Hacker News Reader

I mean lunatic runs within Wasm, which isn't the end of the world, but we compile down to native. Lunatic generally seems to want to be the end-all of concurrency but we fit more into an existing Tokio-based environment.

As far as deployment, do you mean like how we came to build this? Or like how you'd actually deploy it...? Because that's really up to the program being built imo

View on HN

> As far as deployment, do you mean like how we came to build this? Or like how you'd actually deploy it...? Because that's really up to the program being built imo

I mean that deployment of Erlang code on a BEAM cluster is pretty simple, iirc. I wonder if you have a solution that handles that part as conveniently as Erlang/Elixir/BEAM.

Ah yeah we don't have hot-updates like Erlang does (we may never, it depends on how Rust would let us do it). That being said we aren't running on a runtime, so Rust code is native-compiled. You could take down and upgrade a node in a cluster as long as the network protocol doesn't change (or is at least backwards compat), but you wouldn't be able to like upgrade a single actor on a node.

I suspect that it would be possible to somehow implement hot-updates with native, if you're willing to use some container as your "runtime". I haven't investigated this seriously, though, so I may be wrong.

Reply on news.ycombinator.com