PoS Implementers’ Call #67 - 2021-07-01

--- tags: eth2devs description: Notes from the regular proof of stake [Eth2] implementers call image: https://benjaminion.xyz/f/favicon-96x96.png --- # PoS Implementers’ Call #67 - 2021-07-01 [Quick contemporaneous notes by Ben Edgington; fka "Eth2 Implementers' Call"] Agenda: https://github.com/ethereum/eth2.0-pm/issues/224 Livestream: https://youtu.be/FNXk4ScqHn0 ## Client updates **Prysm** Mostly have been working on Altair. Aligned with the latest spec. Done beacon node/validator interactions. Almost there on networking, a few bugs. Very soon will try interop with the Teku devnet. **Lighthouse** Also working on Altair. Recently finished all validator features. Testing and bug fixes. Alpha 8 changes inplemented. Compatible with Teku. Ready for multclient testnet. Big networking refactor coming: discovery and gossip. **Teku** Ready for Altair. Alpha 8 changes are in and working well. Have added batch signature verification in latest release which has improved CPU usage. Added time-based Eth1 head tracking, as per spec. Doing a bunch of work to improve the discovery library. And will finally be able to switch over to the dependent roots functionality after finding the cause of a small performance regression. Have done work to bring our external signer up to Altair spec. Farewell to Meredith from the team :cry: We are hiring :smiley: **Lodestar** Up to Alpha 7 on Altair, and working on Alpha 8. Looking stable on devnets. Updated logo and branding! Made a benchmark tracking tool for CI to identify regressions. This has already proved helpful. Adding more metrics. Improved performance in gossip attestation validation. Getting together a demo of the lightclient work. Serving proof to initialise lightclient with the required info. **Nimbus** Passing Alpha 8 tests on Altair. Updating fuzzing interface so that Sigma Prime can test. Released Nimbus 1.4. Improvements in attestation effectiveness and CPU usage, with caching. Added REST API endpoints. Fixed false positive on doppelganger detection. **Grandine** New client: binary releases available. Cross-platform for Windows, Linux, MacOS. Would like feedback for testing on non-Linux platforms. There is a basic user interface which is helpful for debugging. Looking for ways to open source the code. ### General EF team is going carefully through the Altair testcases and there may be updates yet. Beacon Fuzz: Sigma Prime has reached out to some client teams with issues. Will kick off differential fuzzing in the next few days. ## Altair planning The next two major activies are lkely to be (a) a larger devnet, and (b) fork Pyrmont network. Danny suggests (after a few iterations): - Coordinated devnet at end of next week - Target forking Pyrmont by end of month, with a full decision to made in our next call. After that, set Prater and Mainnet fork dates. Dev teams and EF control over 2/3 of Pyrmont still, which helps with lead time. Note that EthCC is coming up - 20-22 July? Better to aim for week after. Q. from Grandine team: any clients considered having two separate runtimes to handle the fork process? It seems that other clients have not taken this approach. ## Research updates Merge call was recorded and will be shared soon. Merge code will be rebased on Altair soon. The Execution payload needs to be modified with the London fork. Continuing to work on test vectors. [Leo] Metrics standardisation. Finally have a consistent set across all clients. [Pari] It's all on the [Eth2 client metrics repo](https://github.com/ethereum/eth2.0-metrics). ## Spec discussion [Adrian S] has put up [a PR](https://github.com/ethereum/eth2.0-specs/pull/2501) to change the Eth1 data voting default (difference between "unknown block" and "no vote"). This could be a soft fork since there has not yet been a voting period that voted for 0x00 (to be checked). Rationale: this is useful for debugging. [Further discussion ensues...] ## Open discussion [Pooja] Peep an EIP did a session last week with Adrian S on Teku's Altair update work. There is a [playlist](https://www.youtube.com/playlist?list=PL4cwHXAawZxoEw29YmqJtNoFaENUUAREn) of past sessions related to the beacon chain. * * * # Chat highlights From Lion dappLion to Everyone: 03:24 PM : Should we do a long non-finality experiment on Pyrmont after forking? Or setup a new devnet just for that From danny to Everyone: 03:24 PM : yeah, I think it would make sense to leave half of our nodes/validators on for a while From Lion dappLion to Everyone: 03:29 PM : Is any client testing non-finality in CI or has manually test it on an own devnet in the last 3-6 months? From Adrian Sutton to Everyone: 03:30 PM : I split Yeerongpilly 50/50 accidentally for a few days a couple of times but that’s only small and it’s a relatively short period of time so not a great test. Was a very long reorg though. From Lion dappLion to Everyone: 03:31 PM : Other clients have compared block processing + epoch transition times, phase0 vs altair. Same, faster or slower? From Jacek Sieka to Everyone: 03:32 PM : There are new inefficient quadratic calls that need optimization From parithosh to Everyone: 03:32 PM : https://github.com/ethereum/eth2.0-metrics From Hsiao-Wei Wang to Everyone: 03:33 PM : Default to "zero" Eth1Data instead of current state PR: https://github.com/ethereum/eth2.0-specs/pull/2501 From Pooja |ECH to Everyone: 03:41 PM : https://www.youtube.com/playlist?list=PL4cwHXAawZxoEw29YmqJtNoFaENUUAREn