Re: Formation of RDF terms

Dan Brickley wrote:
>
> The basic problem is that the RDF Model doesn't like to grubby itself
> with worldly things like constraints on allowable URI syntaxes for the
> resources named within RDF. RDF syntax(es) can constrain their
> particular concrete encodings of RDF all they like, but all the time we
> go around saying "the abstract model's the thing" we have to stick with
> the possibility that model-only applications may be creating apps that
don't
> live by the constraints of some particular RDF syntactic encoding.

Yet you are already happy to have URIs in a model that can't be serialized
using RDF Syntax 1.0, so what possibly could be the problem providing an
improved syntax that allows us to roundtrip RDF *from* syntax to model and
back to syntax. Of course it might not always work if we don't start with
XML syntax, but it doesn't always work today so no loss.

Let's not let edge (sic :-)) cases have inordinate effect on the syntax ...
already the current syntax forces people to use ugly namespaces with those
funny '#' at the end .... causing most of the XML world to be left
scratching their heads.

> While it feels icky making rules
> about URI string formats at the RDF model layer, perhaps a case might be
> made that the practical benefits outweight considerations of
> elegance...?

We need practical. I don't think use cases with "urn:" URIs ought have much
weight on the RDF XML syntax. They are currently broken and ought not break
the rest of the RDF syntax.

I'm not at all sure this needs involve a change to RDF Model. Simply

1) a rule for the construction of a URI from a QName and
2) the construction of a QName from a URI.

Currently the rule for 1) concat, doesn't provide for a reliable rule for 2)
When either don't work it is not easy to serialize RDF as XML.

My suggestion (again):

    qname: prefix:element ->  URI: namespace#element

Perhaps we can discuss further in Cambridge.

-Jonathan

Received on Friday, 26 January 2001 13:36:42 UTC