Er.. a "microkernel" without an MMU is basically nothing more than a thread switcher (not unlike a variety of "realtime" embedded kernels like threadX and what not).
Argue with Andy and not me and read his book. Andy calls it a message-based uK. Given how they structured it. I would agree.
I think Tanenbaum's point was that MINIX3 is nothing like MINIX1 except in name.
Having talked to Andy about this in person, as well as looking at the code, I think I differ with your interpretation
yes, M3 supports a ton of things M1 did not. But the core API and KPI are supersets.
I also know a bit about what Intel uses for the IPMI support as part of my $ day job. Let's just say this is a great deal that is known outside of Intel and a good bit that is not and/or misunderstood.
My point that started this rat hole was that Larry made a comment about V7 having little value. I know for a fact Larry's observation was not true. And how we use the core Minix subsystem (as a basic V7 platform for a single custom application that allows us to manage the server - as I said to Larry think the LS1-11 on the Vax and the 11/40 on the KL processors) and offered it as a counter-example. It could have been almost anything that you called a thread switcher. A simple V7 was 'good enough' and Minux supplied that for the team.
In fact, I can think of other applications where V6 or V7 is more than enough for a lot of the tasks.