L3VPN WG Meeting Minutes - Philadelphia March 10, 2008 Rick Wilder/Ron Bonica ? chairs Scribes: Cengiz Alaettinoglu and Andrew Lee, no Jabber scribe was available. WG Status Most of the WG work items have been completed except for a few on Multicast VPNs. Particularly, no work item is in the RFC editor queue. There are some new proposals for which the charter does not need to be updated. T. Morin - draft-morin-l3vpn-mvpn-considerations-02 Thomas presented changes to his draft which is now one year old and asked for the next version (-03) of the draft to be considered for WG adoption. Changes were mostly of clarification nature. Eric has provided some feedback few days ago and the authors didn't have time to address them yet. This was followed by a non-technical discussion between Eric Rosen and Thomas and few others whether this document is the right approach at all. Eric doesn?t believe the document is doing a balanced comparison of options and believes a new document is needed. The authors still want a core set of mandatory procedures to be defined to support multi- vendor networks. Thomas agreed to incorporate some of Eric?s comments, but said the authors did not agree with all. Ron questioned whether the number of options in multicast-VPN could be reduced just by having a meeting and coming to consensus. Eric stated that agreeing to disagree makes many options, but if we want to make progress we will have to agree to disagree. The possibility of defining a small number of profiles was suggested as one way to deal with disagreement. There was also a suggestion to just document which combinations of options are not compatible. Yakov Rekhter stated that no IETF documents define profiles. Attempts to advance the discussion of the recently raised technical issues did not go far at this meeting, so must be resolved with discussion on the mail list and document updates by the authors. Ron advised to focus on the parts of VPN multicast where we can reach consensus. M. Napiera - draft-mnapierala-mvpn-part-reqt-01 Maria presented second version of her mVPN partitioning requirements Document. The main requirements are to support anycast sourcing inMVPNs where packet duplication both at the customer networks and in service provider network are avoided. Additionally, the solution should not restrict the multicast tree in anyway such as requiring RPs to be in SP network, or deploying MSDP, forcing customers to switch to source trees. Supporting PIM Bidir with source only branches was also a requirement. There were no comments from the WG. M. Napiera - draft-mnapierala-mvpn-rev-04.txt Maria presented changes to her draft. The changes included both new work and clarifications. The new work includes PIM-SM inter-PE procedures, S-PMSI aggregation and source-specific host reports in PIM- SM. She then proceeded with highlights and summary of its operation. She is not sure what to do with the document next as she is not a developer she cant do much more with the document. B. Davie - davie-tsvwg-rsvp-l3vpn-02 Bruce presented changes to his draft on supporting customer RSVP sessions over an L3VPN cloud. The work now only focuses on admission control aspects of RSVP, not TE aspects. The customer RSVP messages are tunneled over the provider backbone. The main change was the introduction of VPN-IPv4 and VPN-IPv6 VPN address families in RSVP session objects. This enables removing vrf-id and vpn-label objects, supports option B of inter-as L3VPNs, and cleans up RSVP session object semantics. The document is close to completion. Since the proposal changes RSVP, the work will be done in the TSV WG. For this, TSVWG wants a statement from the L3VPNs WG that they support the proposal. Lars Eggert requested a WG consensus from the room as TSVWG would not want to do the work and then have L3VPN WG change to a different solution for CAC. There was a clear consensus in the room to support the solution and provide the required nod to the TSV WG. Ron asked Kenji if Bruce's work meets the requirements Kenji has compiled. Bruce and Kenji agreed that it meets the requirements on the admission control aspect, though the Kumaki document also covers TE aspects. Yakov mentioned that with this work plus the GMPLS UNI RFC (which covers RSVP between CE and PE), the TE requirements may not be too hard to achieve. Bruce favored to move this document forward instead of slowing it down for TE. Bruce did agree to provide the author?s analysis of what is needed for TE, however. K. Kumaki - draft-kumaki-l3vpn-e2e-rsvp-te-reqts-06 Kenji presented his requirements document on Customer RSVP sessions over L3VPNs. It covers both admission control and TE aspects to provide end-to-end QoS. It is intended to support delivery of triple play services. He lists 21 requirements in his draft and went over few of them; the main ones being to avoid scalability issues in the SP networks, and forwarding native IP packets or labeled IP packets between CEs. He said that his document is ready for WG adoption. There was a discussion mostly between Yakov and Kenji on what scalability meant in this context. That is whether it means just the P routers do not maintain state about customer RSVP sessions or PE routers as well do not maintain such state. This issue will be clarified in the mailing list. It looked like some more work is needed and there was not a clear consensus in the room about adopting the document. Ron will bring the issue to the mailing list.