Re: GENERIC kernel "modules" implementation info?

[Matthew Mondor <mm_lists%pulsar-zone.net@localhost>]

On Tue, 22 Dec 2009 11:06:02 -0500
"Greg A. Woods" <woods%planix.ca@localhost> wrote:

> Besides there are going to be many of us building embedded systems who
> will never use modular kernels anyway (even if on platforms supporting
> modular kernels), and perhaps quite a few of us who build server
> platforms who won't use modular kernels in production most of the time
> either.

Indeed, I've been a long-time NetBSD user and have used it for
everything from embedded to servers and desktops, yet am not likely to
ever need a modular kernel in the forseeable future...

That said, I could see some value to kernel modules if they eventually
start to be shipped as part of packages; although I'm unlikely to allow
packages to add kernel modules on my own systems, this might be useful
for the NetBSD Desktop Project or other particular scenarios,
eventually.

> Way back in 1988 through to about 1995 I used a quite common unix system
> which provided the best of both worlds with modular kernels.  It had a
> mode where it would do a system configuration boot using full hardware
> probes and dynamic module loading.  It then saved a static kernel for
> fast, stable, and reliable reboots during its normal production use.
> Perhaps NetBSD will catch up to that old unix system technology yet! :-)

An fully modular kernel with possibility to add back both the image and
wanted modules together using ar(1)/ld(1) or the like? :)
-- 
Matt