Thanks for the reference, I hadn't seen that. I really believe, though, that the solution to interoperability is to use protocol buffer schemas directly; not to introduce yet another schema language.