I've been meaning to put down my thoughts about SystemD for the HPR

community for some while, so here goes.

I want to say that I am not a SystemD hater. When SystemD was a hot

topic of debate, many became irrational over it, but I want to start by

saying that I don't think it's a bad technology. I think it is a rather

good technology. I just don't want it on my personal computer. So I

would like to run things down in this order: what is it (as in, what is

it really,) what makes it a good technology, why I don't want it now

(but might later,) and a few tips for you if you decide that you don't

want it currently.

SystemD Is not an init system. SystemD includes an init system.

SystemD Init was faster than SysVInit, but SystemD Init isn't the

fastest init system, and SysVInit now has a parallelization helper, at

least on Debian.

So, if SystemD Init is not SystemD, than what is SystemD? To

understand this we must first understand something about Linux. Linux

operates under a model where there are root processes, and there are

user processes. These two kinds of processes are usually called

"layers." SystemD is actually a third layer, that can be called a system

layer. So when SystemD is added to a Linux system, that changes the

system so that there are three layers, a root layer, a user layer, and a

system layer. As such, you now ask SystemD to set how the system runs.

This is why SystemD includes things like an init system, because if you

want to change what the system is running, you ask SystemD to change it.

SystemD then messages an appropriate system to implement the change,

like messaging its init system to bring up or bring down a system

daemon. Once you play out this in your head a bit, you really realize

that SystemD acts more like a message passing system in this regard.

So why do I say SystemD is a good technology? Because this can

standardize system control. Without SystemD a fleet of computers becomes

like individual fingerprints or unique snowflakes. If you manage many

computers, as many professional IT people do, you want them to all run

the same, all have the same profiles and general configurations. So if

you have a bunch of computers you are running, you can run a lot more if

they are all run the same way. If your job requires you to run 10,000

webservers, you want them to run identically because it is impossible to

keep an understanding of 10,000 unique configurations in a human

head.

SystemD really shines in its support of virtualization as well. So

to speak of servers, I used to run an email server for a few friends.

Each of us had a userid and number as unix users. The mapping of unix

userids and postfix userids can get confusing when it gets big. Thanks

to SystemD's virtualization work, you can actually put a service like

email into a namespace situation so that it has only the users root and

the daemon user id (like "postfix"), so SystemD greatly enhances

security for server installations. This might help explain its

dominance in linux distributions that have been traditionally

server-centric, such as debian and redhat.

So why don't I don't want it? Well, I've been doing a lot of talking

about professional computer work and corporate work environments, but I

use a "Personal Computer" as a hobby. I've been out-of-industry for

decades now. And when I say "Personal Computer" I'm not talking a

hardware specification, rather I'm talking about "This is my personal

computer where I do things my way, as opposed to my work computer where

I do things my companies way". Dear listener, please remember that I did

the first community show contribution to HPR, and my topic was about

personalization. For me, a hobbyist interested in operating system

experimentation, I don't want a system layer, I want a traditional

unix-like system that operates on a two-layer model and does things my

way, nobody else's way.

So, what advice can I give to those who don't want SystemD now? Wel