Subject: Re: LKM - why don't we use them ?
To: None <>
From: Chris G. Demetriou <cgd@postgres.Berkeley.EDU>
List: current-users
Date: 02/22/1994 01:19:29
> I may be out of line with the core team's opinion here, I don't know,
> but as more and more gets added to the various kernels, wouldn't it be
> an idea to use the lkm stuff ?

people can use them if they'd like...  As an administrator of systems,
if things are going to regularly be used by my kernels, i'd rather
keep them in the kernel rather than to have to frob things at boot,
or whenever.  "just more files to get screwed with."

> File systems are a prime candidate for this, as well as mouse drivers,
> audio etc etc. As long as the driver code is not too dependent on
> "options" in the config files, then there shouldn't be too many
> probelms. I think.

No, there shouldn't.  I've been thinking a bit about how to make the
building of LKM's from kernel modules a bit easier, but i've not
gotten to the point where i'm ready to say that any idea i've had is

> I am of course speaking from a platform of ignorance of the issues, so
> if someone could explain them ... ?

It boils down to this, at least in my opinion:
If something's going to be used all of the time, then it should be in
the 'real' kernel, because that:
	(1) decreases start up time (no modules to load)
	(2) increases security (fewer points of attack)
	(3) increases reliability (fewer points that can fail)

however, for 'random,' not oft-used subsystems, i'd say that LKMs
are 'the thing to use.'