RFE 6405862 changed CompositeType so that it can now have zero items. This follows 5072004, which allows a CompositeData to have additional items beyond the ones that are declared in the CompositeType, and that is the principal intended use of the change. Nevertheless, it seems odd to require that a CompositeType with zero items *must* be associated with additional items in the CompositeData. In other words, there is now a corner case where if the set of additional items happens to be zero the CompositeData will be rejected.

The proposed change is to remove the constraint whereby the items given to either of the CompositeDataSupport constructors must not be empty.