>> Oh, and there's one but in Ted's suggestion.  I prefer to make a
>> subdir first and test running config there; whereas Ted's suggestion
>> pollutes /sys/arch/pmax/conf with lots of twisty .h files created by
>> config.

>No it doesn't.  Config creates all its files in the compile directory.
>If config on the pmax does not, there's a bug in it that needs to be
>fixed.  I'm using config in the way it's been used since time
>immemorial - I've never heard of anybody using it your way before.
>Your way sounds fine, but it's certainly not the canonical way.

Oh, sorry, you're right.  It depends on where and how you run config,
though. (This [ddefault changed when Greg Hudson added the cruft for
building from a readonly source tree in the MIT environment.)

Since then, whether config spits out files in the current directory or
in ../compile/<CONFIGNAME> depends on -b (and possibly on -s).  I have
multiple trees and always use -s, so I prefer to do the explicit 
config -b -s thing.

If you only have one tree, and it's in /sys/, then Ted's suggestion of
the traditional behaviour works just fine.

Hey, why don't we get someone to write a replacement for the CSRG
`configuring and building kernels' section of SMM?