Re: Another view on analysis and ER

Quoth David BL:
> I don't interpret predicates as anything other than predicates. An
> instantiation of a predicate can be used to record a relationship that
> exists between entities. A set of attribute values can sometimes
> identify an entity.

Then I have misunderstood you. I thought this whole discussion was about if it makes sense to look at a predicate and say "that is an entity" or "that is a relationship"---with regard to the predicate as a whole, that is.

I have tried to argue against the notion that most relvars correspond to entities (or entity classes/types, to be more precise), and that relationships correspond to foreign keys (except for many-to-many-relationships, non-binary relationships etc.). If instead you subscribe to the view that all relvars correspond to relationships; and that entities, to the degree that they are present in a relational model/design, are (represented by) attribute values (or combinations of them), then I believe our whole disagreement disappears. This is the view I attribute to the ORM/NIAM method/notation. It seems also to be what Jan Hidders espouses, and since he as far as I can tell consider E/R and ORM more or less the same, it seems likely that other may have this view of E/R as well.

I believe this is an uncommon view of E/R, however. When you see a box named Person with attributes PersonID, Name and BirthDate, you call this box an entity (class/type). And when you design a relational database from it, you'd probably create a relvar Person with the same attributes. To then say that the relvar does not represent the entity Person, but rather the relationship between persons, names and birth dates, is unusual. I agree with it, but it is what I consider to be the difference between E/R and ORM.

If we have a common understanding about this, then I agree with your marriage stuff as well. We have just been using the same words with different meanings.

-- 
Jon