Wikipedia also has an API (although this does little to address tptacek's concerns of database meltdown, unless you cache frequent requests):

http://en.wikipedia.org/w/api.php

http://www.mediawiki.org/wiki/API:Main_page

http://www.mediawiki.org/wiki/API:Query

For example:

http://en.wikipedia.org/w/api.php?action=query&prop=revi...

brings back:

   { 
     "query": { 
       "pages": {
         "1666533": {
           "pageid": 1666533, 
           "ns": 0, 
           "title": "Y Combinator (company)", 
           "revisions": [
             {
               "*": "{{Infobox company\n| company_name = Y Combinator\n 
                     [...redacted full wikitext content of article...]"
             }
           ]
         }
       }
     }
   }

No need to scrape, Wikipedia has full dumps for the articles available for download: http://en.wikipedia.org/wiki/Wikipedia:Database_download#Eng...

Yeah I am aware, but wikipedia changes daily. There would have to be a pretty impressive feature set built on top before many would consider using an out of date wikipedia when they can easily see the latest version.

There is a feed of changes though - both a "realtime" IRC feed that is restricted access and some sort of batch feed (daily? Hourly?)

Hm... for most pages it shouldn't really matter, they don't change drastically within days. For developing or current subjects I guess it would...

This dump is updated once a month, though the API is live.

One problem I've found with the dump is that you still need to render it (unless someone knows of another dump) which has caused us a few problems with the current thing I'm working on.