W3C home > Mailing lists > Public > semantic-web@w3.org > October 2007

Re: XMP [Re: RDF: XULing or Grueling]

From: Tony Hammond <tony.hammond@gmail.com>
Date: Mon, 8 Oct 2007 12:14:36 +0100
Message-ID: <c107aff50710080414x7d6eeea4ka91167a7fc0dde6f@mail.gmail.com>
To: "SW-forum Web" <semantic-web@w3.org>
Cc: "Tony Hammond" <tony.hammond@gmail.com>, "Bijan Parsia" <bparsia@cs.man.ac.uk>, "Bruce D'Arcus" <bdarcus@gmail.com>

Well, of course, I agree with everybody. :)

I think as regards XMP the issues are mainly technical. (And btw,
Bijan, some of my other posts re XMP on CrossTech addressed specific
technical issues, also RDF/XML.) The real problem is how to write
(especially), but also how to read this stuff. I note that the semweb
community is generally stum on the issue of media, and tend to confine
themselves to simple markup documents which are a relatively
straightforward case to handle. I don't see any leading lights here.

Imho, XMP seems to be a pretty good spec for sneaking XML payloads
into arbitrary media files. How to accomplish that is less obvious. If
it means living within the walled garden of the CS suite then we're
badly limited in what we can do being dependent on custom and vendor
panels with their necessarily restrictive view. Tools for reading XMP
packets are available (e.g. Exiftool), but writing is another matter
(and is patchy at best). And then there's PDF. (Easy to say, harder to
do.)

I agree with Bruce that the RDF is crippled. (If I know it's a
resource, then why can't I say it's a resource? Because the XMP
metadata schema says it must be a literal or a group - Bag/Alt/Seq -
of literals, or this, or that.) Otoh, if the intent is to simplify the
RDF profile (for easy UI generation or whatever) then why not go all
the way and lock down to something that can be validated with a DTD or
schema?

I think this (the broken RDF) is by far the lesser problem. Adobe
won't really be interested until this thing is out there in general
use. Though it would appear that they would rather keep it for
read/write by Adobe apps which does kind of circumscribe its world
somewhat. Hence any pressure to change or otherwise revise is minimal
and negligible.

So, maybe the thing is a dodo. A great big white elephant. Can't
write, won't read. But still it's the only game in town for including
arbitrary metadata into media files that I'm aware of. Pity the semweb
community is not really interested.

Tony



On 10/8/07, Bruce D'Arcus <bdarcus@gmail.com> wrote:
> Bijan Parsia wrote:
>
> > Thanks for the pointers. I don't have much to say that I've not already
> > said. The issues you discuss vis-a-vis XMP seem more political than
> > technical. Clearly, there is a requirement for extensibility. Beyond
> > that, it's a bit hard to determine what the actual technical
> > requirements are (no surprise, really). I wonder if aligning XMP with
> > some RDFa profile would make sense.
>
> The political and the technical are hard to distinguish though.
>
>  From what I know, somewhere around 2000, Adobe engineers were looking
> for an extensible metadata framework that could be mapped to a GUI
> reliably, without external configuration.
>
> So they looked at RDF as it existed and said, "ah, RDF literals are kind
> of like a hash, rdf:Seq, rdf:Bag, and rdf:Alt can be mapped to familiar
> programming structures and UIs, let's just create a subset that only
> allows those."
>
> So it's a bizarre sort of subset; not of the syntax, but of the model.
> It leans heavily on literals and blank nodes.
>
> The spec is now effectively frozen by Adobe, which has a ton of legacy
> issues that explain why they have no interest in turning it over to a
> proper standards organization, or otherwise changing it.
>
> Hence, the technical problems are closely connected to the political.
>
> > Actually, I think there's a lot of milage to be gotten out of being less
> > free and easy.
>
> Which I guess aligns with periodic discussions of what has been called
> "RDF Lite."
>
> > One problem we faced at UMD with some of our tools is
> > that they were *too* open and flexible. Photostuff, for example, would
> > build forms for "person" from all the ontologies you loaded. So you got
> > these HUGE forms with dozens of fields. And all you wanted to do was
> > mark up the photo with the fact that it was a photo of your niece.
>
> Norm Walsh had started to work on a cool web app for photo metadata. In
> that case, the UI gets configured.
>
> <http://norman.walsh.name/2006/09/13/photodata>
>
>
> Bruce
>
>
Received on Monday, 8 October 2007 11:14:49 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:18 GMT