Redox OS 0.6, the micro-kernel based Rust-written operating system | Hacker News Reader

Redox OS 0.6, the micro-kernel based Rust-written operating system

phoronix.com

104 pointscassepipe8 comments

Original source has already been submitted as https://news.ycombinator.com/item?id=25533563

So does anyone have any experience with this? How does it compare to, say, Hurd?

Ran it in QEMU a long while ago. Back then it ran on a limited set if hardware, as well as on VMs, although I didn't manage to get that working (thus QEMU). Pretty ambitious. Already has small games and applications written for it. It's quite neat, but lacks the kind of unique selling point that would attract commercial interest. As far as I know there is no role it can fulfill better than Linux. There's the security you get from it being written in Rust, of course, but there has always been less interest in security than you'd think it deserves. I vaguely remember there being some talk about using it for DNS servers, but don't quote me on that.

Anyways, I'm not super up to date on the project, so take everything I'm saying with a pinch of salt.

It’s inspired by plan9 with the whole everything is a url bit. So that is something

In what world does plan9 inspire "everything is a url"?

Just in the sense that there's an "everything is a $foo" opinion, in the spirit of "everything is a file"?

Because at a glance, "everything is a url" seems antithetical to unix or plan9, where how a resource is accessed has deliberately been a transparent feature of the VFS switch, not something encoded into its path.

From the docs: https://doc.redox-os.org/book/ch04-10-everything-is-a-url.ht...

If the redox developers walked away from plan 9 thinking that "everything should be a URL", it seems to me they've completely missed the point.

plan9 is all about composable namespaces, and everything being a sort of polymorphic file path resolved by how the resolving process' namespaces have been arranged.

URLs pull what are really site-specific/usage-specific implementation details up into the path, where they really don't belong when viewed through a plan9-colored lens.

That justification seems rather weak by that some operations don't work on some files.

Of course not all operations work on all types of files. It feels as if they think that one type of “file”, as in the file that is a block of storage, is the “true file”, and the others are “fake files”, whereas the file that maps to a block of storage is no more or less an abstraction than a file that maps to a some segment of a process' memory or a hardware driver.

Would it be different of Unix files were instead named “vectors” as a more abstract name and that some vectors mape to what others might call a “file”, id est a block of storage.

Reply on news.ycombinator.com