Someone on my twitter put it this way: microkernels could see a resurgence if the cost of a context switch into kernel is just as high as the cost of context switching between 2 processes.
In all seriousness, the Nintendo Switch users a microkernel and just got completely p3wned. It's extremely fanciful that we know how to build robust modular microkernels when we can't even get the simple stuff right.
https://en.wikipedia.org/wiki/Tanenbaum%E2%80%93Torvalds_deb...
I'm not saying we should all switch, it's just that it's possible, given my understanding of how we're dealing with the intel bug, that one of the arguments against microkernels just went way. So they might make sense in places they previously didn't.
Yes, it's a form of complexity; just not what tend to relate to problems.
Seriously, completely generic IPC is not something impossible. You may have some to create some large drivers, but the insulation is still there, even for those.
For example, Haiku most probably won't get around to implement KPTI any time soon. But then again, most people consider it more of a fun/hobby OS[0] anyway, so it wouldn't matter that much.
[0] And that isn't meant negatively at all, I actually think the computing landscape would be way more fun again if there were many more projects like it again. Not every OS needs to try and be an everything-OS, and not every OS needs to follow the "always online, everything through the internet" paradigm. If the most interesting use-cases are entirely offline, most of today's popular attack vectors cease to exist.
> This could be an extinction level event for less popular OSes. Intel, and to a lesser extent AMD, may end up causing enormous damage for which they'll pay nothing much.