Building Multi-language, Multi-system Quality Models Using Features and APIs

《软件》杂志对2010年全球500强软件公司进行了排名：其中20家来自加拿大，在全国排名第二，仅次于美国。加拿大的信息和通信技术(ICT)部门包括37,400家公司，其中33,000家开发软件，雇用数十万名开发人员开发和维护这些系统。保守的研究估计，开发人员80%的时间和精力都投入到维护上。因此，对系统可维护性的任何改进都可以在时间、精力和成本方面带来重要的收益。可维护性仍然是软件系统难以捉摸的特征，衡量可维护性的软件质量模型很难在不同的系统上构建和应用。事实上，高质量的车型目前还不够发达。这些困难有三个原因：(1)每个软件系统都是唯一的，因此为一个系统构建的模型通常不适用于另一个系统；(2)现代系统经常结合不同的编程范例，例如面向对象的和功能的，从而降低了它们的可维护性，并使质量模型的构建复杂化；(3)可维护性的子特征定义不明确，因此使用代理度量部分定义模型，通常排除了系统的许多方面。系统的各个方面包括实现习惯用法、设计选择、体系结构风格、用户可观察的功能、内部和外部API以及使用的框架和库。我建议解决这些困难：(1)考虑系统的更多方面；(2)考虑系统的“类别”。在我以前和目前的研究工作的基础上，我将考虑提供类似服务的功能和应用程序编程接口(API)，这些功能和应用程序编程接口可以在不同的系统中找到，这些系统具有不同的实现、设计、体系结构和以不同的编程语言组合实现的框架。我将使用类似的服务，例如Undo/Redo，作为一个起点来研究“类”中的软件系统，而不是作为独特的个体。我还将使用特性、API、度量和设计实例(反)模式来构建关注多语言、多范例系统的可维护性的软件质量模型。我将在大型、多语言、多范例的系统上对这些模型进行评估，比如Mozilla Firefox或SFL Ring。我将进行三种类型的验证：(1)定量和定性地通过(2)开发人员的研究(与参与者一起)和(3)社会研究(例如调查)。因此，我想促进我们对影响软件系统可维护性的方面的理解，并为开发人员提供具体、有用的方法来衡量可维护性，以便他们能够改进它，从而降低维护成本。

Software Magazine ranked the World's top 500 software companies in 2010: 20 were from Canada, the second-highest national percentage after the United States. Canada's information and communication technologies (ICT) sector include 37,400 companies, among which 33,000 develop software, employing hundreds of thousands of developers to develop and maintain these systems. Conservative studies estimate that 80% of the developer's time and effort are invested in maintenance. Hence, any improvement to the maintainability of systems could results in important gains in time, effort, and, thus, cost reduction. Maintainability remains an elusive characteristic of software systems and software quality models measuring maintainability are difficult to build and apply on different systems. Indeed, quality models are currently under-developed. There are three reasons for these difficulties: (1) each software system is unique, thus a model built for one system often does not apply on another, (2) modern systems often combine different programming paradigms, e.g., object-oriented and functional, thus, lowering their maintainability and complexifying the building of quality models, and (3) the sub-characteristics of maintainability are ill-defined, thus models are defined partially, using proxy measures, often excluding many aspects of the systems. Aspects of the systems include implementation idioms, design choices, architecture styles, user-observable features, internal and external APIs, and used frameworks and libraries. I propose to tackle these difficulties (1) by considering more aspects of the systems and (2) by considering “classes” of systems. Building on my previous and current research work, I will consider features and application programming interfaces (APIs) that provide similar services and that can be found in different systems with different implementations, designs, architectures, and frameworks implemented in different combinations of programming languages. I will use similar services, e.g., undo/redo, as a starting point to study software systems in “classes” and not as unique individual. I will also use features, APIs, metrics, and occurrences of design (anti)patterns to build software quality models focusing on maintainability of multi-language, multi-paradigm systems. I will evaluate these models on large, multi-language, multi-paradigm systems, like Mozilla Firefox or SFL Ring. I will perform three types of validations: (1) quantitatively and qualitatively through (2) developers' studies (with participants) and (3) social studies (e.g., surveys). Thus, I want to contribute to advance our understanding of the aspects of software systems impacting their maintainability and provide developers with concrete, useful means to measure this maintainability so that they can improve it and, thus, can reduce maintenance costs.