why I thought "IP based" was badly choosen. draft-ryoogray-mpls-tp-psc-itu-00 10:30 / 10:42 Eric Gray, 15 min no question draft-ravisingh-mpls-el-for-seamless-mpls-01 10:45 / 10:50 Ravi Singh, 10 min Curtis: it is stitching. if you swap you can not change the EL/ELI Lucy: the title of your draft is misleading it say mpls, but it does not apply to mpls-tp Nobo: I'd like a manageability section added Sharam: you can't remove the label if you swap Curtis: this illegal George: to the list draft-chen-mpls-source-label-01 10:55 / 11:04 Mach Chen, 10 min Sharam: I do not think this is good idea. You can use a source identifier in OAM or a separate LSP for the use case you describe Mach: but not for real traffic packets Curtis: source can be inferred from set-up (signalling or management) Lucy: this is too complicated draft-li-ccamp-auto-mbb-te-path-00 11:05 / 11:14 Zhenbin Li, 10 min Sharam: TL/TA is very static. Why do you need to distribute it w/ IGP? Eric O.: I would like to understand why can't you do this using link attributes. draft-xu-mpls-el-capability-signaling-igp-00 11:15 / 11:26 Xiaohu Xu, 5 min George: EL is not necessarily the property of a whole router but only of a line card *** end of Tuesday's session *** draft-chen-mpls-p2mp-ingress-protection-09 15:50 / 15:51 Huaimo Chen, 10 min Greg (on slide 3): A failure on one bfd session will not necessarily cause the other to fail Huaimo: if node fails the two sessions will fail Greg: missing security section in document while source and ingress are in different domains Eric: there are cases you have not explicitly considered. for example what happens if more than 1 link CE3/PE5? draft-xu-mpls-multi-domain-deployment-enhancement-00 16:00 / 16:06 Xiaohu Xu, 10 min Eric O.: if no new technology, what is there to standardize? Xiaohu: this just like seamless mpls Eric O.: it is a bit light Bruno D.: you identify issues. these are solved with seamless mpls and BGP anycast. Please read seamless mpls draft George: Also if summarized bgp adv, how do you know a host has gone? George: will take rest to the list draft-cui-mpls-tp-mfp-use-case-and-requirements-00 16:10 / 16:15 Zhenlong Cui, 10 min Greg: 1:1 and 1:N exist already. is this M:N? Zhenlong: it is M:N by 1 George: are you proposing a new PSC protocol? Zhenlong: no Himanshu: this is M:1. Should be covered under the umbrella of M:N Eric O: will use PSC or a new protocol? Zhenlong: PSC George: please continue discssion on the list draft-hmk-mpls-tp-p2mp-oam-framework-03 16:20 / 16:24 Yoshinori Koike, 10 min Sharam: how can you have a p2p pw in p2mp lsp? Yoshi: should say p2mp-ss-pw Sam A.: a lot of sections are indicated as FFS George: it means OOS. draft-mirsky-mpls-residence-time-00 16:30 Greg Mirsky, 10 min Curtis: there was a discussion in tictoc on the need for internal ttl if node does not support this, it will back hole packet rather than giving a wrong time. if it does support, it needs an internal ttl to avoid loop. George: if going out to RP you can't use this to measure queueing delays. Greg: going to RP is implementation detail if you do not support to functionality it is likely you'll go to RP but otherwise it is likely that packets will be timestamped on ingress and hen kicked out of fast path but not to RP, so can use it for RT measurement George: you said that it can be used for other applications than RT but this is not necessarily true Open microphone Yoshinori: our document is ready for adoption Ross: is an update planned before asking that? Yoshinori: yes Stewart: are you going for P2MP-SS-PW over P2MP-LSP, is that correct? Yoshinori: yes Himanshu: P2MP-SS-PW over P2MP-LSP saves you ingress replication Sam A.: we are working on an update on draft-ietf-mpls-tp-smp-requirements *** end of Wednesday session ***