It depends; if gp has actual things to back-up what they're saying, then I'd actually like to read them. I'm interested in monokernels because of Haiku, but I'm definitely not an expert.
OTOH, somebody releases a whole OS and the response is "Snore! It isn't a microkernel." - Ok, so...why?
Start at "The Paper" here to skip past Linus vs Tannenbaum politics stuff. He describes the common counterpoints and shows with evidence, including existing systems, that they're not as big a deal as people say.
Animats and I think QNX is probably best of commercial ones in balancing all kinds of tradeoffs. It's been used for decades as a self-healing RTOS with good performance. First link is their description of it with second a demo of a product with QNX inside showing how fast it can be on non-desktop hardware.
Open-source one aimed at reliability & legacy compatibility you can play with. Took UNIX decades to get reliable despite all the labor but Minix 3's foundation did it with a handful of people over a few years. That's saying something.
Another FOSS one that aims at high-security integrating many best-of-breed components from CompSci like Nitpicker GUI and seL4 microkernel. First link is descriptive slides with second the actual site. This one is still new so will have bugs.
Finally, it's worthwhile to throw in an exemplary one from capability-security that further isolates things with self-healing properties. Based on commercially successful KeyKOS system on mainframes. No longer maintained but docs and GPL code still available for study or revival. Paper also describes other capability kernels.
So, there's you a few days worth of reading and a few years worth of thinking to do. Hope it helps shed light on why almost every safety- or security-critical system that ever did well in reliability or security was a microkernel-based system. These days, high-assurance is looking at eliminating even it with CPU's with built-in security, compiler techniques for automated safety/security, and DSL's for easy formal verification of OS or system components. Until that's finalized & while using traditional hardware, it's best to build on methods that already worked for decades.
Appreciate it. I try to stay evidence-driven. :) For extra data, Google Gernot Heiser with "L4," microkernel, or OKL4 terms + "evaluation" or "performance." He published lots of comparisons as they put it on lots of phones.
