The Babel of software development: Linguistic diversity in Open Source

Abstract. Open source software (OSS) development communities are typically very specialised, on the one hand, and experience high turnover, on the other. Combination of specialization and turnover can cause parts of the system implemented in a certain programming language to become unmaintainable, if knowledge of that language has disappeared together with the retiring developers. Inspired by measures of linguistic diversity from the study of natural languages, we propose a method to quantify the risk of not having main-tainers for code implemented in a certain programming language. To illustrate our approach, we studied risks associated with different lan-guages in Emacs, and found examples of low risk due to high popularity (e.g., C, Emacs Lisp); low risk due to similarity with popular languages (e.g., C++, Java, Python); or high risk due to both low popularity and low similarity with popular languages (e.g., Lex). Our results show that methods from the social sciences can be successfully applied in the study of information systems, and open numerous avenues for future research.