Question: under what circumstances would I ever want to use this OS, over Linux or MacOS or Windows?
I was hoping Genode would get there sooner, and GNU Hurd should have gotten there years ago.
Capabilities are the computing version of circuit breakers. They make it possible to run code without having to trust it. You can't do that with Linux, Windows, MacOS, et al. MULTICS could do it, but it was deemed too complex, at the time.
A replacement with a more serious architecture based on seL4 is in the works[0].
Even a microkernel has three things that form part of the TCB: the part that interfaces with the even-lower-level stuff (interrupt and exception handlers, paging, memory model, etc), the interface it provides to its clients (syscalls and whatever it provides via syscalls), and the code that glues the first two parts together and implements the microkernel’s internal logic.
The latter two, in seL4, are straightforward and formally verified. The former, on x86, not so much.
AIUI, the situation on ARM is likely better.
I would not trust any system on x86.
>AIUI, the situation on ARM is likely better.
Somewhat, but still not good.
RISC-V is where it's at, when it comes to seL4.
(seL4 participates in RISC-V)
"Sculpt is an open-source general-purpose OS. It combines Genode's microkernel architecture, capability-based security, sandboxed device drivers, and virtual machines in a novel operating system for commodity PC hardware and the PinePhone. Sculpt is used as day-to-day OS by the Genode developers."
I'm hoping that Sculpt 24.04 does the job for me. If I can make sense of it enough to get Free Pascal running on it, or even just a C compiler, and "Hello, World" in a CLI, we're off to the races.
I suppose that's part of the price for high-security.
I don't have an x86 machine, and running it under a VM is too slow.
It’s like asking under what scenarios would you want to drive an ATV through New York. Technically possible, but that’s just not what it was made to do.
Linux is truck/SUV of computer world if I can extend your analogy. It doesn't belong to center of big city, but it is here.
Every other OS has to play catch-up. They either need a killer app or lots of funding, or else they'll be stuck in a specific niche.
Well-designed microkernels have been proven to have only a minor impact on performance, afaik. Especially if said kernel is small enough to fit in a cpu's L1 or L2 cache entirely.
With that in mind, I'd say there's few cases left where users would not want to take a minor performance hit, if that gets rid of entire classes of bugs & vulnerabilities.
Reasons this isn't the norm these days are mostly historic. But going forward, a good midpoint would be popular OS kernels like Linux split into smaller components, while providing same user-space APIs. Along the lines of how XFree86 was modularized into a set of Xorg libraries. Maybe L4Linux could be an example? (dunno how modular that is though).
On such a componentized system, components could be shared among multiple 'OS personalities'. Kinda like a multi-VM setup but finer grained with much more shared code, much reduced attack surface for individual components, while maintaining very strong isolation guarantees between components.