Subject: Re: /etc/default
To: None <current-users@NetBSD.ORG>
From: Rafal Boni <firstname.lastname@example.org>
Date: 07/27/1995 11:06:06
In message <Pine.NEB.3.91.950726225240.6641C-100000@vulcan>, you write:
-> Another system that has an "interesting" setup is HP-UX 9.x, which has an
-> /etc/rc that calls on a number of other scripts that are relatively
-> self-contained. They have silly names, but the concept is good:
-> /etc/netlinkrc # start up the network interface
-> /etc/netbsdsrc # start up low-level daemons, e.g. routed, named
-> /etc/netnfsrc # start up nfs server, client daemons
-> /etc/netnfsrc2 # nfs mounts, automount
Although I used to hate the SysV /etc/rc.d type things, which came
from running a SCO system that about 18 people owned previously to
me (and each butchered different bits), having done a couple of years
of sys-admining some HPUX and SunOS/Solaris machines, I have to say
that the HP system did *nothing* to help sort things out to any
greater degree than the BSD way. (And it changed enough things that
it took some reading of the scripts to figure out 100%).
On my Solaris box, if I recompiled all my add-on deamons (HTTPd,
Sendmail 8, installed a new CUSeeme reflector, ...), all I had to
do was source the proper RC file with a "stop" arg, and then when I
was done installing with a "start" arg, and everything was re-strted
as it should be.
On BSD, and even more so on HPUX, this was next to impossible to do
without extracting the proper lines from the files and typing them
in.... On HPUX, on top of all the usual stuff, you have all sorts of
other gyrations that get dealt with in the files (stuff that's there
to support cluster configs, etc...).
So, if we're changing because we want added functionality, then we
should do the SysV thing... The HPUX method buys us nothing, and
changes the status quo while its' at it...
-> There are others, but one nice thing about this is that the scripts know
-> how to shutdown and restart the various daemons if you simply execute
-> them again. At least, they're supposed to; there are limitations that
-> don't need to be delved into here.
Hmm.... I'm just glancing at the HP machine we used to run at school
and see no easy way for this to happen.... Unless there are another
set of files that handle this.
-> I only bring this up to point out that at least one commercial vendor
-> chose not to embrace the svr4 way of things (there are no run levels in
-> 9.x, either) but still managed to split things up into manageable
-> chunks. That's not to say I prefer it; myself, having /etc/rc.local and
-> /etc/netstart is far more than enough. Local policy dictates adding
-> a few lines to /etc/rc.local to check for a startup file in an
-> nfs-mounted directory with any additional commands (hey, if the server
-> farm is down, they're not going to get any work done anyway ;-).
As much as I love some of the HP boxes from a hardware/speed/...
standpoint, HP/UX has always made me slightly queasy... It neither
conforms to a normal SysV system, nor to a BSD'ish system, and so
royal confusion abounds...
Rafal Boni email@example.com
Engineer, Data Management Banyan Systems, Inc.
(All opinions are my own) 120 Flanders Rd, Westboro, MA 01581