Conversation

Notices

1. Embed this notice
   Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:50 JST Drew DeVault

   Finding the kernels of truth among the bullshit and doing my best to respond in good faith: regarding the scope of a Linux-compatible kernel written from scratch... yes, it's large. But there are exactly zero real-world programs that use that utilize the entire ABI surface of the Linux kernel. You implement the parts that you need to run the software you want and it's useful for that purpose right away. e.g. a kernel that can run nginx is useful even if it doesn't have GPU drivers, for instance.

   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:48 JST Drew DeVault

     in reply to

     And honestly the idea that we *have* to do it in Linux to get anything useful done at a practical scale for kernels is a really fatalistic assumption that spits in the face of just about any serious osdev project and condemns us to live with the design issues of Linux forever -- and believe me Linux has a lot of design issues. Even making a new kernel based on its general design gives us an opportunity to make a ton of sorely needed improvements.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:48 JST Drew DeVault

     in reply to

     I very much dislike the assumption that any serious greenfield osdev work is a lost cause. It's not.

   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:48 JST Drew DeVault

     in reply to

     And for the record, there's nothing wrong with the political game being a necessary challenge to overcome. The toxicity is wrong, but the politics is not necessarily bad on the face of it. The Rust community is not going to get anywhere with the presumption that they're always right and everyone needs to toe the party line about memory safety yesterday or else. The Rust community tends to be bad at politics and makes that everyone else's problem.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:49 JST Drew DeVault

     in reply to

     Yes, a better solution than leaving the Linux community to make your own way in the face of toxicity would be to address that toxicity head-on. And believe me I condemn in no uncertain terms the toxic environment of the Linux kernel. But a complete answer to this problem would be distracting from the goals of Rust-for-Linux, which is to say getting Rust in Linux, and if people are already burnt out I'd rather care for their needs first than put them aside to address the causes right away.

   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:49 JST Drew DeVault

     in reply to

     And it's entirely true that while the political environment of Linux is, on the whole, pretty toxic, that there are legitimate political challenges to overcome with respect to Rust-for-Linux and even absent the toxicity it's going to call for a lot of work which isn't material to the goals of getting Rust in kernel space.

   • Embed this notice
     Drew DeVault (drewdevault@fosstodon.org)'s status on Saturday, 31-Aug-2024 19:25:49 JST Drew DeVault

     in reply to

     Yes, the road I propose is not necessarily the fastest way to getting secure kernels in the hands of users, and it would be faster and easier if we could get Rust in Linux without having to play the political game. But given that the political game isn't going away, we have to consider the other road as an *option*.

   • Embed this notice
     Mark Stosberg (markstos@urbanists.social)'s status on Saturday, 31-Aug-2024 19:45:12 JST Mark Stosberg

     in reply to

     @drewdevault And it's not like it's a binary choice either, like pointing the finger at the single worst cause of climate change and ignoring all others.

     There are a lot of developers in the world. Enough to work on Rust-in-Linux and a Rust-only kernel as well.

     A small, Rust-only kernel could try out some ideas that might actually spend up Rust-in-Linux development.

     Haelwenn /элвэн/ :triskell: likes this.