UML22 — section 13.3.2 – doc ptc/2006-04-02, v.2.1

Issue a: ) in behaviour description (section 13.3.2 – doc ptc/2006-04-02, v.2.1) precise more formally and explicitely which elements can have behaviors, and how the behavior context is defined.

Typically clarification should say something like:

• [A] Any subclasses of BehavioredClassifier (that is: Collaboration, Class, Actor, UseCase) can have a Behavior and its context is defined through the “context” association

• [B] Any subclasses of BehavioralFeature (that is: xxx to be listed xxx) can have a Behavior and its context is defined through the “specification” association

• [C] Additionally, Transitions and States can have a Behavior and its context is defined by the first BehavioredClassifier reached through their “owned” relation

• [D] A Behavior can stand alone and be its own context (e.g. as equivalent to a C/C++ program)

è Is it here necessary to add a context association from the Behavior to itself…? or should we consider that in this case it is always owned by a modelling element (eg a package) that defines its context… and should we explicitly define to which kind of element this can be considered and add these elements to the list of the [C] situation ?

Discussion: The issue is somewhat confusing in its wording when it asks what “elements can have behaviors”. In one reading, only BehavioredClassifier can have behaviors. Probably the issue means to ask what “elements can own behaviors”. It would be not in the style of the UML specification to summarize in a central location such information, as this would conflict with the object-oriented style of the specification, or it would cause a maintenance difficulty. Behavior::context clearly defines how the context object is determined, independent of the type of behavior or its owner. Disposition: Closed, no change