On 03/19/2015 05:17 PM, Keith Winstein wrote: > But what are the merits of an IETF working group performing this kind of > high-risk, high-reward research, versus doing something much more boring > like "writing a specification for VP9 to enable interoperable > implementations, and then iterating on that technology"?
If VP9 could be shown to meet our requirements for performance and licensing, I would be supportive of that path. However, I don't believe that it currently does. In the testing draft, while I specified methods of testing codecs, I did not specify an absolute performance goal. We will need to set one. Opus achieved state-of-the-art performance, which was highly beneficial to its adoption. VP9 does not achieve this yet. In regards to licensing, it is clear that many companies are quite happy shipping VP9 right now. However, others are not nearly as comfortable with it, as seen on RTCWEB. These concerns arise from other, non MPEG-LA rights holders. We need to make sure that we address those concerns in a working group one way or another. One way to do that is just to stay well clear of the patents in question. > Google said in May 2013 that "a draft bitstream specification is well > underway." For whatever reason, they still haven't published it yet. (There > is also no independent implementation of a decoder written by a > non-Google-employee, afaik, much less of an encoder.) ffmpeg's VP9 decoder was written by a couple of non-Google employees: https://blogs.gnome.org/rbultje/ _______________________________________________ video-codec mailing list [email protected] https://www.ietf.org/mailman/listinfo/video-codec
