The 1970s called [LWN.net]

The 1970s called

Posted Oct 10, 2013 12:29 UTC (Thu) by HelloWorld (guest, #56129)
Parent article: Two LSS talks

They want their programming languages back.

Seriously, does anybody else think it's pathetic that we still have to deal with format string vulnerabilities and buffer overflows? It's time that the kernel community acknowledges the fact that ordinary programmers (i. e. other than gurus like djb) can't write secure code in C. Rust will probably be a viable candidate for parts of the kernel code once it reached its 1.0 release. Not to mention that probably most of the code would be way shorter and cleaner in Rust due to superior abstraction facilities.

to post comments

The 1970s called

Posted Oct 10, 2013 21:58 UTC (Thu) by dlang (guest, #313) [Link] (3 responses)

Rust is probably not even in the first hundred 'new secure languages that the kernel should be re-written in to solve all the evils of the world'

the answer now is the same as every other time.

If you think that writing a kernel in a different language is the right thing to do, go write it, and if you are correct and it is superior, you will displace Linux

Linus has said many times that he expects that eventually there will be some other kernel written that will replace Linux. It is possible that you are correct and Rust will be the language that such a kernel is written in.

but given the track record of such claims, I'm not going to worry about it.

The 1970s called

Posted Oct 11, 2013 10:51 UTC (Fri) by HelloWorld (guest, #56129) [Link] (2 responses)

> Rust is probably not even in the first hundred 'new secure languages that the kernel should be re-written in to solve all the evils of the world'
I never claimed that any language will solve all evils of the world, or that the kernel should be rewritten. But a programming language *can* solve silly things like buffer overflows (well, most of them) or format string vulnerabilities.
And let's not mention the fact that you didn't name even *one* argument why Rust wouldn't be a viable competitor. Especially when considering that I was talking about Rust 1.0 that isn't even released yet...

> If you think that writing a kernel in a different language is the right thing to do, go write it, and if you are correct and it is superior, you will displace Linux
So you're essentially saying that the success of a kernel is determined solely by the programming language it's implemented in? Because otherwise your point doesn't make any sense at all.

> Linus has said many times that he expects that eventually there will be some other kernel written that will replace Linux.
I don't think Linus is right about that. Linux is "too big to fail" by now, it's not going to be replaced. The best that one could hope for is an incremental solution to the C problem by allowing, say, some new drivers for non-critical hardware to be written in another language, and then gradually expand from there.

The 1970s called

Posted Oct 14, 2013 12:17 UTC (Mon) by rdc (guest, #87801) [Link] (1 responses)

>So you're essentially saying that the success of a kernel is determined solely by the programming language it's implemented in? Because otherwise your point doesn't make any sense at all.

No it't the exact arguement you make in the next paragraph, its too big to rewrite.

> I don't think Linus is right about that. Linux is "too big to fail" by now, it's not going to be replaced. The best that one could hope for is an incremental solution to the C problem by allowing, say, some new drivers for non-critical hardware to be written in another language, and then gradually expand from there.

The 1980s called they want there micro kernels back //tounge in check

The 1970s called

Posted Oct 14, 2013 22:59 UTC (Mon) by HelloWorld (guest, #56129) [Link]

> No it't the exact arguement you make in the next paragraph

He wrote this:
> If you think that writing a kernel in a different language is the right thing to do, go write it, and if you are correct and it is superior, you will displace Linux
The only ambiguity here is what "it" refers to. It could be either the newly written kernel or the Rust programming language. However, with the phrase "if you are correct" he clearly refers to some statement I made earlier, and I never said anything about some hypothetical new kernel, so he must be talking about the Rust language. So I'm sorry, but what he was saying just rubbish; he probably meant something else but it's hard to tell...

> The 1980s called they want there micro kernels back
The whole point of using a memory-safe language like Rust in the kernel is *not* having to bother with a Microkernel architecture.