> I agree; it is desireable that interface information not be in rc or
> rc.local.  No one should *ever* need to modify rc for any sane
> system; a distribution rc should always just plug in and work.

There's already a whole lot of policy in /etc/rc and /etc/rc.local.
Reading over /usr/src/etc/rc and /usr/src/etc/rc.local, here are the
things that have to change for me:

- "fsck -p" becomes "fsck -f -p".  And it stays that way at least until
  clean bits really work, and probably until check-periodically-anyway
  clean bits go in.

- The "clearing /tmp" code is yanked entirely; this should be done by
  cron, not rebooting.  (Doing it on reboot cleans out /tmp both too
  often - essentially, it means /tmp is useless to humans because
  anything in there may be torched randomly if the machine crashes -
  and not often enough, because if the box doesn't die it never gets
  cleaned out.)

- inetd gets turned off on some boxes.  (Unlike most of the network
  stuff, inetd is not controlled by "if [ "X${foo_flags}" != XNO ...]"
  construct.)

- All the automated motd editing is yanked.  (What gives anyone the
  idea it's appropriate to willy-nilly destroy the first line of the
  /etc/motd I presumably put exactly what I wanted in?!)

/etc/netstart (which is more or less what started all this) needs a lot
of work, too; putting config information in little files in /etc is
gross and klunky.  (If you're going to put stuff there, at least make
sure it's reasonable to share /etc among different hosts!  On one
machine I hacked on things to use kern.hostid as part of the pathname
for those files....)

No, I think startup scripts will _always_ get site-specific edits.  I
wouldn't even try; try to satisfy everyone and you will end up
satisfying noone.  Or else you'll wind up with a configuration language
that's roughly as general as the startup script was to begin with.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu