You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current logic is overly complicated, and burried in gpbftRunner deep enough that unit testing the raw heuristic itself in terms of conformity to basic design goals is cumbersome if not impossible.
The heuristic itself uses a number of parameters from manifest, and it's not entirely clear if we really need so many knobs to tune the instance start.
Refactor the heuristic into a stateless function that takes clear set of parameters.
Add unit tests to assert what the design goals are/should be.
Then simplify the logic and distil to 1 but no more than 2 parameters.
2025-05-02: This could result in better F3 UX for the ecosystem. What would help with prioritizing this is if we get clear ecosystem signal that reducing the lookback from 2.5 minute to 1 minute would make a big difference.
The challenge with this task is that it requires careful testing making it non-trivial. As a result, we'll defer this for the future when we have more engineering bandwidwth and energy for F3 improvements and ecosystem signal that further optimizations here will move the needle.
The current logic is overly complicated, and burried in
gpbftRunner
deep enough that unit testing the raw heuristic itself in terms of conformity to basic design goals is cumbersome if not impossible.The heuristic itself uses a number of parameters from manifest, and it's not entirely clear if we really need so many knobs to tune the instance start.
go-f3/host.go
Lines 389 to 460 in f77d509
The text was updated successfully, but these errors were encountered: