I did some research into Schematron, it being the third schema language often mentioned in addition to W3C Schema and RELAX NG. So far, I'm totally impressed by its elegance, though I'm curious about its processing speed. (Then again, it can't be that much harder to do than processing XSD, right?) Some quick links to resources I found useful:

• Schematron Home
• The reference implementation in XSLT
• James Clark's jinghas support for Schematron as well as Relax NG; it does not rely on the reference implementation, and does some things in Java, some in XSLT
• An interview with Rick Jelliffe