What we heard about Rust's challenges, and how we can address them | Rust Blog (blog.rust-lang.org)

submitted 4 days ago by SorteKanin@feddit.dk to c/rust@programming.dev

13 comments fedilink hide all child comments

top 13 comments

sorted by: hot top controversial new old

[-] onlinepersona@programming.dev 3 points 3 days ago

My biggest issues with rust are

borrow checker barely ever uses the term "borrowed" and I have to relesen it every time I pick up rust
async is absurdly complicated, there are so many gotchas. Streams are actual hell
crates are github only and have no namespacing. Java has the best solution IMO with their reverse domains for example com.github.myuser.myproject.somepackage

It's good to know I'm not alone. But I'll never choose C/C++ over Rust. If I need speed and typing, it's Rust, no questions asked. For everything else, there's Python. I just wish Rust has better GUI frameworks...

[-] Miaou@jlai.lu 1 points 2 days ago

Domain are also susceptible to being hijacked, unfortunately.

Also what do you mean by github only?

[-] onlinepersona@programming.dev 1 points 1 day ago

Domains are, sure but the space is much much larger. With namespaces the space is for example namespace x crate instead of just crate. With domains it's impossibly large. And if you introduce some kind verification with domains e.g putting something in the DNS TEXT entry, or calling a well-known path, or or or, you basically can't squat anything well-known anymore.

As for github, you must have a github account to create a crate on crates.io. to my knowledge there is no alternative crates registry. Point me to one if you do, because I'd be grateful and willing to publish my crates there.

[-] robinm@fosstodon.org 1 points 1 day ago

@onlinepersona @Miaou for github they said for years that they would welcome patches but so far nobody came, nor payed someone to do it for them.

[-] onlinepersona@programming.dev 1 points 22 hours ago

I find it ridiculous that they started with github in the first place, but who am I to judge. It is ironic that in order to contribute a non-github solution, you still need to create a github account to make a PR.

[-] robinm@fosstodon.org 1 points 16 hours ago

@onlinepersona I'm quite certain that if you have a working solution, visible somewhere on internet, you can ping them on zulip/discourse to ask someone to pull your code.

[-] onlinepersona@programming.dev 1 points 15 hours ago

Oh really? 🤔 Every project on Github that I've tried to contribute to from outside has been resistant to it. But maybe I could give cargo.io a try. Zulip you say? Interesting choice... requires creating yet another account since it doesn't seem to be federated...

[-] robinm@fosstodon.org 1 points 13 hours ago

@onlinepersona Given that this subject was already brought up, and regular people from the project said that their is interest for removing the dependency to github (but not leave it, Microsoft is one of the top contributor), I assume that some people would be willing to help the integration a working prototype.

In case you need it:
https://rust-lang.zulipchat.com/

[-] TehPers@beehaw.org 2 points 3 days ago

Your second point has been a focus area for them for years, and we've seen it improve with async fn in trait, async closures, etc. Hopefully we see that continue to improve over time.

The third point has a solution in the works already (at least for crates.io). The accepted (I believe) proposal is that crates themselves act as "namespaces" and you can publish crates like bevy::render or hyper::utils that get loaded under some parent module by the compiler. Publishing crates with names containing :: would be constrained to those who maintain the parent crate (so bevy or hyper in that case).

[-] TheAgeOfSuperboredom@lemmy.ca 10 points 4 days ago

This is cool and it's nice to see some acknowledgement and self reflection.

I actually like async Rust and I feel that they made the right decisions there. It's amazing that I can use similar constructs for a web server as well as a small embedded board. But the low level Futures stuff does get hairy. I still don't really get the relationship between Pin and Unpin, and any time I have to write a Tower Service by hand it feels arcane. Fortunately I rarely have to go down to that level.

The tacit knowledge is interesting. Is there an ecosystem that doesn't have that to some degree? Node and Python feel even worse to me, though I admit I don't work with those as much. Still, if I ever have to look for a Node library, I get worried that I might be inadvertently installing some malware. That said, I am about to sit down with one of my juniors to start on a new Rust project, mainly because I want to help setup the initial crates (tokio, tracing, anyhow, clap, etc...). Efforts like blessed.rs are a good start. Or maybe there are some good "starter kits" using cargo-generate?

[-] silly_goose@lemmy.today 6 points 4 days ago* (last edited 4 days ago)

Feels empty and generic like it was written by an LLM. Why did you even ask developers to fill out a survey? You could have asked chatgpt what users thought about Rust too and saved everyone a lot of time.