xml-dev - Re: Namespace Dereferencing: Erratum

[Dan Brickley <Daniel.Brickley@bristol.ac.uk>]

On Sat, 30 Dec 2000, Sean B. Palmer wrote:

> I should point out a rather large error I made about dereferencing
> namespaces in RDF. Contrary to my earlier statement, not *all* RDF systems
> have dereferencable RDF Schemas at the end of their namespaces.

And that's just fine. Some Schemas are just too big (or expensive) to
make available in this manner. Dereferencing is a privilege not a
right... 

For my (still buggy) experiments with a very large RDF vocabulary see
http://lists.w3.org/Archives/Public/www-rdf-interest/1999Dec/0002.html
WordNet contains a vast number of terms; I've exposed these on a
class-by-class basis rather than as a vast RDF dump. Similar large
vocabularies (eg. UMLS in Medicine) require license agreements btw, so
might be access controlled. You can look up rude words in WordNet by
appending them to the base namespace URI, http://xmlns.com/wordnet/1.6/
eg. http://xmlns.com/wordnet/1.6/Scoundrel

> For
> example, Dublin Core and RSS have no such Schemas, only a prose definition
> at the end of their namespaces.

Take a closer look...

http://purl.org/dc/elements/1.1/ redirects to 
http://purl.org/dc/documents/rec-dces-19990702.htm which is HTML
(not XHTML yet, though that's in the works) with an
embedded RDF Schema describing the Dublin Core property set. There's
also a link rel="meta" in the HTML portion pointing to 
http://purl.org/dc/documents/rec-dces-19990702.htm.rdf where we describe
the 1.1 spec using, er, Dublin Core 1.1. 

Similarly for RSS 1.0, http://purl.org/rss/1.0/ redirects to the XHTML
spec currently living at http://www.egroups.com/files/rss-dev/specification.html
which in turn contains a typed link to an RDF Schema for RSS 1.0
properties and classes,
<link rel="meta" type="text/xml" href="./schema.rdf" />

In both cases, doing a GET on the namespace URI, following HTTP
redirections, and taking note of embedded LINK REL metadata gets one to
the RDF.

So there's something for everyone with DC and RSS; indirection from
stable to less-stable URIs. Human-oriented prose descriptions, and
machine-oriented RDF assertions. The details of how we hang all this off
the namespace URI might need polishing, but the broad approach seems
sane.

Dan