key-frame request handling when scalability mode is not L1T1 for encoder

[Johnny Qiu]

when encoder is configured to work in L1T2, L1T3, etc, Spec should clarify when a key-frame request is configured for current frame, how this should be handled when current frame is not at base temporal or spatial layer.

For example, should encoder simply reset the GOP structure to T0 and generate a key frame at T0; or it should reject the key-frame encoding request; or it should record this key-frame request and continue output a delta-frame, and apply the key-frame request till next base layer frame.

[Eugene Zemtsov]

reset the GOP structure to T0 and generate a key frame at T0

we should spec this

[Eugene Zemtsov]

UA certainly shouldn't reject key-frame requests regardless of configured scalability mode and the current frame layer

[Johnny Qiu]

@Djuffin in this case shall spec require going to T0 immediately, or queue the key-frame request till next base layer frame?

[Eugene Zemtsov]

shall spec require going to T0 immediately, or queue the key-frame request till next base layer frame?

immediately

VideoEncoderEncodeOptions.keyFrame corresponds to the frame given to VideoEncoder.encode() thus its should be effective immediately.

[Bernard Aboba]

Is there is a specification change needed?

If not, can we close this issue?

[Johnny Qiu]

Is there is a specification change needed?

If not, can we close this issue?

Sure. I think this can be closed.