HackMD
Eth2 Implementers’ Call #59 - 2021-03-11
[Quick contemporaneous notes by Ben Edgington]
Agenda: https://github.com/ethereum/eth2.0-pm/issues/208
Livestream: https://youtu.be/s017DQlsCCw
Client updates
Nimbus
v1.0.10 release today. Main things: updated Blst library & SSL lib - new trusted root certs, so please update. Compliant with v1.0.1 spec (new withdrawal creds). Fixed bugs including crash on some h/w configs. Fixed some long processing delays when pruning state before finalisation. Better libp2p peer management. Fixed false positive in doppelganger detection.
New features: improved format of slashing protection db. This version supports both old and new; next will have only new and will reduce disk load for large numbers of validators. Improved automatic config and detection of external IP addresses.
Lighthouse
Release 1.2.0. Several stability and performance improvements. Block production time reduced 4x. Better handling of early blocks. Pruning slashing protection db periodically. Removed pubkey cache file, now in db. Added ability to set graffiti from files. Updated BLS lib.
Teku
Focusing on upcoming hardfork/Altair support: versioned datastructures etc. Forkchoice balance attack mitigation. Looking into block–slot change. Pending config update for Prater testnet.
Prysm
Working on hardfork/Altair prep. Also working on datastructures to prepare for this. More work on standard API. Optimised slasher implementation. Recent release: runtime improvements.
Lodestar
Working on Altair compatibility, refactoring. Profiling and adding metrics to diagnose performance issues. Fixed a long-standing memory leak. Working on PoC light client implementation. Working on weak subjectivity state infrastructure. Would like to discuss how we collectively handle these things.
Altair
64-epoch feature
Feedback from teams suggests that this may not be the ideal optimisation, and may need more thought. Propose excluding from Altair. No objections.
Independent leak-scores feature
This was part of the above, but it is now pulled out as an independent PR. This is a nice feature - Danny plans to merge it today unless there are objections.
Helps network resolve partitions more cleanly.
Fork choice status
Proposed block–slot fork choice change 2197. (Also 2101, randomised timing.) These are not now slated for Altair, and may be abandoned completely. "The solution might be worse than the problem."
EF research team is investigating other approaches. Note that most fork choice changes can be made at any time, outside a hard fork.
Networking
Issue 2183 - there is a fix for this now. PR 2225 will likely be merged today. There are alternative approaches, but this will likely do for now.
Test format updates
A PR from last October introduced compression to reduce test data from ~10GB to ~0.5GB. There is a more recent proposal with further improvements, but compression alone seems ok for now. YAML versions of test data have been removed: it's all SSZ–Snappy now.
Fork choice tests: PR 2202 is a preview of the test format. Follows Alex Vlasov's previous work. Some improvements for easier test data generation. Please check the proposal and give feedback. Will initially build the "get_head" tests to see how it goes for clients.
Prater status
New testnet with significantly more validators. More than 200k. Decided to pre-populate genesis for ease of start-up. Proto created a tool to make a genesis file. PR is ready to go. Custom genesis delay puts it at 16 March. There are 210k validators. 40k per team plus the EF + 10k others. Keys to be distributed before the weekend.
Request to push back one week to the 23rd to have time to get client updates out and set up the infrastructure. NB changing the genesis time affects the genesis state, so be sure to have the correct state in clients. Request agreed: genesis will be 23rd March.
Research Updates
Mikhail
PR for the Merge (executable beacon chain) is up on the specs repo. Please review and discuss!
Spec discussion
Weak subjectivity state distribution. [Aditya] Current assumption is that we share states somewhere, but unspecified. We could look at doing it over the Eth2 p2p network, but the load may be prohibitive for that.
Clients could embed states into releases. None currently does. Teku can read initial states via the command line. Other clients are working on this.
[Mikhail] Dankrad's proposal to defer block execution by a slot on the executable beacon chain. Would be good to have more eyes on this.
Open discussion
[Vitalik] Anyone want to talk about speeding up the Merge? [Danny] Will start a public call in the next couple of weeks. Minimal merge will be the first pass in the specs in order to get to PoS quickly. Might exclude validator withdrawals, for example, and do those 3-4 months later. Mikhail and Danny are working on technical roadmap/project planning docs which will be shared.
Still targeting June-ish for Altair, but not set in stone.
Chat highlights
- From danny to Everyone: 02:03 PM
- https://github.com/ethereum/eth2.0-pm/issues/208
- From Alex Stokes to Everyone: 02:11 PM
- https://github.com/ethereum/eth2.0-specs/pull/2192
(and https://github.com/ethereum/eth2.0-specs/pull/2212 )
https://github.com/ethereum/eth2.0-specs/pull/2222
^ independent leak score - From Aditya Asgaonkar to Everyone: 02:16 PM
- #2101 & #2197 will be dropped from HF1. Most likely will be abandoned as well.
- From Alex Stokes to Everyone: 02:18 PM
- https://github.com/ethereum/eth2.0-specs/pull/2225
https://github.com/ethereum/eth2.0-specs/pull/2202
https://github.com/eth2-clients/eth2-testnets/pull/40 - From Mikhail Kalinin to Everyone: 02:30 PM
- https://github.com/ethereum/eth2.0-specs/pull/2229
https://ethresear.ch/t/commit-to-pre-state-instead-of-post-state-on-the-executable-beacon-chain/8802 - From Aditya Asgaonkar to Everyone: 02:36 PM
- https://github.com/ethereum/eth2.0-specs/issues/2174
