[Quick contemporaneous notes by Ben Edgington; fka "Eth2 Implementers' Call"]
Agenda: https://github.com/ethereum/pm/issues/510
Livestream: https://youtu.be/rYWF7N8tS0g
Chaired by Mikhail Kalinin.
[Pari] No major issues on Goerli shadow fork 2 (with equal client split). Last shadow fork was on Monday, GSF3 which replicated mainnet client split. Geth team is debugging an issue that affected a small subset of nodes, but network is finalising and running well. GSF1 and GSF2 will be deprecated today.
Mainnet shadow fork specs have been shared. Fork to take place next Monday. Pari is running all the validators. But main purpose is to test sync - anyone can do this. Plan to do Mainnet SFs every two weeks or so.
MSF - be aware that we are using the Mainnet chain-id, so be careful if sending transactions to it: txs may also be included in mainnet blocks and cost you real ETH.
[Mikhail] Has been working on a test coverage document. [Terence] is also working on some manual test cases for complex issues. Please feed back anything that's looks to be missing. Prysm has developed a proxy that can manipulate payloads which allows them to test optimistic sync more smoothly in end-to-end tests.
Lighthouse
Merge stuff. Full time person working on testing. Work on post-merge logging to make it smoother. Mostly tidying up and testing. Recent release enables proposer boost on mainnet.
Teku
Stefan Bratanov has joined. Working on MEV-Boost integration. On the Merge, we implemented another check on terminal block detection to take into account the block timestamp - we had had issues in testing due to a terminal block with a future timestamp. Now we ignore a potential terminal block with a future timestamp.
Lodestar
Continue Merge work. Substatial updates to gossip and SSZ coming in a release soon. Iterating on light client PoCs: have a closed-loop demo on Kiln, presented in Dubai.
Prysm
Release next week will enable proposer boost. Everything for the Merge is now in the master branch. Mostly tidying and testing now. Fixing some optimistic sync things. Working on integrating Web3Signer.
Nimbus
Integrating Merge branch into main branch. Looking at MEV-Boost and have early progress on light client support. Putting in support for threshhold signatures in the remote signer. Nimbus execution layer is now able to participate on Kiln!
Grandine
Fixes for the Merge. Web3Signer support.
Geth
Debugging GSF3 issue
Recap of last ACD call decisions.
Mainnet difficulty bomb will start being felt in May, becoming a problem in the following weeks. Would be good to Merge before it is too noticeable - 15-15.5s block times in July; 17s by end of July on a quick estimate. 15s is probably the maximum tolerable.
We might be able to bundle the Execution client releases with a bomb postponement to buy a couple of weeks.
Go/no-go decision on forking testnets to be made on ACD call 29th April. Probably fork a testnet every two weeks. Or whether to delay the bomb for a while.
Clients need to be basically feature complete and production ready by the start of testnet forking.
PR from Danny on timeout for Engine API calls. Does anyone think 120s is insufficient? Timeout being the number of seconds per epoch might also be convenient for progress of the chain. Please review!
Next: LatestValidHash
stuff. These are Mikhail's thoughts. Do we want to support this? Please give feedback to Mikhail.
[Lightclient] Builder API/MEV-Boost has been changed from being a middleware design. It's now a separate service that can provide blocks to the consensus client. How are teams doing with this? It's slightly more work than previously. There are three methods in the builder API. Two types of signing required: blind signing and fee recipient announcement. This approach moves towards being more gossip-based.
[PaulH] Changing names so as not to conflict with engine API was a good move. Should we move the API spec out of the FlashBots repo? A little more credibly neutral.
[Jacek] This all becomes part of the attack surface of Ethereum. Maybe we should require that clients implementing this feature should always have a back-up block available from the normal sources (Eth1 node).
Discussion of reputation systems… need to make progress on having in-protocol solutions for these situations.
[Lightclient] We should be more intentional about the proposer/builder separation direction of travel. [Terence] noted that no-one from FB joins these calls, though they host their own calls. EF could host general "block builder meetings" to help maintain some neutrality. Break-out room calls have been working well as part of ACD.
Discuss further at DevConnect. Aslo see #block-production
on discord to discuss builder API related things.
[Alex] Take a look at Partial Withdrawals PR - looking good; mostly about parameter tuning right now.
There are a couple of pending PRs in the networking spec. Please review!
No call in 2 weeks due to DevConnect. Next call in 4 weeks.
But ACD in three weeks, 29th April, will cover a lot of Merge things, so please join (and on the 15th).