[Modeling] Re: Methodology-Modeling joint meeting

[Massimo Cossentino]

Dear Jim and all,

this mail is forwarded to the Methodology ML since this TC is directly 
involved in the discussion. For its members, a prelude: the discussion 
started with my proposal to Jim of a joint meeting in Barcelona because "We 
should discuss about our needs in modeling MAS models and the modeling TC 
proposed solutions but also coordinating our actions".  We (Methodology TC) 
had this idea in London and there I hint something about it to Jim, 
probably I was not clear enough



At 21/02/2004 16.05, you wrote:
>Dear Massimo,
>
> > (I put Radovan in Cc since he is now the Methodology TC co-chair)
> > from the beginning of our work (in Palermo) it was supposed that our TCs
> > would coordinate their work in order to produce a methodology and modeling
> > language that could be used together. Until today this was actuated only by
> > the presence of some of us in both the MLs or TC meetings. The need that
> > appeared in London was that in the Methodology TC work several possible MAS
> > meta-models have been identified and we now wonder if it is possible to
> > represent them and the correspondent design artifacts with the FIPA
> > modeling language you are developing.
>
>I think I understand.  I am sending this to the full Modeling TC list,
>because I am concerned for two reasons:
>
>1) For the last two meetings, we have worked on converging on a MAS
>metamodel.  Why do you *now* want to present "several possible MAS
>meta-models" now that we have all arrived a  fairly unified one?


I think that things did not go in that way. Methodology TC clearly declared 
its will to explore the problem of refining a FIPA MAS meta-model from the 
beginning of its activity (as you can also note this is clearly stated in 
our web site at the beginning of page 
http://www.pa.icar.cnr.it/~cossentino/FIPAmeth/metamodel.htm). We are not 
late,  contrarily, we are wondering why you did not considered that we were 
working on similar (not coincident, see below) issues.

IMO there could be some misunderstanding about the use of the MAS 
meta-model term. For us (Methodology TC), it is a structural representation 
of what elements will constitute the system to be designed. In such a model 
there will be agents, roles, and communications but also goals or 
requirements or intentions (according to the approach); it includes 
also  the relationships that there will be among those elements. The design 
process, is in our mind, a set of activities that guide the designer in 
instantiating this meta-model in order to obtain the proper solution for 
his problem (Jim: you should already know this point of view since it is 
discussed in the paper we wrote together some months ago).

In our web site you can find three examples of MAS meta-models (ADELFE, 
GAIA, PASSI).

 From the modeling TC discussions, I derived the conviction that it is in 
your scope to create the modeling support for the above discussed 
meta-model and not defining it (this is clearly a methodological concern). 
It seems to me that your results go in that direction and probably all of 
our meta-models could be described with the FIPA modeling language you are 
defining.
Now, with a collaborative intent, we just would like to verify that; it is 
rather strange that you was not looking for an evaluation stage of your 
results.

Anyway, as you should remember, I (and other methodology TC members) 
already represented this problem during the London meetings


>2) Work in the modeling TC occurs between meetings, yet you have not
>proposed any of the "several possible MAS meta-models have been identified"
>for review by the Modeling TC.  Nothing was posted to the Modeling TC
>mailing list nor was it requested to be posted to the Modeling TC site.

As you should remember I proposed all of that to you in a conversation we 
had in London. There we discussed also about the opportunity of a joint 
meeting



>Massimo,  until now this is the first that we hear from you regarding
>"several possible MAS meta-models."

I'm sorry  but in our ML we already discussed about it and I think you are 
subscribed.



>Can you understand how this makes it
>difficult for those of us in the Modeling TC that have been spending a lot
>of time developing the current model?

Again, I think/hope your model is a 'representation' model and therefore 
this should not compromise our 'structural' meta-model, anyway I do not see 
how the identification of a FIPA MAS (structural) meta-model should fall 
into the activity of a TC that is concerned about the specification of a 
FIPA modeling language. I understand the language will be used to represent 
the structural model but IMO the same language (if it is enough general and 
complete) could represent several different models.

Moreover, as a general position, I think that the importance of the 
identification of a FIPA MAS structural model is transversal to all FIPA 
activities and cannot be under the responsibility of the modeling TC (I am 
a member of it too and I hope to have the right to say my opinion). As a 
matter of fact it is not in the modeling TC work  plan while it is the 
first point of the methodology TC activity.

Nevertheless, if the Modeling TC feels the importance of doing it, in some 
way I can probably support you if this is not in contrast with other TCs' 
work and more in general with other FIPA activities.

>FIPA meeting time is much too short
>now for it to be the only working time the Modeling TC has.  Our
>face-to-face meeting time must be spend prudently -- particularly with our
>current workplan schedule.  It short, we do not have time to start over,
>unless we find that our previous work is broken or can be significantly
>improved -- and if that occurs, we must submit a new workplan.


I understand that this happens because this work is already out of your 
work plan and it is becoming too much time consuming for a collateral activity


>So, if you have a proposal, please send submit it to the Modeling TC by 25
>February.  The group will then discuss it *before* the FIPA meeting.  If
>enough people indicate that these new  "several possible MAS meta-models"
>should be considered, then we will discuss them in Barcelona.  Is that fair?


As I already said you can find our MAS meta-models on our web site (from a 
long time)

If you do not naturally feel the need for a verification of your language 
modeling capabilities by applying it to the Methodology TC artifacts (that 
is obviously time costing for all of us), I think we should proceed in our 
works without any other interrupt.


Regards

Massimo