Thread by @TimBeiko

We wrapped up another @ethereum #ACDE today: covered Shapella, EOF, SELFDESTRUCT, EL -> CL state access, local block building & verkle tries 🌳

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

Recap below 👇

https://twitter.com/TimBeiko/status/1636427591191261184

@ethereum First, Shapella is coming! The announcment came out earlier this week:

https://twitter.com/TimBeiko/status/1640722906744487936

@ethereum Note that @ErigonEth has had a version update since the initial announcement:

https://twitter.com/ErigonEth/status/1640823557566472194

2.42.0 is what you want to be running if you use the client!

@ethereum @ErigonEth Lodestar has put out a new (optional) release, and Prysm might do so as well prior to the upgrade. That said, the versions in the announcement for both of these are compatible with Shapella

https://twitter.com/lodestar_eth/status/1641446011553091588

@ethereum @ErigonEth Nethermind has some big updates planned in their first post-fork release: they decided to delay them to minimize any risk with Shapella going live

https://twitter.com/URozmej/status/1640992652140113921

@ethereum @ErigonEth Testing is entering its final stages, with a last mainnet shadow fork scheduled next week and cross-client EVM fuzzing of the latest releases.

@ethereum @ErigonEth Next up, @alexberegszaszi gave an update on EOF - whose latest spec version is ambitiously named 😂

notes.ethereum.org/ZOEv9yyHQdmxTV…

@ethereum @ErigonEth @alexberegszaszi In short, the spec addresses some of the concerns raised about code & gas observability, and tries to remove these as much as possible from EOF contracts. The link shared above highlights all of the required changes, which were spread out across multiple EIPs previously.

@ethereum @ErigonEth @alexberegszaszi A diff with the previous meta spec can be found here: gist.github.com/axic/85c0ed993…

@ethereum @ErigonEth @alexberegszaszi Alex went deep into the details on the call, so I recommend watching it for all the nuance. At a high level, to remove code/gas introspection, contract creation must be significantly altered, and several opcodes are disabled or modified.

@ethereum @ErigonEth @alexberegszaszi Changes could be done to the proposed 4844 transaction type to accommodate the new contract deployment constraints.

@ethereum @ErigonEth @alexberegszaszi Teams were generally positive towards EOF, but skeptical that it could fit alongside EIP-4844 in Cancun. There were proposals to potentially schedule it for the next fork. That said, there was also some pushback related to how this would impact Verkle Tries... more on that later!

@ethereum @ErigonEth @alexberegszaszi Next up, Dankrad shared a new proposal to deal with SELFDESTRUCT: EIP-6780 (eips.ethereum.org/EIPS/eip-6780)

@ethereum @ErigonEth @alexberegszaszi This EIP would follow the rules laid out in EIP-4758 (eips.ethereum.org/EIPS/eip-4758) *except* when SELFDESTRUCT happens in the same transaction as a contract creation. This would hopefully strike the right balance between implementation complexity & not breaking deployed contracts.

@ethereum @ErigonEth @alexberegszaszi We will be analyzing contracts more thoroughly in the coming weeks - if you think EIP-6780 would break something significant on chain, please make yourself heard here: ethereum-magicians.org/t/deactivate-s… 📣

@ethereum @ErigonEth @alexberegszaszi We've agreed to CFI EIP-6780 for Cancun to send a signal that "this is happening", but obviously will only make a final decision about inclusion once more analysis has been done.

@ethereum @ErigonEth @alexberegszaszi Next up, @ralexstokes gave an update on trust-minimized CL access by the EL: notes.ethereum.org/@ralexstokes/n…

@ethereum @ErigonEth @alexberegszaszi @ralexstokes As the doc states, the first part here is providing access to the CL's state roots on the EL, which is what EIP-4788 proposes. That said, to verify proofs, you would also need access to the CL's Generalized Index (github.com/ethereum/conse…)

@ethereum @ErigonEth @alexberegszaszi @ralexstokes Alex proposed three options to address this on the call. There weren't a ton of reactions, so teams will also look into this in the coming weeks.

@ethereum @ErigonEth @alexberegszaszi @ralexstokes Next up, @mkalinin2 asked for updates on implementing the builder override flag in EL clients: github.com/ethereum/execu…

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 As a reminder, this flag would allow ELs to suggest the CL override external builders based on heuristics. This could help in cases of censorship that is detectable by clients (e.g. transactions re-orged out, pending despite a high max fee, etc.).

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 The question was around whether this should be done before or along with Cancun. Some client teams said they can likely start prototyping this soon. Once we have some prototype implementations, we'll merge in the API changes.

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 Lastly, we had an unscheduled update by @gballet on Verkle Tries😄 During the EOF conversation, there was some debate about its readiness relative to VT, so Guillaume walked us through all the latest updates. If you haven't followed Verkle recently, I recommend the full stream 📺

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet First, Guillaume shared there are currently two VT testnets: Beverly Hills, which is running a minimal, EL-only, VT spec, and Kaustinen, which also implements CL changes. Currently, the VT implementations are ~40% slower than MPT, but improving rapidly!

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet Then, @URozmej bombarded Guillaume with questions! Quick summary:
Re-orgs have been tested, JSON RPC performance hasn't. Tests have been run with mainnet-style blocks, but not shadow fork yet. No multi-client devnet forks yet, and a new sync mode will be needed for VT.

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet @URozmej There were then some questions about whether all clients could do this without adopting a similar data model to Erigon. It could be done, but would require relying on these nodes to gossip the VT, which is a risk. Nethermind shared they are working on a similar DB architecture!

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet @URozmej The conversion from MPT to VT is also still a bottleneck, taking 18h on the quickest implementation (Erigon) and much longer on others.

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet @URozmej Finally, Guillaume shared some high level details of the trie transition: some nodes would build the VT, and share it over gossip. Then, clients would maintain both the VT & MPT in parallel. Once the VT+MPT are sync'ed, you can switch the consensus rules from MPT to VT.

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet @URozmej Oh, actually, there was one more question: do zkEVMs make VTs obsolete? @dankrad shared that, in theory, yes, but we're still very far from this happening on mainnet. Provers still require significant resources, which would fundamentally change trust assumptions on Ethereum!

@ethereum @ErigonEth @alexberegszaszi @ralexstokes @mkalinin2 @gballet @URozmej @dankrad And that was it! Next ACDE is set for April 13, right after Shapella 👋! See you then 😄