On Tue, 13 Feb 2001, Dan Brickley wrote:
...
> Yes, until DCMI developers a fancier conceptual model than "resources have
> a bunch of properties such as these 15" we don't have call for much by way
> of syntactic sophistication. A profile of RDF 1.0 syntax can serve
> 'colloquial xml' and 'nodes and arcs' apps equally well.
Exactly...
Now, there will be a point, when there are seqs, alts, reifications and
you name it. When RDF reaches that level of complexity, metadata will
necessarily have to be dealt with as generic RDF. However, what people do
with the metadata in their databases is basically none of our business. If
I take these nodes and arcs and transform them into MARC, in order to
cross-search it with our OPAC, I'm free to do so. Regardless of whether a
future RDF query language will be a success or not, we need to be able to
use the same data in different ways.
Heterogenous metadata, which are just slightly more complex than the data
we are currently discussing, will have to pass through some kind of
semantic normalization in order to be searchable with a finite number of
search boxes in an ordinary HTML form ;)
Regardless of how individual services choose to store and search their
data, as US MARC, or arcs and nodes/triples, XML text files, or whatever,
it is RDF -- semantic XML -- that will be the kind of data that is most
suitable for the normalization process. Having run harvesting bots for
five years, I fear more 100 different XML DTDs/Schemas, than 100 different
RDF Schemas. It is my firm belief that normalization of the former will
incur more manual labour than the latter, even when taking into account
semantic drift and schemas with overlapping semantics which will be what
RDF will give us.
Yours
Sigge
|