Notes on ETHGlobal’s Merge Summit
For the full video, just typein ETHGlobal into YouTube.
– For what is arguably the most important conference for a $2 tn industry this year there was scant attendance. Most of the stream had only 100 viewers, with Vitalik's final presentation "spiking" to 300-500. Meanwhile, the $2 tn walled garden fruity corporation's keynote drew millions of viewers earlier in the week. We are so, so early.
– Everyone was coy about timelines, except Tomasz (Nethermind) surprisingly let slip a target for October 2021. Reading between the lines, I speculate the target is indeed Q4 2021, but everyone's waiting to see how the Rayonism multi-client devnets play out. There's a small chance Shanghai goes before The Merge, which delays it to Q1 2022.
– Most clients are ready for Rayonism. The code changes for the Merge is relatively simple and most clients will be done in weeks, not months. Much of the remaining work is testing, testing, auditing, and more testing. Danny expects official multi-client testnets to commence in a couple of months.
– The "Ethereum 2.0" branding or "Eth2" nomenclature is no more, and going forward it is all just one Ethereum. The two layers are now called execution layer (formerly eth1) and consensus layer (formerly eth2). So, if you see people talk about "Eth2", please correct them. Bonus: harp on about false sequentiality and make Danny proud!
– While the beacon chain seems to be much delayed, the actual engineering and implementation was completed in only 2 years – a remarkable achievement given it took that long for EIP-1559 (a much simpler upgrade) to make it to mainnet. Hsiao-Wei had a fascinating opening talk about how research is iterative and a lot of seemingly wasted time in 2016-18 strongly informed sound decisions for the actual implementations.
– Client diversity is key. Vouch is a very interesting solution that builds redundancy and diversity for validators.
– Both execution layer and consensus layer clients will continue developing in parallel after The Merge, but we may see some synergies, packaging innovations etc. Since Consensys develops clients from both layers, we may end up with a single "Beku" client. (I forget who made that joke – Trent? – it was a good one!)
– While solo staking requires 32 ETH, there'll be a ton of innovation to bring various types of delegated staking with varying magnitudes of decentralization/centralization. It's imperative we get people to move away from Coinbase/Binance/Kraken and onto these more decentralized options. Personally, I'd like to see a protocol where people can run their own validators for 1 ETH or so, making it more of distributed validation rather than delegated validation. No idea how this can be done, but someone will figure it out…
– MEV will live on after The Merge, or rather, VEV. Of course, given there are far more validators and entities involved – 120,000 versus a few dozen mining pools – we'll need new ways to organize things. But the Flashbots team are already working on it. I think we'll need a lot of coordination and ensure most validators are running VEV extraction solutions to ensure the community extracts the value and not opportunistic frontrunners.
– Secret shared validators are live on testnet!
– Alright, after The Merge… first we'll see cleanup fork soon after which will also enable withdrawals.
– Next step is data sharding, which would be the significant feature focus after The Merge. Simultaneously, the execution layer teams will work on Shanghai and Cancun. Data sharding will bring 25x scale to rollups and other L2 solutions that choose to leverage the massive data availability. First we'll have committee-based consensus, with the innovative data availability sampling coming later. Speculation: I'd expect data sharding to go live by late 2022, with DAS in 2023.
– Parallel to sharding, the key execution layer upgrade will be statelessness + state expiry. The current plan is to do them together. Post-state expiry, we'll have static expired states, which would be fairly easy to distribute – could even use IPFS or BitTorrent. There can be protocols to make the UX simple. This will be significant breakthrough for how blockchains work and will enable much higher gas limits and scaling on L1.
– EVMX – 384-bit EVM, and other continuous improvements. Vitalik discusses the many drawbacks of WASM, and believes we can get EVMX to a point where it has most of the functionality of WASM without any of the drawbacks. RIP eWASM. My personal opinion: We'll see rapid innovation with VMs on L2 and a lot of the learnings can then be implemented back on to L1.
– The final piece of the puzzle is SNARK-ing everything! After that, STARK-ing everything for post-quantum. This will make native execution on shards easier. Interestingly, this was the only time Vitalik even mentioned execution on shards, and "if desired". He's very confident about rollups. Speculation: we aren't going to see execution on shards using the prior fraud proof mechanisms ever, and we'll turn on execution only when it's done right. There'll be significant innovation around interoperability and communication between L2s, which will inform how executable shards work. Finally, CBC Casper for a more secure consensus mechanism. These upgrades are 3-5 years out. Beyond this, Ethereum L1 will achieve its final form and might go into "maintenance mode" middle of the decade with all of the innovation moving to L2.
There's a lot more, but these are my important takeaways.
In short, Ethereum remains the only project that's even attempting to solve the blockchain trilemma with innovative cryptographic techniques. Bitcoin has given up on scalability, while all other L1 protocols have given up on decentralization and/or security. I should note that rollups can also be considered as solutions to the trilemma – Arbitrum, Optimism, zkSync 2.0 and StarkNet are the true "Ethereum killers". This will become evident over the next 5 years. If you watch The Merge summit and understand that now, I believe this is what the cool kids call an "alpha leak".