[RndTbl] a different systemd argument
trevor at tecnopolis.ca
Wed Feb 13 03:40:50 CST 2019
On 2019-02-12 Adam Thompson wrote:
> From one of the FreeBSD developers:
> This is the updated version of his talk, and is more cogent and
> helpful than the first version, so just watch this one :-).
> P.S. I still don't agree with some of his fundamental assumptions,
> nor some of his conclusions, but some of what he says makes sense.
>From the name I thought it would be a systemd-bashing. It's the
opposite. He's a hardcore systemd apologist. In fact, it's almost
pure propaganda. His few "arguments" against systemd are strawmen and
not really what most people are saying.
Most of his arguments (for and against) are incredibly superficial,
fallacious, and could easily be destroyed had this talk instead been a
Since this is not really the forum, I'll simply list 2 arguments this
apologist used that particularly irked me:
1. The "right side of history" fallacy: that's not an argument for
anything, ever. That's basically an attempt at peer pressure,
"everyone is doing it", or that somehow resistance is futile so you
might as well convert. That this argument is used in a computer con
tech talk is rather pathetic.
2. The "Old farts and greybeards" argument: Oh no, the new generation
is coming up and they won't like to do it the way the previous
generation did it. We have to change because the people are changing.
Get with the times or get left behind oldster. Oooh, "the world is
With a system as old as UNIX, that is so ridiculous an argument it's not
even funny. UNIX came around when the last of the greatest generation
and start of the boomers were running things. As if the millennials
are the first generation to grow up with UNIX already mature/old? No.
The mid/late boomers and gen-Xers hit the university scene confronted
with UNIX as it was basically since '69. Did Xers throw up their arms
and demand complete rewrites/rethinks of UNIX because their version of
old farts had created it so it must be passe? No, we hunkered down and
learned the system and the history and said "ah, now I understand,
The hubris of this guy to think that something is different about
Lennart's generation that suddenly what was good enough for 40 years
and multiple generations is now in need of change, partly due to the
reason that "it's the wrong side of history" or "we have to account for
the kids coming up today". The kids 10 or 20 or 30 years ago managed
quite fine, but, nope, it's something magical about the kids precisely
40 years later that can't possibly stand to learn the "old ways".
The presenter really blew his chance to proffer some actual, good
reasons for systemd. It's not all bad. Instead he focuses on points
that have only a veneer of legitimacy and as such undermines his entire
talk. (Oh ya, and he still has no idea what the "UNIX philosophy" is;
no, it's not that it's one binary!)
I don't hate systemd. I use it on every system, and as a Fedora guy
I've been forced into it for way longer than every other distro fan.
However, systemd costs me at least 2-4 whole work days every year (for
8 years now) (unpaid since I'm self-employed) where I have to adapt my
existing (and working perfectly) processes, scripts, templates, and
system configs to support whatever it is that systemd has engulfed and
digested that day. And at the end of the day(s) I'm left with
something that, sure, works basically as it did before (or slightly
worse), but never *better*. I can't think of a single thing that works
better *for me or my customers* with systemd.
Except boot times, which I care not one iota about whether they are 1
minute (systemd) or 2 (upstart) or 3 (init), because they are servers
and rarely rebooted. Not to mention, using "speed" as a main argument
for any major software overhaul is pretty weak because next year's
hardware will be twice as fast. I bet booting an old init system on a
brand new Ryzen 7 with NVMe would be faster than systemd started on a
Core2 Duo 8 years ago. (Sure, on laptops you have more of an argument,
but I'm not doing laptops.)
So Lennart and the apologist have eaten several work-weeks of my time
(again, unpaid) for no reason whatsoever (from my point of view). My
favorite thing that endeared UNIX to me was it was almost always 2
steps forward, 0 steps back. With systemd the paradigm has changed
into 3 steps back, then 3.0001 steps forward: on their schedule. If I
wanted that, I'd have used Windows or Mac for the last 25 years instead
Next up from Lennart and gang are the eating of network config, dhcpd
(D!!), and named. That is the ultimate tragedy of systemd. They
didn't know when to stop (hint: it was a long time ago). More time
wasted. Less choice. When *will* it end? When every daemon is
rewritten inside systemd?? At this point, I wouldn't bet against it.
Final thought: he's a hardcore Free BSD guy? Perhaps he's pushing
"systemd is awesome! resistance is futile!" as a way to destroy Linux
to the benefit of Free BSD, so it can take overr zee werld! ;-)
Methinks he should go back to his FreeBSD, and crying that Apple never
gave his OS the launchd he so cherishes.
More information about the Roundtable