While I'm a bit dissatisfied with the XML bit, any excuse to write a parser is a good one for me! I haven't given the spec for their Markup language a close reading yet, but this seems like as good an opportunity as any to fool around with a precedence-climbing recdesc approach. I also like their approach to the [non-?]competition generally - commenting on interesting entries rather than just ranking them.
Yeah. I'm not crazy about the XML thing either (as I noted in a footnote) but all we're using it for is a trivial way to dump an AST that's easy to specify.