Ask HN: Any new company-backed operating systems?
I love low level kernel/OS stuff but loathe the operating systems we have today. Windows, Linux, and BSD are heavily rooted in principles we designed 30+ years ago. These weren't designed with the requirements we have today - like configuration management for servers, live security upgrades, mitigating CPU flaws to name a few. POSIX has served us well but we could probably come up with something much better if we tried.
I want to see something radically different. Whether it's for desktops or microcontrollers I don't mind.
I'm aware of a few hobby projects such as SerenityOS which are cool but still very 'inside the box' in terms of design and don't really have any backing.
7 comments
[ 4.0 ms ] story [ 25.4 ms ] threadReally?
> like configuration management for servers
What do you mean by this? Configuration is standardized on Linux and BSD. With tools like Docker and Nix, it's arguably easier than ever to set-and-forget configuations.
> live security upgrades
Livepatching has been a thing on Linux for close to a decade now.
> mitigating CPU flaws
Microcode updates are a common occurrence, I'm not exactly sure how you'd expect people to mitigate silicon-level exploits with software like that otherwise though.
On a separate note, what you might be looking for is Redox OS. It's not corporately backed, but it does solve real, tangible problems with modern OSes like memory/thread safety.
Configuration management is not a solved problem. I have extensive experience with Puppet and Ansible and used to manage a fleet of thousands. and I can say frankly they suck.
>Livepatching has been a thing on Linux for close to a decade now
Yet nobody does it because the option isn't there. To get livepatching on Ubuntu you've got to go via Canonical and pay for their livepatching suite.
>mitigating CPU flaws
What I actually meant by this is that the design of the kernels we have rely too much on processes theoretically being isolated, even though we now know they aren't. Spectre, Meltdown, Rowhammer, and so on completely break the foundations of every kernel. How can we trust that our browser isn't going to read our disk encryption password from memory? We can't. We built everything with the assumption that can't happen.
[0] https://www.haiku-os.org/
BEAM (Erlang, Elixir) on a RTOS, targetted at IoT.