TSVWG Agenda IETF-84, Vancouver, BC, Canada Wednesday, August 1st, 2012 Room Regency B Time: 1510-1710 local time (PDT) WG Chairs: James Polk Gorry Fairhurst Rolf Winter Notes from David Black ----------------------- * Chairs Agenda Bashing (15 min) NOTE WELL Document Status and Accomplishments Charter & Milestones Review Agenda bashed - moved ITU-T Liaison before diffserv material. Document status reviewed. ------ * Working Group Drafts ** Georgios - RSVP support for PCN (10 min) *** draft-ietf-tsvwg-rsvp-pcn Status report. This draft will need to become an experimental RFC because the base PCN RFCs (6661 and 6662) are also experimental, and they're normative references for this draft. ------ * Non-Working Group Drafts ** James - RSVP App-ID Profiles (10 min) *** draft-polk-tsvwg-rsvp-app-id-vv-profiles Goal: have important voice and video be higher priority than other voice/video (and other traffic). This draft adds traffic type info to RSVP reservations, using the existing template from RFC 2872. This version has been revised in response to a couple of good reviews, James thinks the draft is effectively done. The mmusic WG has a draft that depends on this draft - that mmusic draft is likely to go to WG Last Call in early 2013; James would like this draft to become a WG draft and go to tsvwg WG last Call in the same timeframe. Will take topic of whether to make this a WG draft to the list. ** Al Morton's (10 min) *** ITU-T Study Group 12 Liaison on QoS Classes & markings for interconnection Al Morton is speaking as ITU-T SG12/Q17 rapporteur. SG12 has good service provider representation. Y.1566 on QoS mapping and interconnection. This is about marking traffic on interconnection links between carriers. Appendix contains an informative mapping between traffic classes and codepoints. Intent is to not change DSCP in (inner) IP header, but use these values in a 3-bit codepoint. Use of AF31, AF32 and AF33 on a flow-by-flow basis does not appear to be consistent with definition of AF PHB group in RFC 2597. AF31, AF32 and AF33 are a single AF class in RFC 2597 and are treated as constituting a single PHB. Eliot Lear (as IETF liaison manager): Liaison reply is due in March, there's time to prepare it. He suggests that a small group develop a draft reply and work it forward on the list or in the Atlanta meeting. ** James - DiffServ Standards Problem Satement (30 min) *** draft-polk-tsvwg-diffserv-stds-problem-statement See slides. Comment: Need to get this right, as TCP implementations now send enough traffic to cause problems (they didn't back when DiffServ was first standardized). Class selector codepoints were specified for backwards compatibility - they're not necessarily specified to the same level of behavior as other PHBs. Also, selection of four as the number of AF classes was a somewhat arbitrary choice at the time. A lengthy discussion ensued. Summary of interesting points for moving forward: - We may be looking at a tunnel structure where the RFC 4594 classes and markings are used in the inner IP header (can be viewed as enterprise-oriented), but a much smaller number of classes/markings (e.g., see RFC 5127 and ITU-T liaison) are used in the outer MPLS header's TC bits (can be viewed as carrier- and inter-carrier-oriented). That smaller number of aggregates could also apply to the Ethernet priority field. - The ITU-T liaison can be read as also asking if the treatment aggregates in RFC 5127 are the right ones. This may be much more important than whether the right RFC 4594 classes are in each RFC 5127 traffic aggregate. - In addition, RFC 5127 needs some more detailed work, e.g., as VOICE-ADMIT isn't included, but figuring out whether RFC 5127 has the right treatment aggregates is more important. - The response to the ITU-T liaison is not due until next March so there is time to consider the issues. ** - Update to RFC 4594 *** draft-polk-tsvwg-rfc4594-update ** - Assigning Additional Realtime DSCPs *** draft-polk-tsvwg-new-dscp-assignments These two drafts were not presented - the agenda time was spent on the problem statement discussion. ----- ** Piers - Reactions to Signaling from ECN Support for RTP/RTCP (10 min) *** draft-carlberg-tsvwg-ecn-reactions See slides, doc will stay on hold (not will be a WG doc immediately). This draft is related to the new rmcat work (BoF in Vancouver). ** Piers - Congestion control algorithm for lower latency and lower (10 min) loss media transport *** draft-ohanlon-rmcat-dflow See slides, also related to rmcat. ** Zhu Lei - Flow-based Performance Measurement (10 min) *** draft-sun-tsvwg-flowbased-pm See slides - draft is related to ippm WG. Comment: Can't assume that network doesn't do reordering.