RE: SV: A certain difficulty - lack of action!

On Mon, 28 Feb 2000, Jeff Sussna wrote:

[snip]

> If we're going to be property centric then I would promote the
> predicate to the element name position. You would then have:
> 
> <rdf>
>   <predQname subject="[URI]" object="[URI]"/>
>   <predQname subject="[URI]">[PCDATA]</predQname>
> </rdf>

I know that the aim here is to try to cling to some semblence of
general things written in XML as being "convertable" to RDF.  But I think
it's the wrong way to go.

It seems to me that the cleanest approach to RDF would be to have an RDF
namespace, and to throw away the present overloading of XML namespaces for
RDF's semantic purposes, which is what's happening above.

RDF should have a clean DTD of metadeclarations that can be made in it.
<tuple ...> might be one such declaration, hence things of one of three
possible forms:

<rdf>
	<tuple name="foo" subject="bar" object="baz" />
	<tuple name="foo" subject="bar">baz</tuple>
	<tuple name="foo">
	  <arg val=1>bar</arg>
	  <arg val=2>baz</arg>
	</tuple>
</rdf>

It doesn't get cleaner than that.

I think RDF's approach to XML namespaces is a misuse of XML as it
presently stands.  It's certainly a misuse of SGML, for what that's worth.
It makes RDF difficult to embed into other XML and SGML applications. And
anyway RDF should be handling namespaces in its own way, independent of
whatever direction XML goes so we don't see the XML equivalent of the
Brittle Superclass Problem: RDF overloading XML's stuff in its present
(IMHO goofy) way, then XML moving off in some new direction.

Sean

Received on Monday, 28 February 2000 16:25:44 UTC