From a Toolserver account? just query the db or from a non toolserver use, probably just the API from the wmf servers but if you cause too much load you would get blocked.
The article text can not be fetched from the database on the toolserver.
Using the API would work just as well as Special:export, as long as as many pages are fetched at a time as possible. One request per page would suck.
-- daniel
Looks like the API will do the trick - I can get 50 articles at a time. Someone asked what I'm up to - I'm setting up a cache of pages with hatnotes that link to disambig pages, to begin the process of identifying intentional vs. non-intentional disambig links. I plan to refresh the cache daily, which will definitely involve re-checking hundreds of articles, if not thousands.
- Jason
toolserver-l@lists.wikimedia.org