In Praise of Consensus
The constitution of the debian operating system project says things like “consistent with the consensus of the opinions of the Developers” at various points but doesn’t say how strong a consensus or how the project will test for consensus. I think those were mistakes, breaking a couple of the conditions for consensus.
Wikipedia’s understanding of consensus is even worse. Wikipedia seems to treat consensus as a synonym for unanimity. Its testing methods allow an infinite loop to form where the casual observer can’t differentiate between a controversial proposal and consensus. I think those were mistakes.
These famous-but-imperfect implementations frequently lead to misdirected rants which seem to misunderstand consensus as requiring perpetual bikeshedding. Apache’s implementation is rather better – and it may surprise you to learn that our co-op is mostly run by consensus.
There are two key differences which I feel makes consensus work for us: we’ve set limits beforehand on some decisions where we need to act fast – where not making a decision would usually be the same as making a bad decision – and our methods of testing for consensus are better. We test for consensus with secret-at-vote-time-but-published-after straw polls, or using Crowd Wise by email.
I summarise Crowd Wise as follows: gather all ideas plus option 0 (do nothing) if possible, carry out a de Borda (preference) voting round 1, merge/amend/consolidate ideas, voting round 2 if needed. It does still work better if participants put their ego aside a little and co-operate, but it does put limits on non-co-operators.
Anyway, as described in Xana/ xana2/ bamamba/ Why Russ is wrong, debian isn’t exactly using consensus much at the moment, anyway. Should we try to fix its bugs? Do you know other projects where consensus is working?