[whatwg] Chipset support is a good argument

[Kartikaya Gupta]

On Mon, 6 Jul 2009, Ian Hickson wrote:
>
> On Mon, 6 Jul 2009, Robert O'Callahan wrote: 
> 
> > Specs do generate demand --- by creating author expectation that a 
> > feature will be supported, by adding a well-known brand, and because 
> > test suites get created which vendors then compete on.
> > 
> 
> On Mon, 6 Jul 2009, Silvia Pfeiffer wrote: 
> 
> > I agree: standards generate demand. It is how h.264 hardware support 
> > originated - by making it a ISO standard, the vendors knew there would 
> > be sufficient market demand for it and created the chips.
> > 
> 
> I disagree with both these statements, I don't think they are in fact 
> accurate. Demand can be focused around a specification if one exists, but 
> a specification cannot create demand, and the lack of a specification is 
> not an impediment to deployment. We have seen both facets of this 
> repeatedly demonstrated through the lifetime of the Web, not least of 
> which by HTML itself. Indeed, cutting features that didn't have demand 
> despite being in HTML4 for a decade is one of HTML5's achievements. 

I'm not sure whether specs can create demand, and frankly, I find it somewhat irrelevant to the point at hand. The fact is there is already demand for a single encoding format that will be compatible with as many browsers as possible. The only question is what that format will be. In this case, the spec doesn't need to create demand for anything, it just needs to tell people what that format is.

I believe that if HTML5 specified Theora, that would become a self-fulfilling prophecy, and Theora would become the standard (both specified and actual) of the web. There's already been a huge amount of press (on Slashdot, OSNews, Ars Technica, etc.) about the debate on this list. If a decision were to be made in favor of Theora, that would immediately be broadcast by the same channels and a lot of people (including actual and potential web authors) would be aware of it. A lot of those authors (not major publishers like YouTube, but the long tail that includes everybody else) will not bother to read the details of the decision; they will simply assume that since it is in the standard it will soon be supported by all the major browsers, and they will make their choices and start publishing content with that in mind. Since there is already some browser support for it, those efforts won't just fizzle and die. As there is an increase of Theora-encoded content, support for it will increase (including in terms of hardware) and it will drive a virtuous cycle pushing Theora to be even more widespread.

In a nutshell, what I'm thinking is that putting Theora in the HTML5 spec would not be enough to magically get authors to start generating Theora-encoded content. It is, however, enough to make authors in the process of deciding between H.264 and Theora for their <video> content to pick Theora. It's just a little push, but that little push is enough to make all the difference.

Now, I suppose all of the above may hold for H.264 instead of Theora, but I think it is far less likely given the distribution of vendors and their reasons for being in favor of a particular codec. If you disagree, that's fine. Just replace all instances of Theora with H.264 and read on. I'm also ignoring the Dirac possibility which has been mentioned a few times but doesn't seem to be being actively pursued. If that satifies everybody, so much the better.

So, my questions to you are:

1) Do you agree with my view that specifying Theora for the video element would result in a self-fulfilling prophecy?

2) Do you think that it is better to sit on the fence and not specify anything, thereby forcing authors to either (a) be incompatible with some browsers or (b) re-encode their content in multiple formats? Or do you think it is better to pick a side that has a good shot at winning, even if it means that some vendors may be non-compliant with the spec?

My view with regards to question (2) above is that one way or another, the web will settle on a single encoding format. This can be done the easy way or the hard way. The hard way is to not specify anything, and let authors and vendors battle it out for years at everybody's expense, leaving a trail of carnage and cruft behind that will then need to supported for decades. The easy way is to specify something and cross your fingers. Even if it doesn't work, at worst it will just prolong an already long and bloody battle. The benefits from the best-case scenario make the risk more than worth it.

kats