W3C home > Mailing lists > Public > public-media-annotation@w3.org > October 2009

AW: AW: Definition of ma:fragments and ma:namedFragments

From: Bailer, Werner <werner.bailer@joanneum.at>
Date: Fri, 2 Oct 2009 18:12:56 +0200
To: "Troncy, Raphael" <Raphael.Troncy@cwi.nl>
CC: "public-media-annotation@w3.org" <public-media-annotation@w3.org>
Message-ID: <CD9846F872C7874BB4E0FDF2A61EF09F2E6C00F411@RZJC1EX.jr1.local>
Hi RaphaŽl,

> By fragments, I assume you mean a subpart of the media resource. Your
> ma:fragments property seem then very close to what M3U or ROE (Rich
> Open
> multitrack media Encapsulation) formats specify, correct?
> Do you really mean a list? Does the order matter? Can we have
> duplicates?
> 
> So you would have something like:
>   http://www.example.com/myvideo.ogv ma:fragments ( :fragment1,
> :fragment2, ...) ?

Yes, it is a list. We have not yet discussed the whether the order matters, my opinion is no, as there can be different types of fragments in the list. As the fragments can have different roles (chapters, favourite scenes, ...) they can definitely overlap, and thus there can be duplicates.

> > The API can then be used to query annotations of the fragment, so
> > ma:fragment links different granularities of annotations. These
> > fragments could also be named fragments, with all the issues you have
> > pointed out.
> 
> I still don't understand how these fragments are referred to? By URI
> ref?

Yes.

> > The reason to have ma:namedFragments is different. It addresses the
> > discovery problem that you mentioned, i.e. to determine what the
> > valid range values for each type of fragment identifier is. The ma
> > properties allow to answer this query for temporal, spatial and track
> > fragments by using the appropriate technical properties.
> 
> How? Could you provide an example?

The ma:frameSize and ma:duration properties return these technical metadata and thus give the user agent the range for valid temporal and spatial fragments. It's more tricky for the tracks: we have ma:numTracks, which will allow you to query tracks by number. 

> > The issue I raised yesterday was not related to labels from container
> > formats, but to identifiers of fragments in metadata documents that
> > are accessed using the media annotation API. Assume you have an
> > MPEG-7 document, which describes the shot structure of a video. Each
> > VideoSegment can have an ID, so that, given the media resource is
> > http://foo.com/video a shot could be http://foo.com/video#1, which is
> > an identifier for this fragment and a URI, but not a media fragment
> > URI.
> 
> http://foo.com/video#1 is a valid URI but how do you expect to
> dereference it? 1 is not a valid XML id as you may have guessed.

True, 1 is not valid, let's call it frag1. Thanks for spotting this.

> You're saying that http://foo.com/video#1 identifies a particular shot
> with a video. Why not identifying this segment with the media fragment
> URI that actually defines the temporal boundary of this shot?

This is for sure an option. But then we loose the identifier of that video segment in the source format.

> > In this case ma:fragments should contain http://foo.com/video#1,
> > and when requesting the ma:location property of this fragment one
> > would get e.g. http://foo.com/video/video.mpg#t=10,20
> 
> So you would have;
>   <video> ma:fragments (<video#1>, <video#2, ...) .
>   <video#1> ma:location http://foo.com/video/video.mpg#t=10,20 .
>   <video#1> foaf:primaryTopic dbpedia:love .
> ?

Correct.

Best regards,
Werner
Received on Friday, 2 October 2009 16:13:33 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 2 October 2009 16:13:34 GMT