Blog by Sumana Harihareswara, Changeset founder
How To Improve Bus Factor In Your Open Source Project
Hi, reader. I wrote this in 2015 and it's now more than five years old. So it may be very out of date; the world, and I, have changed a lot since I wrote it! I'm keeping this up for historical archive purposes, but the me of today may 100% disagree with what I said then. I rarely edit posts after publishing them, but if I do, I usually leave a note in italics to mark the edit and the reason. If this post is particularly offensive or breaches someone's privacy, please contact me.
Someone in one of my communities was wondering whether we ought to build a new automated tool to give little tasks to newcomers and thus help them turn into future maintainers. I have edited my replies to him into the How To Build Bus Factor For Your Open Source Project explanation below.
In my experience (I was an open source community manager for several years and am deeply embedded in the community of people who do open source outreach), getting people into the funnel for your project as first-time contributors is a reasonably well-solved problem, i.e., we know what works. Showing up at OpenHatch events, making sure the bugs in the bug tracker are well-specified, setting up a "good for first-timers" task tag and/or webpage and keeping it updated, personally inviting people who have reported bugs to help you solve them, etc. If you can invest several months of one-on-one or two-on-one mentorship time, participate in Google Summer of Code and/or Outreachy internship programs. If you want to start with something that's quantitative and gamified, consider using Google Code-In as a scaffold to help you develop the rest of these practices.
You need to quickly thank and give useful feedback to people who are already contributing, even if that feedback will include criticism. A fast first review is key, and here's a study that backs that up. Slide 8: "Most significant barrier to engaging in onramping others is unclear communications and unfriendly community. Access to the right tools has some effect." Slide 26:
"Contributors who received code reviews within 48 hours on their first bug have an exceptionally high rate of returning and contributing.(And "Github, transparency, and the OTW Archive project" discusses how bad-to-nonexistent code review and bad release management led to a volunteer dropping out of a different open source project.)
Contributors who wait longer than 7 days for code review on their first bug have virtually zero percent likelihood of returning.
Showing a contributor the next bug they can work on dramatically improves the odds of contributing."
In my opinion, building bus factor for your project (growing new maintainers for the future) is also a solved problem, in that we know what works. You show up. You go to the unfashionable parts of our world where the cognitive surplus is -- community colleges, second- and third-tier four-year colleges, second- and third-tier tech hubs, boring enterprise companies. You review code and bug reports quickly, you think of every contributor (of any sort) as a potential co-maintainer, and you make friendly overtures to them and offer to mentor them. You follow OpenHatch's recommendations. You participate in Google Summer of Code and/or Outreachy internship programs.
Mentorship is a make-or-break step here. This is a key reason projects participate in internship programs like GSoC and Outreachy. For example, Angela Byron was a community college student who had never gotten involved in open source before, and then heard about GSoC. She thought "well it's an internship for students, it'll be okay if I make mistakes". That's how she got into Drupal. She's now a key Drupal maintainer.
Dreamwidth, an open source project, started with two maintainers. They specifically decided to make the hard decision to slow down on feature development, early on, and instead pay off technical debt and teach newcomers. Now they are a thriving, multimaintainer project. "dreamwidth as vindication of a few cherished theories" is perhaps one of my favorite pieces on how Dreamwidth did what it did. Also see "Teaching People to Fish" and this conference report.
Maintainers must review code, and that means that if you want someone to turn into a maintainer in your project, you must help them learn the skill of code review and you must help them get confident about vetoing and merging code. In my experience, yes, a good automated test suite does help people get more confident about merging changes in. But maintainers also need to teach candidates what their standards ought to be, and encourage them (many contributors' first thought when someone says "would you want to comaintain this project with me?" is "what? me? no! I'm not good enough!"). Here's a rough example training.
If you want more detailed ways to think about useful approaches and statistics, I recommend Mel Chua's intro to education psychology for hackers and several relevant chapters in Making Software: What Really Works and Why We Believe It, from O'Reilly, edited by Greg Wilson & Andy Oram. You'll be able to use OpenHub (formerly Ohloh) for basic stats/metrics on your open source project, including numbers of recent contributors. And if you want more statistics for your own project or for FLOSS in aggregate, the open source metrics working group would also be a good place to chat about this, to get a better sense of what's out there (in terms of dashboards and stats) and what's needed. (Since then: also see this post by Dawn Foster.)
We know how to do this. Open source projects that do it, that are patient with the human factor, do better, in the long run.