Information Technology Reference
In-Depth Information
notations, but due to its adoption by the community we consider them as major varia-
tions. Thus, we may say that
i*
has three main dialects: the seminal
i*
currently repre-
sented by the wiki definition, GRL, and the language adopted by Tropos.
On top of these three main dialects, we may find many proposals for particular
purposes. Some of them are bound to a particular domain, e.g., security as in Secure-
Tropos [9], or norm compliance as in Nòmos [10]. Others propose very particular
concepts for a particular purpose, like the concept of module or constraint. Finally,
some others propose more fundamental variations affecting the way of modelling, as
the concepts of service, variation point or aspect.
Table 1 presents a comparative analysis of the proposals issued by the community
in the last 5 years. We have carried out a review in the following conferences and
journals for the period 2006-2010: CAiSE, REJ, DKE, IS Journal, RE, ER, RiGiM,
WER,
i
* workshop, and also including the recent book on
i*
[3]. Our goal has not
been carrying out a systematic review but to get a representative sample of the com-
munity proposals in this period as a way to know what the major trends concerning
language variability are. In total, we have found 146 papers about
i*
in these sources
(without including papers talking about goal-modelling, since we are interested in
language-specific issues). From them, we have discarded 83 which are not really
relevant to our goals (i.e., papers not directly related with the constructs offered by the
language). For the remaining 63, the table shows how many of them propose addition,
removal or modification of concepts classified into six different types. It must be
taken into account that a single paper may propose more than one construct variation
and that similar changes are proposed or assumed in different papers. Also it is neces-
sary to remark that most papers just focus on some specific part of the language, in
that case we assume that the other part remains unchanged.
Table 1.
Variations proposed by the
i*
community in the last 5 years (selected venues only).
Each paper increments at most in 1 each column.
Actors
Actor links
Dependencies
Intentional elements (IE)
IE links
Diagrams
New
4
24
10
21
21
19
Removed
8
5
2
1
0
0
Changed
3
1
1
36
43
0
An analysis of this table follows:
-
On actors. The most usual variation is getting rid of the distinction on types of
actors, like remarkably GRL
1
does. Some special type (e.g., “team”) may appear.
-
On actor links. Most of the variants include
is_part_of
and
is_a
but some get rid
of one (e.g., GRL just keeps
is_part_of
) or even both. Of course, having just a ge-
neric type of actor means not having the links bound to specific types like
plays
.
Finally, some proposals use new actor links, like in Nòmos: A
embodies
B means
the domain actor A has to be considered as the legal subject B in a law.
-
On intentional elements. Although all virtually all variants keep the four standard
types (
goal
,
softgoal
,
task
and
resource
), we may find a lot of proposals of new
1
In the rest of the paper, we refer to the GRL implementation supported by the jUCMNav tool,
http://jucmnav.softwareengineering.ca/ucm/bin/view/ProjetSEG/.
