Metadata Discussion

(notes from the discussion)

Transmission network:

In 2006, as part of the Transmission process, we were asked to produce recommendations towards a standard for markup of video metadata. Come up with a metadata standard for searching over different sites.

  • good in that a social network emerging out of it (around 10 projects or so working with it)
  • feed are all atom (not rss)

"would be cool to inject it into the CMML track" format the low level video file information

silvia: brining together different types of metadata good for archiving.

Conrad: If you transcode this file to a different bitrate you need to change parts of the XML...

and|: its actually more for web

Silvia: This is from the user point of view... "information I would like to have as a user". A separate issue is how its implemented. Another thing to add when I was at the w3 workshop .. a lot of people presenting problems with metadata... a few people suggested .. mpeg21, smile, rss etc... it was difficult problem to talk about. The biggest problem about metdata is that everyone thinks they can give some input.

and|: a lot of people that don't know a lot about metadata contributing

ralph: a problem with metadata everyone has their own problem set .. their own title, etc. Itunes for example. In terms of stuffing it in the creative commons people are working on a library to insert into media files. We could convince creative commons people to make the format a bit more general CMS could inject that info in.

and|: if you want people to use it it has to be striped down to the base minimum.

Silvia: another discussion whether it should be in the file or external to the file. the decision that we made is that its used in both ways/ used as a seperate file or muxed into the media file... you have to make it possible for live streams for multiplexing subtitles into the stream.


part of getting ogg as a web standard acceptability concerns.

Ralph: What about multiple languages? Which stream is what language? All in one logical bitstream. Ogg issues with multiple tracks.

conrad: Multiple logical bitstreams... issues with duplicating links and other issues. Maybe if hosted on the server then only send the selected languages.

Michael: Probably going to be hosted on the server anyway.

Shane: Might be ways of putting multiple tracks into a single "page". Silvia: the issue of the clip tag is that there is a limit of 1 caption, desc in that time segment.

Silvia: this is a decision the application should make rather than the spec?

john: distinguish between captions, subtitles, transcripts:

Conrad: captions have extra sound information and abbreviation conventions, subtitles useful text information like tickers, sports scores etc. These should interact.

Silvia: transcript is created by the computer its automatic not necessarily displayed.

Conrad: The problem is they are very close to each other.. it might be useful have one be an annotation to another...

Ralph: why do you need to distinguish? you have the description tag and give it a type and language.

Shane: can we allow the type be a fixed set or extendable

Silvia: we don't have multi-track CMML...

Conrad: each approach (mutil-track vs part of the same presentation) make sense in a different way...

John: what if I want to make different urls for different languages.

Shane: are we going to cascading streams? CMML can't deal with our hierarchical data. At the high level we have a paragraph level, underneath a sentence, underneath words..

Silvia: we have come across this(hierarchical data) before if we just have a higher level identification of stuff.

shane: if we have a physical encoding of hierarchy it gets really hard to rip out a given stream..

silvia:one of the disadvantage of quick time? Ralph sounds like we just need to put this in a skeleton.

Shane: cmml at the text level do we want to allow all this complicated stuff hierarchies etc.

John: it has to be a tool thing...

Ralph: skeleton is a mechanism to identify different streams.

Silvia: should we remove the stream tag from CMML... (in favor of skeleton)

John: is there an XML format for skeleton?

Summary Talk:

We use track attribute to determine which what type like ... this track is English, this is directors commentary etc. the type is promoted to the skeleton so that it can chose the logical bitstream.

Ralph we need caption and desc.. or I guess we only need desc... discussion about what tag would be ideal.. worst case: multilingual sports presentation. track specific types in each 4 types we have a body which is specific to the type. The language is encoded both in the track and in each clip tag.


  • We should all look at the tx_metadata paper and offer constructive criticism :)
  • We use CMML for subtitles
    • Collapse description and captions tag into a "body" tag
  • We want XML authoring format for skeleton
    • We need to develop track roles for skeleton.
Melbourne by mugley
Melbourne at night - Photo by mugley

Logo Design by Pascal Klein