Thread by @TimBeiko

Had another @Etherum #ACDE call today: most of it focused on 4844, with some brief discussion of L2 EIPs, and overall Cancun planning 🏖️

Agenda: github.com/ethereum/pm/is…
Stream: youtube.com/live/s6q5z53SI…

Recap below 👇

https://twitter.com/TimBeiko/status/1652006134142889984

@Etherum On the 4844 front, we first discussed some issues with the input format to the precompile. They use little-endian, which isn't used on the EL (but is the default on the CL), while the output is big endian, the default for the EL.

@Etherum @yperbasis asked if we could harmonize this, ideally using big endian to match the rest of the EL. After some discussion, we agreed to move forward with this! We'll open a PR to the spec and merge it if no objections come up ✅

@Etherum @yperbasis Next up, we spent nearly an hour discussing whether the 4844 transaction type should use SSZ or RLP 😅 I won't go over the entire conversation here. Recommend the livestream if you want the entire nuance, but here is a TL;DR

@Etherum @yperbasis For background, the 4844 transactions are currently spec'ed using SSZ. There is a general sense that the EL would like to move transaction encoding from RLP to SSZ, so this was done as a "first step" towards that.

@Etherum @yperbasis That said, over the past few weeks (months?), we've been trying to figure out whether what is spec'ed in 4844 would be forward-compatible with a full SSZ transition. Etan from @ethnimbus, @lightclients @roberto_bayardo and others have been deep down this rabbit hole.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo There were three potential paths: make some minor modifications to the current SSZ transaction spec, hopefully future-proofing it, doing a full SSZ transition as part of Cancun, or changing 4844 to use RLP encoding.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo The challenges with minor SSZ modifications are that (1) we might get it wrong and need to change the transaction format anyways and (2) we may never (or not for a while) get to the full migration, so be in a spot where the EL only has SSZ for that single transaction type.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo The challenge with the full migration is that it's a much bigger piece of work, and there are still open questions to figure out. It's hard to see how this would not delay the upgrade.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo And the challenge with RLP is that the CL does not use it: so we'd effectively push the same (opposite?) problem over from the EL to the CL 😅

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo Earlier this week, @lightclients proposed a potential solution to this, which would not require RLP on the CL: github.com/ethereum/conse…

After some discussion, it seems like we have a workeable design!

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo It wasn't so straightforward on the call 🙃, but that's what we ended with!

We'll use RLP on the EL, SSZ on the CL, and (already?) have the blob hashes in the CL blocks, which are sent over the Engine API. Each layer can then verify data against them using their own encoding 🎉!

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo Next up, @elbuenmayini raised an issue about a block validity condition that was implicitly implemented by some clients, but not all. We agreed to make this an explicit validity condition, and Mario already has a PR up with the change 🔥

github.com/ethereum/EIPs/…

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini Then, we had a long-ish conversation about the state of the KZG libraries. @peter_szilagyi shared some of his concerns about their design + support across various types of hardware. @potuz1 echo'ed his concerns, saying they are something CL clients have had to deal with for years

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 Recommend the livestream for the entire conversation, but beyond the hardware support, there were some concerns raised about the threading models, the amount of changes being done to libraries and client team's ability to diagnose potential issues.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 The Geth team is therefore adding support for both c-kzg and go-kzg, to allow users to easily switch if either had a bug. There wasn't a clear conclusion to the discussion, but we'll keep it going async in the coming weeks!

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 Then, already 5-10m over time, we had two more EIPs to cover 😅!

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 The first one was a bit weird for ACD: it's a proposal to introduce/standardize CSR on Ethereum L2s: ethereum-magicians.org/t/eip-6969-gen…

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 The conversation quickly moved from the EIP itself to whether ACD should even be involved 😅 A few people shared opinions: @shemnon thinks ACD mostly shouldn't, but perhaps another governance body should? He had shared his view on the agenda earlier, too

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon I mostly agreed, although there may be cases where it makes sense for ACD to provide technical feedback, esp. if the EIP is implemented on L2s with the goal of potentially be implemented on L1 at a later date.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon @wschwab_ from the Polygon team was on and said he thinks there is value in ACD acting as a forcing function for standardization, and highlighted account abstraction EIPs were another good example.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon @wschwab_ @adietrichs wasn't sure if ACD is the right forum, but was open to using it to bootstrap a more separate process.

No conclusion here either, but it's something I expect we'll talk about more and more going forward 😅

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon @wschwab_ @adietrichs Lastly, Charles Cooper, one of the EIP-5656 authors asked if client teams had thought more about it since the last call. We were already ~15 minutes overtime, so it didn't feel like the right time to make decisions about the scope of the upgrade.

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon @wschwab_ @adietrichs That said, there were some positive comments towards the EIP. I proposed that client teams take the next two weeks to think through the additions they'd like to see to Cancun. The list of currently included and CFI'd EIPs is here: github.com/ethereum/execu…

@Etherum @yperbasis @ethnimbus @lightclients @roberto_bayardo @elbuenmayini @peter_szilagyi @potuz1 @shemnon @wschwab_ @adietrichs Hopefully, in the next one or two calls, we can finalize the scope for the upgrade!

That was it 😄 Went a bit overtime today & I appreciate folks sticking around! See you on May 25 for the next ACDE 👋