[Modeling] Modeling an Agent Class- composition

Wagner, G.R. G.R.Wagner@tm.tue.nl
Thu, 26 Jun 2003 10:19:58 +0200

> Hong Zhu wrote (on 19 June):
>> We have a agent that represents a department in a university, 
>> and a number of agents as members of the department. When the 
>> department is destroyed, ...

> Composition seems to work perfectly well to me.

Yes, I don't see any problem either.

And contrary to Hong's appeal, we cannot rely on the UML2 spec for
the semantics of aggregation/composition (especially wrt lifetime
dependency), because it is unclear/confused and not well-defined.

They say themselves:

Semantic Variation Points (p.40 of the current UML2 document)
The precise lifecycle semantics of aggregation is a semantic 
variation point.

Semantic Variation Points (p.75 of the current UML2 document)
Precise semantics of shared aggregation varies by application 
area and modeler.

So we better rely on well-justified theories of the part-whole
relationsip, such as sketched in my paper

Towards Ontological Foundations for UML Conceptual Models

which is also a recommended reading in OMG's RFP for an Ontology
Definition Metamodel.