Information Technology Reference
In-Depth Information
choice at higher levels of abstractions. The parameterization is achieved through
the use of wildcards but this use of text is minimal. Parameterization of point-
cut expressions is important to address scalability issues. Jacobson and Ng use
textual expression to define parameterized pointcut expressions. Note that the
extend relationship for use case diagrams is a visual representation of a pointcut
but does not contain enough information and therefore has to rely on the textual
representation. The binding rules used by Whittle and Araujo and composition
rules used by Moreira et al. are represented in a textual way. Furthermore, these
rules explicitly link one element with another and do not allow parameterized
expressions. Barros and Gomes also use a textual representation of pointcuts
and also explicitly link nodes in UML activity diagrams, not allowing parame-
terized expressions. de Bruin and van Vliet define aspects (refinement maps) in
a visual way but use limited type matching to merge behavior and structure
from the aspect with the base model. Zdun and Strembeck's approach again is
not applicable to this category.
In previous work, Mussbacher et al. [31] compared composition techniques of
several scenario-based approaches to aspect-oriented requirements engineering
and concluded that AoUCM have a flexible and exhaustive composition
technique with significant advantages over the other approaches mentioned in
this section.
In terms of abstraction levels, UCMs are at the same level as the techniques
used by Jacobson and Ng, Barros and Gomes, and Zdun and Strembeck. Note
that the examples in Barros and Gomes represent rather low-level control flow
but that activity diagrams can also be used to describe high-level workflow.
UCMs abstract from message and data details. UCMs can therefore be used
earlier than message-based behavioral models but also contain more information
than UML use case diagrams. Therefore, UCMs are at a higher level of ab-
straction than the work by Whittle and Araujo which is at the message/state
machine level. Moreira et al. make use of some models that are at the same and
some models that are at a lower level of abstraction than UCMs.
Furthermore, UCMs model the whole system making it possible to reason
about interactions between various use cases or scenarios. UCMs have already
been used not only for scenario interaction detection but also for performance
analysis and testing purposes. By applying these research results to the composed
UCM model containing the base and aspects, it is possible to achieve greater
confidence in the model at a very early stage in the development. This is an
advantage over all other techniques mentioned in this section as these techniques
model scenarios in isolation.
This paper reports on the first results of a much larger research goal. In
the long term, we plan to also extend the GRL part of URN with aspects and
synchronize the extensions to the GRL part with the extensions to the UCM part.
The matching and composition algorithms have been implemented and tested
and will soon be available in an ocial release of the jUCMNav tool. A case
study of a non-trivial e-commerce application is also underway which we hope
will further illustrate the benefits of our approach. UCMs have already been used
 
Search WWH ::




Custom Search