WSU researcher developing system to bolster FLOSS project coder communities
The viability of open source software (OSS), which is used by an estimated 78 percent of companies to run part or all of their operations, is largely dependent on contributions from volunteers. Coders who spend at least a year on an open-source project are consider long-term contributors, and they play a critical role in OSS projects as managers, reviewers and mentors to newcomers. However, many new coders looking to join a project are hindered by delayed or unfair feedback on their submissions, and as a result fail to become long-term contributors themselves.
“Since newcomers often have to wait two to six times longer than a long-term contributor to get reviews for their changes, they often become frustrated and abandon their onboarding efforts,” said Amiangshu Bosu, assistant professor of computer science at Wayne State University.
Bosu is developing an automated model called RevRanker to address the issue. He recently received a nearly $175,000 grant from the National Science Foundation to support the project.
RevRanker aims to remove barriers in free/libre open source software (FLOSS) projects which stem from a newcomer’s struggle to identify experienced and capable reviewers. Using a mixed research method, Bosu will build a theoretical understanding of useful code reviews, which will be utilized to train and evaluate RevEval, an automated model to predict the usefulness of reviews.
“Using the RevEval model as well as leveraging multiple historical dimensions of the files under review, RevRanker will be developed and evaluated,” said Bosu.
FLOSS projects are community-driven enterprises that some in the tech profession believe will surpass proprietary software in popularity. Bosu is hopeful that RevRanker will be a tool to maintain, sustain and grow the community.
“RevRanker will also save time of experienced FLOSS contributors, as even long-term contributors sometimes encounter difficulties to identify appropriate reviewers,” said Bosu.
Outside the realm of FLOSS, RevRanker has the potential to make a significant industrial impact by improving the effectiveness of code review, a process that today is mandatory in such well-known software companies as Microsoft, Google, Facebook and VMWare. A recent study reported more than 50,000 Microsoft developers were spending 10 to 15 percent of their time on average in code reviews.
“Assuming that Microsoft pays on average $40 per hour to its developers, the company is spending $12 million per week on code reviews,” said Bosu. “Therefore, even a 5 percent improvement of the code review effectiveness can save Microsoft more than $30 million per year.”
The NSF award number for this project is 1850475.