Re: [rtcweb] RTP Usage: Is RTP Retransmission REQUIRED or RECOMMENDED

["Roy, Radhika R CIV (US)" <[e-mail]>]

Hi, Magnus:

The audio/video performances for both point-to-point and multipoint conferencing must be very carefully articulated, not just what your colleagues did or did not do. All networking parameters have been discussed in list email including lip-synchronization.

So, let us see all the details of networking conditions in the published papers. We can then examine what is this all about.

The only thing that I am pointing out that these types of simplistic emails are not good enough to convincing the technical communities unless all the technical details are presented with precise measurements. Both subjective and quantitative performance parameters are involved along the many subjects who remain in the test environments. 

If anyone thinks that RTP retransmissions is a great tool, it is OK. Let them keep is private as many usages of RTP retransmissions were also discussed via emails.

BR/Radhika

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Magnus Westerlund
Sent: Thursday, June 28, 2012 3:49 AM
To: [e-mail]
Subject: Re: [rtcweb] RTP Usage: Is RTP Retransmission REQUIRED or RECOMMENDED

Hi,

As Individual I like to state my position.

We have a video conference system developed by my colleagues used internally at Ericsson that uses RTP Retransmission for video, not for audio with great success. This is implemented such that we actually allow the video to fall behind the audio when packet loss and retransmission is not able to repair in a timely enough fashion. The benefit is minimal overhead and still no loss induced degradations in the video. Yes, we get degradation in form of frame display jittering and short freezes. But those events that are truly visible are rare over wired networks.

I am personally convinced that RTP Retransmission is great tool in the toolbox when it comes to improve media quality in many use cases. Yes there are scenarios where RTP retransmission is less efficient. Long RTTs (over 200-400 ms) is the primary source of degradations. Compared to FEC it so much more efficient from bandwidth consumption perspective.

I also think it is important that we have some mandatory to implement tool for making the transport more robust now that we have a consensus that we are not going for a FEC solution in the initial specification.

Thus my personal position is that RTP Retransmission should be REQUIRED to implement.

Cheers

Magnus