Four chains spent today deleting things
Serguey Asael Shinder2 min readToday's protocol items across four chains have almost nothing in common as features and one thing in common as work.
Bitcoin's BIPs repository removed a finalizer clause from the description of a Taproot PSBT output field - instruction that had been sitting on the wrong side of the transaction. Ethereum's core developers put a batch of EIPs on the inclusion lists in which the majority remove: storage-clear refunds, the refund cap, bloom filters, SELFDESTRUCT, the deposit and eth1data fields. Solana's SIMD-0558 did not add a capability either; it specified the order in which a syscall fails. And XLS-85d on the XRP Ledger spends most of its substance not on the escrow feature but on the invariant checks and the gateway_balances change that stop the feature creating a reporting gap.
None of that is a roadmap item and none of it will appear in a comparison table.
It is also the work that decides whether two independent implementations of the same chain agree. Every one of today's changes closes a place where a reader of the specification had to guess: what a finalizer does with an output field, which failure fires first, whether an issuer's escrowed supply can be read in one call. A guess that two teams resolve differently is a consensus failure waiting for the block that exercises it.
The Ethereum list is the sharpest version of the same instinct, because removal is the most expensive change a live chain can make - every tool that used the feature breaks at a height nobody can move. Choosing to pay that says the cost of carrying an inconsistency is now judged higher than the cost of breaking what depends on it. That judgement is what maturity looks like in a protocol, and it is unglamorous by construction.