RTP Media Congestion Avoidance Techniques (RMCAT)
=================================================

IETF 101 / London, UK / 21 March 2018

Reported by Colin Perkins and Alan Ford.

## Administration and Working Group Status (Chairs)

   The chairs introduced the meeting and reviewed document status.
   SCReAM has been published as RFC 8298.
   Coupled Congestion Control is with the RFC Editor.
   Shared Bottleneck Detection is with the IESG.
   NADA has completed working group last call, and is waiting for the chairs
   to complete the write-up. 
   The GCC draft is waiting for an update from the authors.
   The requirements draft is with the RFC Editor.

   The evaluation and wireless tests are ready for working group last call,
   but blocked on the evaluation criteria draft which will be discussed later.

   The congestion control feedback draft will be updated to align with the
   RTCP feedback draft, once that is complete. 

   The framework and codec interactions drafts are on hold for now, waiting
   implementation experience and the conclusion of the other work.


## RTCP feedback for congestion control (Colin Perkins)

   Colin Perkins summarised the changes to the draft. 
   Mandate clock source, format, and rate. No objections.

   No report if no packets received for SSRC.
   Jonathan Lennox: why would this be a problem anyway? You wouldn't send
   in this case.

   Give guidance on what sequence number ranges to be included in the
   report. SHOULD NOT overlap.
   Jonathan: if big gap, makes sense to ignore entirely.
   Colin: if you get nothing, fire circuit breaker. If you lose one or
   two, not a big deal. If you lose lots maybe you should ignore. Needs a
   bit more work.


## draft-ietf-rmcat-eval-criteria (Jörg Ott)

   Jörg Ott reviewed the status of the draft and the open issues.

   Jain's Fsirness index: remove the open issue, say nothing about how to
   measure fairness, and consider this done. Zahed agrees with this.
   Xiaoqing Zhu asks if we have anything about fairness for evaluation? This 
   is only this, so if we remove it the draft doesn't define a fairness metric. 
   She suggests we keep the discussion of fairness. Some ongoing discussion.
   Matt Mathis: we don't have sufficient information to define fairness for
   RTP flows. Conclusion: Xiaoqing will send text.

   Loss generation model: suggest to evaluate using independent random loss, 
   but note that other options (e.g., Gilbert-Elliot model) exist and can be 
   used if desired. 
   Xiaoqing Zhu: agree. 
   Mark Handley: notes that TFRC had problems where the losses were correlated
   with the data rate, and this isn't modelled by independent random loss. He
   suggests that modelling the queue would be better than injecting random loss.
   Conclusion: evaluate using independent random less, point to other options,
   and add a note that there are further complexities that evaluation need to
   think about.

   Jitter models: 
   Zahed notes that this has a real impact in their test cases, and they use
   no-reordering bounded PDV. 
   Xiaoqing Zhu noted that NADA has been tested with NS2 using random jitter 
   (no reorder), and for NS3 only due to queuing delay in simulations. Agrees
   with Zahed that jitter influences algorithm behaviour, so we need to say
   something. 
   Conclusion: use no-reordering bounded PDV.

   The chairs noted that there has been some comments about an open issue
   around TCP models. It is now believed this is resolved, but the text needs
   review.

   Next steps: Xiaoqing Zhu will send text, and Jörg will incorporate it and
   make the other updates, and submit an updated draft. This will then go to
   working group last call.


## draft-ietf-rmcat-video-traffic-model (Xiaoqing Zhu)

   Xiaoqing Zhu reviewed the status of the draft and outlined the behaviour of
   the video traffic model. The draft is now aligned with their open source
   code implementation.  She also reviewed the status of the open source code
   implementation (https://github.com/cisco/syncodecs), outlining the codecs
   available there.  Finally, she review the status of the ongoing evaluation
   of NADA using the syncodec synthetic codec suite on emulated wired, WiFi,
   and LTE networks.  They have no open issues, and the authors believe the
   draft is ready for working group last call.
   
   Zahed: he's a co-author, but believes it's in good shape.

   Next steps: chairs will issue working group last call.


## Other Business

   Jonathan Lennox asked if it would make sense to do a hackathon on this
   subject? Maybe - there is some interest. Integrating the RTCP feedback
   mechanism with the candidate algorithms might be a good topic.

                                   - + -