Thread by @TimBeiko

Another @ethereum #AllCoreDevs just wrapped up!

Agenda: github.com/ethereum/pm/is…
Recording:

Recap below 🧵

https://twitter.com/TimBeiko/status/1423680794090950659

First on the call, we had @mkalinin2 present a design document for the Merge Consensus APIs: hackmd.io/@n0ble/consens…

He went over it on the call and answered some questions from other attendees.

@mkalinin2 The document was inspired by what was done in the Rayonism hackathon (rayonism.io), the merge EIP (eips.ethereum.org/EIPS/eip-3675) and even opens the door to L2 interop, as proposed by @protolambda:

https://twitter.com/protolambda/status/1427819892288393218

@mkalinin2 @protolambda We'll have a call to discuss this in more detail next week (github.com/ethereum/pm/is…), and the @go_ethereum is working on a write up about the merge in general they hope to share on that same call 👀

@mkalinin2 @protolambda @go_ethereum Next up, I gave a summary of my conversation with client teams about what went well/bad with London, how we can improve for The Merge, and what their general roadmap is for the next ~6 months. Write up here: hackmd.io/@timbeiko/lond…

@mkalinin2 @protolambda @go_ethereum TL;DR:

Testing 🔩: teams would like to set expectations in advance of how long things should be live on testnets, what to do if a bug happens, etc. so that we don't have to take these decisions on the fly (and under pressure) as we're working on the merge.

@mkalinin2 @protolambda @go_ethereum Consensus vs. Non-Consensus changes⛓: once consensus changes are done, we should leave enough time for tooling/infrastructure to play with clients and help design the non-consensus interfaces.

@mkalinin2 @protolambda @go_ethereum Some teams were very involved for London (a shout out was given to @infura_io on the call!), but it'd be great to see more 😁 @trent_vanepps and I will keep pushing for this to happen.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps Lastly, when asked about their roadmaps, basically all teams agreed that, in addition to the consensus changes for the merge, they had other significant things they need to get done in the next few months.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps Most of these are things that clients will need to be in a better position for the merge itself: performance optimizations, better separations of the consensus engine, etc.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps Given that, we agreed to *not* have a "feature fork" in December. This means that we'll likely push back the difficulty bomb, and perhaps add other "one line" changes, but won't be including new EIPs before the Merge.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps After this, there was some conversation on the call about what to do about the various EIPs which are "pending". We discussed potentially adding them into the first fork after the Merge, but it's something we need to do more serious planning around.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps It is a bit early to make a final call given that we're still early in the merge work itself, but definitely something we will follow up on in the coming weeks/months.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps And those were the two big topics of the call! We then had a laundry list of smaller discussion points.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps First, there was a discussion about limiting the maximum value for account nonces. There was back and forth about using the Int64 vs. Javascript Int (2^52) cap. You can watch the call for the full debate, but we ended with Int64, i.e. EIP-2161 (eips.ethereum.org/EIPS/eip-2681)

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps Main reasons were: Eth2 does this, we don't optimize for JS elsewhere, and JS now has BigInt support. There are other fields which need similar decisions (a lot of them detailed in EIP-1985), and we'll put together a master list for people to review.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps Then, we had @shemnon on the call to discuss EIP-1352, which proposes to reserve some addresses for precompile contracts. Prior to Berlin, this wasn't a huge deal, but now that precompiles are considered "warm" under EIP-2929, which changes their gas costs, it matters.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon Specifically, a lot of L2s use precompiles to store their system contracts, so the gas price associated with those will depend on whether they are considered cold or warm by the EVM upon initial invocation, and making an entire address range warm by default could clear this up.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon Also worth noting that there were some issues on Berlin devnets because of this: some client teams treated the address of an implemented-but-not-activated precompile as warm, while it should have been treated as cold.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon There was some disagreement about the interpretation of the EIP (should the entire range be warm, despite not having a precompile, or should only precompiles within that range be warm?) and it made it clear more discussion was needed 😅

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon You can watch the recording for more nuance, but @shemnon and others will move the discussion to the EIP's @EthMagicians thread: ethereum-magicians.org/t/eip-1352-res…

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians Then, there were a few announcements related to @go_ethereum. First, ICYMI, they have scheduled a security update for next Tuesday:

https://twitter.com/go_ethereum/status/1428051458763763721

🚨 If you use Go-Ethereum, you should update to that 🚨

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians Second, while a PR had been merged to remove eth/65 support, it was reverted to not be bundled with the security upgrade and give the client teams who don't support eth/66 a few more weeks to do so. Both teams who didn't support it now have a PR/branch with implementations 👍🏻

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians Third, EIP-3607, which verifies that the sending account for a transaction is an EOA, was merged into geth. There are some corner cases here, which are described in the EIP's discussion post and will soon have test cases for it.

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians Because the EIP restricts a behavior that we've never seen on mainnet, it is activated as of the genesis block. Other client teams are in the process of adding support, with @nethermindeth investigating any potential concerns with @xdaichain support ⛓

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians @nethermindeth @xdaichain Another announcement was from @AlitaMoore85, who has been working on detailing the various corner cases in different clients' JSON RPC implementations (in order to hopefully unify them!).

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians @nethermindeth @xdaichain @AlitaMoore85 They've put togehter a list of the edge cases here: github.com/ethereum/execu…

They asked clients to review them, and also for any tools who have experience if/else-ing JSON RPC responses based on the node used to chime in 🛠👀

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians @nethermindeth @xdaichain @AlitaMoore85 Second to last announcement: with the merge coming up, we're starting to move away from the eth1/eth2 terminology in favor of execution & consensus layer. Repos, discord channels and forum tags are being renamed 🥳 You can find the list here: notes.ethereum.org/@timbeiko/grea…

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians @nethermindeth @xdaichain @AlitaMoore85 And, lastly, @dedaub has been working on some gas cost analyses for work on state expiry. They've put out a report, which they'll go over in a call next Friday, 15:00 UTC: github.com/ethereum/pm/is…

Report link: docs.google.com/document/d/1s3…

@mkalinin2 @protolambda @go_ethereum @infura_io @trent_vanepps @shemnon @EthMagicians @nethermindeth @xdaichain @AlitaMoore85 @dedaub And that was it 😁! Next ACD is scheduled for Sept 3rd, 14:00 UTC 👋🏻