[Modeling] Modeling an Agent Class- register your opinion

Paola Turci turci@ce.unipr.it
Mon, 23 Jun 2003 16:32:28 +0200

Hi Gerd,

I am giving you a very brief answer since I agree with Jim ... it is time to
work on concrete proposals :)

Gerd wrote:

> > I think that AgentClass could be an unfortunate choice. Actually, for
> > now I have no a good proposal, however since I think that it
> > would be better
> > to extend Classifier, maybe a more appropriate choice could be
> > AgentClassifier.
> Why should this make sense? Classifier is an auxiliary technical term
> to have a generalization of Class, Association, AssociationClass,
> DataType and some other things. Your proposal seems to suggest that
> there is, likewise, a reason to generalize similar things for agents.
> But there is no "AgentAssociation", "AgentDataType" and the like.

I have thought it over and may be there was a misunderstanding.
As it is written in the document we sent few weeks ago (AgentClassAnalysis),
we intend to define an "AgentClass" stereotype that applies to Classifier.
So, "AgentClass" would be a kind of Classifier such as "Class". Moreover we
intend to exploit covariant associations in order to redefine only what we
As far as "data type" is concerned, (from UML 2.0 "data type is a special
kind of classifier, similar to a class, whose instances are values (not
objects)..." I do not understand why we should generalize it.

> -Gerd