Re: [rtcweb] MTI video codec, charter, RFC 3929

[Adam Roach <[e-mail]>]

On 11/7/13 18:57, Stephan Wenger wrote:
> ...a decision for either of the two candidate codecs will be ignored 
> by a significant part of the implementer community.

I'm not sure that's true any longer. Indulge me in a bit of armchair 
game theory application.

Let's start by looking at ground zero: the browsers. Considering the 
public statements that have been made, I think it's fairly safe to 
assume that the following can be taken as given:

 1. Chrome will support at least VP8.
 2. Firefox will support both VP8 and H.264.
 3. Internet Explorer is almost certain to support H.264.
 4. Safari (including Mobile Safari) is almost certain to support H.264.


If the codecs are *only* those called out above, then you'll end up with 
an interesting situation in which IE, Safari, and Firefox can all 
mutually interoperate; Chrome and Firefox can interoperate with each 
other; and Chrome cannot interoperate with anything else.

I'm not going to put plans in Google's mouth, but envision yourself in 
the same set of circumstances: you own a fully-paid H.264 license, 
already ship an H.264 codec as part of your browser, and are getting a 
huge black eye for not interoperating with significant major browsers. 
What do you do?

Okay. Now, shift your imagined role: you're a non-browser implementor, 
and discover the facts on the ground to be those described above. You 
can deploy VP8 and work with two of the four major browsers, or you can 
deploy H.264 and work with all four. Yeah, you might not be in love with 
the structural issues involved in integrating the Cisco library [1]  -- 
and I'm certainly not -- but you're not actually precluded from using 
H.264. What do *you* do?

Which is a kind of long way to get to the fact that I think we're in a 
situation where H.264 is a de facto MTI codec anyway [2].

Now, imagine you're working in an SDO with this particular set of 
writing on the wall. You have a solid decision to choose one MTI video 
codec. You can choose one that's going to be functionally MTI regardless 
of what you say, or you can choose one that is quite probably going to 
be ignored by some non-trivial set of implementors.

So. What do *we* do?

/a

____
[1] There's been quite a bit of handwringing around the impact this has 
on iOS developers, as they cannot download modules at runtime. Bear in 
mind that these are curated applications, with a single, 
well-accounted-for distribution point. These are not libre software by a 
long shot, and they don't need to worry about the implications of third 
parties compiling their own versions. So, these implementors statically 
link in the OpenH264 library, and are exempted for the first 100,000 
instances they ship each year. Beyond that point, they're not "small 
fish" any more. Back when I was one of the owners of the company I 
worked at, "complications arising from selling more than 100,000 copies 
a year" is the kind of thing that I would place on a list titled 
"problems I wish I had."

[2] And to be clear, this is *not* the result I would have preferred. 
But I recognize that this isn't about my personal preferences. I find 
myself in the odd position of actually arguing for the position contrary 
to what I think is ideal, simply because I think it is the only 
practical path to success.