Concern-Driven Requirements for Adaptive Socio-Technical Systems

点击翻译按钮获取中文摘要

The research program's vision is to build the next generation, model-based software requirements engineering environment based on large-scale software reuse to address the challenges of *high-impact, socio-technical systems*. Such systems are characterized by the complex interrelatedness of people with conflicting goals, technology, organizational processes, and the diverse social context in which these systems are deployed. For example, providing affordable health care to millions of uninsured US citizens falls into this category, as does a United Nations program to coordinate vaccination regimes around the world. Furthermore, systems like that need to constantly adapt to the ever changing social, economic, and technical environment to remain relevant and useful. Large-scale systems have to take many concerns into account (e.g., availability, security). Separation of concerns is the ability to concentrate on one concern, while temporarily forgetting about all other concerns, allowing humans to handle the complexities of such systems. However, some concerns, called *crosscutting concerns*, are intricately interconnected with many other concerns, making it difficult to analyze concerns individually. Existing techniques for advanced separation of concerns address this issue by formalizing the relationship of a crosscutting concern with other concerns, which enables reasoning about the impact of a crosscutting concern on a system. Consequently, systems can be analyzed and built by composing a set of reusable concern units, regardless of whether concerns crosscut or not. Nevertheless, *concern-driven development*, i.e., the systematic application of large-scale concern units across software development phases, has not yet received significant attention. Much can be learned from success stories for software reuse, ranging from generic, small-scale programming libraries to planned, large-scale Software Product Lines (SPLs). SPLs are product-oriented, i.e., they describe a family of similar products, e.g., a set of hearing aids. This research pursues concern-orientation instead - leading to *Software Concern Lines* - to maximize the potential reuse benefit, because concerns are applicable to not only one product family but across a wide range of product families. In addition, the need for run-time adaptation requires support when selecting a reusable unit out of a set of similar, applicable ones for the current context. The lightning speed with which changes in context propagate through today's information world means correct selections need to be made quickly for relevant run-time system adaptation. Such support is not yet readily available. This research proposes concern units that include a set of coordinated, heterogeneous, reusable models at different levels of abstraction and use the most abstract, most high-level requirements model possible when composing concerns to adapt to changed context. The research program's objective is to advance concern-oriented modeling languages, frameworks, and tools for requirements engineering of adaptive socio-technical systems, taking social, economic, and technical context into account for run-time system adaptation with the help of reusable concern units. The research program's impact is improved large-scale reuse and hence improved productivity, leading to innovation in and increased competitiveness of the many organizations served by the Canadian software industry. Software Concern Lines are expected to become a viable development choice for the Canadian software industry even in the context of rapidly evolving software development projects, and will allow the software industry to further specialize, bringing software engineering more in line with other engineering disciplines.