I think I find myself taking a position somewhere "between" Rachel and Andy
here... not that I wish to portray them as in opposition, you
understand..... Anyway it seems kind of appropriate as the office where I
work is located somewhere between theirs... ;-)

Andy said:

> The current approach, based solely on RDFS is no more 'open, extensible'
> than a mixed RDFS and relational database approach is it?  Or at least, I
> don't understand why that should be the case.  The registry can still
> import and export RDFS - at least in the medium term (though my personal
> view is that the requirement to import RDFS is of relatively low
> priority).

I think there is an argument that an approach based on indexing distributed
RDF descriptions (and not just RDF _schema_) _is_ _potentially_ (and I
stress the "potentially") more "open and extensible" in the sense that it
exploits RDF's capacity to "allow anyone to say anything about anything".

That (potentially again!) makes the registry become a much more dynamic
"publication environment", not just for the "approved" descriptions of
semantic units "coming out from the centre", but also for "real world"
implementer experience ("application profiles", but also all sorts of other
"annotations" and commentaries) "coming in from the periphery".

Even within the activity "at the centre", it could be possible to create
machine-readable statements about, say, meetings of DC committees and relate
them to changes to descriptions of sets of semantic units. So, precise
information about the introduction of new elements etc could all be
available through the
registry - which becomes a navigator for a micro-"Semantic Web".

And those distributed descriptions (and indeed the descriptions created at
the "centre") might also be (re-) used within other (non-DC, regional,
domain-based) registries, other RDF-based applications, other micro-Semantic
Webs.

_However_ (before I get completely carried away with the excitement....) I
do appreciate that

(i) this is beyond the scope of what is envisaged for the current DCMI
registry

and

(ii) as Andy emphasised in his last message (OK, the reference is to
arbitrary _schemas_, but it applies equally - if not more so - to other sets
of statements),

> My gut feeling is that, while one might be able to infuse arbitrary RDF
> schemas, it will be very difficult to develop user-interfaces that present
> views of differing schemas in a very meaningful way. I.e. to develop a
> useful user-interface one has to make some assumptions about the structure
> of the schemas one is dealing with.

while it may be possible to _express_ all these statements about various
resources in RDF, and to read them into a triples database for
merging/indexing, there is still the problem of creating a user interface
which goes beyond being a generic "hypertext triples navigator" and provides
something a casual reader can interpret.

So while it is true that in the "statements-stored-in-RDBMS" (with or
without import of RDF/RDFS) approach, the things one can say about a
resource are constrained by the database structure (tables for resource
classes, fields within those tables for the properties about those resource
classes [OK, it might be a bit more complex in practice...]), I think we
also have to admit that the "indexing-distributed-RDF" approach would also
have to be constrained in some way (what resource classes are described,
what properties are used to describe them) if the service to be built on it
was to be usable to a non-RDF-aware constituency.

I think this would be particularly true if (as I've discussed briefly with
Rachel and Andy at one time or another), within the
"indexing-distributed-RDF" approach, we sought to provide some sort of
forms-based tools to insulate the publishers of schemas (or metadata about
schemas) from the generation and maintenance of RDF/XML markup.

So I suppose I'm saying that maybe in terms of the specific goal of
providing a registry service which meets the immediate requirements of
managing the DCMI "vocabularies", the constraints imposed by the two options
are perhaps not as different as I admit that at first I imagined, and I can
see the value of Andy's arguments for what I'd call a "pragmatic" approach.

Indeed I'd like to think this "pragmatic" approach may help us to address
some of the "modelling" issues (see e.g. my previous message on
"vocabularies" under the subject "Co-terminous") which apply regardless of
the representational form in which the "schemas" (in the abstract sense) are
stored and/or expressed.

But I'd also like to think that there was a value in continuing to explore
an approach based on the indexing of potentially distributed RDF/RDFS-based
descriptions. Perhaps it is beyond the scope of the requirements for "the
DCMI registry" as it stands, and as such it needs to be re-specified in some
way, but it certainly seems to be a valuable research area and one which
falls within the scope of DC-related activity.

Cheers

Pete
-------
Pete Johnston
Interoperability Research Officer
UKOLN, University of Bath, Bath BA2 7AY, UK
tel: +44 (0)1225 323619    fax: +44 (0)1225 826838
mailto:[log in to unmask]
http://www.ukoln.ac.uk/ukoln/staff/p.johnston/