Re: RDFCore Update

[Brian McBride]

>
> One major area of focus for the WG at the moment is datatyping, e.g. using
XML
> schema datatypes in RDF.  Now would be a good time to let us have your
thoughts
> and ideas on this.
>
 Some of the other posts on this subject have dived right into the XML RDF
syntax.  I suggest getting clear on what the appropriate triples (or
subgraph) would be like, before moving on to the syntax.

To me it seems clear that if you relate a value to a subject by using a
property, you then would want to (optionally) be able to hang an arc off
that value node that attaches some datatype to it.  This takes us back to
the discussion about whether you should be able to do that or not, or
instead model the value with an anonymous resource node that you can hang
the value and datatype onto.  I suggest that some of the other proposals to
encode the typing into a URI should be considered to be equivalent
shortcuts.


(some_subject)---(hasValue)--->( 25.2 )----(datatype)----->(xsd:float)

(some_subject)---(hasValue)--->(  )----(datatype)----->(xsd:float)

\-------(value)---->(25.2)

Given the vagaries of rendering with various fonts, I'm not sure the second
example will display right for everyone, but it is meant to show an
anonymous node with both a datatype and a value arc attached to it.

Now if we take this approach, we can use RDF to assert the range and domain
of the datatype, but if we encode the type into a URI that would be much
harder.

In any event, let's get clear on the model before we start churning out
alternative XML interchange syntax.

This approach would let you plug in other datatypes just by using their
namespace instead of the xsd: (or whatever) namespace (yes, I know the
namespace is really the URI, not the prefix!)

Cheers,

Tom P

Received on Thursday, 18 October 2001 23:44:10 UTC