The Best of Both Worlds: Reusable Software via Unplanned Reuse

自20世纪60年代以来，重用现有的软件构件来开发新的软件构件的想法得到了推广，因为它具有降低成本和提高质量的潜力。经典的软件重用要求我们很好地预测软件系统的未来需求，以便以易于重用的形式将其功能模块化；这就是预期的重用。由于我们预测未来的能力并不完美，因此会出现开发人员无法找到满足其需求的模块化、可重用构件的情况。一种常见的工业反应是复制和修改现有构件的片段；这是意想不到的重用。预期和未预期的重用方法都有优点和缺点。预期的重用使我们能够创建设计良好、文档齐全的软件构件，但成本很高，而且不能保证投资将会收回。意外重用允许高可定制性，但可能导致糟糕的决策，并且无法保证旧系统的重要属性得到保留。这项工作建议将预期和意外方法的好方面结合起来，并支持使用经验反馈来指导可重用构件的设计。我以前的工作使意外的重用变得更加系统、安全，对开发人员来说压力更小。但是，通过自动化过程中必须发生的许多活动，这种情况可能会得到更多的改善，目前这些活动仍然需要人工开发人员进行非常详细的调查。该研究计划的成功将在构建、理解、使用和改进软件构件的方式方面为行业带来积极的转变。对于学术界来说，这项工作是对软件重用问题理解的一个根本性转变。对于HQP来说，该计划是软件工程研究和实践的一个很好的培训基地。

Since the 1960s, the idea of reusing pre-existing software artifacts for the development of new software artifacts has been promoted for its potential to reduce costs and improve quality. Classical software reuse demands that we predict well the future needs of software systems in order for their functionality to be modularized in an easily reused form; this is anticipated reuse. Since our ability to predict the future is not perfect, there occur cases where a developer cannot find a modularized, reusable artifact that meets their needs. A common industrial response is to copy-and-modify fragments of existing artifacts; this is unanticipated reuse. Both anticipated and unanticipated reuse approaches have benefits and weaknesses. Anticipated reuse allows us to create well-designed and well-documented software artifacts, but at high cost and without guarantee that the investment will be recouped. Unanticipated reuse permits high customizability, but can lead to poor decisions and without guarantee that important properties from the old system are preserved.This work proposes to take the good aspects of both the anticipated and unanticipated approaches and combine them, augmenting them with support for using feedback from experience to inform the design of reusable artifacts. My previous work has made unanticipated reuse much more systematic, safe, and less stressful for the developer. But the situation could be improved even more by automating many of the activities that have to happen during the process, which currently still demand much detailed investigation by a human developer.The success of this research program will see a positive transformation for industry in the way that software artifacts are constructed, comprehended, used, and improved. For academia, this work is a fundamental shift in understanding the problem of software reuse. For HQP, the program is an excellent training ground in software engineering research and practice.