Flash, Google, VP8, and the future of Internet video
x264dev.multimedia.cx
x264dev.multimedia.cx
In retrospect, I should've stuck with it; I knew that Adobe's iPhone code would be delayed as hell, but didn't think there'd be enough lead time to make my money's worth off of it. If I would've known they were going to delay it this heavily, I could've had a product out and making money already. Hindsight is 20/20, I guess.
A while back there was an IamA on reddit from a video codec hardware engineer. I found this comment interesting:
We've been asked to support ON2's VP6 and possibly other codecs. Google isn't done entering the video market, and this is part of that push. I can't really divulge anything more sorry.
See http://www.reddit.com/r/IAmA/comments/9q7pp/for_my_fellow_ge...
I can't find the link because twitter search sucks and the Marvell site all seems to consist entirely of Flash and PDFs.
- So far as I know, Apple and Nokia refused to support Theora because they were already heavily invested in H.264, and they didn't want to open themselves to more patent risk than they already had. These reasons apply equally well to VP8. - Mozilla picked Theora because it was the most production-ready patent-free codec. Even if VP8 had zero patent risk, libtheora is already written and tested and portable.
Am I missing something?
There's been huge buzz among Mozilla, FSF, and so forth about the Google purchase of On2 and VP8. They're all basically pleading for VP8 to come out in order to replace Theora. It was really sudden, just over the past few days.
Even if VP8 had zero patent risk, libtheora is already written and tested and portable.
There are things that matter more than just being tested and portable, like whether it's actually good.
One of the barriers to adoption of Theora has been the quality issue; why would Apple or Nokia take the extra risk and cost of going with Theora if they already implement a superior codec? There's a chance that VP8 would help with that, but it's still an uphill battle. It took years for Xiph to take VP3 to something that they wanted to call a stable standard. Google has more resources than Xiph, for sure, but they likely can't just dump VP8 as-is as a new standard. There's also the issue of submarine patents, the hardware issue, just getting this implemented and as widely deployed as H.264, and more, so yeah, it would be tough for this to become a standard.
On the other hand, what the heck else is Google planning on doing with On2? Why would they get into the proprietary codec business, when H.264 seems to be winning out in most places in the market, and they've already bought all the licenses, encoded most of YouTube in H.264, and so on? Of course, those questions also apply to an open codec, but Google has been a big proponent of open, freely implementable standards, so at least that route makes some amount of sense.
But I fully believe them when they say it's patent risk. Just like I believed Nokia when they said it was because Theora was "proprietary" and that it was impossible to implement DRM with Theora.
http://www.w3.org/2007/08/video/positions/Nokia.pdf
In that position paper you can see that Nokia is really keen to push MPEG towards patent-free licencing, but their big strategy to achieve this is to lock themselves into MPEG. Not very smart, but maybe Google can get some momentum going in that direction.
(Related fact: Nokia also argued for Software Patents in Europe.)
Why is x264 GPL licensed instead of LGPL like FFmpeg? It would make x264 the de facto H.264 encoder.
The GPL is used for historical reasons; relicensing would be hard, obviously, because we have to contact everyone who ever wrote a line of code. Ironically we are doing that right now, but for a dual-license scheme (GPL/Proprietary).
The reason we didn't choose LGPL for the relicense is because one of the main devs (not me) refused; he saw releasing the source as LGPL to be effectively throwing away any chance to ever make any money off of his work, while keeping the code as GPL would at least reserve that option for the future. The dual license is a good compromise IMO; commercial applications can support development through license fees while free software gets to use it for no cost.
"psy optimizations are the single most critical feature ... they're the reason that Vorbis beat MP3 for audio"
So in other words, they're not the most critical feature since Vorbis quality and freedom didn't budge MP3 and so, if we imagine a parallel world where x264 does not exist and Google's VP8 was actually noticeably better than the best H.264 (like Vorbis vs MP3 or on par like Vorbis vs AAC) we would find that anti-competitive patent pools would still be forcing us to use their choice of codec.
The best we can reasonably hope for, as we saw with Vorbis, is for a strong competitor to prevent crazy licence fees. And maybe in time for H.265 the EU or China will have enough political clout and common-sense to demand that international video standards must be royalty free.