Library Fission: Improving APIs by observing real-world API usage

现代软件开发严重依赖外部库。我们的假设是，不同的客户端使用不同的库API表面子集，并且有可能使用聚类算法来自动将库拆分为更小、更灵活的库，我们称之为“库裂变”。这样做的副作用是减少客户端暴露在库中的安全漏洞：客户端只受它实际使用的库部分的漏洞影响，而不是库带来的其他所有漏洞。 拟议的资金将使与惠灵顿的维多利亚大学的合作成为可能。我们将与静态分析、经验软件工程和软件可视化方面的新西兰专家合作，开发库裂变技术。这项研究的预期成果是一套新的技术和工具，用于改进图书馆API，这些技术和工具将通过学术出版物和开源工具传播。这些研究成果将有助于图书馆的用户和设计者更有效地工作。

Modern software development heavily relies on external libraries. The proposed collaboration willcreate techniques and publicly-available tools useful to both users of external libraries and maintainers of these libraries.Our hypothesis is that different clients use different subsets of libraries' API surfaces, andthat it is possible to use clustering algorithms to automatically split libraries into smaller,more agile libraries, which we call "library fission". This has the helpful side-effect ofreducing client exposure to security vulnerabilities in libraries: a client is only affected byvulnerabilities in the parts of the library that it actually uses, not everything else broughtin by the library. The proposed funding will enable a collaboration with the Victoria University of Wellington.Working with NZ-based experts in static analysis, empirical software engineering, and softwarevisualization, we will develop techniques for library fission. The expected outcome of the research is a novel set of techniques and tools for improving library APIs, which will bedisseminated through academic publications as well as open-source tools. These researchoutcomes will help both users and designers of libraries work more effectively.