Re: NetBSD Next Logical Step: Microkernel...?

[tlaronde%polynum.com@localhost]

On Sat, Feb 09, 2013 at 10:56:50AM +1100, Martin wrote:
> On Tue, Feb 5, 2013 at 11:27 PM, <tlaronde%polynum.com@localhost> wrote:
> 
> > Be careful not to confuse the advertisement about "microkernels"---I
> > mean the lie about the name proper---and the reality.
> >
> > Plan9 was made by the fathers of Unix at a time when to be published
> > and acclaimed, one had to work on "microkernel". And the Plan9
> > paper has some ironical words about that.
> 
> Just so you and others realize my reference to Plan 9 was purely an
> interest in that OS. Do you have a link to that article i would be
> interested in reading it. :)

Just searcho for "Plan9 from Bell Labs" by Pike, Presotto, ... Thompson
etc. It is in the distribution of Plan9 and probably online too.

> [...] 
> 
> Although while not everyone will agree, you do in a way support my point
> tlaronde, as i am not talking about an overhaul or a traditional
> microkernel like Mach. The fact that Microkernels have there flaws only
> emphasizes my point that pushing the RUMP anykernel to its limitations of
> what it is designed to do, will not only help in examining any flaws or
> faults in RUMP but also pushes the development envelope of kernel design
> (and its always fun to be on the cutting edge of development/ research).

TIMTOWTDI, but whatever way it has to be done ;) As long as one can
select, at compile time, things to add or not, a kernel is already
modular. The one time when might want to be able to add modules is at
discovery/installation time, when not everything is perhaps in the
kernel. But there is a better solution: a minimal kernel, not able to
run all the devices but able to access resources and communicate (this
means a network generally) and able to _identify_ the machine
caracteristics and device (with PCI, this is trivial). This minimal
kernel then send to a vulcan the host description, and the vulcan builds
the needed kernel. It had made this (prototype) a long time ago, and it
was efficient enough for regular installations.

-- 
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                      http://www.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C