Information Technology Reference
In-Depth Information
The
i*
supermetamodel as presented is capable to represent as instances those
i*
models built with any of the variations mentioned or referenced so far. In order to
illustrate this expressive power we show, in Figure 5, an object diagram correspond-
ing to the
i*
supermetamodel. It represents a specific
i*
model selected from [15].
We may observe different usual elements (types of
actors
,
goals
,
softgoals
, etc.) then
some particular elements, more precisely costs in contribution links (both a label and
a quantitative value). We have tested the
i*
supermetamodel with additional represen-
tations including service-oriented
i*
[16],
i*
with norms [10] and the different secure-
oriented
i*
variants [9].
Fig. 4.
The
i*
supermetamodel
It is interesting to remark that, in spite of its expressive power, the
i*
supermeta-
model cannot be considered an
i*
variant by itself. Although it is a metamodel, it just
represents a wide set of possible
i*
configurations but considered by itself, there are
hundred of instances of the
i*
supermetamodel that have not any sense into any
i*
community, e.g., a
belief
decomposed into
resources
. Therefore, the
i*
supermeta-
model has to be considered just a reference framework for supporting model interop-
erability. Nevertheless, it must be mentioned that the
i*
supermetamodel does impose
basic syntactic validity conditions for models to be really considered an
i*
variant.
For instance, it is stated through multiplicities that an intentional element cannot be-
long to more than one actor. Other additional conditions are not shown graphically
but exist in the form of OCL integrity constraints. Just to name one, in the case of
DependencySegments
that arrive to an
IElement
, it must hold that the
IElement
is
inside the boundary of the
Actor
linked by the segment:
context
DependencySegment::IElementInsideActor()
inv
:
self.IElement->notEmpty() implies self.iElement.itsActor->notEmpty()
and self.iElement.itsActor = self.Actor
It must be mentioned that the current
i*
supermetamodel proposed here does not
cover the complete range of constructs that appear in the state of the art, that remain
