Some thoughts on consensus algorithms, why Paxos is so tricky, and a reminder of the Chang/Maxemchuk algorithm.