Interesting article on boot ordering

[Enrico Scholz]

elanthis at awesomeplay.com (Sean Middleditch) writes:

>> minit is reliable, and reliability and LSB-compliance are mutual-exclusive.
>> Current RHL initscripts are the proof that LSB-compliance is not really
>> necessary, so I do not understand why to enforce it.
>
> Generally because LSB is around for a reason.  Whether init scripts
> themselves are important or not, or whether the LSB chose the best init
> script method to standardize on, I can't say, but if Fedora as a whole
> ignores LSB compliance for technical merit,

I do not speak about _dropping_ LSB/SysV style support (at least not at
the current time); I want to support minit beside it; e.g. look at
https://bugzilla.fedora.us/show_bug.cgi?id=35

The main-package has

| Requires: init(ip-sentinel)

and there are -sysv and -minit subpackages which are having both

| Provides: init(ip-sentinel)

The -minit subpackage is a lightweight package while -sysv requires the
full initscript bloat (e.g. glib2, sysklogd,...). A -lsb subpackage
would be yet more heavyweight (e.g. the entire X11-stuff).

There are some problems with this approach (apt is not very clever in
choosing the right subpackage), but they are not unsolvable.


> Surely there's a way to get LSB-style initscripts to run with this
> minit, yes?

A way exists everytime (linking 'run' to /etc/init.d/<foo>, putting
'start' into 'params' and filling 'depends' with the predecessor). But
it would not make sense to go it. minit and LSB/SysV initscripts are
having *nothing* in common.




Enrico