2019-12-05 Eth2 Impl call 29

2019-12-05 Eth2 Impl call 29.txt

2019-12-05 Eth2 Impl call 29
https://github.com/ethereum/eth2.0-pm/issues/108

Nimbus updates

- Core:
  Finished the last SSZ bugs on consensus objects last week so now it can be dumped

  Optimizations: removing quadratic behaviours that reduced our CI time from 50min to less than 20min

  Last bottlenecks: SHA256 (33% time) and BLS
    not blocking for state transition within 6s

- Testnet
  Debugging syncing
  Polish on the instructions: tested from fresh start or from restart

- Tests and CI
  In the process of setting up Jenkins with the goal of
  testing Nimbus on controlled hardware, especially Android and ARM phones
  
  Auto-tracking of long-running tests, especially state transitions
  
  No more Git LFS
  
- Fuzzing
  similar to ncli we have nfuzz, a fuzzing endpoint
  Beacon-Fuzz PR pending, shuffling as a PoC OK
  
- Networking
  libp2p ready for integration in the coming weeks
  Disc v5 next.
  

———

Spec updates

Removal of signing root. (Private) Consensus across most clients for removal, anyone against?
Merge in 3 hours: https://github.com/ethereum/eth2.0-specs/pull/1491

Fix to fork choice
-> work in academic to formalise the fork choice rule
-> hard to test complex fork choice scenarios and formalize
Many: use the same techniques as multithreading interleaving to test fork choice?
—> https://github.com/ethereum/eth2.0-specs/pull/1495

BLS modifications
Will be merged soon
https://github.com/ethereum/EIPs/pull/2333
https://github.com/ethereum/EIPs/pull/2334
https://github.com/ethereum/EIPs/pull/2335

Focus: fork choice test vectors

Trinity Alex
- lots of work on 0.9.2, especially fork choice, libp2p

Danny: reminder, want feedback on naive aggregation

Lodestar: Cayman
Swapped BLS for Herumi WASM: 30x speedup
0.9.x fork choice WIP
DiscV5 WIP

Artemis: Ben
up-to-date with 0.9.2 including naive aggregation
Next focus: public testnet: syncing and deposits
Finish up discv5

Prysmatic: Terence
Single client testnet running smoothly
instances of finality regression
RPC and hardening of RPC server
Focus on 0.9.2

Lighthouse: Adrian
BLS and SSZ optimizations
—> not checking BLS when loading from database
user friendly logging
found bug when syncing lots of blocks from a single node

PSA: testnet down as one boot node was banned (?)

Harmony: Mikhail
updated to 0.9.2
- write up on slashing condition detection:
https://hackmd.io/@sYlY_LZpQIGgFmhdv6vV-A/By897a5sH

Parity: Wei-Tang
Update to the substrate engine
config issues
on spec 0.9.0, trying to join the public Prysmatic testnet
issue with BLS

Tomasz Stanczak (Formal Verification)
Done SSZ, Merleization, BLS
Continuing with fork choice and rest of the spec
target 0.9.2 in 1 month or 2
Will look at Disc v5 and libp2p

———————
Research updates

Vitalik: 

Status of cross-shard transaction: protocol or application layer?
i.e. Eth2 = fat protocol or just a data availability and computation layer

Write-up coming soon

Justin

SNARKS, did a ZK study club presentation on polynomial commitment schemes
https://www.youtube.com/watch?v=bz16BURH_u8&feature=youtu.be

Breakthrough in RSA MPC, successfully tested with 10k participants

Musab (Formal Verification)
Currently building abstraction of the beacon chain module
Goal: safety and liveness prover of the beacon chain specification


Matt garnett - Phase 2
Q on stateless protocols —> ethresearch
https://ethresear.ch/t/remaining-questions-on-state-providers-and-stateless-networks-in-eth2/6585
Next Phase 2 call in mid January

Link to networking minutes from yesterday
 •	https://hackmd.io/@benjaminion/BJ3YqrSTr
 •	https://gist.github.com/mratsim/fef2b0a7c5a335ac6bc61c01592b3fea

Next call in 2 weeks

Q Mikhail: How to reduce persistent database size

Lighthouse:
1. Hot/cold DB, once finalized: cold database, hot database not finalized
    Nimbus style approach: store state snapshots and replay the state transition from the cold DB
    Deduplication of state
    Frequency of snapshots is configurable, heuristics for now (time to fill 32GB on a hard-drive)

Proto: https://github.com/protolambda/eth-merkle-trees/blob/master/typing_partials.md
Jacek: Problem linked to syncing across the network if clients agree that would help.

…
(Internet dropped)
…

Hardening:
Take some single client testnet and scaling them with lots of validators
If you have the scripts —> will see how to scale

Closing remarks?

Next call: TBD (Christmas conflict)