Producing Just Enough Documentation: The Next SAD Version Problem

Abstract

Software architecture knowledge is an important asset in today´s projects, as it serves to share the main design decisions among the project stakeholders. Architectural knowledge is commonly captured by the Software Architecture Document (SAD), an artifact that is useful but can also be costly to produce and maintain. In practice, the SAD often fails to fulll its mission of addressing the stakeholders´ information needs, due to factors such as: detailed or high-level contents that do not consider all stakeholders, outdated documentation, or documentation generated late in the lifecycle, among others. To alleviate this problem, we propose a documentation strategy that seeks to balance the stakeholders´ interests in the SAD against the efforts of producing it. Our strategy is cast as an optimization problem called "the next SAD version problem (NSVP) and several search-based techniques for it are discussed. A preliminary evaluation of our approach has shown its potential for exploring cost-benefit tradeoffs in documentation production.