Subject: Re: Proposed rc.d/rc.conf[.d] changes....
To: Simon J. Gerraty <>
From: Greywolf <>
List: tech-userlevel
Date: 05/08/2000 08:48:28
On Mon, 8 May 2000, Simon J. Gerraty wrote:

# >By the way, it's pretty evident to this wolf that runlevels are going to
# >enter the picture.
# Sorry but now you are being silly.  In all the years that rc vs rc.d
# was debated the _only_ time run-leveles were mentioned was when anti
# rc.d people said:  "I don't want rc.d because run-levels suck".

Okay, I was being a bit pessimistic.  But given that other gunk we have
shoehorned in recently, it really only seems to follow logically that
we're going to eventually shoehorn that in as well.

# Those of us who have implemented rc.d on NetBSD, SunOS etc and used it
# _years_, have done so without run-levels.  They are two very separate 
# issues.

True.  I did not claim otherwise.

# Oh and init.d is not unique to IRIX.  rc.conf.d (or equiv) maybe but
# init.d (or equiv) is used in virtually every rc.d implementation - 
# because it works.

My complaint was not that init.d was IRIX, it was that rc.conf.d's equiv
was.  And every init.d I've seen has had the entire sum of startup scripts,
while the rc.$RUNLEVEL.d has had the ones germane to a particular run level.

And the thing I've been complaining about recently if you'll look at it
is rc.*conf*.d, not rc.d.  I figured that rc.d is a done deal.  I use
it, I've hacked on it, and I still intend to provide a monolithic
RC somewhere.  Yes, I still think that rc.d sucks, but not as bad as
a split rc.conf.

# But please don't confuse the fact that IRIX may have
# init.d with the fact that it has any other miscelaneous feature. 
# One does not mandate the other.

You are quite correct.  The thing I saw as IRIX-like was the config
files which were, say, $prog.conf and $prog.flags.  That precludes any
runtime massaging and it splintered the configuration beyond the limits
of reasonable administrative capacity.  init.d is present on Solaris,
IRIX, HP-UX and, in fact, SVR[0-4], and it was also present, if I am not
mistaken, under System III.

The difference is that the scripts under init.d do not typically contain
the configuration values (at least not that I've used).  They are only
start/stop/stat/reload scripts.

# It is bogus FUD like your statements about run-levels that have helped
# drag this debate on and on.  Please stop it.

That was not FUD, sir.  That was an attempt at a logical follow-on to
the rc.d stuff.  Yes, they're separate -- which would help facilitate
the run-level stuff in the first place.  If they were NOT separate,
it would be a much bigger mess.

Now, please do not pretend that I _advocate_ run levels.  In fact, I do not.
There are those who would, and they will probably continue to do so.
They are actually boggled that we would be implementing the splintered
rc stuff without doing run levels, since both of them are very SysV-like.

# BTW if you do persist with baseless arguments like this don't complain
# if people tune out.

It was, as I said, a logical -- if wrong -- assumption on my part.

# --sjg

BSD... Bitchin'!