Re: [AVT] RTP/RTCP Timestamp and transmission time

[I removed MMUSIC from the cc list, since this question is not 
appropriate there]

On Thursday, Sep 18, 2003, at 09:15 Europe/London, Lorenzo Polidori 
wrote:
> I have two questions about RTCP packets.
> 1- In RFC 3550, section 6.4, it is specified that a sender can set the 
> NTP
> timestamp to zero: in this case, will subsequent RTP timestamp of the 
> same
> RTCP packet be zero, as this value is calculated from the NTP 
> timestamp? And
> in what cases is allowed setting to zero NTP timestamp? Do common 
> clients
> like RealOnePlayer accept this setting?

The RTP timestamp is NOT calculated from the NTP timestamp, but 
represents a sampling of the media clock. Setting the NTP timestamp to 
zero prevents inter-media synchronisation, but does not otherwise 
affect the media playout or the values of the RTP timestamp.

> 2- In the case of unicast scenario, with only one client and one 
> server and
> a fixed bandwidth, is it  possible to calculate the transmission time 
> of
> RTCP packets a priori, and to send the RTCP packet at fixed times, for
> example every 3000 RTP data packets? Doing so, it will be avoided the
> calculation of transmission time for this very simple scenario.

RTCP packet transmission times are randomised, and scale automatically 
with the number of participants in the group. Since it is not possible 
to tell, a priori, now large is the group, it is not appropriate to 
pre-calculate the sending time (even in a unicast session, your peer 
could be a gateway device, bringing many participants into the 
session). The RTCP time calculation is not complex: many 
implementations use the code directly from the appendix of the RFC, and 
there is no reason not to perform the calculation.

-- 
Colin Perkins
http://csperkins.org/

_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
https://www1.ietf.org/mailman/listinfo/avt