News for hackers. Stuff that matters.
Re:Ask HN: Can the same individual accomplish more with prog... by turbinerneiter on Thursday April 14, 2022 @10:45AM (#31027690)
What if you proof the correctness of a compiler? Do you then transiently get all the achievements of all safety critical software that was then programmed with it?

The value of the proof comes from the things it enables. I guess the Rust memory safety guarantees are based on some proofs?

Re:Ask HN: Can the same individual accomplish more with prog... by crocal on Thursday April 14, 2022 @11:08AM (#31027941)
FYI, this has been done: https://compcert.org/
Re:Ask HN: Can the same individual accomplish more with prog... by gpm on Thursday April 14, 2022 @11:16AM (#31028057)
> Do you then transiently get all the achievements of all safety critical software that was then programmed with it?

It's possible to write incorrect code and compile it with a correct compiler, so I don't think so. It's more like a citation that proven correct programs can use to say "I wrote this program correctly (see my own proof), and I compiled it correctly (see citation), thus I have a correct executable file".

> I guess the Rust memory safety guarantees are based on some proofs?

People have retroactively come up with formal proofs for much of it, but I believe the initial design was just based on intuition not a proof. For unsafe code some of the exact details of the memory model are still being worked out too (and thus definitely aren't proven correct).