[ILUG] It's illegal, according to the barrister in PC World

[Rick Moen]

Quoting Smelly Pooh (plop at redbrick.dcu.ie):

> This isn't a a feature of microkernels.  Large, well-funded, mainstream
> monolithic kernel projects such as MULTICS have been known to fail also.

Apples, meet oranges:  Multics flopped for reasons having nothing to do
with the kernel architecture.  One big one was a lasting shortage of
decent development tools.  Another was that many of the hardware
platforms it ran on were badly flawed in various ways.

> BeOS was well designed and I liked it, I haven't used QNX but I'm told
> it's very efficient and flexible....

Yes, I consider these to be the notable exceptions, and I listed them
for that reason.

, and Mac OS X (more specifically the
> underlying Darwin OS) is based on a Mach microkernel.  

Whereas, the Darwin xnu kernel is _not_ an exception of any kind.   The
NeXT development team (and Apple successors) long ago got fed up with
the development and performance problems they were having with xnu, and 
heavily interconnected the original CMU Mach core with the BSD layers
above it.  What has resulted is no longer by any stretch of the
imagination a microkernel, lacking as it does the defining
characteristic of abstraction.

And, even after all that, its perfornace still sucks compared to its
*BSD brethren's.

> What exactly do microkernel architectures have to do before they're
> not failures?

For starters, get written so that they work and perform well.   One
infers that the success of the BeOS and QNX kernel developers in this 
area (regardless of commercial scale) are testimony to their sheer
determination to make a difficult model work.  Most other people long
ago decided that it was way too much work for too little gain.

If you're proposing to write one, good luck to you.

-- 
Cheers,                              "Azathoth need not be present to win."
Rick Moen                                       -- Charles O. Baucum, Jr.
rick at linuxmafia.com