Subject: Re: run levels (was Re: The new rc.d stuff...)
To: NetBSD-current Discussion List <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: current-users
Date: 04/24/2000 20:03:55
In message <>,
Greg A. Woods writes:

>[ On Monday, April 24, 2000 at 17:21:23 (-0700), Paul Hoffman wrote: ]
>> Subject: Re: run levels (was Re: The new rc.d stuff...)
>> rc.d with run levels buys us very little and adds a *lot* of complication 
>> to typical sysadmins.
>That's just not true.

Greg, it *is* true for many of us. If it's not true for you, then I'd
guess you need to recalibrate how you think of many (most?) of the
rest of the members of this list.  (Not intended as a flame; more that
if you keep on with incorrect assumptinos like that, you're just not
communicating effectively with your audience.)

>The extra complication is almost entirely hidden from anyone who is
>simply managing a basic system.

That's not just incorrect, its insulting. The first time I ever ran
into sysadmining Linux, the runlevel goop was a twisty, nasty mess.
It's not hidden at *all*.

>*A*typical sysadmins might find a *different* system more complex until
>they learn it, of course, but that's supposedly a temporary situation.

No, it isn't.  Runlevels continue to be a mess. Different vendors use
runlevels inconsistently. There is no architecture behind them, or how
to use them.  Perry is correct.

>Furthermore a system with traditional "run levels" is only a tiny bit
>more complex than the current rc.d system and if a more intuitive state
>machine (eg. something along the lines of what Peter Seebach has
>proposed) it might actually make the system *simpler* from a truly
>typical sysadmin's point of view!

*If* it was implemented sanely and *without* the abomination of
rc?.d/{K,S}[0-9][0-9] symlinks, perhaps. But from a sysadmin's
viewpiodint,that bears little or no resemblance to traditional
SysV runlevels.

*That* tangled bl**dy abortion is so bad that Linux distributions
include tools to frob the symlinks for you. It's a sick klugde.
Lets bury it, and move on.

>> The folks who need an extra level or two are probably 
>> quite able to write a script that does the mods they need in rc.d for their 
>> specific startup action.
>Nor is that necessarily true.  The people who need finer control over
>the system state are not necessarily systems programmers by definition.
>They could just as easily be junior sysadmins who wouldn't know an
>elegant solution if it kicked them in the butt.

So provide them with a clean mechanism, with examples (shipped in our
standard /etc), and with online documentation and examples of how to
handle, e.g,. Oracle.