Dear Tom,
the discussion at Florence preceded the new drafts about RDF, which
for the first time introduced a way in RDF to refer to datatypes -
I'm deliberately using "refer" here. RDF has no built in facility
to "define" a datatype.
The RDF understanding of datatypes enables the use of some
XML-Schema part2 datatypes as appropriate to be referred to
from RDF as datatypes.
The best readable account i found till now is in OWL drafts
from February 2003 - along with
RDF Concepts and RDF Semantics.
My current position on datatypes and DC is to use the
datatypes which are explicitly listed by OWL - in particular
for dates, language and number representations - and indicate their
use with the URI's assigned by the XML-schema-part2 recommendation.
An inflational use of the RDF syntax, which allows all URI's to be
referred to as datatypes i'm very much concerned about -
Again: RDF offers no facility to "define" a datatype.
Their has been criticism on dc-general on
references to KOS as XML datatypes, which might affect
coding of KOS even in NON-RDF/XML settings.
The core problem in datatyping approaches is the
mapping problem for user defined datatypes, which
conceptually - though not syntactically - go beyond the ideas of xml-schema datatyping.
(DC is a user in this situation of course).
I think reasonable reasoning engines (like a humans
for instance) will find an "inference"
<dc:subject rdf:datatype="&xy;CDU">252</dc:subject>
===>
<dc:subject>252</dc:subject>
hazardous -
Best regards,
rs
|