> I'm quite curious then as why performance of microkernels has always
> measured as "only xx% slower than an equivalent monolithic kernel".
Implemantation of MK is the core problem.
Nothing in the concept of MK has inherent inefficiency.
Infact writing a proof-of-concept kernel which runs faster than equivalent
monolithic one could is an interesting project.
The bottle neck is still IPC, but it is now possible to manage IPC much
faster than believed before - these concepts are just not implemented yet.
(L3 I think had user2user communication 22 times faster than mach, and Exo
30 times). It is possible that we are just not applying the right approach
yet.
There's also the idea of a scheme-based OS that eliminates the concept of
IPC, but this is off-topic here.
--
Chen Shapira
Mental power tended to corrupt, and absolute intelligence tended to
corrupt absolutely, until the victim eschewed violence entirely in
favor of smart solutions to stupid problems.
