On Wed 18 Sep 2019 at 12:06:00 (-0000), Curt wrote:
> On 2019-09-17, Sven Joachim <svenj...@gmx.de> wrote:
> > On 2019-09-17 11:10 -0500, David Wright wrote:
> >>
> >> Well, the only link *needed* is init, hence its dependency on package
> >> init, whose sole function is to keep the number of init configurations
> >> more than zero and less than two.
> >>
> >> The rest of those links just mean that I can read, say, a 60 line man
> >> page for shutdown instead of the 1270 lines of man systemctl.
> >>
> >> So it does seem odd to name the package after some links of
> >> convenience rather than its prime function of associating systemd
> >> (/lib/systemd/systemd) with /sbin/init.
> >
> > The other symlinks are also pretty important, since other software on
> > the system relies on them.  For instance, the kernel invokes
> > /sbin/poweroff and /sbin/reboot to power off or reboot the machine under
> > some circumstances.  Have a look at kernel/reboot.c in the Linux
> > kernel tree.

I'd agree with that, but I'm guessing that run_cmd would deal with a
missing file, allowing the kernel to "force the issue". So I'd still
rank the provision of init as the prime function of the package.

> In the beginning, the symlinks had nothing to do with SysV and one
> expressed bewilderment (or maybe it was rancor?) at the systemd-sysv
> package name and its totally irrelevant allusion to SysV;

I wouldn't like either of those two sentiments attributed to myself.

> then, in a
> certain softening of that radical posture, the symlinked runlevels were
> said to not necessarily be SysV specific (as if the steering wheel and
> tires of a Chevy Camaro weren't specific to the Camaro because other
> cars exist that have steering wheels and tires, too),

AIUI sysv's runlevels are equivalent to systemd's targets, with the
qualification that Debian had less granularity in sysv: rather a lot
was included under runlevel 2, and 3 4 & 5 were ignored.

> and now the claim
> seems to be that, with the exception of init, the runlevel symlinks are
> superfluous,

Superfluous is different from convenience, but I stand corrected in the
use of poweroff and reboot symlinks in the kernel (which has nothing to
do with sysv). BTW I'm not sure what you mean by "symlinked runlevels"
without some elaboration.

> and it is therefore curious "to name the package"
> systemd-sysv "after some links" (kind of admitting by that observation,
> I guess, that those links are indeed related to SysV, but then again
> sort of ignoring, also, systemd's top billing in the package name).

telinit seems to be the only one closely tied to sysv, in that its
arguments use the sysv runlevel names to activate systemd targets.
As pointed out in its man page, it's obsolete, so even "links of
convenience" is a stretch as far as sysv is concerned. Shall I
rephrase my previous observation as "So it does seem odd to name the
package by referencing the init system it obsoletes, rather than its
most important function, providing the /sbin/init link".

Or perhaps I've got it all wrong, and the name of the package is
actually pronounced "systemd, minus sysv".

Cheers,
David.

Reply via email to