[Modeling] Doubt about Roles and Joint Actions

Joaquin Peņa joaquinp@us.es
Thu, 2 Oct 2003 09:55:03 +0200

Hello Marc and all:

> Could you be more specific? Actually, I don't understand where is the 
> point. A role participating into several protocols does not 
> necessarily 
> need to be a recursion of roles. I mean roles for these 
> protocols can be 
> quite different. Why do you want to add recursion?

Ok, I'll try an example:

Consider a high level view of a MAS where in a GAIA Role Model called A
we have three roles: R1, R2, and R3, and three protocols p1, p2 and p3.
In a refined view of the Role Model we may consider a new role model for
every protocol where p1, p2 and p3 are further detailed with finer grain

Role Model A:
	Roles: R1, R2, R3
	Protocols: p1(R1,R2), p2(R....), ... // To short, participants
are shown in brackets

Well, consider now one of the refined role models where we work on
detailing the protocol p1. If in role model A only R1 and R2
participated in p1, refined views of these roles (or new roles, or ...)
appear, namely R1.1 and R2.1: 

Role Model B: 		// A refined role model of p1
	Roles: R1.1, R2.1
	Protocols: p1.1(R1,R2), p1.2(R....), ...

Thus, we may think:

1) That roles R1.1 and R2.1 inherit from R1 (behaviour inheritance in
the sense of Barbara Liskov's paper) 
2) That R1 is composed of two roles R1.1 and R2.1 (the union of
interfaces of R1.1 and R2.1 is the interface of R1). 
3) No relation needed?? Relation at role model level only??
4) XXXX Your opinion?