Everytime we want to add something to all the GENERICs we have
 	to edit over fifty files, then run two different methods to
 	generate the real files for the two ports (amiga and atari)
 	which have tried to move to templates.

 	This is obviously a PITA to maintain.

 	Two obvious options:

 	1) Use MI config include fragments
 	   Could have one file per 'bus', one for options, pseudo-devices,
 	   etc. Or just have one large sys/conf/GENERIC like OpenBSD.

 	2) Use MI m4 include fragments
 	   Similar to option 1), but each port would have a 'GENERIC.in'
 	   from which GENERIC would be built. This would mean after
 	   changing any of the MI files a developer would need to rerun
 	   a make target to update and commit the real kernel configs
 	   but end users (and those who do not care about committing back)
 	   would still have the 'complete' GENERIC and similar files they
 	   have now. See sys/arch/amiga for an example of (much heavier)
 	   m4 usage.

 	In both cases we have the option to split the MI files into
 	entries useful for GENERIC and those for INSTALL type files.

 	What do people think?
-- 
 			   David Brownlee -- abs@absd.org