Execution Layer Meeting 169

[timbeiko]

Meeting Info

• August 31, 2023, 14:00-15:30 UTC
• Youtube Stream: https://youtu.be/GkSjuiqqkKU
• Zoom: shared on #allcoredevs Discord channel shortly before the call
• Ethereum Protocol Calls Calendar subscription

Agenda

• Dencun Updates
  • devnet-8
  • Testnet ordering
    • fix: remove cancun time for holesky eth-clients/holesky#68
• Verkle Trie + State Expiry Strategy
• SSZ EIPs update
• Announcements
  • EELS post, EIP-6780 PR
  • Yellow Paper London + Paris completion grant

[yperbasis]

I suggest to discuss our strategic vision around Verkle and state expiry. To my mind Vitalik's proposal makes a lot of sense. If we follow it, we don't need to translate a chunk of the Merkle–Patricia trie into Verkle with every block, which is IMO in-elegant and complicated. I'm a fan of both Verkle and state expiry and Vitalik's proposal offers a nice path to reach both.

[timbeiko]

Added @yperbasis!

[timbeiko]

My 2c on testnet ordering for Dencun:

• devnet-9: one more devnet, spec frozen, which we stress-test. Assuming we're happy with it, move to testnets, otherwise, devnet-10
• Holesky: fork this testnet first given it's relatively new, and the validator set will be small near genesis.
• Goerli: fork second, given it has the most user activity (apps, L2s, validators), but also is being deprecated. Dencun should be the last fork on Goerli IMO.
• Sepolia: last one before mainnet

In parallel, we can tell validators to use Ephemery to test their setups, but given it doesn't have first-class support in all clients yet, I wouldn't sequence it linearly with other testnets.

[etan-status]

I finally got to update the SSZ eips for transactions / receipts to address the latest feedback.

Previously, we have discussed two approaches, based on SSZ Unions, and a normalized SSZ container. Both of these have now been fused together, to have something based on a PartialContainer. The PartialContainer takes the design space flexibility and the compact serialization of the SSZ Union, but ensures that common fields that are present in multiple cases end up at the same Merkle tree generalized index like in the normalized SSZ container. This essentially combines all the advantages and could also be reusable for other purposes, sunsetting the currently unused SSZ Union and SSZ Optional types.

1. https://eips.ethereum.org/EIPS/eip-7495: SSZ PartialContainer
2. https://eips.ethereum.org/EIPS/eip-6493: SSZ SignedTransaction / PooledTransaction / Receipt
3. https://eips.ethereum.org/EIPS/eip-6404: SSZ transactions_root
4. https://eips.ethereum.org/EIPS/eip-6466: SSZ withdrawals_root
5. https://eips.ethereum.org/EIPS/eip-6465: SSZ receipts_root

I have also created a showcase website at https://eth-light.xyz that allows inspection of both the existing RLP and the proposed SSZ PartialContainer based format. The website is meant for use on desktop / laptop only, it is not mobile friendly. It requires JavaScript to be enabled, but doesn't work in Tor browser due to lack of Shared Array Buffer. A regular Firefox via Tor should work.

To use the website, wait for the latest blocks to be synced in the left sidebar, then click the Transactions button. A list of all transactions within the block is populated. The buttons on the left are for the transaction byte representation, the buttons on the right are for the receipt byte representation. SSZ buttons indicate size after Snappy framed compression as proposed in the EIP. RLP is uncompressed as that's how it is sent over RLPx. The URL can be updated manually to load any transaction by hash to allow inspecting historical transactions. Use #rlp / #receipt/rlp / #ssz / #receipt/ssz suffix for the various views.

Note that when inspecting very old receipts of the 'Homestead' scheme, the default Cloudflare gateway currently has a bug where it incorrectly reports those with a Status code instead of the intermediate post-state root. Switching to an Infura gateway fixes that, but that comes with their logging privacy policy. I have reached out to Cloudflare for a fix, but it is still pending. An earlier bug where they sometimes omitted withdrawals_root from block responses was already addressed.

[barnabasbusa]

Small PR for adjusting cancun time for Holesky: eth-clients/holesky#68

[barnabasbusa]

Holesky: fork this testnet first given it's relatively new, and the validator set will be small near genesis.

1.48M is not really that small.

[timbeiko]

@barnabasbusa I meant the number of operators, my bad!

[timbeiko]

Closed in favor of #857