[MMUSIC] RTP TS/SN NPT rehash

RTP spec is very clear about how time stamps should increase within a session. Since the same RTP session is still being used after PAUSE(i.e. it has the same SSRC and unique end-points) and RTP says thattimestamps as well as sequence numbers should increase monotonically andlinearly with respect to realtime

My first question is how do sequence numbers increase linearly with respect to time when even during streaming their "rate" will be dependent on access unit size, and not real time? In addition, RFC 3550 (5.1) says in its definition of sequence number: "The sequence number increments by one for each RTP packet sent". This text seems to imply that SN will not increase if the session happens to be paused (unlike the TS). There are even examples in the RTSP spec showing the sequence number not increasing during pause.

Since the thread above seems to have been the consensus in the group related to RTP header information as it relates to paused sessions, can someone clarify whether quoted comment may have unintentionally included SN in group of things that increase in this way, or maybe whether there is other text in the RFC that specifies a different behavior for sequence numbers that is being reflected by this comment?

In addition, in 2326bis there is a sentence in C.2.3: "Thus, both RTP sequence numbers and RTP timestamps MUST be continuous and monotonic across jumps of NPT.". While the context of this sentence is describing behavior in light of jumps in the media layer NPT as a result of something like multiple NPT values in a PLAY request's Range header, it seems like this sentence contradicts an implication that was also made in the thread above as it applies to live data, although its not entirely clear given that a live-specific example wasnt given. Specifically, one of the other messages in that thread implied that one difference between VOD and live was that a client pausing the RTSP session will pause NPT for VOD, however it will not pause NPT for live (the live session continues to exist irrespective of the RTSP client's behavior). If a client were to issue a PAUSE on a live session, it seems the Range header in the PLAY response to resume should indicate a jump in NPT. It is clear from the above thread that the timestamps cannot be monotonic in this case (that was the whole point of that thread, if I'm not mistaken), and if so, wouldnt that contradict the sentence in 2326bis? At a minimum it seems sequence numbers and timestamps shouldnt be treated the same way here unless there is a field-specific meaning of "continuous and monotonic".. Perhaps some simple context clarification could fix this?

I think adding a live pause example to 2326bis C.2.4 would be helpful (including description of TS, SN and NPT). It may even be necessary to clarify whether any servers that allow sequence number to increase for paused live data are in fact non-compliant (seems so based on 3550), or whether behavior in live cases is ambiguous enough (eg. what does PAUSE really mean for live content?) to maybe allow this to occur.