Is that the only style of submission you'll get? You implied that your interpreter was going to have to intuit what the submission means. This may be a stupid question but can't you get a meaningful specification for the schemas you'll be expected to consume? From there you just note which schema style was retrieved and use an output reformatter that writes to that style. Also, have you looked at XQuery? I've heard of it in similar contexts except that it may be already designed to handle the problems you're just going to have to invent around.