[lispisok]

Rust is cool but there is way too much dogma around its memory safety and static typing in general being a panacea. Most errors are not type errors. Two days after Cloudfare's little Rust hiccup that took the internet down for a day I saw people posting about Rust "if it compiles it runs".

[JuniperMesos]

I actually don't think this is true. I do think that most programming errors are type errors, in the broader sense of one part of a system making one set of assumptions about the properties of some data, that aren't shared by another part of the system; and that would've been caught automatically by sufficiently sophisticated static correctness checking. I do not think that Rust has a maximally sophisticated type system (nor is it trying to), and while this is reasonable for Rust as a project to decide, I do expect that there will be languages in the future that do more complex things with type systems that might supplant Rust in some domains.

The Cloudflare incident was caused by a confluence of factors, of which code written in Rust was only one. I actually think that Rust code worked reasonably well given the other parts of the system that failed - a developer used unwrap() to immediately crash instead of handling an error condition they thought would never happen; when that error condition did happen the Rust program crashed immediately exactly as expected; and if Cloudflare decided that they wanted to ban not handling an error like this in their codebase, it's a pretty easy thing to lint for with automatic tooling.

[pjmlp]

If it helps finally acknowledging basic stuff like bounds checking matters, great, this from a guy that rather use system languages with automatic resource management.

"A consequence of this principle is that every occurrence of every subscript of every subscripted variable was on every occasion checked at run time against both the upper and the lower declared bounds of the array. Many years later we asked our customers whether they wished us to provide an option to switch off these checks in the interests of efficiency on production runs. Unanimously, they urged us not to they already knew how frequently subscript errors occur on production runs where failure to detect them could be disastrous. I note with fear and horror that even in 1980 language designers and users have not learned this lesson. In any respectable branch of engineering, failure to observe such elementary precautions would have long been against the law."

-- C.A.R Hoare's "The 1980 ACM Turing Award Lecture"

From 1980!

C++26 will finally have hardening on the standard library, something that I could already enjoy in 1990's with Turbo Vision, OWL, MFC, VCL, but was too much to ask for on the standard library apparently, even if compilers kept having each own their approach.

It took governments and companies to start mapping CVEs to money spent fixing them, to finally acknowledge something had to change.

Meanwhile on C land, business as usual regarding Hoare's quote.

[msla]

It's interesting how it's Obviously Impossible to write OSes in garbage-collected languages, and this is proven by the fact successful OSes were written in garbage-collected languages back in the Stone Age, or 1980s, whichever. My current laptop has enough RAM to swallow the entire state of a Symbolics workstation (RAM and disk) without noticing, but it's obviously too wimpy to run an OS written in anything other than C.

(Nitpickers' Corner: "Successful" and "the most commercially successful" are, in fact, two different words. Gots all them different letters an' everything. Therefore, Genera not being as profitable as such Sophisticated Top-Of-The-Line Pieces of Professional-Grade Enterprise-Ready software as MS-DOS doesn't mean Genera wasn't successful.)

[josephg]

Yeah its funny what we can get away with using different design tradeoffs on modern computers.

I've been reading through the SeL4 source code lately. SeL4 isn't a multithreaded kernel. It supports SMP - so, it can use all the cores on your computer. But the kernel itself uses a big mutex. Complex syscalls can't run concurrently.

And you know what? I think its fine. A tiny microkernel like SeL4 offloads almost everything to separate processes anyway. The only things in the core kernel are the bootloader, scheduler and thread cap tables. Device drivers are multithreaded because they all run in separate processes.

Having the kernel itself effectively single threaded reduces a whole class of bugs and complexity, at a (hopefully) negligible performance cost. Its smart.

[pjmlp]

I noticed now that I didn't fully got your point, I do agree and the point with Android userspace also stands.

Better half victory than trying to change the whole world in one day.

[pjmlp]

My android with garbage collected userspace challenging Termux folks, works just fine....

[antonvs]

> Most errors are not type errors.

If you follow good strong typing principles, you can ensure that most errors are type errors. Yaron Minsky’s phrase, “Make illegal states unrepresentable”, captures this. But it doesn’t happen by accident just because you’re using a strongly typed language.

Also, if Cloudflare had run the standard Clippy linter on their Rust code, and taken the results seriously, it would have prevented the issue you referenced. Static checks don’t help if you ignore them.

[wolvesechoes]

This industry pretends to be driven by technical considerations, yet, with some exceptions, is mostly driven by fads, folk knowledge and aesthetic choices.

Folk knowledge may, and often is, grounded in reality and real experience, but let us not forget that most heated debates in programming of today are rooted mostly in tribal logic and fad chasing.

Static vs dynamic typing is a chief example. Empirical evidence that static typing makes some real difference in terms of bugs or safety is inconclusive at best. Yet it doesn't prevent some people from literally shaming those that prefer dynamic languages. Same with OOP - for years it was everywhere, now you may have an impression that it is a sin to ever use it. But now, as much as back then, there is no evidence to support claim that using or not using OOP is "one true way".

Now, memory safety is a real concern, and we can confidently say that we have found ways (exemplified in Rust) to prevent whole class of issues, but suddenly we are in the situation that every single bit of code out there is supposed to put memory safety as a chief concern, no matter if we are talking about some high perf web server, missile control logic, simple script solving Lotka-Volterra equations or simple calculator app.

[pdimitar]

> suddenly we are in the situation that every single bit of code out there is supposed to put memory safety as a chief concern

I don't know why you are clumping together fanatics with the pragmatic folk who finally pioneered, in practice, the idea that memory safety is objectively valuable. Different groups.

> Empirical evidence that static typing makes some real difference in terms of bugs or safety is inconclusive at best

Literally 99.9999% of everything out there is inconclusive. You'll only extremely rarely find something that is objectively 100% true. So this is a disingenuous argument on your part, seemingly used to undermine something that you might not like. As such, it's an unfair discussion technique.

I worked nearly 9 years with Java a lifetime ago. I could not ever see any actual evidence that OOP and global mutability have ever improved any project.

See? This can go both ways.

> This industry pretends to be driven by technical considerations, yet, with some exceptions, is mostly driven by fads, folk knowledge and aesthetic choices.

Always has been, and it goes for almost all industries except those that are heavily regulated.

[josephg]

Rust doesn't eliminate all bugs. But anecdotally, by the time the type checker and borrow checker have humbled me, my programs really do often work the first time I run them. Its quite remarkable.

This isn't a special thing about rust. All languages are on a spectrum of "detect all bugs statically" to "detect all bugs dynamically". Rust programs run correctly "first time" more than javascript, more than typescript. But still less than haskell.

You can still write bugs in rust, obviously. I've written plenty. As you say, so has cloudflare. But strong typing does find a lot of bugs in practice.

[pdimitar]

Yeah, apparently a certain group on HN will forever repeat this.

As another poster said, no checks in the world will help if programmers ignored them. It's a one-liner lint in Rust. Somebody at Cloudflare figured it's too inconvenient to do.

[IshKebab]

> Most errors are not type errors.

With a sufficiently strong type system all errors are type errors! Rust doesn't have that of course, but it does have quite a strong type system so this is a very bold assertion with no evidence.

Rust does have an "if it compiles it works" feel. Nobody means that literally (this should be really obvious). They just mean that once you get it to compile the chance that it works first time is quite high (like 20% maybe?) compared to most other languages where it's more like 1%.

[dcminter]

I don't think your comment deserves the downvotes (upvoted to compensate) but I do think that it's questionable if "Most errors are not type errors" is true.

Rust's culture of pushing things into type checking does eliminate a huge swathe of bugs and I wouldn't be surprised if it was the majority.

The hurdle of negotiating translation between filesystem strings and unicode strings strikes me as a good example of a place where most languages don't protect you from bugs and a strongly typed one does. The downside, of course, is that you have to handle these cases (even if it's to explicitly say "I don't care").

I still create dumbass bugs in Rust, but they are usually simple logical errors that are pretty obvious when debugging.