[whatwg] Google's use of FFmpeg in Chromium and Chrome

[Silvia Pfeiffer]

On Mon, Jun 8, 2009 at 11:27 AM, Chris DiBona<cdibona at gmail.com> wrote:
> Thinking out loud: One thing that was mentioned in an earlier post:
> Vorbis. I am also of the mind that Vorbis is of higher quality/mb/sec
> and statically than is mp3. The only real problem is that people don't
> pirate with it, so the demand isn't there, but I think it is a
> superior codec. For video, I worry that for theora to become 'better'
> than h264, it will need to infringe on the same patents it is designed
> to avoid.

I think you  may be underestimating the potential that is still in
Theora. As Monty described in this May update
http://web.mit.edu/xiphmont/Public/theora/demo7.html, there are still
many improvements to be made on the encoder, but Thusnelda has already
improved heaps without using h264 techniques. I wouldn't give up on
Theora and quality video yet. In particular when most people will not
really notice the difference on a well encoded video using Theora in
comparison to H.264 - and I say that from anecdotal evidence rather
than actual research.

In fact, I disagree with the whole notion that the lack of uptake on
Theora is caused by its lesser quality. I think the real issue that a
user has with Theora is that it doesn't yet have the same kind of tool
support as other codecs, e.g. H.264. However, that was also true of
H.264 not a long time ago - increasing vendor support on h.264 made
all the difference. It would for Theora, too. So, vendor support - and
that includes browser vendors, but also video editors, transcoders,
hosting providers, video players etc - is the real issue.

Lack of vendor support in turn has nothing to do with the slightly
lesser quality of a codec. It's all about costs and returns: about
return on investment on past expenditure on other codecs, about new
cost on the codec (beyond the sheer implementation cost which should
be minimal with Theora), about perceived risk of expenditure on legals
around it (including patents), and about money the new codec support
could bring. A new codec doesn't bring the vendor new income unless it
promises something special to the user and thus creates a market need.
Only if the market is asking for it and the vendor will not be able to
sell his tools without supporting the codec will he be forced to
provide support. We are in a chicken-and-egg situation.

H.264 broke this chicken-and-egg situation by promising a amazing new
quality of picture, which was communicated through marketing and thus
created the market need. This obviously won't work on a codec that is
of equal or slightly lesser quality.

Being royalty-free, unencumbered, and easier to innovate around could
and should be the message for Theora. That married with it being
written into HTML5 as baseline codec would definitely break that
chicken-and-egg situation and create the market need that is required
for Theora to gain vendor support. Amazing demos of capabilities of
HTML5 video available in all browsers because they all support Theora
would blow users away and make them ask for Theora support.

I think we are starting to see this happening: the content published
at Archive.org, Wikipedia and Dailymotion in Theora will have impact,
the native support of Theora decoding in Firefox, Chrome, and possibly
Opera will have impact, and the DirectShow filters, QuickTime
components, encoding and decoding plugins, etc will have impact.

Whether it will be enough to get the all browser vendors on board and
therefore Theora written into the HTML5 spec as a baseline codec, I
don't know. I fear we are on the same path again that we were on with
image codecs and it will take a long time until the new, open and
unencumbered "PNG" (umm: theora) codec is accepted in all browsers.

I'm not even sure that writing it into the standard would make vendors
actually support it, for the reasons above. If everyone had only the
best interest of Web users on mind, it might happen, but that is now
how the world works. I can only hope that it won't take as long as PNG
took, which was introduced in 1996 and got browser support by all
mainstream browsers only in 1999, fully bug-free only in 2008.

Regards,
Silvia.