ogg dev - Sep 2007 - Feedback on XML metadata namespace

Ralph Giles wrote:
> 
> A URI/URL scheme for referencing streams would be a generally useful
> thing. Should we override the query mechanism like the time fragment
> draft? "?oggserial=0x15323421" ?
> 
This is roughly what was suggested the last time around.
A possible refinement is to provide a mapping from identifiers
to IDs so you could say "?oggid=myvideo".  The mapping could
maybe be handled by a fishbone packet.

-- 
imalone

Actually, we have been discussing this kind of addressing for a very
long time and it was half the point of the Annodex project (see
www.annodex.net).

What we have defined for addressing is the following:

* temporal addressing:
http://annodex.net/TR/draft-pfeiffer-temporal-fragments-03.txt
 Any ogg file can be addresses through temporal offsets. This goes
across different logical bitstreams.

* identification of logical bitstreams:
http://annodex.net/TR/draft-pfeiffer-annodex-02.txt
  Skeleton provides all the functionality required to get information
about the logical bitstreams without a need for decoding. The logical
bitstream's serial number is in the fisbone header. In addition, each
of the logical bitstreams gets protocol-type information through the
message-header field. "Content-type:" is mandatory (it has the mime
type as currently being specified by
http://wiki.xiph.org/index.php/MIME_Types_and_File_Extensions. There
is currently not a "name" (ID) specified for the logical bitstream in
the way that Ian suggests, but there is no reason that cannot be
introduced. All we need to do is create a message header field called
e.g. "id:" and make it mandatory. Then we can address logical
bitstreams in skeleton through names as well as serial number.

* named addressing of time offsets:
http://annodex.net/TR/draft-pfeiffer-cmml-03.txt
  For completeness' sake I also want to mention that through CMML we
have defined a means to address time offsets into ogg file (with
skeleton and cmml) by name. This looks e.g. like
http://example.com/sample.axv?id=dolphin


So: in summary, I agree with Ian and we have a simple way to introduce
this kind of addressing using Skeleton. We should probably pick up
editing of the mentioned I-Ds again and include what we learn through
this process.

Cheers,
Silvia.


On 9/9/07, Ian Malone <ibmalone@gmail.com> wrote:
> Ralph Giles wrote:
>
> >
> > A URI/URL scheme for referencing streams would be a generally useful
> > thing. Should we override the query mechanism like the time fragment
> > draft? "?oggserial=0x15323421" ?
> >
>
> This is roughly what was suggested the last time around.
> A possible refinement is to provide a mapping from identifiers
> to IDs so you could say "?oggid=myvideo".  The mapping could
> maybe be handled by a fishbone packet.
>
> --
> imalone
> _______________________________________________
> ogg-dev mailing list
> ogg-dev@xiph.org
> http://lists.xiph.org/mailman/listinfo/ogg-dev
>

On 09/09/2007, Silvia Pfeiffer <silviapfeiffer1@gmail.com>
wrote:
> On 9/9/07, Ian Malone <ibmalone@gmail.com> wrote:
> > Ralph Giles wrote:
> > > A URI/URL scheme for referencing streams would be a generally
useful
> > > thing. Should we override the query mechanism like the time
fragment
> > > draft? "?oggserial=0x15323421" ?
> > This is roughly what was suggested the last time around.
> > A possible refinement is to provide a mapping from identifiers
> > to IDs so you could say "?oggid=myvideo".  The mapping could
> > maybe be handled by a fishbone packet.
>   Skeleton provides all the functionality required to get information
> about the logical bitstreams without a need for decoding.
> There
> is currently not a "name" (ID) specified for the logical
bitstream in
> the way that Ian suggests, but there is no reason that cannot be
> introduced. All we need to do is create a message header field called
> e.g. "id:" and make it mandatory. Then we can address logical
> bitstreams in skeleton through names as well as serial number.
>
What are the pros and cons to mandatory?  It's not required
in (X)HTML for instance, only when the items are significant.
If the serial number mechanism is available then the ID would
only be needed when semantic meaning was required.  (It's
one thing to have id="video", id="soundtrack-en" etc.,
another
to have IDs derived from UUIDs.)

-- 
imalone