Information Technology Reference
In-Depth Information
the translation problem. The existence of many
i*
variants implies the existence of
different semantic-pragmatic communities and the equivalences or mappings among
metamodels (in fact, from the
i*
supermetamodel to variants' metamodels) should be
a matter of a meaning-making process inside that specific community. Just to mention
an example, row 3 and row 4 are proposing two different strategies for dealing with
one specific construct (dependency with an actor as dependee) that is supported in
the departing metamodel but not in the target metamodel. Choosing one or another
depends on the target community.
6 Conclusions and Future Work
In this paper we have dealt with the problem of interoperability among
i*
variants
under a metamodel perspective. We organized the research into 4 questions which we
think have been satisfactorily explored:
-
We have surveyed 146 proposals presented by the community in the last 5 years,
and we have classified them in terms of additions, removals and modifications of
i*
constructs organized into six categories. Thus, we have obtained a quite com-
plete characterization of the
i*
variability to support interoperability goals.
-
We have proposed a framework for the interoperability problem based on an ap-
proach that can be considered consolidated and widespread in the MDE commu-
nity. We have customised this framework about model evolution into the
i*
model
interoperability problem. As cornerstone of this customization, we have defined a
supermetamodel for
i*
that eases interoperability by metamodel containment.
-
Given the framework above, we have classified the surveyed
i*
variation types in
terms of the semantic impact of their translation, having then a general idea about
what types of information loss may happen and to what extent the analyst may
provide information (mappings) to minimize this loss.
-
We have defined a process for translating a model compliant to one metamodel to
another compliant to a different metamodel, and we have demonstrated how it
works by exploring the translations of models built with the OME tool to the
jUCMNav tool.
As a summary, we may say that we have provided a first consolidated step towards
not just syntactic but also semantic interoperability in the
i*
framework. Our approach
may help creating a repository of
i*
models (using the
i*
supermetamodel as universal
reference model), may favour the application of techniques that work over different
metamodels, and may possibilitate the interchange of models between tools.
Our future work spreads along four different axes. First, improving the translation
algorithm which is currently able to deal just with reductions, to tackle increasing
modulo variations. Second, to offer a portfolio of tool interconnections in similar way
to the one between OME to jUCMNav explained here (in fact, we have a more com-
plete case of interconnection among the jUCMNav and H
i
ME [22] tools, described in
[21]). Third, consider not just syntax and semantics but also ontological issues in the
translation process. Forth, digging into more details of Wachsmuth's framework for
proposing translation heuristics depending on the refactoring distance between the
source and target metamodels, allowing thus having some default translation rules
instead of a pure case-by-case analysis (although as remarked at the end of Section 5,
translation will ultimately depend on the community ontological perception of
i*
).
