> 
>  > This is a cool idea, but may cause some problems. Linking may take some
>  > time, and because the kernel isn't preemptive inside everything would block 
>  > during the link time. 
> 
> Um... don't worry about that so much, because it _will_ be changing :-)
> 
Ok, I trust your words :-)

> Besides, if the linking portion needs to do i/o, it will tsleep(), and thus
> yield the processor, etc.  There are already parts of our kernel that
> do computationally intensive things, and we seem to do okay :-)
> 
Simplest way of doing this: Work a while - sleep a while - work a while 
etc... like Win3.11... :-)

I talked to Joel (the linker/symtab hacker) and he said that it would
be simple to fix this. The hard part will be to define a well-behaving
API to the kernel.

A first step on this travel towards a new module handling would be to
fix the symtab handler. (kern/kern_symtab.c) Everybody seems to agree
about this, so it will be ready in a couple of days.

-- Ragge