Hacker News Reader

Comment by pornel | Hacker News Reader

There are projects like Chromium that work extra hard, and yet they keep having memory safety bugs. Do they need to work harder? Get better programmers? Take security more seriously?

Check out slides from this talk: https://www.usenix.org/conference/enigma2021/presentation/pa...

At some point you have to admit that the language is the problem. Eliminating entire classes of bugs by construction is much more effective than testing for them.

For example, sum types prevent most cases of unexpected nulls and uses of returned values without error checking.

Languages can't eliminate all bugs, but the more they can prevent, the more you can focus on testing the rest.

View on HN

I agree that trying to eliminate a class of bugs through language design is a good thing to try, absolutely.

I just mean to say that the idea that you inherently make your software more reliable by "rewriting in Rust" is false. I don't contend that it could have benifits but I think to some extent you swap one set of problems for another and maybe in a language you don't understand as well as the one you're coming from.

I have first-hand experience that yes, Rust does make software more reliable. It noticeably changes the class of bugs you deal with. The low-level bugs like memory corruption or dangling pointers disappear. You're left dealing with mainly high-level issues, like features not implemented up to the spec, and deficiencies in the program architecture or algorithms (which you still have in any language).

It's not some Ying/Yang balance of force or a contract with the devil where getting rid of sneaky Undefined Behaviors must be paid for with another evil.

In C I've struggled with multi-threaded software, and even relatively tame solutions like OpenMP ended biting me. OTOH I haven't had to debug a data race in Rust yet (I've caused a bunch of deadlocks, but these are relatively easy). In C, despite my best efforts, I've occasionally had buffer overflows, leaks, and UAFs. In Rust I've had these only when interacting with C libraries via FFI.

You could probably say that my C bugs were my fault, and I'm just not a good C programmer. Well, I haven't got any smarter, but I'm a good Rust programmer.

Just stopping adding features limitlessly would go a long way

The rate of new big features in Rust is slowing down quickly. The only big upcoming one is Generic associated Types. That’s pretty good for 6-yr old language

Reply on news.ycombinator.com