Re: [Sframe] Partial decodability and IDUs

[Sergio Garcia Murillo <[e-mail]>]

On 20/11/2020 0:22, Bernard Aboba wrote:
> Sergio said:
>
> "I am afraid that we end up another doing a per-codec packetizer, 
> which would be almost the same as just using current packetization 
> formats and explain how to use encrypted data with them. And all for a 
> feature that we are not current using in webrtc."
>
> [BA] Some per-codec logic is required.  How else can the metadata be 
> obtained from the bitstream? But the question is whether it is 
> *required* to packetize IDUs separately, so they don't share fate.  
> I'd say this would depend on the use case. If WebRTC didn't care 
> enough about this to implement it previously, why should SFrame make a 
> difference?


Taking the AV1 dependency descriptor as an example, all of the elements 
are well defined in a codec agnostic way. Being simplistic in the 
example, there is no need for describing how to get the width and height 
of the video frame from an h264 stream, the encoder would be responsible 
for filling the the correct information. Or at least that was the idea 
behind the design.

Best regards

Sergio