Every day we discover new advantages of decentralized systems and peer-to-peer networks . Well designed P2P systems, provide greater resilience against attacks and accidental disruptions. Nevertheless decentralization comes with a price: consensus must be reached by voting or by proof-of-work, and both solutions tend to be slower than traditional centralized decisions. Instant payments requires instant consensus, and that’s something that is difficult to achieve. Two posts ago I proposed a fully decentralized solution that achieves confirmations in around 15-25 seconds. Can we do better? Probably, but that would require a great investment to inter-connect the whole world by fiber optic. Currently the packet delay is influenced by the length of the route and the interaction between the TCP reliability and congestion control protocols. Other important factors are application level manipulation of the data, propagation delay, serialization, data protocols, routing and switching, queuing, buffing, bandwidth and noise. Can we do better with the current Internet infrastructure? I think we can, by designing systems that behave like centralized ones, but can disperse into fully distributed ones if something goes wrong. Nature has many examples of colonies that have a queen, but if the queen dies somehow manage to re-elect a new queen that all obey. This is what the MinCen protocol tries to do: it chooses a master node that decides which is the next block when there are competing chains. Everyone then follows this decision. But if the master node tries to cheat then every miner starts looking for another master node, in such a way that individual decisions converge to a new global selection. Are you designing a new cryptocoin? give a try to the MinCen protocol (preliminary version of the paper available here).