Subject: Re: /etc/inittab (was: /etc/default)
To: None <firstname.lastname@example.org>
From: Steven Plite <email@example.com>
Date: 07/27/1995 16:26:29
> so sayeth firstname.lastname@example.org (Greg A. Woods):
>> More like "from two to a few". If we want(?) to generalize BSD run levels,
>> why not allow for an arbitrary number? I can see it now:
> I'll see your ounce of sarcasm and raise you a pound of flexibility. ;-)
Did I only use an ounce? My sarcasm sprayer must be clogged. ;)
Seriously, I don't understand why SVR only leaves you one level (#4) to
play with. If you're adding complexity, at least make it *useful* complexity.
(Granted, you could redefine some of the "standard" run levels, but that's
asking for trouble.)
> The original 8 states by AT&T seem like enough, but why *not* go to
> something like 256? Who knows what useful possibilities might arise?
> Remember when 16-bit words were enough? ;-)
Exactly. Of course, more than ten run levels would break the SVR
inittab syntax, but who cares?
>> Why not put all the system-defined levels at the beginning [0-x], and allow
>> [x+1-n] to be user defined?
> If it's a totally generic tool, how (or why) do you specify the difference?
By "system-defined" I mean, "as a shipped default". You'll probably want your
OS to ship with run levels for reboot, and single/multi user at a minimum.
But you're right; that shouldn't mean you can't change level 0 from "reboot" to "run Doom".
> BTW, the only reasonable argument I've heard against the idea of run
> levels (i.e. making init a full multi-state machine, rather than a
> two-state one) is that the *potential* for obscure complexity in such a
> critical part of the system is too much to risk.... My only answer to
I think the kernel holds the record for Obscure Complexity in a Critical Part
of the System. ;) (ducks to avoid the barrage of MS-DOS 2.1 manuals)
> that is that I've always felt highly restricted by the old style init,
> esp. when I needed to do important things at critical times in the
> systems life -- SysV init feels like a breath of fresh air in a big open
Hmmm... I picture SysV init as a breath of fresh air in a big open meadow
with my leg caught in a bear trap. (sorry, couldn't resist)
I think the generalization from a two-level to a multi-level init would be a
Good Thing. But I don't consider SysV's init a Good Thing. What
implementation would I suggest? Well, there's always MS-DOS 6.x's CONFIG.SYS
I'll shut up now.
Steven Plite <email@example.com> Open Systems Eng. & Support, Weyerhaeuser
"This is the roller coaster of endless and violent vomit." -- Jason Fox