Hi im OR mapping and have hit a massive fork in the road, do i either follow CJ Dates idea & equate [object] classes to types (domans), or follow the classes to tables (relations) approach.

To shed some light what im doing, im mapping UML based metamodels into the DBMS, so far if i follow date, i maintain object encapsulation, metamodel semantics etc. The other way (class2table) i can properly apply normalisation etc which is not possible with dates idea.

Any ideas?

If your UML models model entities (like Employee, Department, etc.) then they naturally map to tables.

Date's point is that entities like Employee and Department should not really be modelled as "types" in the first place.

Hi Tony,

The metamodel im modeling is a derivetive of the UML-based object model, so im modeling such classes inc: classifier, class, attribute, namespace, element etc - so would these classes be best mapped to tables?

Can i ask in which literature does date states the thing about entities like Employee and Department should not really be modelled as types.

I should think so.

That is (I believe) the view of C J Date and Hugh Darwen as expressed in the Third Manifesto and other writings. They would say that a typical "entity" such as an Employee or a Department maps naturally to a table, but not to a data type (class).