lfs-bootscripts 2.0.2 (long)

Dagmar d'Surreal dagmar.wants at nospam.com
Thu Mar 25 12:59:20 PST 2004


On Wed, 2004-03-24 at 20:29, Nathan Coulson wrote:
> Another problem.
> 
> to clean out /tmp, we had
> ( cd /tmp; find -few commands here -exec rm )
> 
> if it could not cd into /tmp, I imagine it may decide to chew up the root
> filesystem
> 
> Thanks to Anderson for pointing this out.

The contrib section seems to be a step backwards, particularly with
respect to the idea that runlevels should form a stack of strict
layers.  What is being laid out makes every runlevel inclusive of
previous runlevels, which only makes for more problems--like how does
one implement a maintenance runlevel where only the sysadmin can login? 
In the "suspending services" section of the documentation, it neatly
sidesteps the fact that suspending a service creates two extra
unnecessary steps.  One where the service is started when it's not going
to be needed, and the other for shutting it down again.  (It also
baffles me why the author takes a moment to criticize traditional
runlevels, as they have no bearing on the functionality provided by this
new method.)

Initctl is not a part of LFS or BLFS.  I don't know what cracked out
telinit binary accepts service names, but it likely came from the same
place that the initctl mentioned did.  Telinit is supposed to take
runlevel characters as it's argument.  This "+" script designator seems
to be of dubious utility as well.  It introduces yet _another_ place to
stop the booting process, making the runpath from "off" to "running"
more complex.  Individual init scripts themselves should either start
what they're supposed to start, or emit an error.  There's no reason why
these (or the process over them!) should _ever_ stop the booting process
unless it's absolutely, positively an error fatal to the system that
might have the potential for data loss or something by continuing.  The
first time you have to get up at 2:30am and race down to a machine room
to manually restart a boot sequence, the reasons for this will become
obvious.  

This stuff seems to severely reduce our functionality as well.  There
appears to be no way to disable a service without removing it's script
or link.  In the current method, a service can be entirely disabled by
chmodding it's script 000, in this method, if it makes that check, it's
not being mentioned in the documentation.

Anyway, I could go on and on about the bad things I see in this but I'll
stop now.  Whoever came up with the new-boot stuff in contrib needs to
give credit/blame to whatever distro they're copying this from, or do a
_lot_ more research before trying to implement a whole new boot system.
-- 
The email address above is phony because my penis is already large enough, kthx. 
              AIM: evilDagmar  Jabber: evilDagmar at jabber.org




More information about the blfs-dev mailing list