0/ MPLS WG Status (20 mins) Chairs, 20min Adrian: putting Ross Callon as co-chair indefinitely Ice (on mldp-in-band & -recurs-fec): Will be ready for LC soon Ice & Thomas (on draft-ietf-mpls-ldp-p2mp): discussion points resolved. can move forward. George (on the tp-identifiers): one major issue tunnels & lsp have either IP or ITU identifiers on both ends. request to have different types at the two ends 1/ WG documents (mpls) (17 mins) draft-ietf-mpls-ldp-p2mp Ice & Thomas, 2min cancelled, see WG status draft-ietf-mpls-ldp-ipv6 Rajiv, 10min Questions/Poll by Rajiv: how many have read the draft? about 1/3 of the room 3 main questions: should LDP prefer v4 or v6 for session peering if dual stack interface? current: v6 preferred should label space be shared between v4 & v6 bindings current: yes should gtsm be default for ldpv6 poll on 1: ~20 hands equally distributed poll on 2: 10 hands on yes, 1 on no poll on 3 : 10 hands on yes, 0 on no draft-ietf-mpls-lsp-ping-ttl-tlv Sami, 5min [missed]: suppose inter-AS case, should P routers care about ttl-tlv? Sami: no [missed]: need to spell out clearly if you see or not the ttl-tlv 2/ WG documents (mpls-tp) (78 mins) draft-ietf-mpls-tp-rosetta-stone Huub, 5min Loa: aren't we missing the survive docs from which we take definitions? Huub: will check draft-ietf-mpls-tp-oam-analysis draft-fang-mpls-tp-oam-toolset Nurit & Luyuan, 15min -no question/comment- draft-ietf-mpls-tp-cc-cv-rdi Dave, 8min Nurit: any intention to indicate capability not to use the negotiation? Dave: that's the fast start. covered Sasha: do you plan on having your resolution review by BFD WG? Dave: yes draft-ietf-mpls-tp-on-demand-cv Eric, 10min Loa (comment on IANA): you get a pb if pick a value out of a registry and put it in a draft and somebody else does the same. So do not select a value just put TBD. draft-ietf-mpls-lsp-ping-mpls-tp-oam-conf Elisa, 10min -no question/comment- draft-ietf-mpls-tp-csf Jia, 10min Luyuan: you suggest a value of 3.3ms and another one much larger, is this a range or 2 fixed values? George: In fact there is a field but the draft does not say exactly what's in. Luyuan: should be configurable, and you can skip the recommended values [Shahram?]: how is this draft consistent with the requirement that say CSF should be used when no AIS is used. Jia: this is mentioned in the draft ... Ping: why did not you put this in bfd? Jia: you should ask the authors of bfd Nurit: [missed] Sasha: bfd has diag codes to carry AC failure status. could this be used? draft-ietf-mpls-tp-mib-management-overview Daniel, 5min -no question/comment- draft-ietf-mpls-loss-delay draft-ietf-mpls-tp-loss-delay-profile Dan, 10min -no question/comment- draft-ietf-mpls-tp-li-lb Sami, 5min [missed]: does function apply to PWs? Sami: yes. PWE3 group reviewed doc [missed]: why limit to statically provisioned? Sami: if control is here, should be done by control plane. Loa: plan to have a doc that clarifies oam role in cp environments Eric O.: need a new lb mode "everything but oam" (aka leaky loopback) Sami: we can target oam to the node that is in loopback Lou: do you really plan on adding this feature in cp? We must be careful not to duplicate functions Loa: I did not say that we "should add" Sami: me neither George (hat off): this is a data plane function 3/ (30 mins) draft-chen-mpls-p2mp-ingress-protection draft-chen-mpls-p2mp-egress-protection Huaimo, 15min Nurit: which existing solutions are you referring to? looks like a lot of layer mixing Huaimo: existing solution is what we have today. Nurit: this is what I suspected, this is a client/server model this should be spelled out in doc [missed]: how do you protect nodes in the middle? Huaimo: with frr [missed]: there is no p2mp-frr Huaimo: can use it in p2mp context Eric: for links only, not nodes draft-villamizar-mpls-tp-multipath Curtis, 15min George: very detailed draft that need to be more socialized on the list Stewart: agree 4/ Individual I-Ds (147 mins) draft-leymann-mpls-seamless-mpls-03 Nicolai, 5min Asks for WG adoption. Loa: How many have read it? [~20p] Those who had read it how many think it is ready for WG adoption? [approx same number, take it to the list] draft-raggarwa-mpls-seamless-mcast-03 Rahul, 15min -no question/comment- Asks for WG adoption Loa: How many have read it? [quite a few] Those who had read it how many think it is ready for WG adoption? [about the same number. we need to poll it on the list] draft-vkst-mpls-tp-te-mib-00 Tom, 10min Asks for WG adoption Kam: missing node ID if you want to use ICC as global unique identification mean draft-kini-mpls-ldp-hierarchy-00 Sri, 10min Eric O.: looks like OSPF carried in LDP Shane: next hop indirection implemented in hardware. why is that considered not good enough for fast restoration Luca: consider doing aggregation at IGP and advertize label for subnet Loa, Ross: take all discussions to the list draft-kini-mpls-frr-ldp-00 Sri, 10min George: What topology analysis have you done? Particularly rings with asymmetric metrics on links? Are there micro-loop problems with updates being sent when links fail? Sri: No, because the actions are pre-failure, but I we can discuss this offline. Luca: We have RSVP FRR, IP FRR, how is this different why do we need it? Sri: RSVP has a lot of associate complexity. Luca: And this is not complex? How do you constrain the path? Sri: This does normal LDP path setup, not CR-LDP path setup. Luca: You need a tunnel the bypass link? Sri: No, that’s another LDP LSP. Luca: You are using the link state DB from whatever is running? Sri: Yes. Stewart: You mentioned performing computation with links removed, which in hop-by-hop system is a problem. This may cause potential micro loops. George: I would strongly suggest the work is progressed in the RTG WG. Stewart: I agree. Especially given the history of fixing hop-by-hop FRR problems draft-raza-mpls-ldp-ip-pw-capability-00 Sami, 5min -no question/comment- draft-asati-pignataro-mpls-ldp-gtsm-01 draft-asati-pignataro-mpls-ldp-iana-01 Rajiv, 10min Rajiv: Anyone read the draft? [Many] Vero: I want to make it clear; this could only be applied to basic hello and targeted hello. Rajiv: Yes. This is applicable for directly connected neighbours. Vero: We need a security mechanism for LDP. I discuss this later. Luca: Apologies but I did not read the draft. What happens if you reroute the connection? There are some corner cases we handled a few years ago, with local attached Ethernet with one hop and backup connections. This assumes you’re going link only and transport address is link interface IP address. Rajiv: This does not assume that the address is the link interface address, but if you want the LDP session to be preserved, then we literally get into the multi-hop LDP session. If you lose loopback it will break. Luca: Maybe your transport address should be the interface address. Rajiv: If you use loopback addresses, it will break. So that is reasonable. George: Are customers requesting this? Rajiv: Yes. draft-zheng-mpls-ldp-hello-crypto-auth-01 Vero Zheng, 10min Luca: [Slide 2] the problem statement, you mention no security exist for hello mechanism. Most implementations have automatic filtering; only accept packets from a source address that is known. Is this proposal really necessary? Vero: This is a customer requirement. Luca: Well, do you implement the automatic filtering for the hello packets? You should also filter at network edge. Curtis: What Luca is describing is the best current practice filtering, and there is also the KARP WG. In KARP they insist that the protocols have inherent security. Alia: There is an Ops Sec draft that describes filtering to secure routers. You may want to read it. Luyuan: We have RFC5920 for GMPLS and MPLS security, including IP spoofing. It points to BCPs. In practice LDP has been used for 12 years, if you implement the rules then we have not seen a problem. Vero: Where should we take this work? Ross: Take a look at the other documents. CARP is tomorrow morning. If there is consensus to adopt the work, then we should consider which WG to progress it. Scott: On behalf Tom, the mailing list comments from Eric on why this draft is not describing a real problem. Vero: OK, I consider this to be similar comments to previous speakers. draft-zhao-mpls-ldp-multi-topology-01 draft-zhao-mpls-rsvp-te-multi-topology-01 Quintin, 15min Nurit: I read the RSVP proposal and I have a number of concerns for the solution. I do not believe it’s not relevant from a business perspective. There are also inter-domain examples, while this may be technically feasible, is there a business justification? The qok needs to be backed up with business cases. Eric O: We talked offline. I see what you’re doing with LDP. I do not agree with what you’re doing with RSVP. Regarding the two LDP approaches, Eric and George had context labels, which is the label stacking approach you describe. You should consider using the technology so it’s consistent. The Inter-AS scenario is a far reaching use case, not just a new label space for LDP. Jeff: As we know LDP follows IGP, so operational problems may occur. Quintin: We are looking at tradeoffs if no existing solutions exist. We are focusing on LDP. For the mLDP complexity the backup solution only requires default topology so the tradeoff may be worth it. George Polled Room. How many have MT deployed? [Result] How many are interested in this work? [Result] draft-liu-mpls-tp-p2mp-shared-protection-01 Guoman, 10min Matthew: We need some context for this work. We have not defined MPLS-TP P2MP LSPs yet. This was split out of the main architecture draft. Guoman: Yes, the MPLS-TP protection is simple. draft-pan-shared-mesh-protection-01.txt Ping, 10min Ping: Who would support it? [Good support] draft-xhk-mpls-tp-throughput-01 Xiao Min, 5min Stewart: Speaking as an author on loss/delay. Consider restructuring this into applicability. We should use existing measurement tools. Xiao: We could try to find a way. Loa: We need to investigate to see if some of this work should move forward. Stewart: We need a semantic tool, we could use live traffic tools. draft-martini-pwe3-status-aggregation-protocol-02 Luca, 10min Sasha: What happens if I cannot fit within the MTU? Luca: You could break it up into multiple pieces Sasha: Why do we need one more PW control protocol that does not run over TCP, also any congestion avoidance? Luca: Firstly not everything likes TCP for control protocols. This is supposed to avoid a full stack. Sasha: You still need TCP. draft-koike-mpls-tp-temporal-hitless-psm-02 Yoshinori, 5min -no question/comment- draft-fang-mpls-tp-use-cases-and-design-02 Luyuan, 7min Nurit: Maybe you want to use a top down approach to allow the higher-layer make a per-LSP decision. Luyuan: We are more concerned with the operation; the various layers have a protection. Some layers have mechanisms for protection, or routing to resolve issues. draft-kompella-mpls-entropy-label-02 John, 5min Stewart: I would assert that it does not work with MPLS-TP. Link bundling might want to use the entropy label. Does it matter if it’s in there for certain scenarios? Shane: ECMP is not applicable is not applicable to MPLS-TP architecture. Stewart: There seems to be new applications all the time. Is there any harm in allowing the entropy labels? This is not a blocking comment. Shane: John and Kireeti may have some input as well. Curtis: Good work, but maybe better to say nothing at this point, instead of saying no. Stewart: You could say nothing, or state something like “an MPLS-TP node operating in strictly MPLS-TP should not use it”. Curtis: Yes, maybe give some operational advice. Luca: What about MS-PW? Shane: Does not cover that scenario. Loa polled the room. How many read? [fair number] How many read the latest version? [same number] How many people think it should be a WG document? [need to take that to the list] draft-sprecher-mpls-tp-migration-03 Nurit, 15min Sasha: Did you consider publishing as a BCP instead of an Informative document? We should consider making it as BCP. Luyuan: Good point. Curtis: I think this is a good idea for a draft but its missing major sections. It’s still useful for community.