Minutes IETF103: rmcat
minutes-103-rmcat-01

            RTP Media Congestion Control (RMCAT) Working Group

                    IETF 103, Bangkok, 8 November 2018


Reported by Colin Perkins

  The RMCAT working group met once at IETF 103 in Bangkok, Thailand, in
  November 2018. The meeting was chaired by Colin Perkins.

  The meeting started with a welcome, introduction, and a review of the
  working group's status.  SCReAM and SBD were previously published as
  RFCs, NADA has completed working group last call and is waiting on the
  chairs to progress, and the coupled congestion control draft is in the
  RFC Editor queue (but see below). The evaluation drafts have completed
  working group last call with only minor issued noted, and the wireless
  tests draft is ready for WG last call.

  The evaluation criteria (draft-ietf-rmcat-eval-criteria) is not on the
  agenda, but was updated earlier this week. Joerg Ott briefly summarised
  the changes, and noted that the only issue remaining is to complete the
  security consideration. It was suggested that he look at the security
  considerations for eval-test for guidance on what this should contain.
  Once this is resolved, it is expected that there will be a short WG last
  call to confirm the changes are okay, before the draft is sent to the
  IESG.

  The congestion control feedback draft was discussed in AVTCORE earlier in
  the week. There was an attempt to implement this draft at the hackathon,
  with some success, but this also highlighted some issues. Jonathan Lennox
  brought one of those issues for discussion to the meeting: how should the
  feedback mechanism handle duplicate packets? Gorry Fairhurst and Mirja
  Kuehlewind commented, suggesting that the arrival time of the first copy
  of the duplicate should be reported.  Mirja also noted that, if any copy
  of the duplicate was ECN-CE marked, then a ECN-CE mark should be reported
  for the packet. Xiaoqing Zhu confirmed that this should work for NADA.

  The chair noted that the working group milestones are outdated. It was
  proposed to update the milestones as follows:
  - to submit the requirements and evaluation criteria drafts to the
    IESG to December 2018;
  - to submit the first draft of evaluation results and the first draft
    of the standards track congestion control algorithm in July 2019;
  - to submit the standards track congestion control algorithm to the
    IESG to November 2019, along with cc-codec-interactions and the
    framework; and
  - to remove the milestones relating to techniques to detect, instrument
    or diagnose failing to meet real-time schedules, since these are both
    redundant with other mechanisms.
  There were no objections to this plan, so the chairs will made these
  changes.


Update on draft-ietf-rmcat-eval-test

  Zahed Sarker gave an update on the eval-test draft, summarising the
  changes made as a result of working group last call comments. Most
  of these were uncontroversial. Gorry Fairhurst had comments on the
  revised security considerations, suggesting it be split into two,
  and made a little more specific and detailed. Gorry will propose
  text.

  An update will be submitted to incorporate this change, and a short WG
  last call will be issued to confirm the changes are acceptable, before
  the draft is sent to the IESG.


Update on video-traffic-model draft

  Xiaoqing Zhu gave an update on the video-traffic-model draft. She
  summarised the changes made as a result of the working group last
  call. There were no objections to these changes. A short WG last
  call will be issued on the list to confirm the changes are okay,
  before the draft is sent to the IESG.


Update on NADA implementation in Mozilla browser

  Xiaoqing Zhu gave an update on progress with implementing NADA in the
  Firefox browser and testing against Chrome. She summarised the changes
  made, and presented results from local tests on a wireless LAN, tests
  on a real-world path within the US, and tests on an international path
  (Austin, TX to Lausanne, Switzerland). These were using an unmodified
  receiver, sending standard RTCP feedback every second without the new
  congestion control feedback extensions, so performance was limited due
  to this, but otherwise NADA looks to work as expected.

  Magnus Westerlund asked why feedback was so infrequent, given that WebRTC
  supports higher rates. Xiaoqing noted that they're using the default from
  the browser. There was some discussion around how to modify this, and it
  was noted that there was progress with implementing the congestion control
  feedback extension in Firefox at the hackathon. Xiaoqing expressed the
  intent to test with higher rate feedback.

  Julius Flohr asked if the code would be made available. Xiaoqing noted
  that she hoped it would be, once it was complete and cleaned-up.

  The chair noted that it is great to see implementation experiences and
  testing of the candidates.


NADA implementation experiences

  Julius Flohr reported on an independent implementation of NADA in the
  Omnet++ simulator. He described the implementation and test scenarios,
  and presented some initial results. These show that his implementation
  performs very similarly to the NS3 implementation of NADA, indicating
  that the specification is in good shape.

  Sergio Mena asked for clarification on what was meant by the "perfect
  video encoder" used in these tests. Sergio and the chair thanked Julius
  for implementing and testing.


Issue with draft-ietf-rmcat-coupled-cc

  Julius Flohr has implemented draft-ietf-rmcat-coupled-cc-07 in Omnet++.
  He noted that the draft does not consider application limited scenarios
  in case of Active FSE, but does for Passive FSE, and as a result there
  are issues with multiple RTP flows with different priorities when
  application limited streams are present. He highlighted the problem with
  the algorithm, showing that it results in unfair rate allocation, and
  proposed a fix.

  Michael Welzl thanked him for experimenting with this, and noted that
  this is something that was removed to simplify the draft earlier. He
  is happy to make the proposed change to address the problem, although
  he noted a minor nit with the suggested fix.

  Julius and Michael will discuss further and propose text to fix this
  problem. Since the draft is with the RFC Editor, we'll work with our
  AD and the RFC editor to figure out how to incorporate the fix.


Other Business

  Gorry Fairhurst noted that the updated security considerations he's
  writing for eval-test might apply more broadly than the one draft, and
  asked authors of other drafts to review and consider them once available.

  Jonathan Lennox noted that the hackathon was successful, and asked if
  anyone wanted to repeat in Prague? People seem interested, so Jonathan
  and the chairs will follow up to arrange this.

                                   - + -