[clue-tech] upstart

[Nate Duehr]

Dennis J Perkins wrote:
> On Mon, 2008-09-08 at 21:59 -0600, Collins Richey wrote:
>> Interestingly enough,  Gentoo solved the problem of dependency
>> scheduling of init scripts a very long time ago, so now upstart, et
>> al, would appear to be reinventing the wheel.
>>
> 
> They also use a structure different from sysvinit's.  I think it's
> something like /etc/single and /etc/multiuser, but I would need to look
> it up to get it right.
> 
> I'm curious how they solved it.  I know at least one of the BSD's
> redesigned their init program to handle dependencies.  They discussed it
> for a long time.  I'm not sure if there was any true concensus, but they
> then went ahead and implemented it.  Their solution might be standard in
> all BSD's by now.

I still don't get why the startup program NEEDS to handle dependencies.

Any well written app can :

a) Look for the other things running that need to run before it does.
b) Whine and bitch that they're not, if they're not, and go to sleep for 
a little while, and GOTO A later on.

Why does the OS startup program need to be a traffic cop in this regard 
at all??  We already took this further and inittab will keep respawning 
anything that dies... isn't that enough building blocks to make it all work?

Seems like it already is, to me anyway... changing out the startup 
system comes across as just a bunch of kids playing with a toy... not 
professionals coding a system for other professionals.

(Of course, that's part of the REAL root cause... devs don't respect 
users or even sysadmins time, effort, anything... they just code stuff 
and if good, fine... if stupid, they get berated by the users and 
sysadmins.  This helps fuel the loop of "not caring" what users and 
sysadmins think.  Volunteer devs are not building things for others, 
they're building for themselves... if you don't like what I built... 
"tough", they say.)

Nate