Thread by @TimBeiko

.@ethereum #AllCoreDevs starting in ~30 mins!

Lots on the agenda today: possible Berlin blocks 🇩🇪, p2p protocol changes, EVM384 & 1559 updates, and a quick presentation by me about feedback gathered about AllCoreDevs 😁

📆: github.com/ethereum/pm/is…
📺:

https://twitter.com/TimBeiko/status/1352628035124568065

We're on! First, YOLOv3 has launched. Besu and Geth are synced, and Nethermind and OpenEthereum have started syncing 🎉

We still haven't fuzzed the network yet, though. Planning on starting that next 💪🏻

And we're bringing up fork blocks ⛓👀

We've somewhat diverged into discussion potential ways we _could_ go about forking testnets. @peter_szilagyi had an idea about this a while back for "shadow forking" testnets, see: github.com/ethereum/pm/bl…

It seems we have consensus on a Feb 24th testnet fork 🍴🍾!

We'll hash out the specific blocks on each network async.

We're going to aim for a late-march mainnet fork, but we going to decide on the specific block next ACD call 👀

Next up is finalizing the eth/66 spec (eips.ethereum.org/EIPS/eip-2481). There's been a lot of conversation about this on the Eth discord this week. TL;DR all messages are formatted the same way under eth/66, except the ones for the block headers.

eth/66 "wraps" the previous format for each call, but block headers could be done without wrapping it. We need to decide whether we want to have all of them wrapped the same way, or leave the block headers differently.

@shemnon and @mhswende are going into the pros and cons of the various approaches on the call. Won't even try and summarize it here 😅

@vorot93 and @peter_szilagyi also chiming in. If this is stuff you are interested in, strongly recommend listening along. Good insights into how the inner bits of clients work!

So, after a _lot_ of back and forth, we've agreed to implement the EIP as it is today, not accepting the proposed change by @peter_szilagyi 👍🏻

FWIW, if we decide that was the bad approach, we can always change the eth protocol again.

Next up is an update on EVM384, which was shared in the agenda: notes.ethereum.org/@poemm/evm384-…

Two models to price gas are being proposed as part of the update: notes.ethereum.org/@poemm/evm384-…

notes.ethereum.org/@poemm/evm384-…

These models would not only price the new EVM384 opcodes, but propose repricings for the current opcodes.

One thing to note is that these repricings would require a major change to how gas is calculated, where we'd have to deal with gas costs that are fractional (right now, all gas costs are represented by integers).

They also have an ethresear.ch post listing some of the open questions: ethresear.ch/t/surveying-pr…

@mhswende has a comment w.r.t. to EVM384's simplicity vs. precompiles. Originally EVM384 seemed more simple than the precompiles, but with all of these changes, it may not be the case. Aside from that, he's wondering what the other benefits are for EVM384?

The biggest one is user-deployed innovation: with EVM384, anyone could use the "crypto building blocks" to build whatever they want.

A few examples are listed in the ethresear.ch post: ethresear.ch/t/surveying-pr…

More specifically, @Shamatar, says that working with operations >256 bits, but not necessarily 384 is valuable.

And we're wrapping up on this. The next two items on the agenda are updates from my, so I won't be live-tweeting them. I'll give an update afterwards.

Here are the links I'll be going over 👇🏻

First, a preliminary update on EIP-1559 large state testing: hackmd.io/@timbeiko/1559…

Second, going over some feedback I've gathered about AllCoreDevs over the past few weeks: hackmd.io/@timbeiko/acd-…

Quick update on 1559 first doc: one piece of feedback was that we should try and have nodes in separate regions to see if that affects the performance tests. Will do 👍🏻

With regards to the second doc, no objections to the first two suggestions: we'll try and have 1-2 calls to discuss the broader roadmap, and also set longer-term agendas for ACD and see how it goes. Didn't have enough time to get to the 3rd suggestion 😅

And that was it! See you all in two weeks: Feb 19th, 14:00 UTC 👋🏻