Execution Layer Meeting 153

[timbeiko]

Meeting Info

• January 19, 2023, 14:00-15:30 UTC
• Youtube Stream: https://youtu.be/hVeMHoUUZ30
  • Note: calls are now under the @EthereumProtocol Youtube Channel
• Zoom: shared on #allcoredevs Discord channel shortly before the call
• Ethereum Protocol Calls Calendar subscription

Agenda

• Shanghai Updates
  • Testing & devnets
  • Withdrawal roots SSZification & wei/gwei conversion, see:
    • CL-EL withdrawals harmonization: using units of Gwei execution-apis#354
    • EIP-4895: CL-EL withdrawals harmonization: using units of Gwei EIPs#6325
    • Add hexary trie roots for lists in ExecutionPayloadHeader consensus-specs#3078
  • EIP-6049: Deprecate SELFDESTRUCT
• EOF updates
  • Cancun CFI (context)
• Network Upgrade Meta Threads
• Announcements
  • Shanghai/Capella Community Call #1 #708
  • https://medium.com/ethereum-cat-herders/client-diversity-on-ethereum-network-3854b5c3c95f

[protolambda]

Part of the SSZ-ification of the trees in the block is not just the tree commitment, but also how the list items get embedded as leaf values. Current MPT has 32+ long values and thus hashes them, not entirely sure, but seems like this behavior overlaps with the tx-hash behavior. With SSZ depending on the integration, we might lose that: i.e. you can't reconstruct the root of the tree from just the tx-hashes as values (as currently available in the eth_getBlockByNumber(n, false) RPC). And CL definitely doesn't have this (different hash function, hash-tree-rooted opaque-tx bytes).

We'd have to modify the CL to not make a HTR of opaque txs, but of the keccak hashed opaque txs (if the tx hashing even stays consistent between all future tx types)

[ralexstokes]

I think if anything we want to drive more things to SSZ, rather than pulling RLP or MPT concepts into the consensus layer.

This is what I think we should aim for:

1. receipt_root: define a SSZ-encoding Receipt for receipts and replace this root with the hash tree root of a List<Receipt, SOME_BOUND>
2. transactions_root: define a SSZ-encoding Transaction for all transaction types and replace this root with the hash tree root of a List<Transaction, SOME_BOUND>
3. withdrawals_root: define a SSZ-encoding Withdrawal for the withdrawal object and replace this root with the hash tree root of a List<Withdrawal, SOME_BOUND>

Note: the state_root will change in spirit w/ the move to Verkle tries. Everything else has a "flat" RLP encoding that isn't too bad to work with no matter which layer of the stack you are standing on.

Another idea from discussing this with @etan-status is to change how block_hash is computed -- keep data as-is but define a Header format where the block hash is now the hash tree root of this Header object.

And if we really want to go all the way, we can even consider swapping big-endian RLP data to little-endian SSZ data so that the headers match across EL and CL. But I feel this may be too invasive of a change.

[lightclient]

Can we talk about this in the next meeting? ethereum/execution-apis#329

[timbeiko]

Closed in favour of #715, adding your issue @lightclient !