Colin,
[cut agreed]
> Note that sending a sample description together with every text sample > (or set of samples) may result in a situation where the same sample > description is present in the receiver buffer many times, but with > different timestamp values. This is a desing choice to ensure correct > reception of sample descriptions and may cause moderate overhead.
I'm not sure this part's appropriate, since it assumes a particular receiver buffer implementation strategy.
I am assuming that only the basic functionality is implemented: opening RTP packets, reading out units, giving them a TS as the rules describe and storing them in the receiver buffer in TS order. I intended to be as general as possible. The last sentence may lead to misunderstanding. Instead of that one, maybe describing an easy way to avoid this situation is more helpful:
"Note that sending a sample description together with every text sample (or set of samples) may result in a situation where the same sample description is present in the receiver buffer many times, but with different timestamp values. This can be avoided by maintaining a list of currently stored sample descriptions (i.e., their SIDX values) and checking this before writing in the buffer."
_______________________________________________ Audio/Video Transport Working Group avt at ietf.org https://www1.ietf.org/mailman/listinfo/avt