SoylentNews is people
SoylentNews
from the or-at-least-not-be-bothered-either-way dept.
[Ed's Comment: Not wishing to ignite yet another flame war regarding the adoption of systemd, I hesitated before publishing this story. However, although it is not an formal survey, it might still reflect the views of the greater linux user community rather than those who frequent this particular site. There is no need to restate the arguments seen over the last few weeks - they are well known and understood - but the survey might have a point.]
http://q5sys.sh has recenlty conducted a survey finding many Linux users may be in favour of systemd:
First off lets keep one thing in mind, this was not a professional survey. As such the results need to be taken as nothing more than the opinions of the 4755 individuals who responded. While the survey responses show that 47% of the respondents are in favor of systemd, that does not mean that 47% of the overall linux community is in favor of systemd. The actual value may be higher or lower. This is simply a small capture of our overall community.
Although the author questions the results could this be an indication that we're really seeing a vocal minority who don't want systemd while the silent majority either do or simply don't care? Poll results and the original blog post.
(Score: 3, Insightful) by Marand on Friday November 21 2014, @06:06AM
Not worth getting into most of what you said, but a few things stood out and I wanted to comment:
How about you look at the code of bash and init(1) itself and tell me how it looks there since I'm clearly just an admin who doesn't care how things work because I think systemd has potential.
Does anybody actually use bash for init scripts? Debian and its derivatives use dash, which is much smaller than bash, fewer features but safer. I would hope other distros do something similar. Given that dash is 110K and sysv's init is 39K, there's a lot less there to break than in systemd, whose main binary (the one that I believe is the init) alone is over 1MB.
Now, I'm not saying that "well it's got more code so it must suck", but I think it's premature to be pushing everyone to it as an init when it's fairly new, has a lot more code to it, and is just generally less tested.
The comments making this assumption sound as stupid as someone saying that Kwin is bad and uses an ebil binary file format because Kate writes a sqlite file (not that it does, this is just off the top of my head) while at the same time calling Kwin "KDE".
That comparison is only valid if you can run the individual parts of systemd separately without kludges like Debian's systemd-shim. Yeah, it's in a bunch of individual parts, but they aren't intended to be used individually. That's like claiming a jigsaw puzzle isn't a single picture broken into a thousand pieces, it's really a thousand separate pictures. That's only true if the individual parts are useful without the rest.
Considering that it's really easy to see why sysvinit came out of it. Yet apparently people think this is the gold standard which all inits must be held to. Quite the laugh.
I haven't seen many (any?) people arguing that sysvinit is awesome. The argument is generally that it's either simpler and easier to grok, or that it's well-tested in a way that systemd currently isn't, or that it's interchangeable in a way that systemd isn't.
What I mean by interchangeable: I never had to keep a sysv-shim process running just to make dbus or hal happy before. I could change syslog daemons at will and not have to do anything special. For example, I was using a deprecated syslogd instead of Debian's rsyslog for years because I never noticed they changed the default out and init didn't care. I didn't get forced into it, and I could switch out inits the same way. None of that affected me being able to use power management or removable storage, either.
Then systemd shows up and suddenly your syslog and init have to be blessed by Poettering or you can't use power management, removable storage, and some other desktop feature stuff that's worked for years regardless of your init. Is it really a surprise that some people don't consider that an improvement?
The only reason you can use a non-systemd init right now in Debian without being punished for it is because some devs (Ubuntu or Debian, not sure which) made systemd-shim, which is faking what systemd components expects so that you can run the init (and thus syslog) of your choice without the higher-level parts knowing. That shouldn't have been a kludge, it should have been standard behaviour. If that had been the case from the start, there would have been a lot less outcry.
You see, I don't have any particular issues with some of the parts of systemd. I have the logind part of it running (thanks to the systemd-shim package) without having to change my init, and so far I've had no real issues with it. It's just another chunk of userspace crap, akin to having hal or dbus going, and I'm indifferent. If I hadn't had to jump through multiple hoops to get it installed with the shim instead of changing my init abruptly, I wouldn't even have cared about its addition, it would just be another oddball piece of desktop crap that got added in an update.
Likewise, I didn't care about the init side, until updating desktop software started quietly attempting to replace sysv with systemd. The default behaviour for existing sysv users should have been to set up the shim instead. Having more init systems isn't bad, and it could have gradually been brought into mainstream use by making it an installation option or something. I think if it had gone that way, a lot of us wouldn't have minded. It's harder to be pragmatic about something like this when it's shoved in your face and you're told "TOO BAD, GET USED TO IT, IT'S THE FUTURE"
Doesn't matter how awesome it may theoretically be, people are going to be pissed about that, especially the ones that want to approach huge system changes cautiously. And you know what? That's the kind of user that Debian generally appeals to, so this complete change of behaviour and relatively quick adoption of systemd was like poking a hornet's nest with a stick, except the nest was filled with grumpy admins.