[whatwg] on codecs in a 'video' tag.

Henri Sivonen hsivonen at iki.fi
Mon Apr 2 13:07:07 PDT 2007


On Apr 2, 2007, at 21:12, Maciej Stachowiak wrote:

>> Let me add other reasons why Mozilla (for whom, again, I am not  
>> speaking) might want to specify Theora/Dirac:
>>
>> - They have a strong commitment to interoperability
>
> I don't think Theora (or Dirac) are inherently more interoperable  
> than other codecs. There's only one implementation of each so far,  
> so there's actually less proof of this than for other codecs.

Well, at least Theora doesn't seem to suffer from profileitis that  
makes H.264 interop *hard*.

H.264 has two incompatibility axes: profiles and AVC levels. There  
are 11 profiles and 15 AVC levels. Some implementations (e.g.  
QuickTime) support only a some of the profiles. My understanding is  
that QuickTime supports 3 of the 11 profiles. Some implementations  
only support AVC level up to a magic level that you have to know. My  
understanding is that iPods have an arbitrary cap for the AVC level  
and that it has changed with firmware versions.

Some encoders offer profile settings but are silent about AVC levels.  
Some encoders offer a magic iPod preset, presumably for the first  
generation of video iPod firmware. Some encoders expose the AVC level  
and you have to know what the iPod max is. Some encoders offer  
feature checkboxes and you have to know what belongs in each profile  
to avoid overstepping the feature set of QuickTime.

I wouldn't be too surprised if some implementations wanted frame  
dimensions to be divisible by 8 or something like that.

Interop is not impossible. The Google Video team has managed to  
figure out parameters that work for QuickTime, iPod and PSP at the  
same time. But it seems that the recipe is a Google secret and they  
have people figuring this stuff out as their dayjob. For a casual  
content author, figuring this stuff out is insanely hard.

-- 
Henri Sivonen
hsivonen at iki.fi
http://hsivonen.iki.fi/





More information about the whatwg mailing list