How I Fixed Satoshi's Design


There was a really good summary of why Casper's planned sharding is flawed.

Apparently everybody is still oblivious to my solution.

If you shard the blockchain, you've still got to verify it. You can't have shards trusting each other, as that breaks Nash equilibrium (there are game theories other than the one that guarantees the security of the long-chain rule).

But if you have every shard verify every other shard, then you don't have sharding any more.

My hypothetical solution is a statistical one (where the economic interests of all shards become intertwined) combined with eventual consistency where it is required to maintain the Nash equilibrium.

SegWit is (in one aspect but not entirely as afaik it really just centralizes proof-of-work) generally analogous to a similar conceptual idea I had thought of and dismissed, because it relies on the trust that the economically impacted parties will verify before eventual consistency is required, not on the proof that those parties did verify before it was required. The game theory gets quite complex because there are externalities such as shorting the coin. So it is possible I may have a mistake and we will find out once I publish.