Antipatterns for Architectural Knowledge Management

Recent research on Software Architecture has recovered its original emphasis on keeping track of design decisions and their rationales during software development, compiling them under the name of Architectural Knowledge (AK). This knowledge is composed of explicit, atomic decision assets, which relate to each other creating a decision network structure. We argue that relationships in these networks of AK contain valuable information, in particular when they describe negative semantics. We use reusable knowledge, in the form of antipatterns, to exploit and manage these negative semantic relationships systematically. After examining and classifying the kinds of AK relationships, we describe a method that enriches this network by means of antipattern structures. To show the feasibility and suitability of this approach, we provide a proof-of-concept by applying it to an existing process, ATRIUM. A concrete example illustrates our approach in which we use the Excessive Dynamic Allocation performance antipattern against the classic Gas Station metaphor. Results of the use of the presented approach into three different projects with different complexities show both the feasibility and applicability of our method. The combination of this model-driven support and explicit AK makes it possible to go beyond traceability to a more proactive AK management system that may additionally trigger modifications in the final architecture.