Enabling Effective Design Decisions during Software Development

软件设计涉及将软件系统应该做什么的需求转化为满足这些需求的执行系统的活动。在软件设计过程中做出的选择对系统的整体价值和正确性有着深远的影响。尽管在软件开发期间做出了许多设计选择的重大后果，但人们对软件开发人员面临的设计决策类型以及这些决策的后果知之甚少。该研究计划的长期目标是为软件开发人员提供智能帮助，帮助他或她做出明智的软件设计决策，并使先前的决策能够流畅地重新审视和更改。这一研究计划将使软件开发从设计决策相继进行的顺序活动转变为允许软件开发人员探索一系列决策。这种探索将允许增加系统的整体价值，并将可能产生不良后果的选择(如暴露漏洞)降至最低。为了实现这一长远目标，我提出了今后五年研究的三个目标。首先，通过实证研究，我们将识别和分类发生的不同类型的设计决策，并确定各种决策的影响范围。其次，我们将创建技术来识别开发人员在使用开发工具时何时做出决策，我们将设计算法来推荐决策的潜在选择，我们将设计用户界面以有效地向开发人员提供建议。第三，我们将为开发期间所做的设计选择创建表示形式，并且我们将开发支持软件开发的回滚和随后的前滚的技术，以便能够从设计决策点探索和重新访问不同的选择。这个问题领域是我在行业中领导软件开发团队的经历中发现的。这一研究领域的进展可以为软件工程和编程语言提供新的知识，并可以产生技术和软件工具，使加拿大1600亿美元的信通技术产业中的大部分立即受益。为探索设计空间而开发的技术也有可能用于更大的科学界，在那里，设计决策可以直接影响政策和其他知识所依赖的结果(例如，气候模型)。

Software design involves activities that turn requirements for what a software system should do into an executing system that meets those requirements. The choices made during software design have a profound effect on the system, including its overall value and its correctness. Despite the significant consequences of the many design choices made during software development, little is known about the kinds of design decisions software developers face and the consequences of those decisions. The long term objective of this research program is to provide intelligent assistance to a software developer to help him or her make informed software design decisions and to enable prior decisions to be fluidly revisited and altered. This research program will enable software development to move from a sequential activity in which design decisions follow each other to allowing software developers to explore of a range of decisions. This exploration will permit the overall value of a system to be increased and choices that may have undesirable consequences, such as exposing vulnerabilities, to be minimized. Towards this long-term vision, I propose the following three objectives for the next five years of research. First, through empirical study, we will identify and catalogue different kinds of design decisions that occur and we will determine the scope of impact of various decisions. Second, we will create techniques to identify when decisions occur as a developer works with development tools, we will devise algorithms for recommending potential choices for a decision, and we will design user interfaces to effectively deliver recommendations to a developer. Third, we will create representations for the design choices made during development and we will develop techniques to enable the rolling back, and subsequent rolling forward, of a software development, to enable exploration and revisiting of different choices from a design decision point. This problem area came to light during my experiences leading a software development team in industry. Advances in this research area can provide new knowledge for software engineering and programming language and can produce techniques and software tools with immediate benefit to much of the $160B ICT industry in Canada. There is also potential for techniques developed for exploring design spaces to be used within the greater scientific community where design decisions can directly impact results on which policy and other knowledge rely (e.g., climate models).**