4 ms·

> users who will continue to have to deal with kernel panics and other memory issues. I have not seen many of those in my close to 32 years of involvement with

> users who will continue to have to deal with kernel panics and other memory issues. I have not seen many of those in my close to 32 years of involvement with Linux. In the beginning of the project they did occur 'every now and then' but once Linux got past v1.2 or thereabouts they're about as rare as hen's teeth. If Rust ever becomes widespread enough for kernel developers to pick it up with ease it will be a welcome addition but until such a time I prefer for the project to remain C-focused so as not to close out a large fraction of the potential developer base. If that means that some 'competing' kernel gets written in Rust and takes over from Linux that'd be fine by me as well, may the best [1] kernel win. [1] see https://en.wikipedia.org/wiki/Worse_is_better https://en.wikipedia.org/wiki/Worse_is_better

rowanG077 2y ago
I admit it's not often I see a direct panic. But I do see bugs caused by memory issues relatively often. In fact two weeks ago I made a patch for a memory issue in a wifi driver that caused it to drop connections. The long list of CVE is more direct proof though. By far most of them are memory safety issues. Suggesting making a new kernel is just total FUD. This is a multi decade project. It's like saying to a person who want some legislation in their country to change "why not make your own country, may the best country win". It's just so far removed from reality I'm very surprised I have seen this argument more then once.
- flohofwoe 2y ago
  > Suggesting making a new kernel is just total FUD. This is a multi decade project. Linux started as a hobby project of a single student and just a few years later it was already possible to build a full-blown UNIX desktop system with Linux as kernel. Why should that not be possible to replicate today? Especially with modern development tooling and a big and motivated language community?
  - pif 2y ago
    > Why should that not be possible to replicate today? What is missing is the user base! When Linus started his project, many people knew Unix in their professional life and were disappointed with what the market (M$) was proposing for the personal computer world. Today, no such multitude is waiting for the next Linux. Linux may not be perfect, but it's definitely good enough.
    - flohofwoe 2y ago
      Ideally such a Rust kernel project would implement a drop-in-replacement for the Linux kernel, so that Linux distros could select between the original Linux kernel and the Rust 'imposter'. That way the users would already be there. It would just be a detail like GNOME vs KDE, X11 vs Wayland, or glibc vs MUSL.
  - rowanG077 2y ago
    What was acceptable in 1995 is not what is acceptable today. Linux took literal decades + in the billions of dollars to build. It's ridiculous to suggest a rust project can just catch up to that with a few developers and no funding in any sane timespan. If you think it's reasonable to suggest "Just outcompete this heavily and continuously developed 34 year old project which has the backing of the largest technology firms on earth" I really don't know what to tell you. We live in different realities. That doesn't even begin to touch on the problem that linux came at just the right time.
    - flohofwoe 2y ago
      Sure it's foolish, but so was trying to 'outcompete' industry giants like IBM and Microsoft in the early 90s with a hobby kernel. That didn't stop a certain Finn from just writing a kernel anyway. If that project hadn't taken off, then maybe another one, or another one. Replacing any substantial parts of the Linux project with Rust also would be multi-decade effort, most likely more effort than forking off a separate evolutionary branch (which is probably the most realistic option). Such forks happened all the time in the history of UNIX.
      - rowanG077 2y ago
        It was considerably less foolish then, then now. It's not even close. The complexity of the operating system kernels has more then increased 50x over the last 34 years. It's not even remotely comparable. No one wants to replace substantial parts of the linux project with Rust. This is a straw man, and the same bad faith argument the set in their ways kernel greybeards are making. People just want to be able to write green field drivers in Rust.

rowanG077 2y ago

I admit it's not often I see a direct panic. But I do see bugs caused by memory issues relatively often. In fact two weeks ago I made a patch for a memory issue in a wifi driver that caused it to drop connections. The long list of CVE is more direct proof though. By far most of them are memory safety issues. Suggesting making a new kernel is just total FUD. This is a multi decade project. It's like saying to a person who want some legislation in their country to change "why not make your own country, may the best country win". It's just so far removed from reality I'm very surprised I have seen this argument more then once.

flohofwoe 2y ago
> Suggesting making a new kernel is just total FUD. This is a multi decade project. Linux started as a hobby project of a single student and just a few years later it was already possible to build a full-blown UNIX desktop system with Linux as kernel. Why should that not be possible to replicate today? Especially with modern development tooling and a big and motivated language community?
- pif 2y ago
  > Why should that not be possible to replicate today? What is missing is the user base! When Linus started his project, many people knew Unix in their professional life and were disappointed with what the market (M$) was proposing for the personal computer world. Today, no such multitude is waiting for the next Linux. Linux may not be perfect, but it's definitely good enough.
  - flohofwoe 2y ago
    Ideally such a Rust kernel project would implement a drop-in-replacement for the Linux kernel, so that Linux distros could select between the original Linux kernel and the Rust 'imposter'. That way the users would already be there. It would just be a detail like GNOME vs KDE, X11 vs Wayland, or glibc vs MUSL.
- rowanG077 2y ago
  What was acceptable in 1995 is not what is acceptable today. Linux took literal decades + in the billions of dollars to build. It's ridiculous to suggest a rust project can just catch up to that with a few developers and no funding in any sane timespan. If you think it's reasonable to suggest "Just outcompete this heavily and continuously developed 34 year old project which has the backing of the largest technology firms on earth" I really don't know what to tell you. We live in different realities. That doesn't even begin to touch on the problem that linux came at just the right time.
  - flohofwoe 2y ago
    Sure it's foolish, but so was trying to 'outcompete' industry giants like IBM and Microsoft in the early 90s with a hobby kernel. That didn't stop a certain Finn from just writing a kernel anyway. If that project hadn't taken off, then maybe another one, or another one. Replacing any substantial parts of the Linux project with Rust also would be multi-decade effort, most likely more effort than forking off a separate evolutionary branch (which is probably the most realistic option). Such forks happened all the time in the history of UNIX.
    - rowanG077 2y ago
      It was considerably less foolish then, then now. It's not even close. The complexity of the operating system kernels has more then increased 50x over the last 34 years. It's not even remotely comparable. No one wants to replace substantial parts of the linux project with Rust. This is a straw man, and the same bad faith argument the set in their ways kernel greybeards are making. People just want to be able to write green field drivers in Rust.