direct submission / centralized mempool alternative schemes gaining traction. i told people what i think about this when the noise was about baremultisig.
Discussion
tldr if you disable spam vectors in consensus that is perfectly incentive compatible, and has no negative effects on miner decentralization. it just going to cause disproportionate collateral damage to bitcoin's utility as programmable money compared to how much it would hinder scammers.
in some cases it's perfectly fine, in others it's foolish.
That’s not a result of filtering. If it was, it would mean that filters work as intended, and that’s a tough pill to swallow for the other side. The real reason is mining centralisation which is by no stretch of the imagination the fault of filters, and node runners shouldn’t be punished for it by twisting their arm when deciding on their relay policy.
boring ass pedantic arguments don't change the fact that doing anything that incentivizes direct submission is a horrible idea and will harm bitcoin. you simply do not do that as a responsible developer.
A distorted system *itself* creates incentives for rational actors to destroy long-term value. You either try to fix the system to prevent such behaviour, or see it crumble under its own weight. The only question is when, not if. I think we have enough evidence for such systems in the real world to know that I’m not wrong.
to fix the incentive misalignment you need to make spam consensus invalid. then there is no problem.
one example is baremultisig which nobody uses for non-spam purposes. so making creating new such outputs (not spending them) invalid by consensus would be perfectly fine.
To fix the incentive misalignment you need to reduce mining centralisation.
Your nuclear solution will probably work once until a new data scheme finds a way around your change, if not earlier, provided that you need at least 6 months of preparation and farming consensus for the activation.
it's no coincidence that we are where we are.