[NTLUG:Discuss] Any one know how to get rid of systemd

[Steve Litt]

On Thu, 07 May 2015 22:49:32 -0500
Leroy Tennison <leroy.tennison at verizon.net> wrote:

> Another irritation that's taking away control and knowledge.  I'm 
> getting really tired of the programmer-priests.
> 

Hi Leroy,

There are many, many ways to avoid systemd, and I applaud you for
seeking them.

If your desire is to remove systemd from all init and process
management functionality, but you're willing to leave on your computer a
few vestages of systemd, such as udev and logind, then it's easy:
Jumper across systemd with a simple (and vastly superior) init system
such as Epoch or runit. I've used that technique to de-systemd CentOS
itself. If you want to know more about that subject, back in October
2014-January 2015 I wrote a contemporaneous experimental journal of
init replacement activities here:

http://www.troubleshooters.com/linux/init/manjaro_experiments.htm

Several other people have successfully reproduced my experiments.

The remainder of this email describes solutions that purge most or all
systemd code from the computer...

If you don't mind compile-on-install distros, Gentoo and Funtoo are
sans-systemd, and Funtoo has committed to *always* be sans-systemd. If
you like binary package managers I know of two mainstream choices:
Devuan, which is committed to be sans-systemd forever but isn't ready
yet (fork of Debian), and Manjaro OpenRC edition, which is ready right
now and I find excellent.

In the fall of 2014 I did extensive research of sans-systemd distros
that would be sufficient to run my business, and one that passed the
test was PC-BSD (forgive me for calling it a "distro", but you know
what I mean). If you're the kind of guy who can deal with
FreeBSD's ambiguous and twitchy binary package management, then
FreeBSD would be good for you. FreeBSD is PC-BSD without all the
cutesy desktopism. OpenBSD would have come out on top of the heap, but
its accellerated VM capability is badly broken, making it impossible to
run that one "must have" app that won't compile on OpenBSD.

If you're a seriously DIY guy, the world is your oyster. There are a
zillion lesser known distros, with custom package managers or no
package managers, that have nothing to do with systemd.

Leroy, now that you're asking questions like this in the Linux
community, you should get ready for the reflex responses:

* It's better than sysvinit and Upstart
* Surrender, resistance is futile
* Systemd has socket activation/parallel startup/etc
* Systemd has descriptive units instead of those complex scripts
* You're a neckbeard, get with the modern init
* You're a conspiracy theorist/troll/spammer

IT'S BETTER THAN SYSVINIT AND UPSTART
False choice. There are many excellent init systems available. This
false choice worked only because back in early 2014, very few knew
anything nor cared about init systems. Of course sysvinit is ancient,
of course Upstart was license encumbered, but there were plenty of
existing choices. And yes, they *were* "ready for prime time", whatever
that means. And they still are.


SURRENDER, RESISTANCE IS FUTILE
I guess somebody forgot to tell my computers, many of which init with
Epoch, not systemd (or sysvinit). Users of Devuan, Manjaro OpenRC
edition, *too, *BSD, older versions of Debian and *Buntu, those with
roll-your-own inits, and many others have not surrendered, and are
resisting.


SYSTEMD HAS SOCKET ACTIVATION/PARALLEL STARTUP/ETC
Impressive! But this matters only if boot time is your top priority:
Otherwise you can insert a couple well-placed short sleeps in your
init, and get more determinate boots than socket activation will
ever give you. On my Manjaro system, systemd booted in 4 seconds, Epoch
in 10. Is a 6 second quicker boot really worth all the hassle?


SYSTEMD HAS DESCRIPTIVE UNITS INSTEAD OF THOSE COMPLEX SCRIPTS
So does Epoch. And not all init scripts are created equal: Init scripts
for runit and for s6 are an order of magnitude simpler than those for
sysvinit and OpenRC.


YOU'RE A NECKBEARD, GET WITH THE MODERN INIT
To those who think "modern" is a synonym for "good", this makes perfect
sense. For those of us who think more critically, not so much. For
extra credit, when your critic says that the old init system was too
old for [whatever], let him know that there are many, many init systems
newer than sysvinit. When he says those aren't "ready for prime time",
make him define "ready for prime time" in technical terms, and tech
edit him on what he says. Most peoples' knowledge of init systems is
incredibly fuzzy and downright erroneous.


YOU'RE A CONSPIRACY THEORIST/TROLL/SPAMMER
When they can't win an argument on the facts, ad-hominem is their plan
B. Sagely nod in agreement, then turn around so they don't see you
laugh.


Leroy, you want to be my lab partner? In the next few weeks, I'm going
to be using the init system and low level Linux utilities of Suckless
Tools to try to make an init and process management (they're really two
different things although they're usually a package deal) out of the
Suckless Tools. The benefit is I (or we) will finally understand the
init system and process management system down to the Linux command,
and by reading the (incredibly simple) Suckless code, perhaps down to
the call level.

SteveT

Steve Litt 
May 2015 featured book: Quit Joblessness: Start Your Own Business
http://www.troubleshooters.com/startbiz