Subject: Re: [linux-audio-dev] Introducing DMIDI
From: John Lazzaro (lazzaro_AT_CS.Berkeley.EDU)
Date: Thu Dec 20 2001 - 21:48:59 EET
> Dominique Fober writes:
>
> [... algorithm in Appendix B of NOSSDAV paper, does it have clock skew
> issues?]
Note the use of RTCP in this algorithm -- when new RTCP packets arrive,
we check their timestamp to see if they are on time, and if so we update
the ACO model parameters. This on-going parameter re-estimation (about
once per 5 seconds for sessions with small numbers of receivers) handles
the garden-variety "sender and receiver clocks unsynced" issue pretty
well -- in practice on CalREN 2, we only see the "late window exhaustion"
happen for rarer network and host events that mess up local timing.
> Are RTP timestamped packets enough ? I'm not sure ...
Another feature of RTCP can come in handy here too -- each RTCP packet
supplies both the RTP timestamp and its corresponding wall-clock time,
using a 64-bit NTP encoding of time. RTP timestamps are supposed to
be "nominal", not "actual" -- the way RTP systems are supposed to get
a handle on skew, sync, and other issues is to use these NTP/RTP pairs
in RTCP to maintain sync. We didn't use this in our simple "Appendix B"
algorithm, but a more sophisticated algorithm could ...
> However, fixing the limits (the skew tolerance and the corresponding
> time limit) would be useful for the protocol implementation).
This is one possible way to proceed -- another is to take a look at
whether normative "late/ontime" semantics need to be in the MWPP I-D
at all. At Salt Lake, this was another issue that came up -- people
were skeptical how needed this normative text was, given that
the method of deciding late/ontime was totally implementation dependent.
It might be better to simply state that sound playout is implementation
dependent, and offer non-normative references to the NOSSDAV paper, the
RTESP-Wedel.pdf, and perhaps others as examples for implementors to
consider ...
-------------------------------------------------------------------------
John Lazzaro -- Research Specialist -- CS Division -- EECS -- UC Berkeley
lazzaro [at] cs [dot] berkeley [dot] edu www.cs.berkeley.edu/~lazzaro
-------------------------------------------------------------------------
This archive was generated by hypermail 2b28 : Thu Dec 20 2001 - 21:44:07 EET