Re: [office-metadata] meta-field and more...

[Elias Torres <eliast@us.ibm.com>]

Svante.Schubert@Sun.COM wrote on 05/11/2007 03:51:39 PM:

> Elias Torres wrote:
> > "Bruce D'Arcus" <bdarcus@gmail.com> wrote on 05/11/2007 09:37:40 AM:

> > Exactly what I suggested to Svante you would do.
> Elias, I understood your suggestion that the odf:MetaDataFile takes the
> elementURI as NamedGraph
>
> <odf:MetaDataFile rdf:about="uri:elementURI">
>     <rdf:type rdf:resource="http://opendocument.org/Citation"/>
> </odf:MetaDataFile>
>
> That would have the drawback that every element would have at least one
> metadata file.

We are confusing suggestions, I've made many suggestions. But yes that
suggestion is just one of practice, not specification. You don't have to
match URIs for all elements. Just the ones that want their own Named Graph.
We don't have to think that everyone wants their own graph. There'll be a
few writer (generate triples) plugins, but there should be many more reader
plugins.

> A lot of metadata files in case of citations..
> > I stated that you have a
> > need and you would spec out what your implementors would do. You are
> > deciding to do solve your need via rdf:types. What we provide is the
> > extensible mechanism (thanks to RDF) to do so. Now, Svante wants the
> > general solution to that problem (which I believe exists and its
needed)
> > but I think we would be going outside our responsibility and scope. I
> > suggested to Svante that it is a separate spec built on top of ODF
metadata
> > that he and KOffice and others can agree.
> >
> > I think we have a really good and tight (yet general) specification.
Any
> > more specific ontology classes and predicates will be looking for
trouble
> > and decisions that we might wish left for a later time.
> >
> Elias do you believe a single ODF application is not in need of finding
> the binding between a text:meta-field and it's RDF/XML file(s)?
> In case every application is in need of making this binding persistent,
> would it not improve the interoperability to address it in the standard?
>
> Svante

I don't deny the need. I really support it. I just don't want us specifying
it right now. I have worked on a few iterations of this problem trying to
give named graph capabilities to client applications and there's no right
answer yet. I'd be very afraid to over-specify that right now and have
practice show us something different later. Let's see what people do with
our spec as is.

-Elias