Andreas Fackler

Note that even with _g = f_, this would use less bandwidth than the current approach, and you'd never have the added waiting period unless the sender is faulty, in...

In the discussions with the team, I realized that there are several (partly independent) optimization ideas floating around. I'll try to separate them as much as possible. First of all,...

Sounds good to me! :+1: The challenge will be to add all those optimizations without making the code unreasonably complex. Please also double-check the proof and update the module's documentation....

You mean the `self.count_echos(&hash) < self.netinfo.num_correct()` part? We mustn't send a `Ready` unless we have collected _N - f_ `Echo`s, so we need to return a `Step` without a `Ready`...

I fixed what I think was an error in my comment above: We should only call _N - 2 f + g_ nodes "to your left" instead of _N -...

Note that by "_N - f_ `Echos` plus `EchoHashes` in total" above I mean that _N - f_ nodes have sent us an `Echo` or `EchoHash` or both—sorry for the...

@pawanjay176: Please claim your reward from gitcoinbot.

Are you running `cargo run --example simulation --release` from inside the `hbbft` directory, i.e. the directory that contains the [`Cargo.toml`](https://github.com/poanetwork/hbbft/blob/master/Cargo.toml)?

Migrating to logging would also have the advantage of being able to see the production code's log messages in the right place in the test's log.

Let's also `#[cfg(test)] panic!()` if a `Step` with a nonempty fault log is dropped, to make sure our code never loses a fault entry. The user, however, should probably not...