Rust reimplementations

Posted Sep 6, 2024 10:31 UTC (Fri) by farnz (subscriber, #17727)
In reply to: Governance by atnot
Parent article: Whither the Apple AGX graphics driver?

I do, personally, see the value in a Rust reimplementation of a representative entity in each class of code the kernel supports; it's a great way to validate that the Rust bindings cover everything you need to implement something of representative complexity levels, and that they're performant and have no significant usability bugs.

So, GPIO device shows that GPIO device drivers can be written in Rust; Ethernet driver shows that netdevs can be written in Rust; NVMe driver shows that block device drivers using PCIe can be written in Rust; ext2 shows that you can write a filesystem in Rust.

And I expect to see similar as other areas get Rust support - there will often be a component rewritten in Rust to show that the bindings are good enough to support new work, so that you can separate out "this is not good because the Rust bindings are bad" from "this is not good because the design of the new work is not good in any language".