Conversation

2) The auction you describe is equivalent, basically, to a matching engine: 2 MMs alone will make all those arbs close instantly and pay fees equal to the LP's slippage, the same as a matching engine would economically.

3) However, it's _really_ slow. It doesn't just ratelimit arbs--it ratelimits _everything_ if 80% of the network is used on people closing these arbs and everything has a 1m delay waiting to see if someone can arb it.

4) The whole point was to get around using a matching engine for performance reasons, but the replacement will probably cost even more. Beyond that, though, how does this increase liquidity? Generally I'd expect inefficiency to reduce liquidity.

5) "optimizing for liquidity" sounds great but the truth is that, other than not sucking, exchanges can't do much there--it's mostly liquidity providers. Adding large latency and gas fees, though, will scare off LPs, so I think this just makes that harder.

6) Now there are times when having people responsible for something instead of a blockchain is good--e.g. when the blockchain _can't_ do something. But matching engines aren't a great example, because they only require on-chain data.