H264 RTP and packetization mode (no STAP-A in baseline H264 RTP)

Question

There is a spec that literally requires this:

When the SDP negotiation results in the use of the Baseline Profile (BP), a client shall not send Single-Time Aggregation Packet type A (STAP-A) packets, even when the packetization-mode 1 has been negotiated. When accepting the use of the Constrained Baseline Profile (CBP) a client shall support the use of STAP-A packets when packetization-mode 1 was negotiated.

can anybody comment that? Doesn't that sound like complete nonsense? How is that even possible that baseline or constrained baseline decoder could know that h264 was delivered in STAP-A (or at all that it was delivered by RTP).

Answer 1

Althrough you don't provide a ref, this is not total nonsense. STAP-A also means that NAL units are small (since they must fit in a single packet). Thus, less memory is required on the receiver side (RTP layer) AND decoder size (obviously the bitrate will be lower thus required processing power, if each NAL is small). Baseline profile is already very small, but a CBP is even smaller. This spec says simply that if BP is used, then FU-A should be used only (=> large NAL => higher bitrate => more memory/CPU required), and if CBP is used, then STAP-A is allowed too (less load).

In the end, the decoder does not know what was the transport used, but if only STAP was used, the requirements on the processing power of the decoder is less.

I agree however that this should be written in terms of decoder requirements and not network fragmentation requirements.