On Not Designing XML languages
Tim Bray discusses what he calls the "big five" XML languages and argues that whatever you're trying to do, you probably don't need to design a new one, particularly if your application falls anywhere near one of these. I'm inclined to agree.
Language design is hard work, and new languages almost never get any uptake (cf. Esperanto). The same seems to be true for spoken languages, programming languages, and data-description languages.
Tim says that "any non-trivial" XML language is going to have constraints that can't be checked with existing schema languages. That sounds like an interesting problem. Is there a list of such constraints somewhere?