Jain,
--- On Thu, 6/5/08, Jain Johny <jainmjo at gmail.com> wrote:
> The more wiki I read, the more confused I become. :-( I
> just read that XNU
> is monolithic kernel based on Mach(which is microkernel).
> So, is it hybrid
> or monolithic?
I don't know where you read it. But wikipedia says XNU kernel is a "hybrid"
one.
Mach has this feature of running different parallel OSs and Sub-OSs (think of
Hurd's Sub-Hurds and Neighbour-Hurds). These were percieved as performance
overheads by Apple and were removed from XNU. Also, the kernel-mode/user-mode
context switches are avoided. But rest of the features of the Mach are
retained. I suppose this makes it "hybrid". I don't know.
> GNU Hurd is based on GNU Mach. I "think" its a
> pure implementation of micro
> kernel.
Ugh.. I doubt judging the purity of the "design" by an implementation alone.
Is there a perfect program? Is there a perfect design for something? Besides
Hurd people have long been blocked by the age-old design of Mach. Of late,
they've tried L4. I hear now that they're moving on to Coyotes. (Not to
mention the personal animosity between Thomas Bushnell and RMS ;))
> I am sure that micro and hybrid cant scale as much as
> monolithic. But for
Plan 9 is said to be hybrid. And Plan 9 is a distributed OS. That, I think,
is scalability.
Tru64 is an industry standard Unix based on Mach. I don't know about its
scalability. But it's backed heavily by HP and Compaq (along with Open Group
and OSF/1)
> PCs and low end servers, it serves the purpose. In a good
> hybrid design, the
> whole kernel shouldnt fail when one of the subsystem fails.
> But it happens
> in NT. So, I think the NT implementation is not good.
NT means that -- "Not Tested" :D
As I said. Design is one thing. Implementation is another. Microkernels are
so very perfect in their design. But implementation wise, I doubt there's one
that's "usable". ;)
> PS:I dont use email license. Its Shirish who uses it. And I
> am for the first
My mistake. Pardon me.
