As far as I can tell, the UML standard does not provide guidance for creating traceable logical and physical data models. By traceable, I mean the ability to have separate models which are linked in a way that allows a modeler to trace from either direction where a modeling object is used in either model.
So this begs the question:
If UML tools support only a single row in the Zachman framework at a time, which one?
I'm guessing that theoretically the answer could be:
Whichever one you want it to support....
In my experience, though, UML models are prepared solely for a physical view of a specific system/application. I'm not writing about what the standard says, but how these standards are applied in real life.
Could it be that these models reflect the rows in which most developers work? Or is it something inherent in the modeling techniques themselves? And what about Conceptual models? Or any of the other rows in the Framework?