[theora] Indexing Ogg files for faster seeking

[Ivo Emanuel Gonçalves]

On 9/24/09, Chris Pearce <chris at pearce.org.nz> wrote:
> (...) and we won't break existing decoders which expect the skeleton
> to be encoded in  the current format. Best to not break skeleton if possible.

Not necessarily.  In fact, I predict the opposite.  Skeleton is a very
recent addition in the Ogg world, and most current decoders don't
interact with it -- they play the Theora/Vorbis/whatever ignoring the
Skeleton because they don't recognize it.

Moreover, there are one or two buggy parsers out there that actually
refuse to read the entirety of the Ogg file because they expected the
file to begin with a Theora or a Vorbis stream.

When you add yet another stream to the mix which, if I read the spec
properly, starts before Skeleton, when decoders are just now
understanding that the initial stream is the one that pin-points all
of the content within the Ogg... well, it will get messy.

However, were this index idea an expansion on Skeleton, you would
remove the unecessary complexity of adding yet another metadata stream
to the mix and build on what exists.  Win-win, I say.

-Ivo