massa-bot
commented
2 years ago In GitLab by @damip
If we want to get rid of the controller/worker pattern, we need first to define a viable alternative that:
- still provides a stable interface to every module so that the development of different modules can be effectively isolated
- is demonstrably faster without adding extra complexity
- behaves properly in case of congestion (backpressure when needed, timed message drops when relevant, priorization etc...)
As for mixing multiple languages, the true benefits, as well as the extra interoperability complexity need to be evaluated thoroughly before proceeding.
In GitLab by @yvan-sraka
Rust is a great programming language, but it's hard to find developers/contributors confident with it...
As we will as ultimate part of refactoring described in https://github.com/massalabs/massa/issues/1232 rewrite
massa-nodeand the core business logic of the app and get rid of the current custom controller/worker fuzzy pattern and its performance cost.Could we think of using a more high-level garbage-collected language like ECMAScript/Python/Haskell etc... and keep Rust where it only really makes sense, like in
communication/consensusmodules.The purpose would be to stay as KISS as possible, implement the modularity with feature-based building and the separation of concerns by sticking to small internal APIs (what FFI will force us to do).