Feds want devs to stop coding 'unforgivable' buffer overflow | Hacker News Reader

Feds want devs to stop coding 'unforgivable' buffer overflow

theregister.com

15 pointschrisjj9 comments

This is why I think it's so crucial to move Rust into the Linux kernel. It's not about new-school-cool vs old heads. It's not about getting off someone's lawn, it's about solving a really basic but important problem. Sure we can just get not into accidents and be rid of seatbelts, airbags, crumple zones, and ABS, but it's not a real world.

Any solution that just adds to C, will make it a new C, that will take years to prove itself and still require relearning and push-back.

I'd love a real world suggestion that does it better. The best I can imagine is annotations on C that are checkable formal methods... but that just sounds like reinventing the borrow-checker.

I'd just be happy if organizations stopped doing business with software vendors that still mandate disabling DEP, ASLR, and the dreaded "Please disable your antivirus before installing".

You'd be able to code memory bugs into an OS kernel written in Rust. It's the same problem as a JavaScript JIT - the fundamental operations it has to do are unsafe, so it doesn't matter what language it's in.

Among other things, you need more secure hardware to make progress.

Able to, sure. As probable? Definitely not. Practically all kernel exploits are accidents in buffer handling.

(Getting machine-readable syscall descriptions that describe the rules for pointers to buffers and their lengths in ways that can get safe code generated from them would be a huge win, on top of "just use Rust".)

Source: https://www.cisa.gov/resources-tools/resources/secure-design... (3 points) https://news.ycombinator.com/item?id=43033450

The headline sounds like there's a specific kind of bug called an unforgivable buffer overflow - was hoping to find out what it was.

So much a joke, the "buffer overflow" that is a good perfect "straw man", when most current world issues are: intentional defects like backdoors, business logic failures, ...

A lot of coding needs to run on hardware for which there simply isn't a memory-safe alternative.

Can the mainframe over at the IRS swap-in Rust for all that legacy unsafe COBOL code?

Some memory-safe languages are capable of code generation. I think the ones that aren't should be.

Admittedly, I don't know any compiler backends that generate COBOL.

Reply on news.ycombinator.com