Hi everybody,

Actually, we don't have one - we've often talked about it but it's never bubbled to to top of the stack.

Typically (as we're biased :-)) we've tended to go the other direction.

Hi,

Do you mean, that we create model serialization's XSD schema? If it is, I think it is different. I mean this schema is essential, but before DSL is coming up, all of us who think about model aspect use XSD as language for describing metamodels. So I guess most of us have a lot of such metamodels. And now I sure there will be tendencies of moving from old XSD metamodels to DSL.

Yes, I totally agree - folks will have many XSDs, some of which might make interesting starting points for DSL models.

I was just poitning out that internally we've tended to have DSL models around maybe a lot longer than our customers, so we don't have as much backlog in our group.

There are a lot of entertaining structures possible in an XSD that are don't translate trivially to a more straighforward O-O model like ours, so working out the heuristic seems to be the hard part. I think its relatively easy to code up once you have a set of rules. Potentially just the same set of rules that xsd.exe uses woudl be better than nothing.