Can't start decoding a VP9 stream without a keyframe, but this breaks spatial scalability structures

[GoogleCodeExporter]

The VP9 decoder will reject any stream that doesn't start with a keyframe, even 
if the stream starts with an intra-only frame.

This makes random-entry in spatial scalability cases less useful.  The use case 
is that an encoder is generating a stream with several spatially scalable 
layers; a middlebox is routing all the layers to high-definition receivers.

When a new low-resolution receiver joins the session, the middlebox needs to 
generate a new intra-encoded frame in the base spatial layer for the receiver.  
However, if this frame is a keyframe, it means that the higher spatial layers 
can no longer use reference pictures from earlier than the keyframe.  Thus, the 
ideal would be to have the base layer intra-encoded, but not be a keyframe; 
however, this will not be valid for the low-resolution receiver's decoder.

Original issue reported on code.google.com by jonat...@vidyo.com on 2 Oct 2013 at 2:12

[GoogleCodeExporter]

Original comment by fgalli...@google.com on 17 Oct 2013 at 10:21

• Changed state: Assigned

[GoogleCodeExporter]

Fix in progress:
https://gerrit.chromium.org/gerrit/#/c/70029/

Original comment by agra...@google.com on 6 May 2014 at 11:58

• Changed state: Started

[GoogleCodeExporter]

New patch pending final review:
https://gerrit.chromium.org/gerrit/#/c/70439/

Original comment by agra...@google.com on 8 Jul 2014 at 10:55

[GoogleCodeExporter]

Requires an additional patch:
https://gerrit.chromium.org/gerrit/#/c/70842/

Original comment by agra...@google.com on 15 Jul 2014 at 5:02

• Changed state: Fixed