On Fri, 2003-03-14 at 10:28, Dave Beckett wrote:
> With my RDF Core WG membership hat on - thanks, that's useful
> feedback. I personally thought it was appropriate (if not perfect -
> long story) for DC datatypes and will be using it myself since pretty
> much all RDF applications are DC ones.
While we're on the topic: I *would* find it useful if the *same* literal
could be placed as the object of several statements, allowing longer
texts (more than just atoms such as dates) to be reused.
If a Literal could also be used as the *subject* of statements, I guess
the encoding of rdf:Datatypes within the literal object would be
unnecessary. You could just use an rdf:type statement, which would be a
more coherent approach. This would allow Literals to carry other
information as well. I guess this must have been discussed and resolved
in the Core WG already, though.
For this reason and others, I agree with other comments (Pete etc.) that
vocabulary values should not always be encoded using Literals, but
sometimes using URIs. This is certainly the case for "finite"
vocabularies, where each single terms carries a very distinct meaning.
Then there is a need for describing that resource too using different
properties.
We do use this approach rather much in LOM, for properties such as
dc:subject and the like. In my experience, it makes maintaining and
using the vocabularies much simpler that using literals.
/Mikael
--
Plus ça change, plus c'est la même chose
|