Ten Rules for Open Source Success


This is the Golden Rule, taught to me by Isabel Drost-Fromm. Build community, not software. Without community your code will solve the wrong problems. It will be abandoned, ignored, and will die. Collect people and give them space to work together. Give them good challenges. Stop writing code yourself.

Seeking upfront consensus is like waiting for the ideal life partner. It is kind of crazy. Github killed upfront consensus with their fork/pull-request flow, so you've no excuse in 2015. Accept patches like Wikipedia accepts edits. Merge first, fix later, and discuss out of band. Do all work on master. Don't make people wait. You'll get consensus, after the fact.

Educate yourself and others to focus on problems not features. Every patch must be a minimal solution to a solid problem. Embrace experiments and wild ideas. Help people to not blow up the lab. Collect good solutions and throw away the bad ones. Embrace failure, at all levels. It is a necessary part of the learning process.

Be aggressive about documenting contracts (APIs and protocols) and testing them. Use CI testing on all public contracts. Code coverage is irrelevant. Code documentation is irrelevant. All that matters is what contracts the code implements, and how well it does that.

Use your power to enforce rules, not bully others into your "vision" of the project's direction. Above all, obey the rules yourself. Nothing is worse than a clique of maintainers who act special while blocking patches because "they don't like them." OK, that's exaggeration. Many things are much worse. Still, the clique thing will harm a project.

Maybe you noticed that "be innovative" isn't anywhere on my list. It's probably point 11 or 12. Anyhow, cultivate a positive and pleasant mood in your community. There are no stupid questions. There are no stupid people. There are a few bad actors, who mostly stay away when the rules are clear. And everyone else is valuable and welcome like a guest who has traveled far to see us.