[Modeling] Interaction protocol and method calls

Wagner, G.R. G.R.Wagner@tm.tue.nl
Tue, 18 Mar 2003 18:50:20 +0100


> I also agree with Renato. It seems to me that putting the 
> interaction protocol information together with agemt 
> implementation details such as method calls violates the 
> autonomy of the agents themselves. The specification of the
> interaction protocol should not engender any constraints on 
> the part of the agent designer beyond the support of the 
> interaction; in other words, the agent implementor should 
> be free to use whatever methods he deems proper for
> his agent implementation.  

You make an important point. But the question raised
may also be viewed at the level of design modeling,
not being concerned about implementation issues,
if "method" stands for "action", in general. Since
interaction protocols model only communicative actions,
the question then is how to include non-communicative 
actions in an interaction diagram?

-Gerd

 
> Renato Levy wrote:
> 
> > I respectfully disagree, and personally favor a set of 
> diagrams, perhaps
> > with hyper links betweeen each other. I very afraid of 
> putting too much
> > detail in one diagram, and loose its main meanning in the 
> bulkness of the
> > notation.
> >
> > >Marc-Philippe Huget wrote:
> > > > I just received a good question that must be considered 
> even if I
> > > > have my thoughts about it. The person represents agents 
> as a set of
> > > > objects:
> > > > one for communication, one for means-end reasoning, one 
> for scheduling,
> > > > etc. She wants to write an interaction diagram that presents the
> > > > exchange of messages between agents but as well, the 
> method invocation
> > > > between the agent as whole and the objects constituting 
> it. The aim is
> > > > to link messages to actions triggered by these messages.
> > >
> > >I agree that it is important to be able to do this, although it
> > >depends on what use is being made of the interaction 
> diagram.  If the
> > >diagram is being used as an external representation of the system
> > >(e.g. to aid debugging, as suggested in the AAMAS 2002 paper by
> > >Poutakidis et al., 
http://doi.acm.org/10.1145/544862.544966), then the
> >connection of messages to actions doesn't seem to be required.
> >However, if the diagram is intended to be a fully detailed
> >specification of an interaction protocol then it seems important to
> >indicate what information is contained within the message and what it
> >is that the agent should do with that information (e.g make a
> >particular type of decision).  I have a paper discussing this issue at
> >http://CEUR-WS.org/Vol-66/oas02-16.pdf, although it uses Petri net
> >notation, not AUML.
> >
> >This requirement could possibly be dealt with by making connections
> >between different kinds of AUML diagram, but I think it would be
> >better if there were a way to define an interaction protocol and the
> >connection between messages and actions in a single diagram.
> >
> >- Stephen
>
> Renato Levy
> Principal Scientist
> Intelligent Automation, Inc.
> 7519 Standish Place, ste 200
> Rockville, MD 20855
> phone: (301)294-5241
> fax: (301)294-5201
> WWW: http://www.i-a-i.com
>
> _______________________________________________
> Modeling mailing list
> Modeling@www.fipa.org
> http://fipa.org/mailman/listinfo/modeling