Rough Consensus And Running Code As A Working Culture

Decisions are settled by broad agreement and by demonstrated implementation rather than by voting or by executive choice.

Two ideas define how these groups decide things. The first is that agreement need not be unanimous but must be broad enough that remaining objections have been genuinely heard and answered rather than outvoted. There is no ballot in which a majority simply prevails over a minority holding a serious technical objection. Chairs judge whether the sense of the room has settled, which is imprecise and sometimes contested, but which pushes participants to persuade rather than to count heads.

The second idea is that a proposal that has actually been built and shown to work carries more weight than one that exists only on paper. This is a powerful filter. It rewards people who do the work, it exposes problems that look invisible in a specification, and it makes it harder to standardise something elegant that nobody can implement. It also biases the process towards those with the resources to build things, which is a real limitation on the merit based ideal the culture claims for itself.

Together these produce a process that is slow, argumentative, and surprisingly durable. Slow because persuading everybody takes longer than instructing them. Argumentative because disagreement is the mechanism rather than a failure of it. Durable because a convention that survived that process usually has few obvious defects left, and because everybody who adopts it did so voluntarily and therefore has no grievance about having been made to. Adoption without compulsion is the only kind available here, so the process is built to earn it.