description: Notes from the regular Eth2 implementers call
# Eth2 Implementers’ Call #41 - 2020-06-11
[Quick contemporaneous notes by Ben Edgington]
## Testing and Release updates
There are a couple of non-substantive clarifications in fork choice PRs going up. Probably a release in the next week to include them.
Lighthouse has integrated the fork chioce tests. Still looking at how to release them.
[Proto] Testnet [validator keystore tooling](https://github.com/protolambda/eth2-val-tools), trying to work around all keystore incompatibility issues. Zrnt updates. [E2DB](https://github.com/protolambda/e2db/) - open source explorer, sits on LH API to see forkchoice and blocks. Extracts that info to a database. Offers good analytics.
Beacon Fuzz updates. Prysm now has native Go builds. Allowed progress in integrating Prysm into Beaconfuzz. Found 2 bugs as a result. Found panic in Lighthouse ENR crate. Also integrating Lodestar: 4 bugs found. Helping with BLS fuzzing for Eth1/Geth. Working towards dockerising Eth2fuzz. Confident that can get fuzzers running on Google infrastructure: free fuzzing for OSS projects.
[Aditya] [Small bugfix](https://github.com/ethereum/eth2.0-specs/pull/1886) to fork choice rule. Found by Alex Vlasov.
Prysmatic will be launching Onyx testnet (see below).
[Afri] Text update:
* Schlesi is history. I purged the last validator and bootnode.
* Witti is fairly stable, 100k slots, liveness almost perfect, 75-90% validator participation. Not aware of any major issues.
* Altona might happen in the next one and a half weeks, only one client ready for the 0.12 spec yet. we might be able to launch with 4 clients this time, I'm confident Nimbus will be able to provide a genesis validator too. that said, going forward, I'll invite the client teams for genesis, so they have a chance to run their own nodes upon launch. if altona works out we might be able to prepare for an official, public multi-client testnet.
### Validator key hygiene
We need to start training users on good key management today. E.g. don't store withdrawal credentials on staking node. Air-gapped generation.
PSA: start shifting docs and processes in favour of good practice.
[Mamy] Nimbus has asked auditors for advice on secrets management. EF is also getting advice on management for hobbyists.
[Raul] Prysm is looking at having basic commands vs power-user commands. Lots of questions from users on how to handle/store/move keys.
Keystore vs. wallet. [Proto] How to deploy 1000s of validators of different clients. Initial tooling geared towards Prysm, which is wallet-centric (`ethdo`). Lighthouse is keystore-centric, loading from filesystem directories. Secrets are managed separately. Teku is doing similar things. We should try to systemise these things. [Carl B] Want to update EIP-2333 to the new salts from the BLS standards. [One-liner PR](https://github.com/CarlBeek/EIPs/commit/690ddcc12c06f471acb4c06c1d6897d09338e489). Concerned that we don't end up with multiple different implementations. Discussion on EIP 2334 on key derivation paths. Preference is for hardware wallets to export keystores. This is not possible today, and they may not have the hash-power required to produce a keystore. Clients depending only about keystores is the maximally flexible and simple option. Don't want to tie withdrawal and signing keys too closely together.
Discussion points to cover in some forum:
- Evolving EIPs/standards
- Guidance on use of keys/keystores in clients
Question, do clients need access to deposit data/withdrawal keys to start up?
Do we need a standard for keystore and secret management in clients? Propose calls with stakeholders to discuss. [Jacek] Need to define minimum necessary; clients can do other things if they like.
[Carl] Anyone opposed to using unicode passwords for encrypting/decrypting? This is hairy, but there are formats that are consistent. From i18n point of view it might be important. It's also a good defence against dictionary attacks. But current practice mostly excludes unicode. BIP39 allows for a unicode password. Needs further discussion. **Danny to organise call next week. Also, create Discord channel.**
## Client updates
Version 0.12 update is going well: BLS is done, Gossipsub 1.1 is done, fork choice, state transition and validator changes are done. Just finishing off the last networking changes.
We've had a big focus on our memory footprint and managed to reduce the remerklable memory use for hot states, and we've just merged a big rework that limits the number of states stored in memory and regenerates them as needed.
Other than that, mainly fixing bugs, reducing the noise in our logs, and improving usability.
A few exciting changes. Block and attestion processing 10-50x faster (but not full remerklable). Validator client split in progress. First PR on 0.12.x, including BLS. Can toggle between 0.11 and 0.12 at compile time. EIP 2333 is in PR. Audit will start in 10 days, will announce soon. Grafana monitoring updated. On Witti - you can start Nimbus on your phone, it can sleep overnight and resync to head! Impreved guides for installing Nimbus, in particular on phones. No Go required any longer. A few bugs exist, e.g. Windows 32-bit, ARM64, which are under investigation. Also an issue with forkchoice: protoarray forkchoice doesn't lead to finalisation.
2/3 through the 0.12 update. Fork choice is done. Gossipsub 1.1 is in progress. Working on tooling to spin up ephemeral testnets. Initial sync is stable, but not yet switching over from sync to gossip of blocks.
Spec updates to 0.11.3. Optimisations to be able to run mainnet config. Efficient forkchoice: looking at protoarray.
Working on a proxy contract for the main deposit contract to verify BLS signatures. Get in touch if interested in discussing.
Implementing 0.12.1. Tests passing, but still some work needed. Fork choice is now in its own crate. Raised some interesting issues with the spec. Made memory usage available to Eth2stats. Natively supporting Raspberry Pi thanks to external contributor. Discv5 rebuilt, resolves CPU usage issue (not merged to master yet). Can use multiaddr rather than ENR. Trail of Bits has completed first round. Nothing critical so far. Selecting another firm for a second independent review.
At v0.12.1. Topaz is shutdown. Launching Onyx. Need 4000 more deposits, may start Saturday afternoon. Doing some multi-client experiments: Teku agrees on Genesis time. Will try Lighthouse next.
Now builds natively in Go. Reworking validator key management.
Working on Eth2/Eth1 integration, deposit workflow. Updating to 0.12.1. Looking at updates to the API. Will add capability to run tests automatically.
## API working group update
Update from [agenda discussion](https://github.com/ethereum/eth2.0-pm/issues/158#issuecomment-641305001):
* merged: /node namespace endpoint
* open PRs: /beacon/state and /beacon/pool endpoints
* in progress: /beacon/blocks
* missing: /configendpoints (a bit tricky, it's hard to keep params up to date), update /validator endpoints
[Danny] Expect convergence on a v1.0 within a week or so. Keep an eye on the Discord channel for discussion.
## Research Updates
Looking at verkle trees and other alternatives to merkle trees for state-storage. See Ethresearch post. Kate commitments have been the leading contender, but any changes lead to big recalculations. Verkle tree is like a Merkle tree constructed from Kate commitments.
Lighthouse can now use multiaddrs for discv5. Since LH was refactoring everything anyway, this wasn't a big problem to implement. Amount of work required is implementation-specific. UX around using multiaddrs probably makes it worth it.
## Spec discussion
## Open discussion
* * *
# Chat highlights
From protolambda to Everyone: 03:11 PM
: E2DB I mentioned earlier: https://github.com/protolambda/e2db/
*testnet* validator keystore tooling, trying to work around all keystore incompatibility issues: https://github.com/protolambda/eth2-val-tools
From Carl Beekhuizen to Everyone: 03:17 PM
From Alex Stokes to Everyone: 03:46 PM
From Aditya Asgaonkar to Everyone: 03:50 PM