Re: Jessie and Systemd integration
On Sat, 20 Sep 2014 09:45:30 +0900
Joel Rees <joel.rees@gmail.com> wrote:
> On Sat, Sep 20, 2014 at 5:28 AM, Steve Litt
> <slitt@troubleshooters.com> wrote:
> > On Fri, 19 Sep 2014 21:08:45 +0400
> > Reco <recoverym4n@gmail.com> wrote:
> >
> >
> >> 4) Why re-implementing systemd? Writing your own init is much more
> >> fun, as [1] shows to us.
> >
> > What other things, besides firing up /dev/tty* and running all
> > daemons, does an init program absolutely have to do?
>
> Well, I'm sitting here looking at the results of
>
> ps wwaux | less
>
> on a wheezy box. (And I'll check it on openbsd when I got back from
> running an errand for my wife.)
>
> And I'm thinking, not much.
>
> Managing threads doesn't need to be part of init.
>
> softirqs don' have to be part of init.
>
> Watchdog doesn't need to be part of init. (Who'd've thunk it?)
>
> Managing swap doesn't have to be part of init.
>
> > For
> > authentication, is there any way to use a non-systemd connected
> > PAM, or any other mechanism?
>
> My goodness. That's there as a separate daemon as well.
>
> > I'm not asking about replacing what the current init or systemd does
> > --- I'm asking for the absolute minimum that PID 1 must do.
>
> The only other thing that I am aware of is the collection of dead
> processes, as the grandaddy of all processes.
>
> I'm thinking that may mean init has to be the backstop for signals, as
> well. But I believe that actually handling the signals that hit the
> backstop is done elsewhere.
>
> Which, I think, is at the core of the disagreement on systemd.
>
> Do we use a streamlined init, forcing interprocess communication to be
> well-defined and explicit?
>
> Or do we use a larger init, more capable of fielding less well-defined
> and more implicit, meaning "easier" interprocess communication
> techniques?
>
> (Unpacking all the implications of that question here will be viewed
> by some as "fud" or "scaremongering" or such, so I'll refrain at this
> point.)
Joel,
Check this out:
http://code.dogmap.org/svscan-1/
This guy's actually done it and gotten it running.
SteveT
Steve Litt * http://www.troubleshooters.com/
Troubleshooting Training * Human Performance
Reply to: