IRC meeting

[Matthias Urlichs]

Hi,

Shawn Rutledge:
> On Nov 26, 2007 2:01 AM, Scott James Remnant <scott at netsplit.com> wrote:
> > disagree with your ratings as well (e.g. I would say that a monolithic
> > init is good, and plug-in based is bad).
> 
> Why do you think so?
> 
Dunno why Scott thinks so, but my opinion is -- what would you plug into
/sbin/init that you can't do equally well by talking to it once the
system is up sufficiently? Or by using another, existing back-end
(system dbus? HAL?)? Or, if that doesn't work for some reason, by just
coding the feature into it?

"plug-in" in general doesn't mean anything. Can you be somewhat more
specific about which features you'd like to plug into /sbin/init and why
the additional complexity would make sense?

Plug-ins are a lot of work. You need to define a complete set of methods
the plug-in needs to / may support and what callbacks you present.
You need a plug-in loader and have a sensible fall-back if the plug-in
doesn't load or, worse, causes a fault while it's executing.

That in addition to the fact that anybody who says they know exactly
what a new init should be able to do and how it should go about it
is lying. We've learned that the hard way.

-- 
Matthias Urlichs   |   {M:U} IT Design @ m-u-it.de   |  smurf at smurf.noris.de
Disclaimer: The quote was selected randomly. Really. | http://smurf.noris.de
 - -
No animal should ever jump up on the dining room furniture unless absolutely
certain that he can hold his own in the conversation.
		-- Fran Lebowitz