[Modeling] Comments on interaction diagram modeling doc

[James Odell]

Lin,

> Possibly yes.  How do you differntiate between 1 and 2? Are the agents
> instances or types? Or can they be either?

The UML 2.0 format is effectively <agentName:agentRole>, where you can
specify either or both variables.  For example, you can say
- "Mario:Employee" where Mario is playing the role of employee;
- "Mario:" where Mario is inteacting (and can be though of as playing some
unnamed role; or
- ":Employee" where some unnamed agent is playing the role of employee;

 
> When you say "UML can support" I'm unsure what you mean. UML doesn't
> have agents does it??? (I am not a UML expert and certainly am not on
> top of the latest things :-)

UML, as it is currently written, does not use the term "agent".  Instead, it
calls them "active objects."  In FIPA AUML, we can just substitute the agent
for active object with minimal change to the UML metamodel.  Furthermore,
UML 2.0 does not use the term "role" here per se, but it does imply the
name.  (If you want to know why, I could give you the historical
background.)


> I doubt we would ever explicitly show the different roles of an agent
> by using different lifelines, tho maybe. More likely we would just
> identify the agent type and combine 1 and 2.

So, now you can express multiple roles for a particular agent via multiple
life lines. For example, three such life lines might be:
- "Mario:Employee" where Mario is playing the role of employee;
- "Mario:PropertyOwner" where Mario is playing the role of a property owner;
- "Mario:Customer" where Mario is playing the role of customer;

Does this make sense?


-Best regards,

Jim