[nequo]

The problem is that neither Linus nor the other prominent maintainers will live forever. C was the right choice in 1991 but today the landscape is different enough that its shortcomings, for the younger generations, are painful to ignore.

So saying yes to Rust, or to some other language that is not filled with foot guns and could also work in the kernel space, is not only a matter of kindness but a matter of long-term strategy for the kernel.

[snvzz]

>The problem is that neither Linus nor the other prominent maintainers will live forever.

I agree. But likewise, Linux's design is far from the state of the art. Maybe there's value in letting it be what it always has been. It is not realistic to try to migrate millions of LoCs written with poor structure, to properly structured Rust, with barely any support from the pre-existing developers.

At some time, it makes sense to move on to something better. This hypothetical system be written in a different language. It could be Rust, it could be C23, or something else entirely. But what it will definitely be is better structured. It would have clean, versioned APIs. And drivers will no doubt run in user space.

[iknowstuff]

I’ve become convinced that complex things in tech either ossify and get stuff built on top of them, or get Ship of Theseus’d. It’s extremely rare for an outright replacement to replace something overnight. Linux is not going to be suddenly replaced by something incompatible, but I’m concinved it will become more oxidized over time. If not under Linus, then under an initially fully compatible corporate fork which would ultimately become more popular. So there’s some pressure on Linus to move it along .

[snvzz]

This is typically the case.

But the argument being made is that Linux is particularly ill suited for an in-place conversion to a better design.