On 1/3/2012 10:07 AM, Randell Jesup wrote:
On 1/2/2012 1:56 PM, Wesley Eddy wrote:
Hello all, following up on our discussion in Taipei, I have proposed a working group in the TSV area to work on RTP congestion control. This is scheduled to be discussed by the IESG on 1/5 in "internal review".
The draft charter is at: http://www.ietf.org/iesg/evaluation/rmcat-charter.txt
I would welcome participants in this list to comment on this, particularly if/when it moves to "external review", as I would expect the folks currently on this list to be some of the main participants in the proposed WG. I had previously had this reviewed by the TSV Directorate, and got fairly good responses and strong interest expressed from many of the directorate in participating as well.
Happy New Year!
Thanks Wesley.
First a nit:
"Other algorithms developed in light of experience with TFRC are felt to be motivated." Perhaps instead: "Experience with TFRC has motivated demand for alternative algorithms."
While I very much do not want to do anything to delay progress on this, I should note that most aspects of such congestion control mechanisms would not be specific to RTP, though some implementation details might be. We should attempt to define these algorithms such that they'd be applicable to any flow with similar characteristics and requirements (certainly other media flows), while providing specifics of how they apply to RTP flows.
Good point; I will try to reflect this in the next revision, though I think we do need to clearly have RTP/RTCP as the target for a concrete instantiation of the algorithms.
Some aspects that are developed in this process might be useful in other similar domains as well, such as congestion control of multiple TCP flows between a pair of hosts. I would not explicitly include TCP in this effort at this point to avoid feature creep, but we may want to have a dependent/follow-on effort for TCP.
I should note that for WebRTC/RTCWeb, we hope to be able to share a congestion domain with the data channels, currently expected to be handled via SCTP, though the details of how this will occur have not been decided.
Thank you; this is a very good point, and I will attempt to reflect it in the next revision. -- Wes Eddy MTI Systems