All Core Devs - Execution (ACDE) #206, February 27, 2025

[nconsigny]

ACDE 206

• Feb 27, 2025, 14:00-15:30 UTC
• 90 minutes
• Stream: https://youtube.com/live/tlezpGztpi8?feature=share

Agenda

• Pectra
  • Holesky Debrief
    • Chain status (proposals, slasheable validators, etc.)
    • Next steps
      • Syncing from unfinalized checkpoints
      • Start sync on FCU go-ethereum#31254
      • Fork ID checks dependent on (standardized?) configs
  • Sepolia fork timing
  • Client releases
    • https://github.com/sigp/lighthouse/releases/tag/v7.0.0-beta.1e
  • devnet-7 and/or new testnet
  • Mainnet requirements
    • Builder API testing
  • ACDC retrospective
• Fusaka
  • Update EIP-7594: include cell proofs in network wrapper of blob txs EIPs#9378
  • Disabling EOF to help PeerDAS testing
  • EOF PAY opcode
  • Fork deadlines
    • Change dates to Mondays to allow Tue/Wed for review before ACDs
• EIP-4444 progress updates
• Stateless clients for MPTs

[github-actions]

Discourse Topic ID: 22896
Zoom meeting created: https://us06web.zoom.us/j/86939519578?pwd=5U3qfDob0dwgOfXgcApqsZOKvUC5Qa.1
Zoom Meeting ID: 86939519578
Discourse topic created/updated: https://ethereum-magicians.org/t/22896

[pipermerriam]

Can we get a slot for teams to give updates on their 4444s progress. May 1st is coming

[kevaundray]

Want to discuss asking execution teams to start thinking about adding "stateless clients" for MPT for zkvm's and also fielding zkvm questions.

• Geth currently has an implementation of this that does most of the work
• Reth has code in various zkvm repos that does most of this

One desiderata that geth and reth currently does: The code is ideally modularized from the rest of the codebase and not tightly integrated to the whole codebase

[ralexstokes]

I won't be able to make the call but we should discuss:

• the ability for ELs to disable EOF at the Fusaka hard fork, so that we can test PeerDAS in isolation
• the spec change to move PeerDAS proof computation into the EL (from the CL)
  • Update EIP-7594: include cell proofs in network wrapper of blob txs EIPs#9378

[potuz]

Given what happened in Holesky I'd like to briefly discuss a path to prioritization of syncing from an unfinalized checkpoint, hopefully this Thursday will only be to signal the need to CLs and confirm that this is trivial from the EL perspective.

[leeederek]

Given what happened on Holesky and as a follow up to the timelines mentioned in #1265 (comment), I'd like to get an update from teams on whether those timelines are still accurate or if we believe it should change.

[wjmelements]

I plan to discuss the EOF motivation for including the PAY opcode in Osaka.

slides

[philknows]

We should discuss whether or not we should startup a new public testnet that we can use for testing client releases (I'm sure many other client teams were relying on this network as well for beta testing). With the current state of Holesky and the unlikely finalization of it for at least 2.5-3 weeks in a happy case, we don't have a test network that replicates or exceeds Mainnet's validator set and/or decentralized enough distribution of nodes.

Assuming this might be pectra-devnet-7? We should probably clarify whether this will be the expected long-lived testnet should Holesky fail.

[timbeiko]

@pipermerriam @kevaundray @ralexstokes @potuz @leeederek @wjmelements @philknows I've added your items to the agenda, but given the Holesky situation, it's possible we may run out of time to cover everything tomorrow.

[realbigsean]

Lighthouse latest version to use on Holesky not mainnet https://github.com/sigp/lighthouse/releases/tag/v7.0.0-beta.1

More information about configs that may be helpful here, but none are required sigp/lighthouse#7040

[ralexstokes]

i won't be able to attend, but did want to flag that we should also take into account testing the Builder APIs/MEV pathways on testnets, esp on one with a high validator count and all of the fun new Pectra features

eg if we only did serious builder testing on sepolia before mainnet, we may be missing some useful live data

[github-actions]

Discourse Topic ID: 22896

• Action: Updated
• URL: https://ethereum-magicians.org/t/22896

[shemnon]

the ability for ELs to disable EOF at the Fusaka hard fork, so that we can test PeerDAS in isolation

Probably the most sensible place is to prevent EIP-7698 contract creation transactions and not deploying the EIP-7873 precompile/predeploy that would deploy EOF contracts. This keeps EOF contracts from entering the system without any notable changes to the other sections of the EL client.

Would a CLI flag or config file entry for each client suffice?

[LukaszRozmej]

Point for Holesky retrospective: We should design all conensus configuration in a way that it should change ForkId's, so not only EIP activation should do that, but also things like DepositContractAddress. Then this could be easier spotted cross-client where ForkId's don't match.

[fjl]

Geth ideas for improvement after Holesky failure:

• Start sync on forkChoiceUpdated instead of waiting for newPayload (Start sync on FCU go-ethereum#31254)
• Standardize EL genesis/fork configuration file. It's time. For historical reasons, EL clients are not 100% aligned on the config format, which basically means we all maintain the fork configurations ourself. However, there are only two formats left in use: 'geth style' (used by Geth, Erigon, Besu, Reth, EthereumJS) and 'chainspec style' (used by Nethermind). AFAIK Nethermind has an adapter for 'geth style'. While the chainspec format is technically superior, since it allows activating individual EIPs, it might be good to just agree on the simpler format and make it the official one. That way, fork configuration for testnets can be committed to as part of hard-fork definition.
• We will also review our mechanism for caching bad blocks on the engine API.

[fjl]

Geth statement about fork timing:

• We should go ahead with the scheduled Sepolia activation. Scheduling of mainnet activation needs to wait until the fork has passed successfully on Sepolia. We should also allow for CL teams to implement more mitigations for the issues found in Holesky split.
• About Geth readiness: we continue getting reports about minor regressions related to Prague/Pectra across our feature set (like in RPC APIs), and have been fixing these. We are on track to have everything working for the mainnet upgrade.

[poojaranjan]

Holesky Debrief - Recording

[github-actions]

Discourse Topic ID: 22896

• Action: Updated
• URL: https://ethereum-magicians.org/t/22896

[wjmelements]

Hopefully there will be more time to discuss Osaka EIPs in the next meeting.