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

[Maik Merten]

Dave Singer schrieb:
> You miss the point.  MPEG defines levels exactly so that bitstreams can
> say "you need to be level X to be able to play this" and players can
> implement "up to level X" and interoperability is well-defined and
> assured.  Levels *improve* the interoperability, not make it worse. 
> Without formal level definitions, you never know for sure what you can
> 'get away with' in your encodings if you want it to play on a set of
> devices.

In practice you usually don't know what devices your is media going to
be played on so the fact that the profiles exist is often of little
value when encoding content.

As end user you may end up creating content for your iPod that doesn't
play back on your PSP and/or vice versa (I don't know if the profiles
match or not, so that's just an example), so obviously the levels did
*not* improve real-life interoperability here. You just end up finding
out that "H.264" is not "H.264" the hard way.

Without profiles you may run into a brick wall regarding decoding
resources - but do people really assume a DAP will decode HD video? In
many cases people will just assume all coding features are supported by
the decoder (that's easier to achieve for not-so-complex formats like
MPEG4 Part 2 and Theora or H.263 or H.261) and it often will "just work".

I can see the value of profiles in such a complex scheme as H.264 (after
all it *is* way harder to support all its features compared to earlier
schemes), so profiles are just necessary to make sure vendors can at
least reliably express where their solutions will fail. That doesn't
change the fact that there are H.264 files out there that will play on
device A but not on device B. So the profiles add another dimension to
interoperability issues (there are already many others).

Maik Merten