On Wed, Apr 06, 2005 at 12:22:25AM -0700, Pete Zaitcev wrote:
On Tue, 5 Apr 2005 12:05:19 +0200 Axel Thimm
<Axel.Thimm(a)ATrpms.net> wrote:
> Having an init.d only with a modprobe is the wrong design.
What makes you say that? How is it different from having an
/etc/modules.d/ivtv?
It is not a service and get's loaded early enough.
I would say that reflexive, knee-jerk reaction "let's add
.d" to any
problem is a recipe for wrong design which produces
redundancies.
Name one :)
Don't forget that it's easy to introduce entities, not so
easy to
get rid of them.
So let's make sure this is a sensible one, which IMHO is very much.
> And there are other modules that need to be before any init.d
scripts,
> like capabilities modules whose absence will break named etc.
The named's number is 55. Surely a slot to load capabilities before
named can be found.
That's a kludge and tends to break. Do you really want /etc/init.d to
get filled with fake services that all start with 00_*?
So far I did not see a good reason to keep /etc/rc.modules around
in this thread (with a possible exception of pcspkr, because it
plugs into HID; but even there a smart kernel patch ought to help).
Who cares about pcspkr, that's cosmetics. No, a mechanism is needed,
if you don't like /etc/rc.modules.d suggest something else, but not
a fake mapping into the services. We could probably map all of
modprobe.conf into /etc/init.d, too. ;)
--
Axel.Thimm at
ATrpms.net