description: Notes from the regular Eth2 implementers call
# Eth2 Implementers’ Call #47 - 2020-09-03
[Quick contemporaneous notes by Ben Edgington]
## Testing and Release updates
Danny has been working on Phase 1 testing.
We are moving closer to a v1.0 Ph0 spec release. Scoring parameters for Gossipsub 1.1 need more work, which is underway using the Protocol Labs Testground simulator. Expect that to be done within 1.5 weeks. Propose a networking call next Wednesday.
Testnet: Medalla looks good right now. Plan to do a dress-rehearsal network launch, with a 2-3 day short-lived testnet. Getting Launchpad fixed up for that. Penultimate week of September is pencilled in for config params to be advertised, ~21st. Do the launch 1 week later.
Should we plan to do a practice hardfork on Medalla at some point? Could do it either before launch, or before any planned hardfork on Mainnet. May not be an immediate priority. There is a range of ways we could approach a hard fork: e.g. blacklist a block; change some parameter...
[Proto] Few issues with the attacknet, but now back up. Working to revamp network testing and reviewing some clients.
## Client updates
Getting good feedback on DoS vectors and fixing them. Wrapping up issues from Quantstamp audit. Implementing the Eth2 API - plan to have it supported for Mainnet. Looking at weak subjectivity sync. Fixing up slash-protection and the slasher. UX for client exits. Fixed a few interop things.
Fixed a big memory issue: prune unfinalised states. Fixed a stack overflow in fork choice. May replace fork choice now. Working on adding Phase 1 types and parameters. Mainly trying to get synced on Medalla: goal is to propose a block. Eth1 processing needs more work.
Bug fixes. Validation of Eth1 data improved. Failed proposals due to fork handling fixed. Huge improvements in attestation effectiveness. Plan to do flood publishing to help further. Implemented EIP 2335 updates. And have implemented the slashing data interchange format.
[Danny] in Gossipsub 1.1 flood publishing is likely to be the default
Audit: phase 2 of audit, cryptography and consensus. Some memory exhaustion attacks identified. Changing to use bounded buffers and dropping peers or blocks.
Couple of issues on Medalla. Nimbus needs frequent restarts. (1) Log file fills disks, will move default to less verbose. (2) DB fills disk. Will add pruning. (3) Memory usage issue, probably async timers in libp2p.
Still maintaining multinet script. Can test a small network with Lighthouse and Prysm and finalise. Encourage Teku and Lodestar teams to add their clients. See Eth2 client multinet repo.
Fixed a deadlock - see write up on blog. Fixed a pruning bug. Implementing http API. Looking at attestation inclusion. UX fixes for ease of use. Gossip 1.1 is in master and now defining parameters. Slasher found some unique slashings. Implementing slashing protection interchange format. Two audits starting in October.
Announcement: tabling the Eth2 efforts on Trinity for now. But open to any open source contributions. Focus on Mainnet for the time being.
## Research Updates
Phase 1 changes. Formerly a crosslink must have come from an immediately prior slot. There are a couple of [PRs](https://github.com/ethereum/eth2.0-specs/pull/2024) up to be more inclusive. Some changes to accounting for validator activity and to not store attestations. This is a simplification for crosslink accounting.
[Leo] Building a crawler using Rumor. First implementation is done, that does discovery. Next step is to add gossip.
## Weak subjectivity sync
Needs to be in place soon after mainnet launch as a priority. Alternate histories are an annoyance at a minimum and could make sync difficult.
Easiest approach is to have a finalised root specified, and reject chains that don't include it. But starting from a given state is better UX.
[Aditya] Low priority but high reward for now. Could specify this outside the p2p spec as a client-side tool.
[PaulH] Lighthouse will probably do sync from scratch and check it hits the given block for now. There are difficult edge case otherwise.
[Terence] Should we all agree on the command-line flags and format for the input checkpoints? General agreement that this is a good idea.
[Aditya] This will also help to sort out a badly forked testnet/mainnet like we saw a few weeks ago.
Aditya and Danny will work on this next week.
Call next Wednesday: Gossipsub v1.1 update; parameter choices; DoS vectors; monitoring tools.
What else? Nope.
## Spec discussion
## Open discussion
Nope. Short and sweet.
* * *
# Chat highlights
From terence to Everyone: 03:25 PM