A Software Engineering Framework for Context-Aware Service-Based Processes in Pervasive Environments - Advanced Design Approaches to Emerging Software Systems - page 125

is alleviated from defining the context features as
it is the role of the context modeler to define the
context feature model that could serve as much
applications as possible. In addition, unlike their
approach, using the proposed context modeling
approach we can describe context at arbitrary
different levels of granularity which is necessary
to serve different applications' needs.
Similar to our work, (Sheng et al. 2009) pre-
sented ContextServ, a platform for simplifying
the development of context-aware Web services,
which adopts model-driven development where
context-aware Web services are specified using
ContextUML, a UML based modeling language.
ContextServ offers a set of automated tools for
generating and deploying executable implementa-
tions of context-aware Web services. However,
Apto approach takes a step further. Typically, ac-
cording the separation of concern principle, the
application developer has to focus on the core ap-
plication business logic and then define separately
the customization and business rules, and weave
them to the core application. Apto separates the
system modeling into four separate models; and
it is able to generate according to the acquired
context information the customized version of the
process. Moreover, context-aware Web services
should continue to work even in the absence of
context information. Unlike their approach which
is tightly dependent on context, Apto adopt the
basic process logic idea which will be generated
in absence of context.
(Grassi and Sindico 2007) presented a model-
driven and aspect-oriented approach to deal with
context-aware adaptation in the design process
of an application. Similar to our approach, they
consider the adaptation as a crosscutting concern
with respect to the core application logic; and
their approach facilitates the plugging of different
adaptation strategies within the same basic applica-
tion, tailoring it for different contexts. However,
as they leverage the idea of aspect weaving, their
approach may not be flexible enough to accom-
modate the deletion or changing of application

tasks; which means that the approach could only
be used to accommodate adding different tasks
according to different contexts.
(Muller et al. 2004) propose “AgentWork”, an
interesting approach for workflow adaptation to
customize the hospital cancer treatment workflow
to suit each patient's medical profile by adding and
deleting tasks in the running workflow instance
according to the predefined extended ECA rules.
The adaptation in this approach provides dynamic
and automatic workflow adaptations and suggests
and implements a predictive adaptation strategy.
Apto, on the other hand, takes another approach
so that adaptation can be applied to processes
modeled and developed without an adaptation
possibility in mind and independently of specific
usage contexts.
VxBPEL (Koning et al 2009) is an adaptation
language that is able to capture variability in pro-
cesses developed in the BPEL language. VxBPEL
provides the possibility to capture variation points,
variants and realization relations between these
variation points. Defining this variability informa-
tion allows capture of a family of processes within
one process definition and switching between
these family members at run-time. Unlike Apto,
VxBPEL works on the code level and the variants
are mixed with the process business logic which
may add complexity to the process developer
task. Further, unlike the generative approach of
Apto, VxBPEL is specific to the BPEL language.
Another interesting work that is similar to
our work is the Provop approach (Reichert et
al. 2009), which provides a flexible solution for
managing process variants following an opera-
tional approach to configure the process variant
out of a basic process. This is achieved by apply-
ing a set of well-defined change operations to it.
However, Apto deviates from Provop in that it
uses the MDD approach and defines the evolu-
tion fragments as evolution model elements not
as change operations.