***************************************************************** Minutes of the MPLS WG Sessions of IETF 81st Meeting. ***************************************************************** Recordings can be foud at: http://www.ietf.org/audio/ietf81/ For Monday morning session: http://www.ietf.org/audio/ietf81/ietf81-206b-20110725-0855-am.mp3 For Wednesday afternoon session: http://www.ietf.org/audio/ietf81/ietf81-206b-20110727-1256-pm.mp3 ***************************************************************** Admin & WG Status Chairs, 20 min Loa (on draft-ietf-mpls-mldp-in-band-signalling): no feedback given to Martin's e-mail Ice: ready for WG LC Loa (on draft-ietf-mpls-tp-csf): on hold, looks like we have all we need in other drafts draft-ietf-mpls-lsp-ping-mpls-tp-oam-conf-02 Elisa, 5 min Ross: How many people have read? Out of these how many think ready for wg lc? much smaller set of hands draft-ietf-mpls-tp-li-lb-02 Sami, 10 min no question/comment draft-ietf-mpls-tp-security-framework-01 Luyuan, 5 min (09:40) 09:22 > 09:27 Greg M.: I'll send editorial comments to the list. Reference model 2b. It mentions that the trusted domain is SPE SPE is always understood as a single node. Not sure this is the intent. Luyuan: yes it is. Greg M.: On models 1a and 3; not much difference between the two except a PE router in the middle of the drawing. Luyuan: difference is on inter-provider Vs. single draft-ietf-ccamp-mpls-tp-rsvpte-ext-associated-lsp-01 Fei, 10 min Deborah (as CCAMP co-chair): CCAMP asked that it be presented here, also. want to make sure that the mechanisms developed in ccamp cover the reqs developed in mpls wg both provisioning models (single ended and dual ended) are covered. will be discussed tomorrow in ccamp no question/comment Loa (to Deborah): in what form do you want the response from mpls and what is the time-frame? Deborah: Draft is maturing, would like inputs very soon Loa: we (mpls & ccamp chairs) will produce a mail towards the list George: mpls folks can also go to ccamp list directly Deborah: I'll ask authors to ping both lists draft-ietf-mpls-tp-te-mib-00 Sam, 10 min Kam: In many places you equate ICC & Global Node ID These are not the same Sam: ok Greg M.: In intro you do not mention that Sections are in scope Sam: they are not Greg: only mention bidirectional lsp not unidir nor p2mp Sam: we support all types Greg: then it should be mentioned in text draft-vkst-mpls-tp-oam-id-mib-00 Sam, 5 min Ross: do you want to come up with update and then ask for WG adoption? Sam: yes draft-raza-mpls-ldp-olf-00 Kamran, 10 min Ross: do you intend to ask for wg adoption now? Kamran: after revision Tom: Did you think about the case of really messing up with filtering? Kamram: user has to be cautious, no different that with other policies (Wee Lee?): no real need to introduce such heavy filtering mode. use on-demand Kamram: but also drawbacks in that case, as indicated in slide Jeef Haas: support comment that DoD is the right place to do this. Acee: in e-BGP, routers are typically under different admin control while here, ldp routers would be under same admin control, so less of a requirement to do the dynamic filtering in ldp Luca: would have to agree with these comments do not see the need to add this feature to ldp Kamram: the bindings get multiplied by the number of neighbours, that is getting a lot Eric R.: DoD is definitely the proper technology for that draft-beckhaus-ldp-dod-00 Maciek, 10 min Eric R.: nice document but I do not think we want to use DoD for that ?: DoD is already well documented, no need for another draft describing it also in your uses cases you should not look only to DoD but also DU and mixed DoD+DU Maciek: DoD Vs. DU is covered as part of negotiation. Also, the idea of LDP DoD for seamless is not new! Jeff Haas: Convergence is an issue. DoD has that issue, problem is not necessarily filtering nor DoD Issue is informing the origin of LSP that network is interested in that set of things Suggestion to both sets of co-authors: consider the idea of filtering, consider the idea of doing it on demand but also consider the idea of transitivity of filtering Sami: do you plan on adding filtering to DoD? Maciek: no Luca: any req for fast convergence? Maciek: link/prefix/fec DOWN event are covered in draft Luca: but is there a requirement to come up quickly? Ross: take this offline draft-ezy-mpls-1ton-protection-00 Eric O., 10 min Ghani: you mention lock but what about mis-connection? Eric: I do not see how can there be mis-connection Greg M.: more generic than 1:1, did you consider doing M:N as it is even more generic? Eric: we could do, but we focussed on satisfying requirement seems you have case in mind, please talk to us Kam: there is 808.1 (common aspect of linear protection) docs should be consistent Eric: we looked at 808.1 Jia: seems new state machine compared to 1:1. Eric: same state machine, George: there are more states, this is a bigger machine Jia: changes in PDU also Eric: yes Jia: knowing that 1:1 is not finished why not merge everything Eric: 1:1 is very near to rfc Jia: should nevertheless make sure we have same mechanisms Dave A.: struggled with use case Eric: you are questioning 1:N in packet world? Dave A.: yes George: there is a requirement Ross: will you ask for adoption now? Eric: at next revision Ross: encourage people to read it draft-fuxh-ccamp-delay-loss-te-framework-00 draft-fuxh-ccamp-delay-loss-rsvp-te-ext-00 Xihua Fu, 15 min Acee: originally your draft was specifying TLVs and so on specific to g709 while this document is now more framework style Are you backing off wrt to requesting an exact match with G709? Xihua Fu: ... Greg: did you come across any requirement towards oam and p&r as, for example, switch-over is source of packet loss? Fu: yes Greg: did not see any requirement on protection can provide text on this topic for you to consider for your next revision Ghani: what about impact on delay/loss of the protection path Vs the working Fu: protection needs to also meet the SLA draft-villamizar-mpls-tp-multipath-te-extn-00 Curtis, 5 min Eric O.: you cover multiple parallel links, not multiple paths Curtis: yes, but you could if you sere to bundle 2 LSPs for example draft-raza-mpls-ldp-applicability-label-adv-01 Kamran, 10 min no question/comment draft-napierala-mpls-targeted-mldp-01 Eric R., 10 min no question/comment *** end of monday's session *** draft-kini-mpls-frr-ldp-01 Sri, 10 min George: have you presented it to RTGWG? Sri: it is today. George: you are not on the agenda but there is plenty of time. ?: do we also have to assume that the backup LSPs must be installed in the forwarding table Sri: yes, but in most case this is shortest path lsp, so ne new installation George: take pending questions to the list draft-bhatia-mpls-rsvp-te-bidirectional-lsp-01 Lizhong, 10 min Adrian: we discussed feature creep in rsvp-te from gmpls, at that time mood was that we should not feature creep we should switch over just want to make sure wg does this eyes wide open Lou (to chairs & AD): where do you want to draw the line of features that you bring from gmpls into mpls? in other words, why not use gmpls George: my take is that we will be developing signalling for mpls-tp in CCAMP and that looks like the same thing, except possibly for php. there should be a single effort Lou: I guess I agree w/ you George draft-martinotti-mpls-tp-interworking-02 Dave A., 5 min Daniel Cohn: this is something WG should be working on Luyuan: this needs to be aligned with use case and design considerations document Loa: if authors interested working together, please go ahead Nurit: this document should also integrate the migration doc draft-li-mpls-mt-applicability-requirement-02 Li Lianyuan, 5 min no question/comment draft-zhao-mpls-ldp-multi-topology-02 Quintin Zhao/Li Lianyuan, 5 min George: how many service provider interested in deploying? George: how many people in the room are interested in pursuing this? George: take it to list draft-chen-mpls-6pe-mib-02 Gang, 10 min George: how many read? Loa: few Loa: need to coordinate w/ L3VPN if we go ahead draft-fang-mpls-tp-oam-consideration-02 Lianyuan Li, 5 min (13:50) 13:40 > 13:42 no question/comment draft-fang-mpls-tp-use-cases-and-design-02 Luyuan, 10 min (14:00) 13:43 > 52 no question/comment draft-rkhd-mpls-tp-sd-03 Daniel Cohn, 5 min (14:05) 13:52 > 13:59 Kam: serious concern if SD is triggered/declared by lower layer; should be in same layer George : Please continue your discussion on the list draft-koike-mpls-tp-temporal-hitless-psm-03 Yoshinori, 5 min no question/comment draft-cheung-mpls-tp-mesh-protection-03 Taesik, 10 min Ping: deja vu all over again. this does not work. packets can not be assumed to be delivered Taesik: this is a transport network George: We're out of time. Take it to the list. draft-chen-mpls-p2mp-ingress-protection-03 draft-chen-mpls-p2mp-egress-protection-03 Ning So, 15 min ?: in egress protection case: do you assume the egress router knows where the penultimate hop so that he can compute path there ?: for inter area/as so we need PCE support Greg: more scalable to pw redundancy? Ning: from my perspective, with network in mind, better that pw but need to hear of your case Eric O.: for the PE-CE coordination wouldn't you need to coordinate w/ L3VPN Ning: chairs? chairs: need to thing bout it draft-farrel-mpls-tp-mip-mep-map-04 Rolf, 10 min draft-ali-mpls-inter-domain-p2mp-rsvp-te-lsp-06 Rakesh, 10 min ?: any scenario where crank-back would not converge? Rakesh: there are some indeed ?: how do you handle that Rakesh: data plane solution is much better thus signalling extensions to take advantage of it Lou: not explicit in draft what holes there are in exist signalling you should document them better George: strongly advise the authors to have a discussion w/ lou & john draft-pdutta-mpls-mldp-up-redundancy-00 Wim, 5 min Ice: we have a similar doc, presented to rtgwg. Wim: yes, we discussed. our doc is mldp. while yours is pim suggest to merge draft-ldp-ila-extension-00 Alton, 5 min Rob Andersson: when too many changes you do not to restart, you reinitiate sessions Alton: sender will re-advertise all its local label bindings ?: so you now need to checkpoint the version numbers. what is the impact on scalability? Alton: you can aggregate draft-pan-shared-mesh-protection-02 Ping, 5 min Nurit: shared mesh is very important but you look to trivial cases Greg: you make exactly the same assumption that previous draft on protection i.e., message are delivered reliably while they are not unless you have an ack George: I agree with Greg George: How many read? (good show) George: how many like it? (same) George: need socialization. would like to avoid multiple drafts