Eth2 Implementers’ Call #48 - 2020-09-17

[Quick contemporaneous notes by Ben Edgington]

Agenda: https://github.com/ethereum/eth2.0-pm/issues/181
Livestream: https://youtu.be/5GYF6gCIbGE

Testing and Release updates

v0.12.3 coming in a couple of hours. Please take a look at modifications to messageID in gossipsub.

Proto has been working on something like Hive for Eth2: easy network spin-up for testing.

Q. Merging fork choice spec tests? Current plan is to leave them as-is, but to use Proto's framework to test via the network interfaces of clients. Revisit in a week or two. [Jacek] This sounds more complex than it need be, since fork choice is just a function. Ideally should test it in isolation. [Mamy] Interesting edge cases occur that may be difficult with a network. [Danny] With only Rumor plus a node then exact, predictable scenarios can be fed into the node. [Proto] It's not just the algorithm that matters, the integration is important as well. E.g. tracking the attestation pool, latest attestations, gossip, etc. Not just a state-machine. Client unit tests already take care of the basic scenarios.

Spadina

Plan is to launch a brief 3-day testnet to run through Genesis. Config is merged. Launchpad due for release on Tuesday. 1024 minimum validator count. Genesis planned for September 29th.

Client updates

Nimbus
Several networking fixes, especially sync and memory leaks. Peer management improvements. Gossip v1.1 is ready and in test.

Slashing protection is merged. Attestation aggregation in place. DB pruning (offline) is done, will make on-line. New sync algorithm which is much improved.

Started last phase of audit: focused on validators, secrets, keystores, slashing protection, RPC, commandline, gossipsub, documentation.

Teku
Improving sync algorithm to use multiple peers. Implementing standard REST API. Implementing weak subjectivity things. Fast BLS library now runs on ARM.

Fixed some network vulns in discovery. Discovery v5.1 updates.

Lodestar
Running into memory issues on Medalla. Now mostly fixed, can now prune states much more intelligently by breaking up block storage by checkpointing. Now have state regeneration.

Put fork choice into separate package. Found some bugs around attestation balances. Refactored Eth1 data fetching. Gossipsub v1.1 update in progress. Plan to have all this merged for Spadina.

Prysm
Fixed a few attack vectors, and more robust core logic. Implemented validator exits. Integrated Blst library. Full speed on Eth2 API implementation. Started weak subjectivity sync. Peer scoring. Fixing up IPv6.

Lighthouse
Two new team members. All standard API endpoints are implemented and tested. Will have a code freeze for audit. Checks for weak subjectivity.

Plan 0.3.0 version that will have some breaking changes in 2 weeks. v0.2.9 will be updated only with critical fixes.

Working on a remote signer implementation.

Updating discovery to v5.1. Improving sync protocol to be more resilient to non-finality.

Research Updates

Aditya
Update on the weak subjectivity guide. Implement basic version first: take checkpoint as command line input and confirm it is in the synced chain. Can also input a state and check it is within the WSP.

Danny
Phase 1 stuff: working on reward accounting and micro-incentives for duties. Simplification of reward calculations.

Leo
Will post a report of gossipsub study.

BLS v4

There is a small incompatibility with our practice in the latest BLS spec v4 version: banning of the zero public key.

For mainnet release we will likely bump to this version. So clients need to update accordingly. Blst library will update accordingly.

We will not be using wrappers around the standard libraries.

Likely to be a PR destined for v1.0 spec release. Should we update Medalla? We will probably eliminate Medalla around mainnet launch, and run a new testnet with the mainnet spec. Not planning to update Medalla. [Afri] Why not do both? Note that there are not yet any zero public keys on Medalla. Action: Danny to document what the upgrade process might look like for Medalla, and we can decide.

[Carl] Note that EIP-2333 is being updated to take into account new keygen method in BLS v4. This affects key-recovery primarily; need to hunt down old versions if you need to do this.

[Mamy] Also note auditor comments on input validation in EIP-2386 on keystores/wallets.

API Status

See above for updates from client teams. Any other issues? Everything is moving forward.

[Proto] Beaconcha.in is moving from the Prysm API to the standard API.

Networking

Proto is working on the test runner (as above). Should soon be able to run multi-client network tests.

Spec discussion

Phase 0 EIP suggests lowering penalties for mainnet launch. Reduce coordinated slashing penalties by factor of 3; reduce leak penalties by a factor of 4; minimum slashing penalty reduced to 0.25 Eth. Plan to increase back to spec params within 5 months, via a hardfork.

v1.0 spec to be released in a couple of weeks with these params. There will be a new test suite, and we will run some multiclient test networks. Will also include BLS v4 params.

Open discussion

Start prepping releases for Spadina!

---

Chat highlights

From danny to Everyone: 03:03 PM

https://github.com/ethereum/eth2.0-pm/issues/181

From protolambda to Everyone: 03:12 PM

sorry I missed the first few minutes, helping design to new block explorer features today with the beaconchain team. And they are moving to new standard API :)

From Aditya Asgaonkar to Everyone: 03:20 PM

https://github.com/ethereum/eth2.0-specs/blob/weak-subjectivity-guide/specs/phase0/weak-subjectivity.md

From Leo BSC to Everyone: 03:23 PM

https://github.com/leobago/BSC-ETH2/blob/master/gossipsub-benchmark-tools/Testing_Docs.md
https://github.com/leobago/BSC-ETH2/blob/master/gossipsub-benchmark-tools/port_collision_issue_report.md

From Carl Beekhuizen to Everyone: 03:25 PM

Wrappers: https://github.com/ethereum/eth2.0-specs/pull/2068

From Carl Beekhuizen to Everyone: 03:30 PM

https://github.com/ethereum/EIPs/compare/master…CarlBeek:eip2333_bls_v4

From Mamy to Everyone: 03:31 PM

https://github.com/ethereum/EIPs/pull/2386#issuecomment-693575270