C4S: Coordination-centric Change and Consistency Support

The research proposal addresses change impact analysis in development efforts of complex, safety-critical software (e.g., air traffic control systems). Such software typically exhibits a long lifecycle, involves various engineers (beyond pure software developers) across the entire development lifecycle and is subject to countless changes in the scope of maintenance and evolution activities. Under such conditions proper impact analysis needs to go beyond knowing which artifacts are affected (traditional traceability). Successfully implementing a change requires awareness which engineers to involve and why, how these engineers should be coordinated and communicate, and how these engineers should correctly propagate changes. Hence, what is an efficient and effective way to support engineers in change impact analysis and consistency maintenance? Pure top-down specified processes are insufficient to provide actionable guidance. This proposed work, therefore, investigates methods and techniques for bottom-up determining the underlying communication, coordination, and joint work relations among engineers, artifacts, and tasks. Monitoring such low-level events enables a holistic bottom-up activity view that constitutes the basis for extensive change management support: (i) Dependency Awareness and Consistency Support: provides sophisticated artifact dependencies and recommendations for maintaining artifact consistency beyond core development activities. (ii) Coordination Know-how Learning: generalizes and extracts insight into which coordination structures prevailed, how much collaboration occurred, and which expertise was involved, thus enabling to deduct how these properties affect the duration, quality outcome, and resource utilization of various development activities required for change implementation; and (iii) Coordination Guidance: provides actionable recommendations for a given change management instance based on previously learned best practices. Key research questions address the challenges of dealing with incomplete, incorrect, and uncertain information in detail.

In software engineering projects, a significant amount of development effort is directed at improving the product by fixing errors, introducing new features, or making the product more flexible, secure, or efficient. Each effort of changing the software product might have side effects on some not directly involved part of the software that needs to be evaluated. A change impact might require someone else also to change another part of the software. Change impact analysis therefore comes in two forms: what will likely have to change (which code or other artifact), and who will do the change (respectively who needs to coordinate to correctly implement a change). Change impact analysis support thus addresses artifacts and engineers. This project therefore asked the question how these two aspects can be addressed together. Effective change impact propagation needs both. This project developed techniques, models, and prototypes that aim to identify hidden dependencies between engineering artifacts (not limited to source code but also including requirements, architecture models, test cases, and variability models). The novel aspect of these techniques is that they don't determine a priori where the dependencies are and what needs to be done exactly when, but observe what the engineers do and then identify who needs to talk to each other, when are artifacts ready so that the next engineer can begin his/her work. The work in this project was evaluated with open source software projects as well as industrial software from a safety-critical domain. The achieved results have also pointed to the benefit this approach potentially can realize in other domains beyond software systems such as the engineering of hardware-centric products: for example, connected machines in the producing industry.