[Sorry Gravis, I could find no shorter way to say this]


On Wed, 25 Feb 2015 15:49:34 -0600
"T.J. Duchene" <t.j.duchene@???> wrote:

> On Wed, 2015-02-25 at 13:11 -0800, Go Linux wrote:
> > This excellent analysis of the systemd debacle was just posted over
> > on FDN. Should be required reading IMO. Enjoy!
> >
> > http://forums.debian.net/viewtopic.php?f=20&t=120652&p=570371
> >
> > golinux
> >
> I must respectfully disagree. I find the analysis to be very biased
> toward one side of the discussion,

And the author tells us that. Now I'd like you to admit that you're
very biased toward the other side of the discussion. I'm proud to say
that I'm biased in the same direction as the author. So is the vast
majority of this mailing list, whose project was created in order to
choose one's init system without trashing the entire OS.

> as well as creating their own
> definitions to fit their side.
>
> If something replaces init, it is by definition "an init system".

So then, if I replace your car's radio by replacing the whole car, it
is by definition a "car radio"?

> Whether it does more or less than the previous init is immaterial to
> that simple fact.

I find no credible element of truth in the preceding sentence. But
anyway, disregarding the definition of "init system", the author is
dead bang right on:

* Debian isn't other distros
* no one—has ever articulated a value proposition for systemd that
adequately addresses its implementation costs.

About "Debian isn't other distros", he characterized the situation
exactly right, plus the fact that when Debian moved, all the Debian
descendents moved with it (except a couple that were born to exclude
systemd, like DNG). And, his assertion was even more right back in
September, when many of the brains behind DNG were helping out with
Debian.

About value proposition vs cost: 90% of the value ennunciated by
systemd fans boil down to "it boots faster", because any benefit
achieved by socket activation and the like could be simulated by
strategically placed sleep statements in any other init. And keep in
mind that if boot speed and reliability are truly important to one, one
would be unlikely to start the number and type of services that would
be problematic to boot speed. AND, although I've gotten systemd to boot
in 4 seconds on a spinning platter, it took 30 seconds after that to
get into the Desktop Environment, because a lot of boot tasks including
networking happened in the desktop environment. AND, I got Epoch to
boot in 7 seconds, and runit to boot in 11 seconds, on the same
hardware, and they both took less time to get to the GUI.

The other 8% have to do with making the GUI responsive to changes in
the system, and vice versa. Nice, but not essential, and not worth a 15
major component monolith tied together with thick, not well documented
interfaces. Not only that, but there are plenty of other ways to get
that feature without gumming up the system by eliminating advantages of
interchangeable parts.

That leaves the 2% benefit of cgroups, whose benefit boils down to,
when all the bullfeathers are removed, reaping zombies. Zombies were an
irritation to all of us, but we've lived with them for 15 years, and
their removal certainly doesn't justify a software V'ger.

SteveT

Steve Litt                *  http://www.troubleshooters.com/
Troubleshooting Training  *  Human Performance