PCN Meeting Notes Friday morning Mar 27/09 Scott Bradner chairing minutes by Tom Taylor and Kwok Ho Chan Architecture Status -- Phillip Eardley ============================== draft-ietf-pcn-architecture Phillip Eardley: He is currently clearing IESG DISCUSS comments -- primarily editorial, clarifications Lars Eggart (as AD): three new ADs have to state their opinions. Outgoing AD comments no longer count -- incoming ADs will review and decide whether they hold that position. Because the changes have been substantial, new ADs will be encouraged to review from the start. Suggests Phillip talk to broadest commenter. Tom Taylor: pcn should have a common "elevator pitch" for PCN at the start of every WG draft (Abstract and Introduction). Kwok Ho Chan noted that there is material in the Problem Statement draft that was used to start the WG can that be reused for this purpose. Tom has action to draft. Marking Behaviour -- Phillip Eardley ============================ draft-ietf-pcn-marking-behaviour Phillip Eardley: Would like to go to WGLC. SATOH Daisuke has called for delay. SATOH Daisuke: Without a third code point, termination and admission are inaccurate. He said that he had ideas for improving accuracy, but needs more time (measured in weeks) to simulate them. Phillip Eardley: have to make a decision at some point. We already cut off another proposal that had asked for more time and had not come through within the period given. Lars Eggart: does Daisuke believe current approaches don't work at all or work less accurately than can be achieved? SATOH Daisuke: Only have Single Marking according to the intended standard. Phillip Eardley: operator has deployment choices. Daisuke is concentrating on a single case. Bob Briscoe: Daisuke's proposed marking behaviour would preclude migration to three-code-point marking. Giorgios Karagiannis: as target of last year's closure, he accepts that it is time to put something on the ground and experiment with improvements after we have experience. Lars Eggart: we don't intend to have multiple marking behaviours. Chair: reasonable to go to WGLC. This discussion can be part of the WGLC discussion. Baseline Encoding -- Phillip Eardley ============================ draft-ietf-pcn-baseline-encoding Tom Taylor: he put some of the encoding discussion into the behavior draft Phillip Eardley: Talking about "What's Next" slide. Code point transitions are specified in each encoding document. Tom noted that also summarized in the edge behaviour "core assumptions" sections. Re ECN tunneling draft: thoroughly rewritten -- more report later. Last call discussion: Gorry Fairhurst noted that lots of editorial nits, but OK technically. Chair: will take to WGLC after nits addressed. Edge behaviour drafts -- Tom Taylor (for Anna) ============================ Tom Taylor: Will be submitted as WG documents after next revision (in line with discussion results). Bob Briscoe: these drafts are informational drafts and they are informational on some experimental ideas Ken Carlberg: he have a draft on Emergency and that scares people, hence Ken's draft will pull out the notion of "Emergency". Tom Taylor: the decision point should be in the loop for flow terminations Tom has action to write up draft on "How to write an edge behaviour draft", incorporating by reference "Per domain behaviour" draft (reference to be obtained). CL and SM drafts should be updated to include per domain behaviour. Bob Briscoe: template draft should reference Kathie Nichols's PHB template RFC 3086 Three state encoding -- Phillip Eardley =============================== Phillip Eardley: Time to look at experimental extensions. Toby Moncaster: "Basic 3 state" and "Extended 3 state" encodings from the baseline draft presentation. "Extended" is used to carry ECN marking across the PCN domain. Bob Briscoe: "3 in 1" based on premise that tunnel behaviour is suitable. No tunnels, or tunnels end in your network or at its edge. PSDM has dual mode. Overwrites ECN. Proposal is that all three become WG documents. Giorgios Karagiannis: supports. Wants to work on HOSE model after basic set done, hence would need the "Affected Marking" codepoint. But this would be after the initial set done. Gorry: are the two DSCPs related or just arbitrary? Giorgios -- should be related, same PHB. Kwok Ho suggests AF. Lars Eggart: logically related obviously. But (general agreement) no need for a numerical relationship. Second DSCP would have to be reallocated, but evidence of successful experiment would be strong argument in favor. Bob Briscoe: are we aiming to have a few experimental proposals and then narrow down to one result, or just leave it to local use? Lars Eggart: if we go to multi-domain, experimental is going to be interesting because of potential clashes with other experiments. support in room that PSDM, Moncaster 3-state, and 3 in 1 become WG drafts, aiming at Experimental. Jabber comment (Steven Blake): Wiki available for public notification of PCN experiments. Further comment on use of Pool 1 vs. Pool 2, which will not be allocated until Pool 1 is exhausted. [Follow-up note to list] Lars Eggart: PCN should also start a wiki page on IETF site indicating which DSCPs are used for PCN experiments. Tunneling of ECN -- Bob Briscoe =============================== draft-ietf-tsvwg-ecn-tunnel-02.txt Bob Briscoe: Layered encapsulation. Intended to be Standards Track. Summary on charts. Proposal: At ingress, bring ECN tunneling in line with IPSec. At egress: use two wasted combinations of inner and outer codepoints. Absolutely no backward compatibility issues. Gorry: OK on merits. But assumes ECT(1) is more severe than ECT(0). Is this the intention? Seems consistent with others' suggestions. Kwok Ho: ECT(1) usage new? Gorry: ECN nonce doesn't have this concept, but original ECN did. Chair: basically this is a TSVWG discussion. Bob Briscoe: Discussion on TSVWG list: don't necessarily need to drop illegal combinations. Gorry: actually now distinguishing (0) and (1). Have to think through the implications. Kwok Ho Chan will review the draft. PCN Encoding Comparison - Kwok Ho Chan ====================================== Kwok Ho Chan: Charter work item. Tracks reasoning leading to current positions Question is whether this draft should be continued or closed? Chair suggestion: output as Informational now, can update later. Make it a Working Group draft. No further points. Finished at 10:45.