What is the intended future of LKMs? I see that there are changes afoot in
-current to try to enable emulations to be loaded as LKMs, but are there
any plans to try to actually move stuff out of the kernel and into LKMs
instead?

It would probably be a Good Thing if more kernel reconfiguration could be
done without having to recompile the thing.

Perhaps one day, the kernel file can be a tarfile containing a core kernel
and enough LKMs to get the system talking to the root device... all the
bus probing code could get the device IDs and try to load:

/lkm/<bus type>.<device id>.o

...for the driver.

Making LKM versions of kernel components a default rather than an option,
such that the kernel sources only build a basic core without any device
drivers or network, with the rest in seperately compileable modules, would
probably also make kernel development much easier due to "cleaning up" the
interfaces between things.

It would make for a lot less kernel configuration... most of the kernel
config file would be replaced by the simple existance of all the device
driver LKMs, with a config file read during boot (probably tucked into the
"kernel tarfile" since the kernel may need this information to bring up
the filesystem) that basically modloads non-pnp modules, giving them their
device iobase/irq/drq/etc. as parameters.

I'm probably not the first person to raise this idea, but what do people
think?

ABS

-- 
                               Alaric B. Snell
 http://www.alaric-snell.com/  http://RFC.net/  http://www.warhead.org.uk/
   Any sufficiently advanced technology can be emulated in software