Dear Pete,
datatypes appear as a bit subtle in RDF.
My current understanding is as follows:
One can declare a class given by a URIref as a rdfs:Datatype -
but an RDF(S) declaration as such (!) one may/should view as
inappropriate to
serve as definition of a datatype.
Some indication for that:
A datatype is supposed to have a value space, a literal space
and a literal-to->value map -
RDF does NOT provide
tools to define these constituents or
a standardized way to point to a tool,
which would provide this input.
The XMLSchema-2 built in (primitive) datatypes are mentioned to meet the
requirements: definition of value and literal space and literal-to->value
map.
At least currently in my understanding XMLSchema on the other side
does not provide a standardized way to create a URIref for a user defined
datatype.
Such a meachnism might become incorporated in a future version of
XMLSchema - i have no real information about that.
The RDF Semantics draft is saying:
A 'datatype-aware' RDF engine SHOULD be competent to recognize at
least the rdfs:XMLLiteral datatype and the set of all XML Schema primitive
datatypes.
A datatype entailment one could make in RDF (turned into somehow human readable text):
Given the statements: 1. xsd:date denotes a datatype
2. The resource, denoted by URIref A,
has been modified during "1999-10-26"^^xsd:date
Then there exists a resource, lets call it x, such that the resource, denoted by URIref A,
has been modified during x and x is a xsd:date -
no pointer is specified from x to "1999-10-26"^^xsd:date -
{I've gotten a hint somehow indicating,
that it might be tolerable from an RDF view point
to add (!) the inference
[x has 1999-10-26 as rdf:value. Thereby interpreting rdf:value
as pointing to a literal representation - so probably
one is allowed to use the undecorated string also as
a dumbdown value}
-------------
In effect it seems to me that RDF as such can put datatypes on record
by decorating character sequences in a specific way - that's it.
In the case of date related stuff there are as you know
quite a few primitive datatypes:
duration, dateTime, time, date, gYearMonth, gYear, gDay and gMonth
In case i wanted to use RDF datatyping right now (!) in that area,
i would prefer to use one of the xml-schema built in primitives
to help with interoperability.
Please consider this view as preliminary.
W3C's RDF primer draft in fact proposes such a use in the example:
<rdf:Description>
<ex:creation-date rdf:datatype="http://www.w3.org/2001/XMLSchema#date">1999-08-16</ex:creation-date>
</rdf:Description>
------------
One natural question is, why we should worry about the humble "dcterms:W3CDTF"
at all in the future- why not deprecate it's use?
------------
It's a bit confusing, that the W3C RDF validator doesn't have an option to test the
new syntax construct, which is not compatible with traditional RDF M&S parsing (!).
There are a few issues still with RDF datatypes....
--------
In my view an important new syntax development is the introduction of nodeID's for
bNodes.
I'm experimenting with that.
-------
Still another possibly important feature is the new collection parsetype.
Till now i hadn't really time to look for that in it's relation with OWL.
-------
A little dark i found the wording on fragment identifier and
the proposed MIME type application/rdf+xml -
An innocent - non-native English - reader might get the erronous impression,
that there is currently no valid RDF in the world.
Cheers,
rs
