[annodex-dev] captioning and cmml

[Silvia Pfeiffer]

Hi Ralph,

On Dec 9, 2007 9:35 AM, Ralph Giles <giles at xiph.org> wrote:
> On Sat, Dec 08, 2007 at 05:54:18PM +1100, Silvia Pfeiffer wrote:
>
> > If you look at the DTD, there is an ENTITY defined that is called
> > i18n. That ENTITY is used by just about all elements in CMML and is a
> > default collection of attributes. It includes the language. Therefore,
> > both options are already possible with the existing specification of
> > CMML 3.1. Which option is being used really is up to the authoring
> > system, since both have advantages and disadvantages.
>
> Thanks, I didn't see that you already have a lang attribute. So having
> each language in a separate CMML stream already works fine.

Even each caption could be of a different language. This is
particularly important for video/audio files that come in more than
one language where the language changes constantly. So, yes, this
situation is covered.


> I gather the idea in a single stream is to define separate track
> attribute values in the enclosing <clip> elements and have the player
> select based on that. However, the only way to choose a track selector
> is to go by the lang tag on the first few clips. Unless I'm missing
> something else?

There is a "track" attribute through which clips can be identified to
belong to different annotation tracks. So, you could define one that
is called "french_captions" one that is called "english_caption" and
then the clips can be separated on these two tracks. Of course if you
don't use that track attribute on the clips, they all stay with the
default track.

> I think we need a way to specify the lang attribute for the track as a
> whole. What about adding a way to "<import>" a particular track
> attribute value as an explicit child of a <stream> element?

I don't understand this.

> Having a place to hang track-level metadata would also let us add a
> "contentrole" attribute to define additional semantics, like
> distinguishing dialog translation from commentary from interpretation
> for the visually impared...it's also important to know which of several
> tracks that match your user's language preference to use as a default.

Is the "track" attribute enough for this? I would think so, but I may
be missing something.

Cheers,
Silvia.