HackMD
Eth2 Implementers’ Call #35 - 2020-03-12
[Quick contemporaneous notes by Ben Edgington]
Agenda: https://github.com/ethereum/eth2.0-pm/issues/132
Livestream: https://www.youtube.com/watch?v=orVYfqP_YuQ
Testing and Release updates
v0.11 released last night. Includes DoS hardening on the network. Also a few things in the state-transition spec: more easily isolating chains; adding proposer index to blocks to allow easier verification without having access to all the state. This is the stable multi-client target. Not expecting any further changes to state transition beyond small fixes.
Likely will be some more changes to network spec as we gather feedback.
Next testing priorities are fork choice and networking.
Zrnt has been updated to 0.10 (including BLS). Test vectors are out for v0.11.0. Rumor network testing tool: scripting approach for setting up testing scenarios. Should be really easy to script network tests now.
Ph0 audit results will be release asap.
Phase 1 is making good progress. In a couple of weeks would be a good time to begin prototyping this.
Client updates
Lighthouse
Research on peer reputation. Contributed ENR crate to open Ethereum. Improving Discv5 in a number of ways.
Validator UX survey wrapping up soon. See Sigma Prime Twitter.
Trialling Remerklable. Optimising BLS for better block processing.
Prysm
Interopping with Lighthouse on their EthDenver network.
v0.1 of slashing service is active on Sapphire network, and succeeded in detecting a double vote.
New SSZ for Prysm is significantly faster. Uses a code generator. State-generation service in progress. Noise interoperability. Working towards v0.11
Nimbus
Priorities for March:
- Interop
- Start audit process
- Eliminate libp2p daemon
Almost at 0.10.1. Some bugs in BLS. Progress on attestation aggregation.
Discv5 can interop with Geth, but not yet with Lighthouse. Improvements to libp2p stability. Log analysis tool.
Performance: found bug in libp2p integration. Traffic is now 5x improved.
Teku
On progress,
- Discovery v5 is in, modulo a bug or two
- Deposit processing is all done
- Sync now seems in good shape
- Our RPC implementation is making great progress. As a heads-up, there's an interest within PegaSys on working to standardise RPC across clients. An effort started a while ago, but seems to be on the back-burner. There is strong interest in this from Infura, as varying RPC interfaces is a major pain point for them, so they may get involved. Anyway, look out for more news on this.
On performance,
- We've finally merged in a massive PR from Anton to implement Proto's Remerklable scheme. We see a 4-times speedup in block processing (200ms), and a 4-times slowdown in epoch processing (1200ms), so it's a win overall. We're now working on improving this in the context of epoch processing.
- 98% of our block processing time is signature verification, so that's the next target. We'll likely move to Herumi pretty soon.
- A big chunk of our sync time is database operations, so we are working on that as well during this sprint.
Plans,
- Finally, we're ready to merge 0.10.1 into our master branch. We've been delaying this as we've been testing our sync against Prysm's Sapphire network, but we think the time is right, on the day that v0.11.0 appears!
Trinity
Stability improving. Discv5 - interopping with Geth and Lighthouse. Updates towards latest spec release.
Lodestar
Discv5 almost done. Doing some work on storing data more efficiently (i.e. not every state). Results from some audits: EIP2333-5 BLS work. Results fairly good, will share soon. Plan to sync some public testnets.
Nethermind
Update to 0.10.1, including BLS. Good progress on other aspects.
Research Updates
TX/RX
Two recent write-ups (links in chat)
- Safety of Eth1 follow distance
- Non-sequential receipt cross-shard transactions
Vitalik
Polynomial commitments post (replace state roots). Becomes easy to have short witnesses, and makes requirements for statelessness much more palatable. Relies on some very recent research by Aztec.
Block bodies, and maybe receipts, would be the easiest starting point for using these. Could be used for shard block data. Shorter proofs, data availability checking "for free", better proof of custody schemes. In the longer term could be used for the state as well.
Now more optimistic about being able to use zkSnarks over a full virtual machine.
Quilt
Looking at feeding Eth2 research on statelessness into Eth1x stateless research.
Networking
Productive call last week. Notes
Some meetings and planning was done at EthCC with Dmitriy around networking. If your discv5 is ready for testing, Proto/Felix can test it against the Go implementation.
Danny will decide whether to do a call next week or later. No views expressed on the call.
Spec discussion
Nope.
Testnets
March is multiclient experimentation and getting to v0.11. April will be multiclient testnets.
Open discussion
Was hoping to do another interop event, but not feasible due to the virus 
Nothing else - everyone is too busy and heads-down 
Chat highlights
- From danny to Everyone: 02:03 PM
- https://github.com/ethereum/eth2.0-pm/issues/132
- From protolambda to Everyone: 02:09 PM
- link to Rumor, ongoing work on testing branch, https://github.com/protolambda/rumor
- From danny to Everyone: 02:17 PM
- https://ethresear.ch/t/fast-verification-of-multiple-bls-signatures/5407
- From Joseph Delong to Everyone: 02:23 PM
- Write-up covering the saftey of Eth1 follow distance
https://ethresear.ch/t/on-the-way-to-eth1-finality/7041
Write-up covering non-sequential receipt cross-shard transactions
https://ethresear.ch/t/appraisal-of-non-sequential-receipt-cross-shard-transactions/7108 - From danny to Everyone: 02:25 PM
- https://ethresear.ch/t/using-polynomial-commitments-to-replace-state-roots/7095
- From Me to Everyone: 02:31 PM
- Networking call notes: https://hackmd.io/@benjaminion/rk2OEQ64L
