Wiki2text

Too bad about the complex template formatting getting in the way of parsing. It sounds like a hard problem to solve from the raw XML too, forcing one to take steps like rendering the documents and extracting the rendered output, solving the template issue since it's now all rendered, but then instead losing the logical structure and instead getting a display-oriented structure.

I also think this is counterintuitive from a project that is about opening the world's information to the masses. You'd think a problem with heavy use of templates -- possibly and ironically making especially the most popular and requested articles harder to parse -- would have been solved by a cleaner version in the dump besides this XML-with-a-Turing-complete-template-language. It sounds like it makes Wikipedia data interchange really tricky, raising the question "who are they dumping for?"