Software of Things: Composition and Integration

我们对软件的依赖与日俱增。软件系统本身变得越来越大，越来越复杂。与此同时，软件开发周期缩短，客户希望新产品能够比以往更快地推向市场。 在这样的现实中，软件构建变成了一种组合活动，开发人员从内部和第三方库中集成代码，从早期项目中“借用”的代码片段，以及在线找到的代码。这种借用的代码可以包括高达90%的典型软件系统。 以可预测的方式集成独立开发的代码片段是一项具有挑战性的任务：开发人员通常对每个软件片段的功能了解有限，更不用说它们的组合行为了。无法推理软件功能及其相互作用使软件开发过程复杂化，缩短了上市时间，并降低了所生产软件的质量。事实上，有几个众所周知的灾难是由软件重用、组合和集成问题引起的，包括阿丽亚娜5号火箭爆炸、KCG交易系统故障以及Therac-25治疗机造成的大规模辐射过量。 受这些挑战的激励，这里提出的研究计划旨在为组合软件的有效和可靠的开发奠定基础。该程序与现有工作的主要区别在于，它不专注于指定可重用软件组件或具有定义良好的接口的子系统的方法。相反，它利用形式化方法和程序分析领域的最新进展来开发用于理解和组成现有的、已经可用的软件片段的范例。 具体而言，该计划探讨了以下方面的发展： 1)一种形式化的模型，用于根据软件的概念特征和对开发人员的逻辑意义单位来表示软件。通过从现有代码工件中提取关于特性、特性规范（先决条件和保证）和特性依赖性的信息来填充模型。 2)推理复杂组合软件属性的分析技术。这些技术的目标是，例如，识别由于在其原始上下文之外重用特征而导致的意外副作用以及由于相关特征的独立演化而导致的不一致。 3)将组合模型和分析技术集成到现有开发工具和实践中的概念，包括版本控制系统工作流（例如，的Git）。 这项工作的成果预计不仅将推动科学的发展，而且还将缩短上市时间，提高我们软件的质量，为软件行业以及加拿大和全球经济做出重大贡献。

Our dependence on software continues to grow every day. Software systems themselves become increasingly larger and more complex. At the same time, software development cycles shrink and customers expect new products to be delivered to the market faster than ever. In such a reality, software construction becomes a compositional activity, where developers integrate code from in-house and third-party libraries, snippets of code “borrowed” from earlier projects, and code found online. Such borrowed code can comprise up to 90% of a typical software system. Integrating independently-developed code fragments in a predictable manner is a challenging task: developers often have a limited understanding of what each individual piece of software does, let alone how their composition behaves. The inability to reason about software functionalities and their interplay complicates software development processes, prolongs the time to market, and reduces the quality of the produced software. In fact, there are several well-known disasters caused by software reuse, composition, and integration problems, including the Ariane 5 rocket explosion, the KCG trading system failure, and the massive radiation overdoses inflicted by Therac-25 therapy machines. Motivated by these challenges, the research program proposed here aims at building foundations for efficient and reliable development of compositional software. The main distinction of this program from existing work is that it does not focus on approaches for specifying reusable software components or subsystems with well-defined interfaces. Instead, it leverages recent advances in the fields of formal methods and program analysis to develop paradigms for understanding and composing existing, already available software fragments. Specifically, this program explores the development of: 1) A formal model for representing software in terms of its conceptual features logical units of meaning to developers. The model is populated by extracting information about features, specifications of features (prerequisites and guarantees), and feature dependencies, from existing code artifacts. 2) Analysis techniques for reasoning about complex compositional software properties. These techniques aim at, for example, identifying unexpected side effects caused by reusing features outside of their original context and inconsistencies caused by independent evolution of related features. 3) Concepts for integrating the compositional model and analysis techniques into existing development tools and practices, including version control system workflows (e.g., that of Git). The outcomes of this work are expected not only to advance the state of science but also to reduce time to market and increase the quality of our software, making a significant contribution to the software industry, and to the Canadian and global economy.