[einarvollset]

Robbert is indeed one of (if not the) smartest men you'll ever meet in relation to fault tolerance in distributed systems. However, if you're looking to build a fault tolerant consensus algorithm, you owe it to yourself to check out randomized consensus (which predates Paxos, and really is the same algorithm - just simpler)

[DenisM]

Google search for "randomized consensus" turns up a paper comparing several such algorithms. Which one did you have in mind?