[svlug] Writing start-up scripts...

[Dagmar d'Surreal]

On Fri, 16 Mar 2001, Rick Moen wrote:

> begin Dagmar d'Surreal quotation:
> 
> > After a lot of anguishing about how the heck I was going to design the
> > init scripts for a five interface box I was running about a year ago, I
> > came to the painful conclusion that putting configuration information into
> > /etc/syscconfig/network was pretty much the only way to go.
> 
> I call it painfully incompetent.  Trying to understand what the hell the
> init scripts do becomes a protracted exercise in dereferencing
> variables.

You know, in some parts of the country parents teach their children to not
deliberately insult strangers.
 
> Sometimes I get so annoyed at it that I rip all that crap out and
> replace it with the straightforward ifconfig and route statements that
> God Intended.

...and that's great if you have a box that never changes, or you're just
configuring one machine and not expecting to replicate your work to
others.  Of course if you happen to replicate your work to more than one
machine, writing everything into one shell script with hardcoded values is
_inexcuseably_ incompetent.

> > Trying to work out an elegant way to be able to bring up and put down
> > (heh) interfaces in a manner that would take into account both firewalling
> > tables and the possibility of transient interfaces (modem being a good
> > example, 802.11 via PCMCIA being even more evil) as well as network
> > services bound to their respective interfaces, using
> > /etc/sysconfig/network/${INTERFACE_NAME}/${SEVERAL_DIFFERENT_FILES} was
> > the only way I could conveniently access organize the data so that one
> > dinky little shell script (only about 40K) could parse and deal with it
> > all.
> 
> "ifconfig [interface] down" works in one direction, and standard
> ifconfig and route statements in the other.  No?  You honestly prefer to
> parse configuration files?

No, I happen to occasionally need to bring down interfaces for extended
periods of time and don't feel like leaving all the old firewalling rules
and whatnot taking up memory in the meantime.  I also occasionally set up
_groups_ of machines to test theories and mechanisms and don't feel like
spending a half hour or more on each one writing one-use scripts to set
everything up.  _Some_ of us understand that modularity in design allows
one to leverage previous work against current problems, and that putting
together a ton of one-time use solutions is simply wasteful.

...and there's not a hell of a lot of "parsing" involved in sourcing a
file that contains a small handful of rather obvious variable
assignments.

I've had enough of arguing with you and Raffi about this.  You can either
believe me or not (and you can definitely believe that I don't give a
*damn* which choice you make), but insulting me and making a *lot* of
assumptions that go far beyond just what I say in order to try to prove me
wrong makes you look like a pair of self-righteous _idiots_.