100% yes. I really feel bad precisely for people in your situation. But, there’s a good reason why you see this async spam on HN:
> (I am not very familiar with Rust)
Once you get past the initial basics (which are amazing, with pattern matching, sum types, traits, RAII, even borrowing can be quite fun), you’ll want to do something “real world”, which typically involves some form of networked IO. That’s when the nightmare begins.
So there’s traditional threaded IO (mentioned in the post), which lets you defer the problem a bit (maybe you can stay on the main thread if eg you’re building a CLI). But every crate that does IO needs to pick a side (async or sync). And so the lib you want to use may require it, which means you need to use async too. There are two ways of doing the same thing - which are API incompatible - meaning if you start with sync and need async later - get ready for a refactor.
Now, you (and the crates you’re using) also have to pick a faction within async, ie which executor to use. I’ve been OOTL a while but I think it’s mostly settled on Tokio these days(?), which probably is for the best. There are even sub-choices about Send-ness (for single-vs-multithreaded executors) and such that also impact API-compatibility.
In either case, a helluva lot of people with simple use-cases absolutely need to worry about async. This is problematic for three main reasons: (1) these choices are front-loaded and hard to reverse and (2) async is more complex to use, debug and understand, and (3) in practice constrains the use of Rust best-practice features like static borrowing.
I don’t think anyone questions the strive for insane performance that asynchronous IO (io_uring, epoll etc) can unlock together with low-allocation runtimes. However, async was supposed to be an ergonomic way to deliver that, ideally without splitting the ecosystem into camps. Otherwise, perhaps just do manual event looping with custom state machines, which doesn’t need any special language features.