[annodex-dev] first multitrack description draft

Sat Mar 22 08:09:21 PDT 2008

Hi Michael,

On Mon, Feb 18, 2008 at 4:13 PM, Michael Dale <dale at ucsc.edu> wrote:
> so I have been implementing ROE as the data transport format for
>  mv_embed... so an mv_embed "embed" line could be something like <video
>  roe="roe_url"/>  ... or <div class="mv_embed" roe="roe_url">...
>
>  In talks with Conrad at FOMS/LCA concluded with <video
>  src="roe_url=media_url" being the ultimately goal...
>  This would work in the existing browser platform by having javascript
>  issue a HEAD request on the video src and get the ROE url from an extra
>  line in the response header of the media stream. It would use this url
>  to pull in the data in cases where the browser/plugin did not expose the
>  muxed metadata. (ie everything but oggplay atm)
>
>  but implementation is a bit tricky: Server side components are not yet
>  ready, the javascript header request only works on the same domain as
>  the javascript is running (can't do cross-site DOM js injection of
>  header requests without a server side proxy)  Its a bit more complicated
>  to setup when media servers are not the application servers...And
>  finally it adds a bit of complexity to the proxy/caching setup since
>  your grouping /small/ data that changes a lot
>  (wiki/transcripts/translations/tags) with /big/ data that does not
>  change often (audio/video streams)
>
>  But it will still work as a transport format for emedding video while
>  other issues get sorted out... here is what I am exporting now for roe
>  files:
>  http://mvprime.cse.ucsc.edu/wiki/index.php?title=Special:MvExportStream&feed_format=roe&stream_name=Senate_proceeding_08-01-07&t=0:00:00/0:20:00
>  and this is what its exporting for CMML (as linked to in the text tags):
>  http://mvprime.cse.ucsc.edu/wiki/index.php?title=Special:MvExportStream&stream_name=Senate_proceeding_08-01-07&feed_format=cmml&tracks=anno_en
>
>  I wanted to have people take a look at it before I re-write the
>  javascript xml parsing and assoc interface ;) probably in a day or two...

I have just looked through the files and added the "img" tag to the
ROE header specification at
http://svn.annodex.net/standards/roe/roe_1_0.xsd.

I am not clear about your use of "node_count" in the text elements.

Also, may I suggest using "mediaSource" rather than audio/video/text
as element names? I believe the content-type attribute signifies loud
and clear what it is.

Also, I have introduced a attribute called "inline" for the
mediaSource element, so you can add inline="true" for the CMML
inclusion.

Other than that, it looks a lot like the ROE specification, so I am
happy. What issues did you have?

>  some initial suggestions/questions:
>
>  1) it would be good to have a simple indicator of a default track for
>  display... The profiles shown in the example cmml stuff ...seems a bit
>  too smile-ish ;P ...I am not sure how to use them... so I added
>  /default="true"/ attr to video tag? ... but it could be done with
>  style="display:true"  ...

This applies only to switch statements, right? How about we add a
"default" attribute to the switch statement and then add the id of the
tag that is default? I have added this functionality for the moment,
including the examples. I have even made the "default" a required
attribute because it is so important.

>  2) I added "title" attribute to to <text> tags to describe the layer a
>  bit.. but If we can think of a way to put the longer description in
>  there that would be cool...

No, I think this is right. I also have that in the ROE spec.

>  3) ... I did not know how to put the <cmml: prefix for all the cmml tags
>  without crashing firefox's xml parser. (probably just importing the
>  namespace incorrectly) ... If anyone has ideas there let me know.

I believe the header to the file has to look something like:
<?xml version="1.0"?>
<xs:schema targetNamespace="http://www.xiph.org/roe"
	    xmlns:xs="http://www.w3.org/2001/XMLSchema"
	    xmlns:cmml="http://www.annodex.org/cmml4.0"
	    xmlns:html="http://www.w3.org/1999/xhtml"
	    elementFormDefault="qualified"
	    attributeFormDefault="unqualified">

But this is just what it should be theoretically:
http://www.xml.com/pub/a/2000/11/29/schemas/part1.html?page=8

>  4) would be cool to have official namespaces URI for the xml documents
>  to point to.. where are those?

Namespaces are as follows:
ROE: http://www.xiph.org/roe1.0
CMML: http://www.annodex.org/cmml4.0

>  5) for <video tag would it be good to include the "params" tag that we
>  had with cmml for informing the client of the video resolution/ file
>  duration etc?

Each mediaSource element now has a param tag, so you can use that for
any mediaSource.

>  thanks for all your work on these specs and associated software etc :)
>  Its nice to see it all start to come together ;)

Slowly, slowly. You're the only one that has software yet. ;-)
But I'm glad the specs are progressing.

Cheers,
Silvia.

>  peace,
>  michael
>
>
>
>  Silvia Pfeiffer wrote:
>  > Hi all,
>  >
>  > I committed a first version of the multitrack media XML description
>  > language into the annodex svn. It is for now called ROE (Rich Open
>  > multitrack media Encapsulation) and lives at
>  > http://svn.annodex.net/standards/roe/roe_1_0.xsd, examples are at
>  > http://svn.annodex.net/standards/roe/examples/ .
>  >
>  > Check out the notes from our FOMS discussions at:
>  > http://wiki.annodex.net/wiki/MovieDescriptionLanguage
>  >
>  > A few design decisions that were made in this first draft:
>  >
>  > * The use of XML schema made it possible to import onther XML schemas;
>  > this one uses the XHTML XML schema for some elements and attributes.
>  >
>  > * SMIL wasn't used after all. It would have been possible to use the
>  > following tags: switch, seq, par, and ref, but it would have required
>  > major changes to the attribute sets and the content models for these
>  > to work in the ROE context. Instead, switch, seq, track and
>  > mediaSource were defined respectively for these and adapted to ROE needs.
>  >
>  > * Introduced a head and body element to separate the profiles from the
>  > core document structure.
>  >
>  > * The profile element now has a subelement called "partial" which
>  > contains the ID of the selected track and potentially the ID of a
>  > selected alternate media source for a track.
>  >
>  > * The "par" element that we used earlier was replaced with the more
>  > explicit "track" element. This makes it easier visible where the input
>  > streams go and it makes it impossible to cross over data between tracks.
>  >
>  > * All the media sources have been aggregated into a generic
>  > mediaSource - unless we find a compelling reason to re-introduce
>  > audio, video, textstream, they are not necessary.
>  >
>  > BTW: specs typically start numbering at 1.0, so do not take that
>  > version number as a sign of quality.
>  >
>  >
>  > Feedback very welcome!
>  >
>  > Cheers,
>  > Silvia.
>  > ------------------------------------------------------------------------
>  >
>  > _______________________________________________
>  > annodex-dev mailing list
>  > annodex-dev at lists.annodex.net
>  > http://lists.annodex.net/cgi-bin/mailman/listinfo/annodex-dev
>  >
>
>