I think IBM DFDL already has this label. I think we need to standardize it as part of DFDL.

Seems DFDL schemas, to make compilation feasible, and avoid many unnecessary warnings and spurious weirdness, must have a way of deciding what top level elements are intended to be Root elements.

Some global element decls might have clearly upward-and-outward relative path expressions inside them, and such elements are unsuitable for use as root elements. We don't want to issue diagnostics for these unless the user expresses intent to use one of them as a root. In that case they have made a mistake and need a diagnostic.

I believe one could sidestep the issue by only creating elements for things that could be roots, and use global types or global groups otherwise. But we have users who really want to use lots of global elements and element refs for their own reasons (they want the schema to have a certain regular structure for non-DFDL reasons)

We definitely do see a need for many elements of a schema to be possible roots. This is particularly important for message-oriented formats where there is no natural root element that surrounds all the messages because they are arriving on a non-bounded stream that is effectively open "forever".

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com

Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy