My next question is about buffering and latency. RFC 1890 specifies a clock
rate for ulaw (8000Hz) and jpeg (90000Hz). With each RTP packet, you include
a timestamp which is the sampling instant of that data. If my timestamps are
incremented according to the RFC, my audio and video are synchronized, but
there is about a 5 second latency, like maybe Quicktime is buffering up 5
seconds in case of network burpage.
If I "lie" about the jpeg timestamps, and increment the stamp by, say, 4000
each second instead of 90000 each second, the video runs without perceptible
latency, but the audio is still lagging. If I try to lie about the audio
rate, I get no audio at all. I can't seem to lower the latency on the audio.
My goal is to get both the audio and video to have low delay. Does Quicktime
always buffer the stream for some set number of seconds, or is there something
I can do with the stream to tell it not to do that?