Could you please suggest what the semantics of mapping a single field to multiple columns means? Are you thinking of a multi-column foreign key usage for references (relationships) or do you see it applying to primitive values as well?
Certainly multiple columns are needed for compound foreign keys. But being able to define multi-column mappings is also important to vendors who allow their users to create custom mappings. For example, one of Kodo's custom mapping samples maps a java.awt.Point to two columns -- one for the X coordinate and one for the Y coordinate. I don't think the DTD should make expressing mappings like that unnecessarily difficult.
