[Modeling] Roles [was Comments on interaction diagram modeling doc]

James Odell email@jamesodell.com
Wed, 19 Mar 2003 17:56:59 -0500

On 3/19/03 5:22 PM, "Lin Padgham" indited:

> I agree with Misty that agent types are not necessarily known in open
> systems, so there we need to use roles.  However AUML should also be
> suitable for designing closed agent systems and there I believe we
> also need a notion of agent types. So just defining roles is not
> enough. 
> Possibly it is not necessary to distinguish in interaction diagrams
> and we could understand either role or agent type - where the
> difference is really to do with granularity and at what level of
> detail people wish to express things?  If we do take this approach
> though, it may be necessary to address questions around what freedom
> designers/implementors have in following a protocol, if they combine
> several roles within a single agent. This arises from the fact that
> protocols are both a design specification tool, but also a
> standardisation mechanism. We may need to address these 2 uses
> separately?

What is the difference between a role and an agent type?  This is a really
important thing that we all need to solve and standardize.  As someone
working with this issue for several years, I can say that it is not easy to
reach a consensus.  We need a formal distinction.