[annodex-dev] sequence description list question
zohar.babin at Kaltura.com
Wed Jul 23 12:21:41 PDT 2008
The current Kaltura SDL can is described (partially) here:
we're looking to solve few problems, and use a more standardized format
for convention sake (make the SDL more familiar and easy for other
One of the main things that is solved by using ROE in the location tag
is multiple streams for the same video/audio file (ie, ogg/flv/avi/3gp..
versions or Spanish/English/Chinese... languages).
By basing the SDL on an extended version of xspf we can achieve a
standardized format of describing a sequence while also maintaining an
"open for extensions" and robust xml description of the sequences.
I think there are 2 main topics that xspf specs lack to be able to
support Kaltura sequences like representations:
1. Multiple tracks (of audio/video/overlas etc.) (what's the reason for
xspf to force only one tracklist?).
2. Support of arbitrary plugins with plugin specific saved data - each
plugin can be implemented in different ways, just like a video can be
encoded in different codecs. It's saved data should be free and open to
be represented in any way the plugin developer chooses (ie, if it's
binary use CDATA).
From: annodex-dev-bounces at lists.annodex.net
[mailto:annodex-dev-bounces at lists.annodex.net] On Behalf Of Silvia
Sent: Wednesday, July 23, 2008 5:36 PM
To: Nils Dagsson Moskopp
Cc: Annodex; playlist at lists.musicbrainz.org
Subject: Re: [annodex-dev] sequence description list question
Hi Michake & Nils,
On Wed, Jul 23, 2008 at 2:19 AM, Nils Dagsson Moskopp
<nils-dagsson-moskopp at dieweltistgarnichtso.net> wrote:
>> We have been looking at open sequence description formats and it
>> seems xspf could be extended to fit our needs. By using the extension
>> tag we can embed most the necessary xml needed for filters,
>> transitions, effects etc.
> why not use SMIL <http://www.w3.org/AudioVideo> ?
I am wondering about the aim of the sequence description that you are
If it is just to display a playlist, xspf is perfect. You could consider
doing a mediaRSS http://en.wikipedia.org/wiki/Media_RSS which is being
used for syndicated media content and has a few more fields than xspf -
similarly iTunesRSS. These RSS derovatives work in players such as the
Democracy player (what's it called now?). But if you want to KISS, go
I would try to avoid SMIL where possible, because it is very complicated
and really a format for creating interactive multimedia presentations
(i.e. Flash animations) rather than just for simple media sequences.
>> Additionally the location tag could be a ROE xml file so that clients
>> could negotiate the media type/quality/language of the given media
> better use the Dublin Core namespace
> <http://dublincore.org/documents/dcmi-terms/> !
Dublin Core is a metadata standard. ROE is a xml format for describing
the structure of a multitrack media file and which tracks are
additional, which are alternatives etc. Dublin Core won't solve
Michael's problem. ROE does, so I would suggest using ROE here.
I am curious: what do the current kaltura sequence descriptions look
like? I followed the link and watched the video, even got to the
keyframe view, but wasn't able to see a xml representation of the
Assuming you want to build a simple description language for video shot
sequences where all you want is the start-time, end-time of the clips,
and transition effects, you should consider using a list of temporal
CMML is really there to describe the source video, so if you wanted any
timed text along the clips, CMML can do that - and it would continue to
stay with the video even if only a part is referenced through a temporal
ROE on the other hand is there to describe a multitrack video with all
sorts of files (overlays and audio tracks), so you could use it as a
sequence describer where the individual clips are specified through
Hope this helps.
annodex-dev mailing list
annodex-dev at lists.annodex.net
More information about the annodex-dev