Further Reading - Model-Driven and Software Product Line Engineering - page 215

architects create decision models during domain engineering.
Decision models are used during application engineering
as the main artifacts to assist in the product derivation.
Decision models, as used in [ATK 00, BAY 00, FOR 08,
DHU 08],are intended to close the gap between variability at a
conceptual level (variation points and variants) and variability
at the implementation level (concrete core assets). Bayer
et al . [BAY 00], in the PuLSE approach, use decision models
as a means to realize domain and architecture decision. Many
authors have been now exploiting this notion, such as [ZIA 06]
who designed it with the abstract factory pattern. [FOR 08]
present decision models as a means for dealing with software
variability and views on decision models that are supposed
to help in variability management. Some mechanisms for
supporting the process of decision modeling and resolving
decision models have been introduced. In general terms,
every SPL approach based on MDE introduces a form of
decision model or similar artifacts in order to assist the
derivation process (see Chapter 4). For instance, Wagelaar's
approach [WAG 05, WAG 08a, WAG 08b] uses platform
dependency constraints and superimposition variants, an ATL
facility, in order to assist the derivation process; Loughran
et al .'s approach [LOU 08, SAN 08], defines a language to
capture variability and decision models.
MDE and AOP provide good opportunities to automate
the product derivation, thus improving greatly the
production chain. We propose the use of explicit decision
models in the context of MDE as a mechanism for
composition of transformation rules based on feature
configurations [ARB 09]. This mechanism can be used in
conjunction with transformation languages, which provide
facilities to compose transformation rules. One of the strengths
of our approach is that it provides a model-based strategy
for defining decision models. It can be, thus, the base of
a platform independent strategy. In particular, we used
the oAW modeling framework and the Xtend and Xpand