Regarding microkernels, I opine that idea extremely important. Consider UML, for instance. It's a Linux microkernel, basically. BTW, existance of UML is an experimental fact, not opinion.
Researchers of golden age of classic microkernels, who attended USENIX Symposium on Micro-Kernels and Other Kernel Architectures (no joke, that was the title, I've got the proceedings) failed to realize two little facts about microkernels (ok, two little opinions of mine):
- Performance differential cannot be erased. Ergo, classic and other microkernels are to be used where their well known advantages overwhelm performance disadvantage (e.g. QNX is the best example of it). C is "slower" than assembly too. Not to mention Python.
- Applications do need decent (rich) execution environment, not bare hypervisor.
In respect to Billy shelf, I am having extremely hard time to believe that a hole may be drilled wrong in them. They all are made by a pattern drilling, for crying out loud. I respect Raph so much that I went around the house to examine our own Billy's. It may be remotely possible to offset the pattern vertically, depending what tooling was actually used. Hmmm... I guess we should never underestimate the corruptive ingenuity of a sweatshop worker. I did get a defective item once from Ikea, it was the closet organizer with poor welding. But I never saw a hole of component mismatch.
On a related thought, a great Russian Jewish physist Leo Landau was said to send his wife a telegram from a resort: "TOOTH POWDER WONT OPEN STOP PLEASE ADVISE STOP", followed by another one "TOOTH POWDER OPEN STOP". (If you do not know what a telegram is imagine e-mail sent over a 50.5 (fifty point five) baud serial, printed on a bi-color label tape, and delivered to you by a carrier boy).