Conclusion - Model-Driven and Software Product Line Engineering - page 242

engineering. Model-driven engineering techniques and tools
have the potential to increase the productivity and quality of
software engineering processes significantly. The integration
of both approaches requires a tight coupling and tuning of
the derivation chain. To develop complex and modern systems,
we can use several points of view, which can be captured
in different models. This implies capturing the variability in
different models conforming to metamodels, configuring the
product, and defining an automated production chain relying
on a multi-staged process. Examples of variability models
are feature models that organize the client requirements
whereas architectural models are concerned with the structure
of software. The derivation process is a complex task, and
it requires sequences of transformations with a precise
scheduling. Decision models are artifacts specifying base and
specific transformation rules for scheduling the rules for
both commonality and variability. Decision models capture
transformational aspects of the baseline derivation process to
derive specific variable applications.
This general MD-SPL process is effective, but it is not
always flexible enough for fine-grained configuration. We have
to go further to configure differently several instances of the
same kind of artifacts. The purpose of the FieSta approach
is to extend MD-SPL processes to cope with fine-grained
variation and configuration. The fine-grained variation arises
when we need to define variable artifacts of the same kind.
Coarse-grained approaches do not allow the configuration
of different instances of the same metamodel. To solve this
issue, the concepts of binding and constraint models are
introduced. Binding models allow the capture of the links
between a model and a variability model, thus, enabling the
fine-grained configuration of model elements. On the other
hand, the constraint model specifies precisely the semantics
of the bindings using cardinality and structural dependency
properties. Metamodels were created to support the creation
of constraint, binding, and decision models. The baseline
scheduling of the rules is modified according to valid feature