Seems like, as one of the OAI missionaries, I need to say something
here.
Sigge, I sympathize with your frustrations here and certainly share your
desire to find a common format for sharing metadata and expressing
relationships among many vocabularies. I agree strongly with you that
RDF provides a plausible foundation for this - as DanB knows, we even
have a research project, Harmony
(http://www.ilrt.bris.ac.uk/discovery/harmony/), that is exploring this
issue and I am committed to RDF as the basis for doing this.
Now back to OAI and why we used XML schema, with a bit of proselytizing
to the DC-architecture group. Our concern in putting together the OAI
stuff was three-fold:
- data verification
- base functionality
- deployability
Regarding the first (data verification), we wanted to have a technical
infrastructure for which we had some hope of conformance tests. Our
goal is to have a registry of OAI protocol supporters and ensure some
integrity of the registry. By defining XML schema for each of our
protocol responses and for the metadata some of them contain, we can
partially achieve this. This is simply not possible or in the scope fo
RDF schema as currentlyh defined.
Yes, it would be nice to also be able to express vocabulary
relationships. But that leads to the base functionality goal. With all
the talk about relationship of DC to other vocabularies, we've yet in
the DC community to really show the usefulness of simple metadata for
simple discovery (see, of course,
http://www.dlib.org/dlib/january01/lagoze/01lagoze.html). Bill Arms
recently stated the goal very nicely "to try to prove that weak metadata
with powerful indexing (e.g., the kind of thing google does) might be
better than deep metadata with mediocre indexing (and at a much lower
cost). So, I admit in OAI to saying vocabulary relationships can wait,
what we want now is a defined way for exposing simple DC, and also
exposing metadata in other vocabularies. We can build the
infrastructure for defineing the relationships among them as another
layer.
All of this would be a moot point if both goals (simple metadata and
complex vocabulary relationships) were available in some deployable
technology. However, and DanB has heard me harp on endlessly about
this, RDF is simply not there yet as a deployable technology. Right now
I can write an OAI protocol based on XML and XML schema and use readily
availble off-the-shelf technology (e.g., XML-spy) to make it all work.
It would be wonderful if the data validation of XML schema and the
semantic expressiveness of RDF schema could be combined (and I recommend
that the interested read a paper by Jane Hunter on this subject
(http://www.cs.cornell.edu/lagoze/papers/WWW10/Schemas.html). However,
that combined capability is simply not available in any really
deployable form. One can't propose to communities like libraries,
museums, and the like that they spend their time with an forever
increasing list of alpha tools and standrads when what they want is a
deployable method of exposing and sharing metadata. I am the first to
admit that we by our choice of XML schema we sacrificed some
functionality, but that is the choice one makes when promoting a
deployable product.
That said, I close with my personal caveat to the DC-architecture group.
Yes, we absolutely need to keep one eye on the longer term goal of
having infrastructure that coordinates dc type vocabularies into a
broader metadata architecture. But, we also need to come up with
technologies that people can use now in a stable and deployable fashion.
Let's not loose the second goal. Being a veteran of years of "we need
to use RDF to deploy DC" discussions I am of the opinion that that
obsession is one reason for the dilemma we are in now (lots of talk but
precious little understanable and deployable technology).
Carl
> -----Original Message-----
> From: Rachel Heery [mailto:[log in to unmask]]
> Sent: Tuesday, February 13, 2001 8:36 AM
> To: [log in to unmask]
> Subject: Re: Deliverables required pre-DCMI tools meeting
>
>
> On Tue, 13 Feb 2001, Sigfrid Lundberg, Lub NetLab wrote:
>
> >
> > What Open Archives are using on their server is none of my
> business. And
> > neither is what Z39.50 target transfers to its origin.
> These are both, as
> > I see these XML objects used internally in specific
> applications, and they
> > follow certain agreements between those involved.
> >
> > People's internal data formats are out of scope for DCMI
> architecture.
> >
>
> Sorry, I disagree, in this case we are talking about an
> exchange format,
> not just what is on people's servers. All the data providers
> that want to
> be OAI compliant will need to produce records according to the OAI DC
> schema, and those that want to be Z39.50 Bath compliant need
> to transfer
> records according to the Bath Profile DTD (thinking about it
> Bath profile
> specifies a DTD rather than schema).
>
> An DC XML schema would define such an exchange format, once,
> so not every
> time someone wants to mandate DC as a format they have to re-do a XML
> schema.
>
> In addition, ok, such a schema might well be used
> 'internally' but I think
> it would be useful there too. And for example in a registry,
> we might want
> to refer to a DC XML schema with one definitive URI. And a
> DTD for those
> who like them.
>
> Rachel
>
>
> >
> > Sigge
> >
>
>
> --------------------------------------------------------------
> -------------
> Rachel Heery
> UKOLN (UK Office for Library and Information Networking)
> University of Bath tel: +44
> (0)1225 826724
> Bath, BA2 7AY, UK fax: +44
> (0)1225 826838
> http://www.ukoln.ac.uk/
>
|