> What FOSS operating systems (and not even Linux specifically) offers is freedom to choose. But the effort of using some alternatives can be significant.

True. On the other hand, Linux (the kernel) does provide the foundation. You can replace most of components above it and you will still have hardware support for a wide range of hardware. Even if you were to develop an entirely new kernel, the Linux kernel can serve as a proxy for documentation for undocumented or poorly documented hardware.

I realize that places a disproportionate amount of emphasis on the device drivers when the kernel offers so much more and an incredible amount of effort goes into the ecosystem outside of the kernel, but the reality is that most alternative operating systems grow beyond the academic toy phase. Part of the reason is the difficulty in attracting potential users, including developers, if those users have to use said operating system in a virtual machine.

Linux isn’t a unicorn here. Sure FreeBSD has pretty good driver support. In fact back in the 00s I had fewer problems with FreeBSD WiFi drivers than Linux.

I don’t know what the driver story is like for all of the other main BSDs but DragonflyBSD is a really interesting platform with some real innovations in the kernel.

Technically XNU, the macOS kernel, is open source too. Though there doesn’t seem to be much community effort behind that.

There’s also OpenIndiana. Which itself has some really novel ideas.

Then you have smaller projects that have compatibility layers, for example Haiku can run some Linux and FreeBSD systems. ReactOS and Windows driver support. But even if you exclude them in favour of larger, more stable projects, there’s still plenty of options out there.

I do also think you’re placing far too much emphasis on the kernel here. For example some platforms don’t guarantee ABI compatibility between releases, instead requiring their standard libraries being used as the kernel entry point (eg Darwin instead of XNU, or libc on some UNIXes). This is why people often talk about GNU/Linux, Darwin or NT as more than just a kernel.

And that’s before we address the massive elephant in the room that is all the abstractions that sit on top of the userland. For example how so many applications these days are basically running inside a virtual machine called a “web browser”. These days a modern browser is almost like an OS in its own right.

Don’t get me wrong, I’m not taking anything away from Linux nor the Linux kernel. But its main strength is its momentum rather than it being unique.

Unfortunately Apple doesn't accept pull requests for XNU. They're more "source available", to put it under the old .NET Framework terms.

Unless I’m mistaken, XNU’s license (APSL) doesn’t forbid it being forked and reused in other projects (FOSS license compatibilities aside).

For example like how OpenIndina was forked from OpenSolaris (CDDL) when Oracle dropped OpenSolaris development and support.

APSL has been approved by the FSF which other “Source Available” type licenses aren’t.

Personally though, I’d still rather people supported (for example) FreeBSD rather than Darwin/XNU.

That's true, but I agree, why would anyone make a XNU-based OS? Upstream wouldn't accept your forked changes and if you wanted to implement any of the upstream changes over time, that would get difficult as the fork diverged. I really don't think there's anything special in XNU. What exactly does it do well vs. BSD, Illumos, or NT?

100% agreed on BSD, though I'd opt for OpenBSD myself. Either way works!