Hacker News Reader

Comment by PeterWhittaker | Hacker News Reader

Sure, we can do those things in C, but they are easier and faster to do in Rust, and safer, largely because of Rust's enums (algebraic data types, not as simplistic as C's) and traits, and because of crates: a lot can be done in Rust by gluing together a few well maintained crates.

The only language that comes close, AFAICT, is python.

Add the fact that Rust's memory safety brings thread safety for free (once you've grokked Rust's memory management model) and yiu get quickly written type safe memory safe thread safe high perfomance code.

View on HN

There's aggressive language advocacy, and then there's passing into parody. I think you're getting perilously close to the latter here.

Anyone have any idea what makes Rust evangelists / apologists so religious about it?

It's not like it's the only high-performance memory safe language. It wasn't the first. It's not the highest performance. What am I missing?

1. It's highly opionated so there's a cult-like element of an ingroup/outgroup, e.g. "we don't have OOP, OOP is bad! if you use it, you are an outsider and we can do things better in Rustland here!"

2. It's posed as an alternative to C/C++ which have many historical problems. It's very easy to construct a strawman to show Rust is so much better than those ancient languages. ("here's this trivial buffer overflow bug Rust could have prevented!")

3. It's low-level enough that people feel 1337 for using it but not low-level enough to actually require low-level skills to use it. This is why you can see how in many cases, the "solution" to the borrow checker is just cloning everything - because they have zero idea about how to make performant programs.

4. It doesn't have many established libraries/frameworks compared to other languages, so "I wrote X but Rust" is a viable thing. In other languages, this wouldn't stand because there are already established libraries so interest in new ones is not as high + there are certain quality standards you have to meet.

5. It has certain high-profile features which are easy to explain why they are better - option types and slices for example. These Rustaceans are usually not exposed to functional programming before - so they start to embrace simple forms of functional programming. It makes Rust programs look fairly different compared to similar languages, which is another differentiator. It's easy to showcase the advantages of these features in code snippets, which makes it easy to evangelise.

We don't have a lot of place for religion in my industry, and safe, performant low level code is everything. Writing and debugging such code in Rust is just easier and faster than in C.

Two years ago I was casually learning Rust by night while by day I was extending a scheduler I had previously written in C. I kept wishing I could switch, because I kept having to handcraft in C things I would have gotten for nearly free in Rust.

Now, we're designing the Rust replacement for that C program, and it is taking all my will power not to start coding it: that job falls to someone else, and I am wistful about it. Not because I am going to miss the C code, but because it would just be so much fun to do in Rust.

Not sure you are trolling us or not, but there is a bit of a contradiction between "here is a huge learning curve to Rust, one that I am still climbing," in the comment above and "is just easier and faster than in C".

It is certainly true that you have a lot of things you can easily have at your fingertips with cargo in Rust, which you might need to first identify or even handcraft (if you have no existing framework to rely on) in C. On the other hand, I am not sure this is a good thing. I am a bit skeptical about the cargo / NPM style of package management.

> but there is a bit of a contradiction between "here is a huge learning curve to Rust, one that I am still climbing," in the comment above and "is just easier and faster than in C

I see no contradiction when you realize that no one has ever been proficient enough in C to write code that is free from UB… The problem with C is that cannot realistically be written by humans, so pretty much any language (except C++ …) is going to be “easier than C”.

Wild claims. So many huge projects have been successfully written in C by humans. There are patterns to solve all the built in solutions provided by Rust, how do you imagine Rust got them in the first place?

One tiny example of many here is the Linux kernel.

I'd say we didn't have many options back then except for C and no more than 4-5 others. So we did the things you quoted _despite_ C's drawbacks, not because of them.

I can very easily imagine doing the same things with Rust, with small amount of exceptions (that the core team is working on even now).

And yes Rust learned from the mistakes of the languages before it. Is that a bad thing?

Java and C# runtimes are getting memory safety at the cost of RAM usage in runtime, to lesser extent CPU overhead in runtime. Rust is getting memory safety at the cost of cognitive load of developers. Due to this tradeoff, by design Rust is extremely hard to use. Some people still learn the language despite the learning curve, then fall to sunk cost fallacy.

Another thing, it’s obvious people who designed Rust liked functional programming, so Rust attracts FP programmers. In the online communities, functional programmers are known for over-the-top evangelism for decades now.

For a brief history of the origins of rust evangelism, I recommend you consult this article: https://en.wikipedia.org/wiki/Cargo_cult

Just when the rust hype train was getting annoying, the anti-rust hysteria becomes even more so.

Besides, cargo cult is simply the wrong concept to point out here.

I think it may just be an attempt on a pun on the name of rust's package manager, 'cargo'.

indeed

While I agree that Rust advocates can be annoying at times, your last paragraph confuses me: which language do you have in mind that is both memory-safe and higher performance than Rust?

Rust is legitimately special: being memory safe without any kind of garbage collection (I include ref-counting here). That's what enables its deployment in stuff like QEMU or the Linux kernel.

Also, rust is legitimately a nice language. I get it that people get annoyed over the hype, but then they start acting like people like it just because they're morons who can't think for themselves.

Agreed. I feel gaslit by the gap here. I too find the hypesters annoying, but it's fun to hack on Rust code usually. Usually...

There is the theory that Rust people constantly have to reassure themselves that the language is great and the future, so that can bear the suffering from using it.

If you say so. Considering that you can define the basics of a memory and thread safety for free JSON parser in a single enum in Rust, and derive a lot of utility functions, where C would take a lot more code, and the code would be riskier, well, I'm just going to keep enjoying working in the language. It has a lovely balance of power and conciseness.

There is a huge learning curve to Rust, one that I am still climbing, but I haven't been as excited learning and working with a new language since 1987 when I learned C afters years of working with Pascal, Fortran, and a number of procedural job control systems.

In the years since, I've written large complex programs and systems in C, JS, bash, and PowrShell, and so far Rust feels like my new forever home, my defauly go to.

I'm glad of this, because my industry is heading to memory-safe, compiled languages, and the others I've experienced don't work for me. It's aesthetic, mostly, I guess, but Rust just makes sense.

And there you go, you've passed the line. In response to a suggestion that maybe you could tone it down a bit, you just repeat the party line again, because a slight opening has been spotted and by golly you're not going to let that pass!

You're hurting more than you're helping at this point.

And, just in case you're having trouble following along, as the evidence rather suggests that you are, note that replying with yet another paeon to Rust will be really missing the point.

Well, it's nice on this side of the line.

There is no party line, and there is no tone. There are over 30 years of experience with C, not all of it pleasant, though most of it productive, and, comparatively speaking, amazing recent experiences with Rust. I do have to shift my thinking from how I would solve problems in C, because the memory management paradigm is so different, but, thanks in part to deep dives into scoping in other languages in pursuit of pernicious bugs, once I make that shift, problems are easier to solve in Rust: The language and the compiler, hand in glove, do a lot of the work for you.

You can carp all you like, and that is all it is, carping, not criticism, we'll just happily work along getting a lot done in less time.

If you want to advocate for things, you should learn more about social cues. Blatently missing them this badly hurts your advocacy and pushes people away. This is not unique to the Rust community, and it isn't universal to the Rust community by any means, I know you're a minority, but you people who just beat on people like this are hurting far more than you are helping.

To me it seems like you are passing a different line, into downright anti-rust hysteria, as if it is some sort of taboo to talk positively about it, or argue about its merits.

Your earlier comment, that his advocacy passes into parody, also seems silly to me. Perhaps you don't believe Rust's merits, and that's fine. Perhaps you don't believe the grandparent, and that's fine too. But either way, his argument was fair.

Edit: I later realized that my comment is mostly just an attack that doesn't really add anything to this thread. I apologize for that.

To add something to it, I'd just like to second that Rust is indeed very easy to write after learning to deal with the borrow checker. Much, much easier than C, and I say that with 15 years of writing various applications in C (with all of the autotools rituals, pthreads, Glib gobject nonsense, GTK/GDK/cairo, etc).

I had a fun little code battle with a colleague, and here's a comparison of Rust [1] vs Python [2]. Although Rust is certainly nowhere near as readable as Python, it was just as easy to write, due to the type-system saving on debugging/learning time.

[1] https://gist.github.com/andreivasiliu/a4a2a6007d1659799b1f80...

[2] https://gist.github.com/lemon24/252ee6aa122a2b38395b096f96cb...

The argument is fair... in a different context.

Monomaniacally monopolizing the topic is anti-social behavior.

I mention this because apparently you could use a bit of clue on that as well. You're actively pushing people away from Rust at this point when you resist the clue stick this hard.

I see. Yes, I will admit that I see the context as being perfectly fine.

It started as a misguided comment that takes a tongue-in-cheek statement but believes it is unironic, followed by a comment saying it was tongue-in-cheek, there is some truth to it, but that truth is hard to believe, and then finally PeterWhittaker's comment, which argues that it should not be that hard to believe after all.

All in a thread that is mostly about Rust.

> Sure, we can do those things in C, but they are easier and faster to do in Rust

...I don't know what to say to this except "no?".

I'd have to question what priors someone making that claim is bringing and where they got them from.

> The only language that comes close, AFAICT, is python.

With well over a decade in Python experience and having regularly defended it against many claims of unsuitability for many tasks, I have to call BS on this and say that no, that is not even remotely correct. Of all the possible use cases where Python is an easily dismissed but very defensible choice this is NOT one of them.

> easier and faster to do in Rust, and safer, largely because of Rust's enums (algebraic data types, not as simplistic as C's)

Python enums are largely equivalent to C enums, except you can use string constants as values. Otherwise there's little ergonomic difference.

> and traits

Python Protocols (less anything that came before) aren't even remotely comparable to Rust traits. And neither affords any major advantage to red team style development.

And while memory safety is great for mission critical pieces of a larger system, and language ergonomics are essential for readable and maintainable code, I don't know what either of these possibly has to do with writing exploits.

> shellcode, a web server, and a phishing page in

Python would at least be a better choice than Rust for all of these, but so would have been Perl 20 years ago.

I wonder, what is it that causes the hackernews users' inability to comprehend sarcasm?

Well, it not being sarcasm at all might be a start, in this particular case.

oh my god I just looked at the other replies and it's real

I'm not sure what to say

Apologies, my reference to python was unclear: I was referring to the python module system and the strength and depth of community contributions, not to the syntax and semantics of the language.

In both cases, the communities do so much good work that often someone else has at least a partial solution to problème-du-jour, so it can be relatively easy to get at least a prototype or PoC, and learn from it, and sometimes one gets a complete solution to at least part of the problem.

Reply on news.ycombinator.com