Concern-Oriented Reuse

点击翻译按钮获取中文摘要

Reuse can increase the quality of software, while reducing development time and cost. Successful reuse techniques range from class libraries to components, services, frameworks, and Software Product Lines. Domain-specific knowledge can be reused by embedding it in domain-specific languages (DSL), models, model transformation tools and code generators based on the Model-Driven Engineering (MDE) paradigm. While each technique in isolation has had important success, the ultimate goal of reuse, namely the industrialization of software production, has not become reality.******The objective of this research program is to elaborate a theory of reuse that aligns the strengths of existing reuse technologies within a unifying framework. The vision is to apply advanced techniques for separation of concerns, i.e., aspect-orientation and feature-orientation, to define an encompassing reuse paradigm: concern-oriented reuse (CORE). A concern encapsulates everything about a development issue: 1) models at all relevant levels of abstraction, 2) a description of the choices and options that a concern offers to developers together with their impact on non-functional requirements, 3) an interface that describes how to customize the reused artifacts to the needs of the developer, potentially with a DSL specifically designed for manipulating the concepts of the concern, and 4) consistency rules and model transformations that automate proven multi-view development strategies and MDE generative approaches. Designed with support for evolution and collaboration, CORE will enable world-wide contributions from a growing user community to an open library of reusable concerns.******The potential impact of CORE on software development (time, quality and cost benefits) is considerable. When choosing to reuse a solution from a concern, developers do not have to look through lengthy informal documentation, but benefit from the decision rationales provided for each solution variant. Composition of the reusable artifacts with application-specific models and code is safe, supported by tools and concern-specific DSLs. This decreases the reuse effort, makes reuse less error prone, and even accessible to non-experts. Finally, the availability of a reusable models in the concern library makes MDE a viable development choice. ******The envisioned common framework would make it possible to combine best practices from each reuse technique. A large scale adoption of CORE would allow software engineers to become concern specialists. Software providers could sell concerns, maintaining and evolving the concern over years, e.g., by adding new languages, features, and platform support, as they become relevant. Ultimately, libraries, reuse, and specialization would provide a clear structure to software development, and as a result align the practice of software engineering closer to what is done in other engineering disciplines.