Hello, I read the mails about RDFa and Microdata and idea came to my mind. There is a technique called content negotiation and also LinkedData (see [1])
Basically the Web server returns the content based on what the client requested. For Wikipedia this would mean: if the browser (it normally does) sends a request to Wikipedia with "Accept: text/html" in its header the normal html view is returned (the one we know) else if some application would access Wikipedia and wants the metadata it can send "Accept: application/rdf+xml" in the request header and would get a 303 redirect.
Wikipedia already exists as RDF in DBpedia, so it could redirect for example there, returning RDF of the page.
Example: http://en.wikipedia.org/wiki/Leipzig with "Accept: text/html" shows the normal page.
http://en.wikipedia.org/wiki/Leipzig with "Accept: application/rdf+xml" redirects to http://dbpedia.org/data/Leipzig (see here for a human friendly version in a browser at http://dbpedia.org/resource/Leipzig )
This would be easy to implement and much less intrusive as writing markup manually.
There is also a synchronized version of DBpedia (DBpedia-Live with a 5 minute update delay), which is still in Beta [2]
Regards, Sebastian
[1] http://www4.wiwiss.fu-berlin.de/bizer/pub/LinkedDataTutorial/ in 2.1 there is an image [2] http://dbpedia-live.openlinksw.com/resource/Leipzig