PGV spoke thusly:
">
I cite mine. Numerous failed microkernel ventures, many of which ‘gave
a process to a driver’ none of which were commercially viable.
If you’re referring to Mach… that’s an entirely apples to oranges
comparison."
Mach, Chorus, Spring, and on and on and on…
Gee I forgot to mention Multix, which wasn’t even a microkernel.
What you are proposing is essentially identical to the microkernel
architecture, unless you are proposing pushing all drivers into one ringx
space, in which case you haven’t solved the problem. Hardly apples and
oranges, more like ‘been there, done that’.
But I’m still waiting for any evidence from you at all that supports your
claim that ‘this’ (whatever you really meant by it,) would have a negligible
performance impact. Please do post it soon.