Hacker News Reader

Comment by mijoharas | Hacker News Reader

I was about to say macOS uses a microkernel, but on checking it seems that it's hybrid. Also, the source isn't available, so might not be what you're looking for.

View on HN

XNU is opensource....

https://opensource.apple.com/releases/

Click on the release (Monterey 12.5) and scroll down. xnu-8020.140.41 is there..

Links to: https://github.com/apple-oss-distributions/xnu/tree/xnu-8020...

oh! I was not aware of this. Thank you.

>Also, the source isn't available, so might not be what you're looking for.

It is available, and it's called darwin-xnu:

https://github.com/apple/darwin-xnu

a mix of Mach and a FreeBSD kernel:

https://i0.wp.com/www.linuxandubuntu.com/wp-content/uploads/...

And a really good yt presentation:

24C3: Inside the Mac OS X Kernel

https://www.youtube.com/watch?v=-7GMHB3Plc8

> 24C3: Inside the Mac OS X Kernel

> https://www.youtube.com/watch?v=-7GMHB3Plc8

https://fahrplan.events.ccc.de/congress/2007/Fahrplan/events...

It's based on a microkernel (mach) but is hybrid, and the source is available https://github.com/apple-oss-distributions/xnu/tree/xnu-8020...

For the time being, after all kexts are forbidden, replaced by the respective userspace drivers, the real difference will only be arguing about semantics, as for all practical purposes from 3rd parties, it will be like a proper micro-kernel OS.

https://support.apple.com/guide/security/kernel-extensions-s...

They'd need to replace Mach with something less inefficient first.

Mach is a pre-Liedtke microkernel. It's slow at IPC. Which is bad when IPC is what it has to do the most in a proper multi-server system.

Regardless of what we think, Apple has been quite clear at WWDC what the end goal is, no more kexts allowed when there is a corresponding userspace framework for the respective driver class.

This the the best thing in commercial software, decisions happen, and there are no forks botching the final outcome by disgruntled devs.

so ELI5 here, but aren't modern computers fast enough to overcome many of its original shortcomings?

I'm not a software engineer, but out of curiosity.

Not only they are, many hardcode anti-microkernels folks kind of forget their beloved monolithic Linux kernel nowadays runs on top of type 1 hypervisors (which is basically a tiny micro-kernel), and then there is this whole thing with Docker and Kubernetes deployments for every executable.

Ah. So there's no dire need for Apple to improve Mach's performance.

I wonder if the computer industry in general tend to overlook at performance, that's how I feel about the hardware community, where a tiny differences in benchmarks always taking the crown.

There is also GNU Mach from the venerable HURD.

Reply on news.ycombinator.com