Document which information should be available to SFUs in case audio/video content is encrypted

[youennf]

Let's document in the draft the properties we identified that would be useful for SFUs. Two SFU use cases seem to be identified:

• SFU content adaptation
• SFU media recording

For video, there is:

• at which packet SFU can switch
• resolution (could be generalised to stream 'quality', for instance frame rate, bit rate as identified by video-layers-allocation00)
• codec specific information like profile_idc. Maybe this is something that could be skipped?

For audio, there is:

In particular, for OPUS, I check the TOC of the opus frame to know the audio frame length and be able to correctly set the timestamp in recording. Also for recording, I was planning to extract in-band fec in case of packet losses and recreate previous frames as the players do not support PLC.

[murillo128]

Also, for video it is needed to know what recovery mechanisms should be applied in case there are some missing packets for a frame.

• The frame is discardable and no need to RTX it
• The frame is required by future frames, so better RTX it
• The frame requires previous frames that are lost so a LRR/PLI is needed
• The frame requires previous frames but the stream will auto recover in case of loss and no need of LRR/PLI is needed.

(All this info is covered by the dependency descriptor)