Re: Definition of ma:fragments and ma:namedFragments

Dear Werner, all,

Let me jump into this discussion :-) I agree that the current 
definitions of ma:fragments and ma:namedFragments are pretty fuzzy.

Some clarifications: it *might* (see warning) be possible to specify 
within a Media Fragment URI a name referring to a fragment of a video. 
For example, assuming the producer of the video 'foo.flv' has encoded 
and marked up the video with cue points [1]. The video contains markers, 
referring to a time interval and noted "love scene". These cue points 
refer to the time offset 42m15s till 46m57s in the flash header.
A media fragment URI referring to this fragment would be:
http://www.example.com/foo.flv#id='love%20scene'

Assuming now that the media server understands this media fragment URI, 
it could serve this range request with just the bytes corresponding to 
this time interval. The UA will play this sequence but display the full 
length duration of the source video so that the user is aware he is just 
watching an excerpt of it.

Warning: this is the ideal scenario, but I first wrote "might" since it 
is unclear at the moment how the "named" dimension will actually work in 
the media fragment URI spec. The WG is still working on it.

> When working on the implementation of our mapping prototype we came
> across an issue with the definition of the ma:fragments and
> ma:namedFragments elements. My understanding was (and this is
> supported by the examples in the table) that both contain lists of
> identifiers conforming to the media fragment URI specification. I
> still think that this should be the case for ma:namedFragments, i.e.
> listing the defined name fragments for a resource.

The problem of named fragments is that:
   - you need container formats that deal with named sections or 
chapters (see [2])
   - you need to know which named sections have been defined in a 
particular video file, i.e. discovery problem

> However, in some metadata standards it is possible to assign
> identifiers to fragments (e.g. shots), and during mapping it seems to
> make sense to keep the identifier of the fragment (which could be any
> URI), and in addition provide the ma:locator property for the
> fragment, which would really be a media fragment URI (e.g. the
> temporal fragment representing the shot).

Are you sure about that? Do you know which constraint the various 
container formats put on how you can label the fragment? Are you sure 
you can use anyURI? Our reference is currently:
   - The QTText for .mov files
   - The MPEG-4 Part 17 for .mp4 and .3gpp files
   - The MPEG-21 spec
   - The CMML and ROE format for .ogv files
   - The cue points spec for .flv files
   - we need to complete the list for .mkv, .mxf and .asf

> To summarize: the current specification of the ma:fragments and
> ma:namedFragments properties is not very precise, it talks about
> fragment identifiers. I think that we should be clear whether we mean
> media fragment URI compliant ones or any URI. From the problem above
> it seems that it should be any URI for ma:fragments (which does of
> course not exclude media fragment URIs) and media fragment URIs for
> ma:namedFragments.

I'm not sure I understand the distinction you make. I don't understand 
either why there are two properties in the first place? Perhaps an 
example of how the property will be used would help ...

   Raphaël

[1] 
http://help.adobe.com/en_US/Soundbooth/2.0/WSA5A1DDFB-6BE2-4486-BE0C-A10CEEF119ADa.html
[2] 
http://www.w3.org/2008/WebVideo/Fragments/WD-media-fragments-reqs/#fitness-table

-- 
Raphaël Troncy
EURECOM, Multimedia Communications Department
2229, route des Crêtes, 06560 Sophia Antipolis, France.
e-mail: raphael.troncy@eurecom.fr & raphael.troncy@gmail.com
Tel: +33 (0)4 - 9300 8242
Fax: +33 (0)4 - 9000 8200
Web: http://www.cwi.nl/~troncy/

Received on Thursday, 1 October 2009 15:33:51 UTC