using wikipedia URLs as URIs to identify resources for RDF - semantic web - purposes.

This looks like a good idea to me too! And it could be as simple as:

@prefix wp http://en.wikipedia.org/wiki/

or for example another language: slovenian. wikipedia has inter-language 'sameAs' links

@prefix wpsl http://sl.wikipedia.org/wiki/

James Tauber's blog

this gives us a namespace with lots of terms already precisely named, defined and explained too. statements using it can be very expressive. and undefined terms can be easily added by anyone.

for example:

wp:VAX wp:Endianness wp:little-endian .

which reads "VAX is a little-endian architecture"

 wp:World%20War%202    event:startYear    wp:1941 .

"the second world war started in the year 1941"

1. RDF
2. Wikipedia namespace
3. ...
4. Profit!

Stability: such a namespace should ideally be based on a stable release of Wikipedia (1.0?) where terms would never change. however wikipedia is for now still without a stable version. and no a random snapshot won't make things much easier.

This is not a very controlled vocabulary, but i think primarily its large size and the intrinsic process of resolving conflicting terms that goes on in Wikipedia make this a rather precise way of referring to many concepts, people, places, events and times, factual and fictional, future and past. and it's already done, mostly.

For OpenDDB, similarly one can use Freshmeat and SourceForge URLs.

 @prefix fm http://freshmeat.net/projects/ 
 @prefix sf http://sourceforge.net/projects/
 
 <http://freshmeat.net/projects/debian> foo:MemberCount 1000.
 fm:fedora_project owl:SameAs <http://www.fedora.org/>. 

a strange fact - most of <wp:Dublin Core> terms are undefined in wikipedia.