Ethereum block-level access lists reach their Last Call deadline

A review of how nodes organize their work

The official EIP-7928 document lists November 1, 2026, as its Last Call deadline. This Ethereum proposal, created on March 31, 2025, and developed by several contributors, still shows Last Call when checked. It describes block-level access lists, or BALs, identifying the parts of state used. Today’s date is therefore a review milestone, establishing neither final adoption nor a speedup already available on mainnet.

The problem is practical. To process operations in parallel, a node needs advance knowledge of which accounts and storage slots they will read or modify. Optional lists attached to some transactions do not provide a sufficient guarantee. The proposal would require a block-level inventory of actual accesses and the values resulting from changes. Intended benefits concern disk reads, validation, and state-root computation.

An inventory that must match execution

The mechanism would add a cryptographic hash of the list to the block header. Detailed data would be encoded separately and transmitted between the consensus and execution layers through the Engine API. A node would then check that the supplied inventory exactly matches accesses produced during execution. Missing, incorrect, or unjustified entries could invalidate the block. The inventory is not simply a hint that a validator should trust.

The list covers accessed addresses even when they remain unchanged, as well as storage slots and changes to balances, code, or counters. It also distinguishes operations before transactions, transactions themselves, and post-execution operations. Rules enforce deterministic ordering and prevent duplicates. This precision is intended to let different software produce the same representation, which is necessary for unambiguous hash comparisons.

Potential gains with additional costs

The text describes parallel reads and reconstruction of state from recorded changes without replaying every transaction for that reconstruction. This does not remove the requirement to validate blocks. It cites an analysis of 1,000 October 2026 blocks in a 60-million-gas setting, reporting an average compressed list size of approximately 92.1 kibibytes. The sample illustrates additional data costs rather than guaranteeing performance for every use case.

The proposal also acknowledges the risk of spurious storage reads imposing unnecessary input/output work before detection. A block-gas-related limit bounds the number of items, while clients must avoid rejecting valid lists. The changes affect block structure and the Engine API and would require a hard fork. For operators, the proposal remains a tradeoff involving parallel processing, data transmission, and verification. Passing the November 1 date does not turn this standards work into a completed deployment.