Philosophical view of the term "constraint&qu

[Bill Egge]

As I was making objective definitions of UML concepts, I ran across something weird.

The term constraint as it is applied to an association.

On objective inspection of actual associations, the meaning of constraint simply defines the relationship bewteen 2 classes.

So I wondered why it was not simply called, "Defined Association" rather than an "Association Constraint".

In fact, if there is any association between 2 classes it must follow the law of identity and thus all associations are "defined", or at least can be.

There is no such thing as an association without an identity.

The word "constraint" gives the idea that this is a certain kind of association, a constrained one as opposed to one that is not constrained.  But there is no such thing as a non constrained association if constrained is meant to mean the identity of the relationship.

So, the term "constrained" association has no context and makes no sense.

Maybe the UML should call this an "Association Declaration" or something like that, but not a constraint.

[mikewhit]

If you did not place a constraint on the association, that would imply that your model was concerned with all instances of the associated classes.

But by specifying a constraint, you are specifying that only the subset of associations meeting the constraint are valid/permissible within your model.

Or have I missed the point ?