Stories
Slash Boxes
Comments

SoylentNews is people

posted by n1 on Wednesday October 22 2014, @02:04PM   Printer-friendly
from the head-meets-desk dept.

The Initfinder General (uselessd author) discusses why the systemd debate is so heated. He points fingers at both sides and the story has little to do with who's correct, but everything to to with why no one can agree at all.

[...] I saw the same systemd debate unfold again. I’ve seen it countless times already, and there was virtually no variation from the archetypal formula. You have two ardent and vocal sides, roughly classified into an opponent/proponent dichotomy, neither of which have anything enlightening to say and both with their own unique set of misunderstandings that have memetically mutated into independent ideas that poison virtually every debate of this nature.

Read on for a look at the fuel behind everyone's favorite flamewar.

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 3, Insightful) by Blackmoore on Wednesday October 22 2014, @02:08PM

    by Blackmoore (57) on Wednesday October 22 2014, @02:08PM (#108669) Journal

    really - can i just get the honest facts about how systemd works, and what it is doing; without any of the god damn drama about how it got into the distributions?

    I dont give a crap about the politics, or the developer. I just want pros and cons about how it is written, functions and hooks into the packages.

    • (Score: 2) by Sir Garlon on Wednesday October 22 2014, @02:25PM

      by Sir Garlon (1264) on Wednesday October 22 2014, @02:25PM (#108680)

      TFA is not bad, though it needs to be taken with a grain of salt. On the plus side, he does expose the major points of controversy. On the minus side, he tips his hand more than little by using words like "stupid" to refer to one side of the debate and and phrases like "This is not meant as an indictment on [another side]" for the other.

      --
      [Sir Garlon] is the marvellest knight that is now living, for he destroyeth many good knights, for he goeth invisible.
    • (Score: 3, Funny) by Horse With Stripes on Wednesday October 22 2014, @02:26PM

      by Horse With Stripes (577) on Wednesday October 22 2014, @02:26PM (#108681)

      Sure. Here you go:
      systemd is: a black hole consuming tradition, user choice & control, transparency, reliability and the soul of Linux.
      anything else is: puppies, sunshine and freedom, along with some unicorns impaling systemd supporters.

      Any other questions? ;-)

      • (Score: 3, Funny) by jimshatt on Wednesday October 22 2014, @07:55PM

        by jimshatt (978) on Wednesday October 22 2014, @07:55PM (#108847) Journal
        Would you mind putting that in a chart for me?
        • (Score: 1) by Horse With Stripes on Wednesday October 22 2014, @08:53PM

          by Horse With Stripes (577) on Wednesday October 22 2014, @08:53PM (#108874)

          Sure. Here you go:
          ———
          | v | = Linux using systemd
          | ^ | = Linux using anything else
          .
          ———

          • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:01AM

            by Anonymous Coward on Thursday October 23 2014, @06:01AM (#109037)

            I see an hour glass. The upper part of a hour glass is where the sand goes away, the lower part is where the sand accumulates. If the sand represents the Linux distros, it would mean that Linux distros move from systemd to anything else. Sadly, the opposite seems to be happening right now.

            It's time to turn the hour glass around.

            • (Score: 1) by Horse With Stripes on Thursday October 23 2014, @08:30PM

              by Horse With Stripes (577) on Thursday October 23 2014, @08:30PM (#109355)

              It was a poorly envisioned ASCII Art chart with an arrow pointing down for systemd and an arrow point up for anything else. Though your interpretation is completely understandable, I would have had to (unfortunately) swap systemd and everything else. The hourglass could also have been seen as 'Windows' ;-)

    • (Score: 3, Interesting) by Anonymous Coward on Wednesday October 22 2014, @02:27PM

      by Anonymous Coward on Wednesday October 22 2014, @02:27PM (#108683)

      http://www.freedesktop.org/software/systemd/man/systemd.html [freedesktop.org]

      The *idea* of systemd is a good one. System startup and process maintenance is a big deal. The devs of systemd have basically said to the rest of the group 'you are being babies shut up and just do what we say'. That rubs everyone wrong. So it becomes a huge strawman attack. Basically because neither side wants to listen to the other and pick the best ideas of both.

      The idea we are still using bash scripts to tie our systems together is not good. Pretty much every other OS out there at one point figured out 'here are our consistent ways of starting our daemon processes'. There is a place for scripts. But doing it everywhere seems rather brittle. Having to deal with 5 different ways different linux distros start up is not what I consider 'fun'. I consider it rather annoying. I can do it. But why should I have to? I like the idea that it is flexible enough to pull it off. But in practice it leaves much to be desired. Most of the scripts end up looking about the same anyway. A clear programmers refactoring low hanging fruit of 'lets make a set of functions out of this'.

      Systemd on the other hand instead of making things more clear has decided to obfuscate what it does. There is 0 reason the logs should be in binary. It leads to a second set of tools that do the same thing as the existing tools but have to be maintained 'out of tree'. So instead of using tail/cat/grep to go thru logs I now need another utility to unwide it so I can THEN use those and that tool basically will end up doing the same thing as tail/cat/grep eventually.

      Systemd is a wonderful opportunity to straighten out the startup/shutdown mess. Making it much easier for package maintainers (a good thing). So it is easier to take a debian package and fire it up on a redhat system and very little tweaking would be very cool. Instead of grab the whole source blob compile it and HOPE that the person who made 'make install' knew about the particular nasties in your flavor of linux. Oh and hope someone out on the internet has made init scripts for this source blob.

      • (Score: 5, Interesting) by morgauxo on Wednesday October 22 2014, @03:20PM

        by morgauxo (2082) on Wednesday October 22 2014, @03:20PM (#108712)

        I must admit I don't know that much about Systemd. I know that I like that Linux has (at least until now) used scripts to control it's startup process because I can change them. Bash scripts are so pliable, if you know bash you can make it do ANYTHING.

        Even lower level than startup scripts, sometimes I just set up machines to be apliances by using inittab. I know it's considered "wrong" to do it that way but it's great when you want something to be automatic AND adjustable by someone who knows very little about computers and nothing about Linux.

        For example for a radio station I set up an audio streamer which starts alsamixer instead of a tty on vt1, the streamer on vt2 and icecast on vt3. Yes, I know that this isn't the "right" way to do it. But it just works and I don't even get support calls! I don't have to add an X server and Desktop environment to something which should basically be a server and yet even a DJ who knows almost nothing about computers can just turn on the monitor and go to town adjusting audio levels. By running the streamer and server as VTs rather than init daemons I can switch over to them, get real time displays of everything that is going on and even type commands to control them.

        Starting with an ordinary Linux distro and then munging the startup like this is much much easier than rolling out a proper custom embedded solution. That was one specific example but it's really the flexibility to do ANYTHING that the original software authors didn't intend that have made me really apreciate Linux over the years. I worry that all of this push to make Linux more like other Destkop-only OSs Linux will lose this characteristic. So, how hackable is SystemD? Does it turn a Linux distro into something that is more or less dedicated to being a regular Desktop or does it still provide the flexiblity I have come to love?

        Anyone care to leave informed comments?

        • (Score: 2) by Arik on Wednesday October 22 2014, @10:49PM

          by Arik (4543) on Wednesday October 22 2014, @10:49PM (#108937) Journal
          Who told you this is not the right way to do it?

          Where on earth did you get that idea?

          This is exactly how it's supposed to be done. And this is exactly what the cabal does not understand (or does not want to understand.)
          --
          If laughter is the best medicine, who are the best doctors?
          • (Score: 2) by Grishnakh on Thursday October 23 2014, @01:50AM

            by Grishnakh (2831) on Thursday October 23 2014, @01:50AM (#108982)

            I thought systemd allowed you to continue using init scripts if you chose, so if you wanted to use an init script for a particular service rather than a config file (as with most distro-set-up-services), you could do just that.

      • (Score: 3, Interesting) by arashi no garou on Wednesday October 22 2014, @08:06PM

        by arashi no garou (2796) on Wednesday October 22 2014, @08:06PM (#108852)

        The above is pretty much my take on it. Great idea, terrible execution, horrendous treatment of anyone who questions anything about it. I like the idea of all GNU/Linux distros settling on one init system, but systemd is not the right init system, especially since it continues to grow into its own self-contained OS grafted onto the Linux kernel and GNU userland. It could have been the right one; there are a few great ideas in it, and a few good ideas. But the bad ideas have grown like a cancer, infecting more and more packages along the way. If it continues at this rate, pretty soon you won't even be able to have a basic kernel+command line GNU/Linux distro without systemd as a dependency.

        Of course, it could easily go the other way if the systemd devs ever start actually listening to the GNU/Linux user base at large. They still have time to roll back, to scale it back down to just an init system instead of the Hydra that it has become. But I'm not holding my breath.

    • (Score: 4, Informative) by zocalo on Wednesday October 22 2014, @02:33PM

      by zocalo (302) on Wednesday October 22 2014, @02:33PM (#108688)
      Only by wading into the swamp of the SystemD flamewars, extracting the facts from that morass, and then making up your own mind. There are a few sites that attempt to address this, but unless anyone knows of one I've missed they are either exclusively pro-SystemD (here's why the haters are wrong in a long list of some facts mixed up in a lot of FUD) or exclusively anti-SystemD (SystemD sucks because of this long list of some facts mixed up in a lot of FUD). A neutral and balanced assessment of the pros and cons would be welcome, but it's probably too late to change course at this point.

      There was a post made on the Fedora list a short while back by one of the steering committee (or whatever they are calling themselves these days) about SystemD adoption and it was basically an admission that they agree that not everyone is happy, want to listen to their feedback, but all such discussions get doused in flames so fast that it's pretty much futile. In the mean time they, like most other Linux distros, are taking the path of least resistance to the tools that are taking advantage of SystemD functionality and marching ahead with SystemD so they can support those tools with the least pain. From what I can tell about the newly revived discussion on the Debian lists, it seems a similar outcome there is all but inevitable - SystemD might officially be an "option", but Debian is going to install it by default and resolving all the dependency issues it's removal will no doubt entail is your problem, not theirs.

      While SystemD does have it's good points when it works (which, to be fair, is most of the time), I've written this off as futile - Linux is going down the SystemD route, albeit kicking and screaming in the case of some distros. The old init.d just isn't cool anymore, Upstart has lost what momentum it had, and it's just a matter of time before the number of tools that have either been assimilated by SystemD or require SystemD's presence to function reaches the point where they fold. At this point, if you really don't want SystemD, your choices seem to be between holding your nose and learning to live with it, moving to increasingly niche distros amid dwindling tool support, or look at different flavours of *NIX (we went with BSD). At a pinch, there are also LTS versions of pre-SystemD releases (e.g. Debian Wheezy), but even they won't last forever.
      --
      UNIX? They're not even circumcised! Savages!
      • (Score: 2) by hubie on Wednesday October 22 2014, @02:48PM

        by hubie (1068) Subscriber Badge on Wednesday October 22 2014, @02:48PM (#108698) Journal

        It seems to me that people who say they have, or are, migrating to a BSD are doing it more out of emotional rather than technical reasons, at least at this early point in the game. Would that be a fair assessment? Or perhaps it might be more fair to say that the people who make a point to broadcast the fact that they've migrated are more inclined to have done it for emotional reasons.

        • (Score: 5, Informative) by zocalo on Wednesday October 22 2014, @04:05PM

          by zocalo (302) on Wednesday October 22 2014, @04:05PM (#108740)
          I've posted here this before in another SystemD discussion, but our decision process was this:
          1. We were unhappy with RHEL6 (which was, and at the current time still is, running on the bulk of our servers) in general. This prompted us to look at alternatives.
          2. Because we'd had more than a few issues with SystemD (trying to deal with compromised/broken systems where its IPC related communication and binary logfiles were a huge hinderance), and the way it is often packaged by distros forcing us to either install both parts of SystemD and additional packages we won't use, or to compile our own packages (not a major problem, but an extra hoop we'd prefer to avoid, we included SystemD-free Linux distros and non-Linux distros like BSD in our candidate list.
          3. We then did a lot of testing and benchmarking.
          4. Once we'd got over the BSD learning curve (which wasn't that hard), what we found was that we could build a like-for-like BSD system just as quickly as we could build an equivalent Linux system and because there was less cruft (fewer dependencies) hardening was easier and quicker. Also, despite SystemD's claims to improve startup performance, many of the BSD systems booted to an operational state faster than the SystemD based Linux distros, and performance/resource usage was better pretty much across the board although that's more clearly down to general Linux build options and other baggage than SystemD exclusively.
          5. Given all that, switching to BSD where practical to do so was pretty much a no-brainer. The remaining systems (some are staying with Linux for compatibility & tool support reasons, plus some of our staff prefer Linux on their workstations) are probably going to remain on either RHEL, Debian, or the few other flavours some staff prefer for their personal systems, SystemD not withstanding.

          I'll concede that the decision to include BSD at stage #2 might have been motivated in part on emotional grounds, but the decision to migrate most of our systems to BSD was absolutely made on technical grounds. That it will let us jettison so many SystemD installs, and potential issues troubleshooting the same, is purely an incidental emotional side benefit. :) As I noted, I think the "discussion" is done and it's just a matter of time before the decision is down to Linux with SystemD or some other flavour of NIX altogether, so pointing out that the latter is a viable option and might bring other benefits (YMMV, naturally) is a reasonable point to make.

          --
          UNIX? They're not even circumcised! Savages!
          • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:41AM

            by Anonymous Coward on Thursday October 23 2014, @06:41AM (#109050)

            I encourage you to document and detail all your experiments and findings in some kind of blog, specially those regarding booting speed.

            What's really needed to reduce the load of drama is a little bit of well done engineering and testing (AKA facts).

            Thank you for sharing.

        • (Score: 3, Interesting) by VLM on Wednesday October 22 2014, @06:31PM

          by VLM (445) Subscriber Badge on Wednesday October 22 2014, @06:31PM (#108818)

          at this early point in the game

          that would be the inaccuracy.

          If you think Debian Jessie will release with systemd on time, you are far more optimistic than I am. I mean it could happen, but this is probably going to be one of those "multiple years between releases" that Debian is famous for.

          I've had some interesting problems running Debian testing with systemd. It's pulseaudio all over again. Either it works and thats nice or it doesn't and there's nothing you can do but switch to something else. And the "something else" is being taken away. Other than freebsd, of course. So thats where I'm going.

          I like the debian fork theory, but I don't think its sustainable. Best thing ever is systemd being unportable to *bsd. Music to my ears.

          Also, lets just be honest, freebsd is pretty cool. Its different enough to be interesting, maybe frustrating, but it still drips with awesomeness. I should have screwed around with the BSDs more over the last 20 years. Freebsd's init BTW is pretty cool, its like sysvinit turned from a caterpillar to a butterfly. Pretty cool how it works. Now that should get ported to linux...

          • (Score: 1) by GeminiDomino on Thursday October 23 2014, @12:21PM

            by GeminiDomino (661) on Thursday October 23 2014, @12:21PM (#109134)

            Freebsd's init BTW is pretty cool, its like sysvinit turned from a caterpillar to a butterfly. Pretty cool how it works. Now that should get ported to linux...

            Isn't it already? I thought that's what Pat used for Slackware.

            --
            "We've been attacked by the intelligent, educated segment of our culture"
      • (Score: -1, Troll) by Anonymous Coward on Wednesday October 22 2014, @03:12PM

        by Anonymous Coward on Wednesday October 22 2014, @03:12PM (#108708)

        http://planet.debian.org/ [debian.org]

        What do you see there in the pictures?
        Faggots, feminists, and fat women with bolts through their faces.

        FUCK the free software "community"
        It was great before the mid 2000s.
        Then the faggots and women discovered it and kicked everyone else out
        (ample calls to police too). They are scum.

        These people need a beating for taking over.

        • (Score: 2) by Blackmoore on Wednesday October 22 2014, @03:49PM

          by Blackmoore (57) on Wednesday October 22 2014, @03:49PM (#108729) Journal

          this is exactly the OPPOSITE of what i am looking for.

          I dont give a rats ass about WHO or What "political reasons". that's a god damn strawman argument.

          I care about - FUNCTION.

          does systemd HAVE to initialize and behave as a monolithic code?
            -- if this is only a design issue why can we not generate a fork that is not a monolith?

          For what technical reasons does it latch into so many base libraries?
            -- is this a case of a designer/outside force imposing control over open source code in other projects? How do those projects feel about this effect on the code?

          What alternatives are left available? what distributions / windows managers / options are going to be left?

          • (Score: 5, Informative) by zocalo on Wednesday October 22 2014, @04:43PM

            by zocalo (302) on Wednesday October 22 2014, @04:43PM (#108761)
            The SystemD code technically isn't monolithic. It's often packaged as a single blob, but it actually consists of dozens of binaries and it should (I haven't tried) be possible to package it up into multiple packages some of which are optional - you'd need to take that up with the maintainers of your preferred distro(s). That the various modules communicate via IPC and the log files are binary (unless handed off to a more traditional syslogd) does make it seem like it is a monolithic black box though, particuarly when the system (or SystemD itself) blows up for some reason. Resolving the IPC/blackbox issues could be addressed with a fork, but I suspect that would be a huge amount of work in what could (I've never looked) be some very complicated code.

            SystemD actually doesn't latch onto many base libraries; SystemD itself only needs the Kernel and some of the effectively mandatory core libs to be present. The dependency problem is mostly in the opposite direction - tools and applications that have been compiled to required SystemD so that if you attempt to remove SystemD you get a long list of packages that require it be present. In many cases that's just the default compile options chosen by the maintainers of the distro, although in some cases SystemD functions are hardcoded into a tool in question and it simply won't work unless SystemD is present, which is down to the tool developer(s), not the SystemD developers. To address your query about imposition of control, some of that *might* be down to supporters of SystemD like Red Hat using their huge influence with the development of other tools to drive SystemD adoption, but that's pure speculation, and how the developers feel about this is subjective - so trying to interpret that would likely be a strawman.

            SystemD alternatives? Upstart and OpenRC are the main two, plus the original init, of course. The problem with that is that Linux distro maintainers are really between a rock and hard place with SystemD adoption. As more and more third party tools utilise the functions and hooks into the base OS provided by SystemD, it becomes more and more onerous to find alternatives or recompile them to not require SystemD (which may require custom patches). The path of least resistance is to adopt SystemD and be done with it, and because so many distros have - for whatever reason - adopted SystemD more developers are going to feel that it's OK for them to require SystemD be present, even if their tool won't work on the few distros that are still holding out. Currently the holdouts include pre-SystemD LTS releases of distros that are still supported - e.g. Debian Wheezy, plus Ubuntu (although not for long), PCLinuxOS, Slackware and Gentoo - and probably a few others.
            --
            UNIX? They're not even circumcised! Savages!
            • (Score: 2) by digitalaudiorock on Wednesday October 22 2014, @06:30PM

              by digitalaudiorock (688) on Wednesday October 22 2014, @06:30PM (#108817) Journal

              The SystemD code technically isn't monolithic. It's often packaged as a single blob, but it actually consists of dozens of binaries and it should (I haven't tried) be possible to package it up into multiple packages some of which are optional - you'd need to take that up with the maintainers of your preferred distro(s).

              This page does a pretty good job of debunking a lot of this spin from the systemd folks, and the above is the #1 fallacy:

              http://judecnelson.blogspot.com/2014/09/systemd-biggest-fallacies.html [blogspot.com]

              • (Score: 2) by zocalo on Wednesday October 22 2014, @08:40PM

                by zocalo (302) on Wednesday October 22 2014, @08:40PM (#108865)
                That article does make the point that SystemD is both modular and monolithic, for pretty much the same reasons I did if you'd included the next sentence after the one you quoted; lots of distinct binaries that are so tightly coupled they are at the point that they are heavily interdependant - and when it blows up and needs troubleshooting then it might as well be a single binary for all the difference what modularity it does have makes. It *is* however possible to completely disable some parts of SystemD (that might break other stuff that has been coded to depend on it though), hence my suggestion that it may be possible to break those parts out into optional packages. Sure, you need the core SystemD PID1 replacement no matter what, but that's kind of reasonable if you take the view that everything else serves to extend that.

                So, it's not a single monolithic binary, but neither are the components as independant as, say, the binaries in coreutils but overall I'd say it is slightly more modular than monolithic, but it's really a matter of personal opinion and semantics about how you make that call - I certainly don't have any issues with people who declare that it's one or the other. There's definitely a lot of fear of the unknown in connection with SystemD, and while I don't like the current implementation or on-going assimilation of other services very much I do think the general idea is sound and there is some very nice functionality there. If the developers were to address the modularity by making it easier/possible to swap out the various modules for alternates, and improve the ability to debug in the event of issues then I'd be a lot happier about it. I'm not holding my breath though.
                --
                UNIX? They're not even circumcised! Savages!
                • (Score: 2) by Hairyfeet on Wednesday October 22 2014, @11:45PM

                  by Hairyfeet (75) <{bassbeast1968} {at} {gmail.com}> on Wednesday October 22 2014, @11:45PM (#108953) Journal

                  "Sure, you need the core SystemD PID1 replacement no matter what"...and THAT is what is pissing people off! One of the big selling points of Linux is there is NO "one way" to do anything forced upon you by on high,want to strip it down so small you can run it on a $1 SOC? You an do that, want to use Python to control boot? again not a problem...until systemd gets in the mix then you HAVE to take it (and a shitload of ever expanding shit they keep sticking into systemd) because it is coupled too damned tightly to easily drop.

                  Its the same damned problem that we saw with his last work PulseAudio, sure it SOUNDS like a fricking great idea...until something goes wrong, then its a fricking mess. Say what you will about ALSA but it generally wasn't that hard to figure out what went wrong with it when it crapped out. Maybe its just me, but I think you should look at a person's past work to predict future performance and when the thing most likely to cause a Linux system to shit itself on update and fail the Hairyfeet Challenge is Pulse? Really don't make me feel all warm and fuzzy at the thought of giving the guy control over the boot process.

                  But I'll let you guys fight this battle, its looking like all my horses gonna be out the race anyway. All my Linux admin friends have moved to OSX or *BSD, Windows 10 looks to be the next XP (and will probably be just $30 or even free) so the days of having to do the Hairyfeet Challenge trying to find a Linux capable of replacing Balmer's Folly is over....just in time from the looks of it, as it appears Linux is gonna shoot itself in the foot like it did with KDE 4 and Pulse...good luck with that!

                  --
                  ACs are never seen so don't bother. Always ready to show SJWs for the racists they are.
                  • (Score: 2) by zocalo on Thursday October 23 2014, @08:13AM

                    by zocalo (302) on Thursday October 23 2014, @08:13AM (#109084)
                    Nice way to take something *way* out of context there. The whole thing was that you need the SystemD PID1 binary no matter what *IF* you are are adopting SystemD, which would kind of be the whole point - especially given the bit about the other binaries in the package being effectively extentions to that binary or tools to manage it. You certainly wouldn't need it if you were building a distro around some other init system like Upstart, so there are multiple options available - I listed three others earlier, all of which are mature enough to use. As an aside, that's also three more options than we have to interact with Windows boxes via SMB and to join AD domains, which the community generally seems quite content about. At best you could level some kind of complaint about not being able to unbundle the binaries such that you could use systemctl to manage processes started with Upstart, but in that case you might as well start complaining that you can't use, say, rndc to control MaraDNS or PowerDNS too.
                    --
                    UNIX? They're not even circumcised! Savages!
                    • (Score: 2) by Hairyfeet on Thursday October 23 2014, @10:09AM

                      by Hairyfeet (75) <{bassbeast1968} {at} {gmail.com}> on Thursday October 23 2014, @10:09AM (#109100) Journal

                      Perhaps YOU might want to read up a bit about what is going on, about how udev is next in line [freedesktop.org] to be system'd, or how his dream is the whole base OS tied to systemd [0pointer.de]. You can find this and more in this most informative article [blogspot.com] which uses the man's own words against him, if Lennart Poettering has his way Linux will be running on TOP of systemd, not the other way around.

                      But again I'm out of the race, the Balmernator is gone, Win10 is the new XP, OSX is kicking, which sadly means its time for Linux to shoot itself in the foot. Its a damned shame but it happened when Lennart Poettering crapped Pulse just as Windows was going to crap (I hope Balmer at least sent him a fruit basket), it happened with KDE 4 being shoved on users when even the devs said it wasn't ready, and now systemd is gonna make a big old mess all over things...ya know, if I didn't know better I'd swear somebody is pulling an Elop in the Linux camp!

                      --
                      ACs are never seen so don't bother. Always ready to show SJWs for the racists they are.
                      • (Score: 2) by zocalo on Thursday October 23 2014, @12:07PM

                        by zocalo (302) on Thursday October 23 2014, @12:07PM (#109131)
                        I'm well aware of Lennart's plans to assimilate other low-level (and not so low-level) services into SystemD and that Udev is next on the block. That is my main concern with the whole thing, although not so much with Udev as I am with the likes of NTP, DHCP and so on. I wouldn't have a big problem with that if a number of conditions were met:
                        1. The assimilated daemons were all fully modular and could easily be completely swapped out for alternates, if a user wished. Obviously if the user wishes to run a tool that requires a given SystemD component when they prefer an alterative, then that's up to them to determine the best solution - being able to run the two in parallel as can be done with Rsyslog, or even with the SystemD component present but disabled
                        2. Ideally, I'd like to see SystemD packaged up as "systemd-core" for the main PID1 and supporting tools, then a whole bunch of other packages for the rest, e.g. "systemd-logd", "systemd-ntp", "systemd-whatever-else-lennart-wants", etc.
                        3. The code was generally of better quality (it is getting there).
                        4. The inter-process communication was less opaque when debugging.
                        5. The developers were not such prima-donnas when confronted with issues in their code, or that their code has caused elsewhere. Those that claim a lot of the SystemD hate is actually Lennart hate do have a point, but that issue wouldn't exist if the developers were more accomodating when dealing with others.

                        Sadly, it's not that easy for me to just wash my hands of this, although I wish I could. We might be in the process of moving the bulk of our systems to BSD, but we still have a lot that require Linux for some reason or another and while it's not a done deal quite yet, we're resigned to the likely inevitability that eventually all those Linux systems will run SystemD unless we are prepared to really hack them around (we're not). When SystemD works, we're actually quite OK with most of what SystemD is doing and how it's doing it, but in the mean time the best we can do is keep pointing out what we don't like and hope it gets better.

                        --
                        UNIX? They're not even circumcised! Savages!
                  • (Score: 2) by tangomargarine on Friday October 24 2014, @03:31PM

                    by tangomargarine (667) on Friday October 24 2014, @03:31PM (#109602)

                    Windows 10 looks to be the next XP (and will probably be just $30 or even free)

                    WTF. Citation needed? The first result on Google for "Windows 10 price" is an article dated October 15 (i.e. 9 days ago) saying that Microsoft hasn't said boo about pricing. Second is from September, saying the same thing.

                    --
                    "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
                • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:09AM

                  by Anonymous Coward on Thursday October 23 2014, @06:09AM (#109039)

                  Sure, you need the core SystemD PID1 replacement no matter what

                  One component to rule them all and in the darkness bind them.

        • (Score: 0, Offtopic) by velex on Wednesday October 22 2014, @05:21PM

          by velex (2068) on Wednesday October 22 2014, @05:21PM (#108784) Journal

          Stop. Just stop.

          It was great before the mid 2000s.
          Then the faggots and women discovered it and kicked everyone else out

          Did you ever consider that there are some “faggots” who are pretty pissed off every time the stereotype you're creating gets applied to them simply because they had the misfortune of being assigned the same gender as you at birth by some womyn-born-womyn who can't see past her cisfemale privilege?

          Did you ever consider that there are some “faggots” who are considering moving to BSD to get away from systemd? That use OpenRC? That have no intention of ever following the systemd Linux from Scratch?

          That think Poettering is a creep and a white knight and are just as sick of him and his backroom dealing as you are?

          Did you?

          Do you know how much hatred feminism and womyn-born-womyn direct at folks like me just for daring to get mistaken for a womyn-born-womyn by strangers and at homosexual men because they aren't good little sex objects? Obviously anyone who doesn't date women must be a sexually frustrated dweeb who just “good with women!”

          Hell, I got ma'amed this morning while clearly wearing men's clothing and not even attempting to mask my voice somewhere I go pretty much every morning. Do you know how much hatred and rape hysteria is out there because I might be extended cisfemale privilege by accident and infiltrate womyn-born-womyn only places? They invite drag queens into womyn-born-womyn only spaces, but it'd be the fucking end of the goddamned world if I entered one by mistake.

          Oh, this will make your head asplode. I regularly vote Libertarian and I may be renewing my membership soon. Factor that into your fantasy that “faggots” are communists that want to destroy families. Yeah, I'm going to achieve my feminist-communist goals by voting Libertarian and contributing money!

          Until you understand this shit and get yourself out of the delusional world of the MRAs, stop.

          I might as well hit submit after typing this rant, but what the fuck good could it possibly do? karma to burn.

          • (Score: 2) by Jeremiah Cornelius on Wednesday October 22 2014, @10:50PM

            by Jeremiah Cornelius (2785) on Wednesday October 22 2014, @10:50PM (#108938) Journal

            The above may be decoded into non-jingo Inglysh with the Google Translate option for "GenderJargon".

            Dress how you like and sleep with who you want. None of my business. Nonetheless, the coupling of white, middle-class men does not register super high on my list of world problems.

            I reserve that status for PID1 flame wars.

            --
            You're betting on the pantomime horse...
    • (Score: 3, Insightful) by Anonymous Coward on Wednesday October 22 2014, @02:46PM

      by Anonymous Coward on Wednesday October 22 2014, @02:46PM (#108696)

      Simple: Systemd is init done the Microsoft/Apple way.

      There are two types of People using or developing Linux:

      1. People who like Linux for being different from Windows/Mac. They are not amused about Linux being morphed into a more Windows/Mac-like experience. For them Linux is an alternative to those systems in a technical way. That is, they like Linux because it is different.

      2. People who actually like Windows/Mac, except for the fact that they are not Free Software/Open Source. They would prefer Linux to work more like those other systems. For them Linux is an alternative to those operating systems only due to the licensing. That is, they like Linux although it is different.

      Basically, you can sum it up as this:

      Imagine, Microsoft would one day release Windows under the GPL, donate all related patents to the public, and resolve anything you don't like about Microsoft itself. Imagine further, through a heroic effort, all software that runs under either operating system, including outdated earlier versions, were also made available under the other, with the same quality, free of cost for those who already own them on the other OS (so there would be no software availability/lock-in reason to prefer one OS over the other).

      If you still would stick to Linux, then you belong to group 1, and probably hate systemd.

      If you would change to Windows, then you belong to group 2, and probably love systemd.

      • (Score: 2) by fadrian on Wednesday October 22 2014, @04:03PM

        by fadrian (3194) on Wednesday October 22 2014, @04:03PM (#108738) Homepage

        How about group 1 and don't care either way? I've been using UNIXes of various flavors since the late 1970s. Technology changes. BFD. It's a big evolutionary stew. And it's messy.

        Basically, this is why you fork - to put a stake in the ground as to where you stand. This gives the distros/community a chance to either (a) choose one (b) choose the other or (c) route around the damage by modifying their distros/systems so that a choice can be made. And, in the end, the people who use the software decide. It's an elegant solution, if you think about it. Bloody, but elegant.

        And the best thing is that there's microwave popcorn.

        --
        That is all.
        • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @04:47PM

          by Anonymous Coward on Wednesday October 22 2014, @04:47PM (#108766)

          Basically, this is why you fork

          And the fact that nobody's doing that here proves that most of the anti-systemd crap is just temper tantrums. If systemd really is bad then lots of people are going to want an alternative without it, but if nobody is willing to work on it, it shows that systemd really isn't all that bad.

          If you want truth, look at actions, not words.

          • (Score: 3, Interesting) by velex on Wednesday October 22 2014, @05:38PM

            by velex (2068) on Wednesday October 22 2014, @05:38PM (#108791) Journal

            Potential Debian fork. [theregister.co.uk]

            You can read other comments here about people moving to BSD.

            Me? I'm rolling a Linux from Scratch. I want to use Wayland and E19. (Well, it'll be xorg for me until Wayland support is more complete—might even crack out my C skills and contribute to move that along.) I'm using a BodhiLinux as the host system, and it's pretty good. I just don't want to get too used to it when it get sucked into systemd. Binary logs do not want. Binary configs? Meh, whatever. Binary logs? Fuck Beta^H^H^H^Hsystemd. I'm perfectly happy with metalog.

            So, don't get mad, just fucking fork it. And you're done.

          • (Score: 3, Insightful) by curunir_wolf on Wednesday October 22 2014, @07:54PM

            by curunir_wolf (4772) on Wednesday October 22 2014, @07:54PM (#108845)

            From Jude's Blog [blogspot.com]:

            Fallacy #6: "Systemd is open source, so if you don't like it, you can fork it!"

            And, some people are working on that, despite the fact that systemd is a large moving target.

            However, this statement usually misses the point. If you're talking to a non-systemd user, chances are that person is more interested in having systemd not encroach on his/her systems. In this context, you're telling them to fork systemd...to not have to use systemd.

            Fallacy #6.1: "I wish those systemd detractors would stop whining and make their own alternative!"

            This also misses the point. Most systemd detractors and non-systemd users are perfectly happy to leave systemd alone and let it do its thing for the people who want to use it. Whatever they have going for them is perfectly fine for their needs.

            The problem is that systemd doesn't want to leave them alone. Lennart Poettering's vision for systemd is for it to subsume the whole base OS. He's also advocated for GNOME's hard-dependency on logind, and is of course responsible for logind having a hard-dependency on systemd-as-PID-1. When non-systemd users lament that Linux is supposed to be about choice, systemd users will often fling this link at them (as if it proved anything).

            The point is, a lot of non-systemd users don't need an alternative, so there's no point in telling them to go work on one.

            --
            I am a crackpot
            • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @11:02PM

              by Anonymous Coward on Wednesday October 22 2014, @11:02PM (#108942)

              Who said anything about forking systemd? Your "Fallacy #6" is pure strawman. People are crying that systemd is being forced on them by the distros, so fork the distros. "B-b-b-b-but they should be the ones forking, not us!" is on the same maturity level as "But he started it!"

              • (Score: 2) by SlimmPickens on Thursday October 23 2014, @04:16AM

                by SlimmPickens (1056) on Thursday October 23 2014, @04:16AM (#109010)

                Linux is at the crossroads of abandoning Unix philosophy and you're saying it's immature to argue against it?

                • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @04:48AM

                  by Anonymous Coward on Thursday October 23 2014, @04:48AM (#109014)

                  If it is at a crossroads, its only because nobody is willing to do more than bitch and whine. If nobody is willing to work on systemd-less forks of distros, not even a single one of them, then it must not be that important an issue.

              • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:15AM

                by Anonymous Coward on Thursday October 23 2014, @06:15AM (#109041)

                OK, I'm coming to you and tear your house down. Don't complain. If you want a house, you can just build a new one. It's not as if there would be any cost associated with that, right?

      • (Score: 2) by Kromagv0 on Wednesday October 22 2014, @04:45PM

        by Kromagv0 (1825) on Wednesday October 22 2014, @04:45PM (#108762) Homepage

        What if like me you would just keep using each OS for what I currently use it for since I wouldn't have to wast the time installing and setting everything up on the one it doesn't currently run on. I suppose that would probably put me closer to group 2 since all I care about is the right tool for the job (what ever one makes my life easiest).

        --
        T-Shirts and bumper stickers [zazzle.com] to offend someone
      • (Score: 1, Insightful) by ghost on Wednesday October 22 2014, @04:47PM

        by ghost (4467) on Wednesday October 22 2014, @04:47PM (#108767) Journal
        launchd [wikipedia.org] is init done the Apple way. Apple wouldn't have anything to do with a turd [documentingreality.com] like systemd.
        • (Score: 2) by hemocyanin on Wednesday October 22 2014, @11:23PM

          by hemocyanin (186) on Wednesday October 22 2014, @11:23PM (#108950) Journal

          Well, I thought you were using "turd" metaphorically -- till I clicked the link. Very literal of you sir.

        • (Score: 2) by Common Joe on Thursday October 23 2014, @01:49AM

          by Common Joe (33) <{common.joe.0101} {at} {gmail.com}> on Thursday October 23 2014, @01:49AM (#108981) Journal

          Parent has a link that is NOT WORK SAFE.

          ghost, if you post a link like that, you should warn people. The higher ups get upset when us mere mortals click on links like that.

      • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @07:13PM

        by Anonymous Coward on Wednesday October 22 2014, @07:13PM (#108835)

        I'm sorry to inform you but I don't fit neatly in your dichotomy. Further I think your categories are quite strange to be honest.

        * I hate the win/mac black box hands off approach. I love to tinker and think it's ridiculous for somebody to claim they know better what I need and want than I do myself. And there is an infinite number of choices...

        * On the other hand I think free software is the greatest idea of the century if not of all time. I hate the greed and exclusion that drives proprietary software. The pointless repetition of work. The animosity and the secrecy.

        In your hell freezes scenario I wouldn't probably do either or but instead a hybrid, the best of both worlds. So I'm not informed on systemd. I don't know too much of the subject but am so far cautiously pessimistic.

        And it's sad when people say they don't care for politics but just for technical things. Guess what, politics is a technical thing. Obviously one those people have very little knowledge of. You can hack it too and yes, it very much affects the life of everybody on this goddamn rock. Pull your cowardly head out of the sand because the life goes on around you.

        • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:28AM

          by Anonymous Coward on Thursday October 23 2014, @06:28AM (#109044)

          I love to tinker and think it's ridiculous for somebody to claim they know better what I need

          In other words, you should not like systemd. Because systemd is exactly in the "I know better than you" camp.

          And you seem to have misread my post: I didn't mean to imply that the members of the first group don't value Free Software. Note my use of "only" in the second point.

          From what you posted, I'd put yourself firmly in the first group. You like Linux because it is different. You also like it because it is Free, but not only for that reason.

          And of course, the option of building something new is always there. There was an implicit assumption that you had to choose between existing alternatives; building something new does not tell in which camp you are, unless you specify how what you'd build works.

    • (Score: 2) by bzipitidoo on Wednesday October 22 2014, @03:08PM

      by bzipitidoo (4388) on Wednesday October 22 2014, @03:08PM (#108706) Journal

      This is what I would like to see. Just the facts.

      My own personal experiences with systemd are negative. I used Arch Linux, until they switched to systemd. I tried to stay with Arch, tried to follow the lengthy and complicated directions they provided for modifying files and directory structure by hand, but must have missed something. Or maybe their instructions had a mistake. Whichever, my system ended up unbootable. So I tried to reinstall Arch from their latest installation CD, which at that time was the first Arch installation media that defaulted to systemd. Arch had thrown away their fairly nice interactive installer, and so it was necessary to copy all kinds of commands into a terminal.

      My next run in with systemd was when I needed to look at the system logs. They weren't in /var/log/ any more. Took me a bit of time to run them down, but it turned out they'd moved the logs to a subdirectory of /var/log. And systemd had compressed them. I wanted only to look at the last few items, and was annoyed that this forced me to wait some 30 seconds for decompression. That's no big deal if you need to look only once, but when you're tweaking the system and checking repeatedly, a 30 second delay each time gets very tiresome. I also had to discover what command to use, since less and tail could no longer display the logs in a readable way. zless? No, it seems systemd uses some other compression method. What method? I still don't know. The magic command to view the logs was "journalctl". Some documentation of all this would've saved me some time.

      I think all that fun with the system logs had more to do with the choices the distribution maintainers made. They were the ones who made compression of all the logs, even the most recent, the default setting. Ubuntu 14.04 uses systemd, and has the logs set up the customary way.

      Lastly, the attitude of the Arch maintainers was offputting. When I questioned the switch to systemd in the forums, I was told to STFU because the decision had already been made and I didn't know what I was talking about anyway.

      Arch also tried to unify /bin and /usr/bin. Put all the binaries in one directory, and made the other a symlink. Anyway, I switched distros. Dumped Arch Linux, and moved to Lubuntu until I find a distro I like better.

      I don't know if systemd is good or not. Arch did a lot to make me think systemd was not good.

      • (Score: 5, Insightful) by metamonkey on Wednesday October 22 2014, @03:46PM

        by metamonkey (3174) on Wednesday October 22 2014, @03:46PM (#108727)

        My problem when I first encountered systemd was the lack of documentation.

        I hadn't been following open source development for a few years as my job required me to use photoshop so I was running macs for everything but servers. I needed to set up a box and I just grabbed the latest Fedora (this was maybe 3 years ago) because I had always used Red Hat in the past (like, early 2000s) so I figured it would be similar. Well I go to add a custom script to /etc/init.d and nothing happens. I get some message about systemd from somewhere or another and I had no idea what it was. I figure, "no big deal, I'll just google around, get a HOWTO and be up and running in no time." Uh, no. It took like 4 hours to wade through this thing because the only guide I could find was half full of placeholder headings, and the content and organization of the config files was non-obvious and mostly undocumented. I eventually figured it out, but it was just stupid. They changed something that was, if inelegant, simple and widely understood and replaced it with something completely different and poorly documented.

        It hasn't gotten much better. To be honest the problem is just that I don't trust it. It's still so "under development." They're constantly adding not just new features but whole new (I'm looking for a word that means "whole bunch of areas of influence" like their recent announcement that they're replacing the consoles) and changing shit around. It's not stable. And I kind of what whatever's running at PID 1 to be rock solid. They need more development time. They need to be the ones forking the major distros and building a better system, showing us how much better it is, and we'll flock to it. But they're just massively changing critical shit in major distros with an unstable piece of software. No thank you.

        --
        Okay 3, 2, 1, let's jam.
      • (Score: 1, Informative) by Anonymous Coward on Wednesday October 22 2014, @06:22PM

        by Anonymous Coward on Wednesday October 22 2014, @06:22PM (#108814)

        If you liked Arch and don't want systemd you'll probably love Gentoo. It has Arch's rolling release model and makes no attempt to force systemd on you.

        • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @03:25AM

          by Anonymous Coward on Thursday October 23 2014, @03:25AM (#109002)

          Or Funtoo, which is a gentoo derivative that actively avoids systemd.

      • (Score: 3, Interesting) by VLM on Wednesday October 22 2014, @06:41PM

        by VLM (445) Subscriber Badge on Wednesday October 22 2014, @06:41PM (#108821)

        Lastly, the attitude of the Arch maintainers was offputting. When I questioned the switch to systemd in the forums, I was told to STFU because the decision had already been made and I didn't know what I was talking about anyway.

        This cultural issue summarizes systemd quite accurately. The same people who had that attitude toward all desktop environment development are now taking over init and the OS in general. When a DE sucks, thats ok, no one uses DEs. I just need a program launcher, as long as it can light up a terminal, an emacs, and a web browser, like 99% of humanity I don't care how much they F up the GUI. Which I probably don't use anyway. Modern Gnome and KDE suck and everyone hates them or works around them thats why I use xmonad. Although awesome is a good WM also. And I used ratpoison for awhile.

        And now those type of people are in charge of the init and the OS. Its gonna be ugly, really ugly. "I don't like unity" "STFU I am in charge and you will like it" works just fine as a cultural model for a desktop environment nobody uses, but its a real mess applied to an OS. A culture shock.

        "What, we can't just product tie everything to death?"

        "How come these ignorant savages don't understand we are civilizing them?"

        "How dare they not like changing everything, why we do that in our shitty desktop environments all the time and look at the amazing buy in we've got from the public, almost 1% sometimes, how dare these guys owning 50% of the server market dare to have an opinion unlike my market leading opinion" (well market leading among shitty DEs that no one likes, anyway)

        • (Score: 2) by Magic Oddball on Thursday October 23 2014, @07:01AM

          by Magic Oddball (3847) on Thursday October 23 2014, @07:01AM (#109061) Journal

          I agree about the attitude, though to be fair, the KDE team hasn't had it -- in 2008 they were open about 4 not being ready, the distro maintainers replaced 3.5.x anyway, and the angry users were told off by the "new = better/shiny!!!" fans (not the KDE devs, turned out they were working to fix all of the issues users complained about).

          In any event, if "everybody" hates DEs, then why on Earth would the various versions of GNOME, KDE, Xfce, LXDE, and Enlightenment all be extremely well-known with tons of users, while the launcher/window-manager-only options obscure with relatively few users?

          It just seems a little silly to say (to almost quote you) "everyone hates DEs so I use [WM/launcher]," instead of just "I like launchers/WMs rather than DEs like most people, so I use [name]." That way could get a person curious enough to look into the option you prefer, rather than just being confused by your comment, given it's not obvious that you're being sarcastic or facetious (if you are).

          • (Score: 2) by tangomargarine on Friday October 24 2014, @03:40PM

            by tangomargarine (667) on Friday October 24 2014, @03:40PM (#109606)

            I took refuge in XFCE but dread the day that that gets taken over by the UX Windows guys as well.

            Inevitably the guardians of every awesome thing let down their guard for a moment and then the awesome thing gets shot in the head.

            --
            "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
        • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @03:49PM

          by Anonymous Coward on Thursday October 23 2014, @03:49PM (#109222)

          Ahh, the "desktop environment". It started with some people decided the didn't like how some applications had different buttons and other UI widgets, even though this has always been true on every OS to varying degrees. Every time they saw a Motif [wikipedia.org] window next to a Tk [wikipedia.org] or XForms [xforms-toolkit.org] application their inner neat-freak took over. Many of us wondered why they didn't just used GNUstep [wikipedia.org] given that they obviously liked the OSX^H^H^HNeXTSTEP style. (I assume their heads might have exploded if they ever had to use a program that used Xaw [wikipedia.org])

          They started several years of flame wars about how there was some need to unify various desktop features. Concepts like the "window manager" never really fit with their plans, and now with this systemd they have spread their poison of never learning from those that came before them into far more critical areas. This always seemed to me that these people were just n00bs, and might come around in time - that their serious design failures was just a sign of a severe lack of experience.

          I was forced to rethink that interpretation after reading about some of the GNOME 3 drama. The sudden removal of features and other bizarre design choices in GNOME 3 suddenly made sense, after reading [gnome.org] a [gnome.org]emails [gnome.org] by "user experience designer" for GNOME, Allan Day. (paycheck from RedHat, of course...)

          When approached by someone who was working on re-adding some lost features implemented as their new-at-the-time "GNOME Shell extension", some very revealing arguments were used against the idea: (emphasis mine)

          Facilitating the unrestricted use of extensions and themes by end users seems contrary to the central tenets of the GNOME 3 design. We’ve fought long and hard to give GNOME 3 a consistent visual appearance, to make it synonymous with a single user experience and to ensure that that experience is of a consistently high quality. A general purpose extensions and themes distribution system seems to threaten much of that.

          ...

          So, I would very much like to hear about how this web site will relate to our core design goals.

          When it was pointed out that Firefox seems to be able to handle people installing arbitrary extensions, the final shoe dropped: (again, emphasis is mine)

          The point is that it decreases our brand presence. That particular user might understand what it is that they are running, but the person who sees them using their machine or even sees their screenshots on the web will not. The question we have to ask ourselves is: how do we make sure that people recognise a GNOME install when they see one?

          Ahh. That explains a lot. These people are not interested in Free Software. They aren't interested in the engineering challenge of making usable tools that solve complex problems.

          No, the people pushing GNOME 3, systemd, are marketroids [catb.org]. They see software as a way to push branding, not tools, and the person that runs the software not as a customer, but as an advertisement. In the end, they just want their own little kingdom of "users" they can control and "monetize".

          As usual, The Onion was way ahead of us, and provided the perfect caricature of this style of asshole, "The Power Of Selling Out: Your Customers As Political Capital [youtube.com]" - it's worth watching just because it's the Onion with their usual high-quality humor.

          Myself, I'm going to just stick with my working, stable, and simple copy of e16 [enlightenment.org] that I have been using for about two decades now. It's fast, it works - and it isn't written by the marketing department.

      • (Score: 1) by SDRefugee on Thursday October 23 2014, @09:08AM

        by SDRefugee (4477) on Thursday October 23 2014, @09:08AM (#109091)

        ...Ubuntu 14.04 uses systemd...

        Umm.. I'm on Ubuntu 14.04 and unless systemd runs pid 1 as /sbin/init, you're wrong... I played around with Fedora 20 a while back and got a mouthful of systemd running as pid 1, plus the binary logging, and frankly it was a BIG PAIN IN THE ASS.. I'm more than content to stay with Ubuntu 14.04 until 2019 at which time I'll see which distros have eschewed going down the systemd rathole....

        --
        America should be proud of Edward Snowden, the hero, whether they know it or not..
        • (Score: 2) by bzipitidoo on Thursday October 23 2014, @02:57PM

          by bzipitidoo (4388) on Thursday October 23 2014, @02:57PM (#109198) Journal

          Seems you're mostly right. I'm on Lubuntu 14.04. Checking, I see "init" for pid1. Is that the old Sys V init process, or is something new reusing the name? I also see 2 systemd processes: "systemd-udevd" and "systemd-logind". Not all the way systemd, but has parts of it.

          I also see several "upstart" processes. Well, Ubuntu is developing Mir to replace X, rather than waiting on Wayland, so it makes complete sense that they would also go their own way on the init system, developing this Upstart system rather than jumping into systemd.

    • (Score: 3, Informative) by caseih on Wednesday October 22 2014, @04:05PM

      by caseih (2744) on Wednesday October 22 2014, @04:05PM (#108741)

      Pottering has written quite a bit about how systemd is architected and how it functions, and why. An early post is http://0pointer.net/blog/projects/systemd.html. [0pointer.net] One of the most brilliant things he came up with was using a very simple, and backwards-compatible API to let systemd create all the unix sockets services need all at once, before the daemons even run. This allows daemons to be started in parallel and synchronize themselves all without needing any dependency tracking logic, due to the kernel's own socket buffering and blocking reads. This is called "socket registration." Daemons do have to be modified to take advantage of this, but it is backwards compatible in the sense that the binaries run just fine on systems without systemd.

      There are also numerous tutorials and discussions you can google for. The best way, though is to fire up a virtual machine, install the latest Fedora or Arch, and start playing with it. Fedora has pretty good documentation online about systemd, setting up your own services, and using systemctl. Personally I like systemd but I hate having to type these stupid *ctl commands... they don't exactly roll off the fingers. Pottering likes these ctl commands.

      I disagree with the article that desktop linux is driving systemd for the most part. I consider it far more useful in a server environment, compared to the way things were done before. RH has been developing and driving systemd, and their focus is almost exclusively on servers and enterprise. Fedora's desktop focus is nice, but not that essential to Red Hat. As one example of why systemd benefits the server, over the years I've had numerous problems with server boot getting stuck on various race conditions involving NFS, network connectivity, and OpenLDAP. Back in RH 5 days there was a bug in nss_ldap that led to a chicken and egg problem when the server was an LDAP server and also needed to have local users come from LDAP. With systemd, if there was such a race condition, it would at least not block the rest of boot.

      One thing the article forgot to mention was that syslog is very much still in distros that use journald, though one could in theory run without it, but no one wants to that I know of. And some organizations are probably required to have it. RHEL7 has a standard syslog in the standard place. Journald is there, and you can use it, but it's not configured to be persistent.

      • (Score: 3, Insightful) by emg on Wednesday October 22 2014, @04:35PM

        by emg (3464) on Wednesday October 22 2014, @04:35PM (#108756)

        Our servers spend six minutes in the BIOS, then boot about thirty seconds later, starting up a few services which then run until they're rebooted.

        Switching to systemd is utterly useless to us, and probably actively harmful due to the extra complexity.

        • (Score: 2) by caseih on Wednesday October 22 2014, @05:46PM

          by caseih (2744) on Wednesday October 22 2014, @05:46PM (#108794)

          How is it actively harmful? What extra complexity is this? That's the whole problem with these discussions, and that's part of what the article was taking issue with. 1000-line shell scripts are plenty complex yet you are apparently comfortable with them despite their pitfalls and problems. Udev was already complex, and it works, and is essential on servers these days.

          • (Score: 3, Insightful) by emg on Wednesday October 22 2014, @06:21PM

            by emg (3464) on Wednesday October 22 2014, @06:21PM (#108813)

            Because, with init scripts, I know exactly what they're doing. With all that extra complexity of systemd trying to work out which services to start and stop... I don't.

            • (Score: 2) by digitalaudiorock on Wednesday October 22 2014, @09:33PM

              by digitalaudiorock (688) on Wednesday October 22 2014, @09:33PM (#108904) Journal

              Because, with init scripts, I know exactly what they're doing. With all that extra complexity of systemd trying to work out which services to start and stop... I don't.

              Absolutely. Not only that, even kernel developers have commented regarding their concern about the large size of the systemd executable itself. An face it, almost nothing about it has any place on a server (or for my money even on a desktop), and is bound to be a huge attack surface.

              I've been saying all along that it's just a matter of time before the realities of this come to light, like endless reboots of critical servers for all the systemd security fixes etc...that is at least the ones that are in fact found by the right people, and don't get you rooted outright.

              ...and it can't be said enough times...how are people not repulsed by those binary logs...I get ill just mentioning them.

          • (Score: 2) by FatPhil on Wednesday October 22 2014, @09:33PM

            by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Wednesday October 22 2014, @09:33PM (#108903) Homepage
            show me a 1000-line startup script, and I'll show you a package I'm less likely to let on my system than systemd.

            The fact that fucknuts can write shitty startup scripts for sysv init isn't an anti-sysv-init argument.
            --
            Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
            • (Score: 2) by caseih on Thursday October 23 2014, @03:25AM

              by caseih (2744) on Thursday October 23 2014, @03:25AM (#109003)

              Don't know of any that long, to be honest. But I do know of some that are over 300 lines long. And /etc/init.d/functions is 800 lines, but it's a library of course, and we assume it's bug free.

        • (Score: 2) by VLM on Wednesday October 22 2014, @06:55PM

          by VLM (445) Subscriber Badge on Wednesday October 22 2014, @06:55PM (#108829)

          My numbers are probably the other extreme of yours, given a truly huge NAS and a huge overpowered virtualization farm, I can light something up in seconds, its nuts how fast these things are now a days. The cost to me is labor troubleshooting something of infinite complexity that I'm assure won't break because their code never contains mistakes. But it has to be infinitely complex because otherwise it would be slow (WTF?)

          So freebsd it'll be. A little bitter about that. 20 years with linux and I wouldn't mind leaving so much if I wasn't being forced.

      • (Score: 2) by tangomargarine on Wednesday October 22 2014, @04:51PM

        by tangomargarine (667) on Wednesday October 22 2014, @04:51PM (#108769)

        I consider it far more useful in a server environment

        Oh look, they do exist!

        --
        "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
      • (Score: 2) by caseih on Wednesday October 22 2014, @05:31PM

        by caseih (2744) on Wednesday October 22 2014, @05:31PM (#108788)

        What's interesting to me is that while a group of people started the boycott systemd web site, these same folks have forked systemd into uselessd. This indicates to me that the idea behind systemd is completely sound, but they just want a more modular implementation. Somewhere in all the noise this gets lost. The idea of systemd is a good one, but you can certainly take issue with the implementation. Seems to validate many of the ideas that systemd has brought ot the table. Also uselessd is, apparently, aiming for more cross-platform portability so that they can bring a lot of this stuff to other operating systems like freebsd, which are lagging behind in the desktop environment area, as they lack a lot of the facilities needed to interact with hotplugged devices, at least in ways that Mac, Windows, and Gnome 3 users have come to expect.

        • (Score: 2) by FatPhil on Wednesday October 22 2014, @09:42PM

          by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Wednesday October 22 2014, @09:42PM (#108908) Homepage
          > The idea of systemd is a good one

          Nope.

          I'm fundamentally opposed to its design and architecture. Almost all of its *features* are to me *badness*, and I want less of them, not more.

          Upstart at least had good intentions, but was so much the opposite of sysvinit, it was unable to do some of the simplest things that sysvinit could do (and which I wanted/needed to do all the time every working day (as a kernel developer)). So I was violently opposed to upstart too.

          Pretty much every "solution" to a "problem" is to me a problem created where there was none before.
          --
          Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
        • (Score: 2) by sjames on Thursday October 23 2014, @03:14AM

          by sjames (2882) on Thursday October 23 2014, @03:14AM (#109000) Journal

          There is a good point. It really isn't that a more advanced init system is unwanted. It is a matter of poor design as far as I am concerned. It would be a real shame though if the mindless rush to systemd blocks a superior solution (perhaps uslessd) by being much too codependent to replace things in a sane and controlled manner, even for testing.

      • (Score: 2) by velex on Wednesday October 22 2014, @06:17PM

        by velex (2068) on Wednesday October 22 2014, @06:17PM (#108811) Journal

        If you like it, more power to you.

        I once wrote a start-up script that fired off a makefile with something like -j10 (parallel) I think it was that in turn started services, brought up network, etc many moons ago. Never published it though. Socket registration sounds kinda neat, but would I be correct that it's still prone to human error the same as editing a makefile since it requires the daemons to be modified?

        • (Score: 2) by FatPhil on Wednesday October 22 2014, @09:46PM

          by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Wednesday October 22 2014, @09:46PM (#108910) Homepage
          What happens to socket activation when the daemon itself can't activate? You've detected the existence of a service, not knowing it was just a facade, connected to it, and you find you actually conntected to a call centre in Indonesia that services everyone badly.
          --
          Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
          • (Score: 2) by caseih on Friday November 07 2014, @03:26AM

            by caseih (2744) on Friday November 07 2014, @03:26AM (#113736)

            This is a very stale story, but you probably would like an answer. The simple answer is that any failed service is handled by systemd and the socket is shut down, after trying to restart it according to instructions in the service config file. A service daemon could go into an endless loop and never respond on the socket. But that could happen on any init system.

            • (Score: 2) by FatPhil on Friday November 07 2014, @11:29PM

              by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Friday November 07 2014, @11:29PM (#113953) Homepage
              That's the view from the service's side. I don't care about that, I care about what the client who connected sees.

              What happens when the strange bloke puts blue tabs in your drink? It's fine, he can still order a taxi for you, and get you back to his place.
              Erm, no, that's not the side of the interface that I care about. It is outcome for the the party being wronged, and lied to, that is important.

              Yay, I just equated systemd with a rapist!
              --
              Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
              • (Score: 2) by caseih on Monday November 10 2014, @05:02AM

                by caseih (2744) on Monday November 10 2014, @05:02AM (#114409)

                Once a service is marked as failed, it's stopped. Though I made that clear. Plain and simple: service stopped == no socket, just as on any init system, even though sockets can be created and destroyed by systemd on behalf of daemons should they choose. There's a very small window of time that a client can connect to the socket before the service is fully initialized, but that can also happen with daemon-initialized sockets too. Clients will block until the service is up and running, or fails and the socket is shut down. So no, there's no issue here that you seem to be raising. These issues that you raise have in fact been thought about.

                • (Score: 2) by FatPhil on Monday November 10 2014, @08:10AM

                  by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Monday November 10 2014, @08:10AM (#114442) Homepage
                  And unsatisfactorily solved.
                  --
                  Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
                  • (Score: 2) by caseih on Monday November 10 2014, @03:58PM

                    by caseih (2744) on Monday November 10 2014, @03:58PM (#114525)

                    Fair enough. You asked for facts, I provided them.

                    • (Score: 2) by FatPhil on Monday November 10 2014, @08:02PM

                      by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Monday November 10 2014, @08:02PM (#114605) Homepage
                      Indeed you did. Thank you for that. What are desirable features to one person may be mis-design to another. I respect your right to consider them desirable features.
                      --
                      Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
      • (Score: 2) by sjames on Thursday October 23 2014, @03:04AM

        by sjames (2882) on Thursday October 23 2014, @03:04AM (#108998) Journal

        Well, that illustrates one common point very well. Your link is a 404. (just remove the ending .)

        But more to the point, nothing in that explains the IMHO horrible design/architecture choices everyone is complaining about.

        For one, why can't systemd allow itself to be other than pid 1? Why is logind made dependant on systemd when Ubuntu has demonstrated it need not be?

        It is revealing right before the FAQs section, they 'ask' that daemon writers don't do any of the best practices for daemons and then claim systemd works just fine if you do them all. So then why shouldn't I do them, to tick off *BSD users? To make sure it only works right in systemd? If it works just fine anyway, shouldn't I do all of those things to maximize inter-operation?!?Who the hell advocates against inter-operability in the Linux world?

    • (Score: 2) by FatPhil on Wednesday October 22 2014, @08:59PM

      by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Wednesday October 22 2014, @08:59PM (#108881) Homepage
      It's an *init* system which apparently cares about what directories users put their "documents", "music", "pictures", and "videos" in. And many other things too. That is so many millions of miles away from making any sense that my mind can do nothing but boggle about its [elided, so as to not bias you]. Init shouldn't need to know about much more than where / is, everything else can be worked out by performing initialisation based on configuration under /.

      And for those who demand evidence:

      http://cgit.freedesktop.org/systemd/systemd/tree/src/systemd/sd-path.h#n61

                      /* User resources */
                      SD_PATH_USER_DOCUMENTS,
                      SD_PATH_USER_MUSIC,
                      SD_PATH_USER_PICTURES,
                      SD_PATH_USER_VIDEOS,
      --
      Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
    • (Score: 1) by GeminiDomino on Thursday October 23 2014, @12:17PM

      by GeminiDomino (661) on Thursday October 23 2014, @12:17PM (#109133)

      Are you a South Park fan? Remember Cartman's "Dawson's Creek" Trapper Keeper?

      --
      "We've been attacked by the intelligent, educated segment of our culture"
  • (Score: 1, Funny) by Anonymous Coward on Wednesday October 22 2014, @02:16PM

    by Anonymous Coward on Wednesday October 22 2014, @02:16PM (#108674)

    Systemd is not my favorite flamewar topic.
    I'll be using the superior emacs on a KDE flavored android phone to create something GPL3.

    • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @03:30PM

      by Anonymous Coward on Wednesday October 22 2014, @03:30PM (#108717)

      Die emacs user!!!!
      NEOVIM shall rule the world!

    • (Score: 2) by isostatic on Wednesday October 22 2014, @07:42PM

      by isostatic (365) on Wednesday October 22 2014, @07:42PM (#108841) Journal

      I don't mind the emacs people, despite their inferior editor we can both happily get along. I can !apt-get purge emacs, they can ctrl-meta-although-shift-start-ru6 to get rid of vim.

      Systemd doesn't allow this choice though. What should have been an incremental, backwards compatible improvement to init (flag scripts with requirements, have init, or init2 parse the scripts in startup and fork them in parallel when ready) has turned into a monster.

      • (Score: 2) by FatPhil on Wednesday October 22 2014, @09:49PM

        by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Wednesday October 22 2014, @09:49PM (#108912) Homepage
        As a long-term emacs person, all I can say is...

        ...today I quite happily used vi (not even vim), for at least an hour, as that's the only terminal-based editor on my phone, and I wanted to do some either python or perl scripting - but I'm not going to tell you which one ;-)
        --
        Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
  • (Score: 2) by fadrian on Wednesday October 22 2014, @02:22PM

    by fadrian (3194) on Wednesday October 22 2014, @02:22PM (#108677) Homepage

    You have two ardent and vocal sides, roughly classified into an opponent/proponent dichotomy, neither of which have anything enlightening to say and both with their own unique set of misunderstandings that have memetically mutated into independent ideas that poison virtually every debate of this nature.

    So the article is saying that the beings discussing the issue are human. How... special.

    --
    That is all.
  • (Score: 2) by jasassin on Wednesday October 22 2014, @02:31PM

    by jasassin (3566) <jasassin@gmail.com> on Wednesday October 22 2014, @02:31PM (#108685) Homepage Journal

    Remember, Lennart Poettering, with great power comes great responsibility.

    --
    jasassin@gmail.com GPG Key ID: 0xE6462C68A9A3DB5A
    • (Score: 2) by caseih on Wednesday October 22 2014, @03:45PM

      by caseih (2744) on Wednesday October 22 2014, @03:45PM (#108725)

      Wish I had mod points to mod your post appropriately as funny. Well played, sir.

  • (Score: 2) by skullz on Wednesday October 22 2014, @02:31PM

    by skullz (2532) on Wednesday October 22 2014, @02:31PM (#108686)

    It is because the other guys are stupid and won't listen to reason.

  • (Score: 1, Insightful) by Anonymous Coward on Wednesday October 22 2014, @02:58PM

    by Anonymous Coward on Wednesday October 22 2014, @02:58PM (#108701)

    We like unix and how linux was.
    They "depreciate" everything we like by fiat and force us to use their crap by subverting key positions in distros etc.

    • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @05:43PM

      by Anonymous Coward on Wednesday October 22 2014, @05:43PM (#108793)

      The word is deprecate, you illiterate idiot.

      • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @06:06PM

        by Anonymous Coward on Wednesday October 22 2014, @06:06PM (#108803)

        Hmm seems you understood what was written (also it was in quotes), must be some literacy there.
        What did samual clemens say about the proper spelling of words?

        • (Score: 2) by tangomargarine on Friday October 24 2014, @03:46PM

          by tangomargarine (667) on Friday October 24 2014, @03:46PM (#109612)

          *Samuel Clemens

          --
          "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
      • (Score: 2) by melikamp on Wednesday October 22 2014, @08:54PM

        by melikamp (1886) on Wednesday October 22 2014, @08:54PM (#108875) Journal
        "Deprecate" is a verb, so the structure "the word is deprecate" is not well-formed grammatically. You probably wanted to say

        The word is "deprecate", you illiterate idiot.

        or

        The word is "deprecate," you illiterate idiot.

        if you are mentally stuck in the last century, but your own illiteracy got in the way.

        Captain Hindsight Out!

    • (Score: 2) by VLM on Wednesday October 22 2014, @06:47PM

      by VLM (445) Subscriber Badge on Wednesday October 22 2014, @06:47PM (#108827)

      You forgot to mention after all the screwing around and debugging and retraining, its doesn't do anything I care about any better, although it does make the things I actually do, harder and more complicated.

      But we're told its "better" and the only thing that matters is the desktop users that don't exist and don't care about what it does and who cares about you server guys and its doesn't matter if everything takes 10x longer to debug because our code never has bugs.

  • (Score: 1, Interesting) by Anonymous Coward on Wednesday October 22 2014, @03:10PM

    by Anonymous Coward on Wednesday October 22 2014, @03:10PM (#108707)

    That summary is bullshit. Anti people, including me, do not wish to cram sysVinit or any other init system down those who do not want them. Pro people insist on cramming down systemd on everyone, and this is reflected in the design of systemd itself, with its ever growing scope and hard dependencies in other packages.

    • (Score: 3, Insightful) by theluggage on Wednesday October 22 2014, @03:22PM

      by theluggage (1797) on Wednesday October 22 2014, @03:22PM (#108714)

      The anti-systemd people had me with "binary log files". It is as if a million nerds all cried out in horror, and were then silenced.

      • (Score: 2) by digitalaudiorock on Wednesday October 22 2014, @06:34PM

        by digitalaudiorock (688) on Wednesday October 22 2014, @06:34PM (#108819) Journal

        The anti-systemd people had me with "binary log files". It is as if a million nerds all cried out in horror, and were then silenced.

        + 10^100.

        The fact that there are Linux users out there who aren't completely horrified by a switch to what is effectively a Windows Event Log astounds me...even if is wasn't subject to database corruption that they don't consider to be a bug.

    • (Score: 2) by Sir Garlon on Wednesday October 22 2014, @03:52PM

      by Sir Garlon (1264) on Wednesday October 22 2014, @03:52PM (#108730)

      I am not even clear on what about the init system was broken or why they "fixed" it. I do know that I have spent more hours of frustration with systemd than I have with all other software in the past two years combined. If this is the cure, what could the disease have been, that the cure is not worse than the disease?

      --
      [Sir Garlon] is the marvellest knight that is now living, for he destroyeth many good knights, for he goeth invisible.
      • (Score: 4, Insightful) by emg on Wednesday October 22 2014, @04:31PM

        by emg (3464) on Wednesday October 22 2014, @04:31PM (#108752)

        Things like:

        1. Starting up services in parallel.
        2. Starting up services in the correct order.
        3. Starting up network-dependent services only when the network is available.
        4. Restarting services that crash.

        All of those are problematic with the current script-based init system. But systemd has bloated far beyond fixing those kind of problems, and its tendrils are infiltrating the entire operating system to the point where it can't easily be removed.

        That wouldn't be a big deal, except this is being developed by the same guy who wrote Pulseaudio, which took years to reach the point where it wasn't the first thing you uninstalled after a new Linux install. You won't have the option of uninstalling Systemd if it's as problematic as Pulseaudio was.

        • (Score: 2) by fnj on Wednesday October 22 2014, @05:49PM

          by fnj (1654) on Wednesday October 22 2014, @05:49PM (#108797)

          1. Couldn't care less.
          2. Yes, absolutely, SysVInit ordering is clumsy.
          3. This is just a part of dependence/ordering.
          4. Do. Not. Want.

          So one meaningful point out of four. For the rest, spot on.

        • (Score: 2) by Arik on Wednesday October 22 2014, @08:13PM

          by Arik (4543) on Wednesday October 22 2014, @08:13PM (#108854) Journal
          "1. Starting up services in parallel."

          Bad idea. Yes, you can boot slightly faster. The problems that you open yourself up to are not worth it (unless you are that guy that's spinning up and back down thousands of VMs on the server farm - if you are that guy then maybe you need this, but you should expect to install it custom instead of foisting it on the rest of us.)

          "2. Starting up services in the correct order."

          This is what my current init system does, and systemd (potentially) does not.

          "3. Starting up network-dependent services only when the network is available."

          See 2. above.

          "4. Restarting services that crash."

          Completely out of scope for the init system.

          Not saying it's a bad feature - it's just in the wrong place.
          --
          If laughter is the best medicine, who are the best doctors?
          • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @11:21PM

            by Anonymous Coward on Wednesday October 22 2014, @11:21PM (#108949)

            1. ...
            The problems that you open yourself up to

            Would you mind elaborating on those problems? I know nothing about the issue but would like to be informed.

            2. ...
            3. ...
            This is what my current init system does, and systemd (potentially) does not.

            And the current init system potentially does not as well. Everything can potentially fuck up. Potentially != Definitely. Objecting due to preferences is fine, but at least be honest that its only your personal preference.

            "4. Restarting services that crash."
            Completely out of scope for the init system.

            Again, pardon my ignorance, but isn't starting, and thus restarting services a function of initialization? Or is that the function of something initialized by the, er, initialization system?

            • (Score: 2) by pe1rxq on Thursday October 23 2014, @12:11AM

              by pe1rxq (844) on Thursday October 23 2014, @12:11AM (#108961) Homepage

              You don't just restart crashing services.
              You analyze the crash and debug the application.
              By blindly restarting services you are just begging for a corrupted system.

        • (Score: 2) by Bot on Wednesday October 22 2014, @10:35PM

          by Bot (3902) on Wednesday October 22 2014, @10:35PM (#108932) Journal

          Those who don't dig gentoo or slack: voidlinux uses runit, eudev, seems a good starting point for people knowledgeable enough who want to try life without systemd and without sysvinit. I am configuring one laptop with void and E19, but in general replacing sysvinit still makes little sense for most desktop people, who wait more to get a dhcp lease anyway.

          --
          Account abandoned.
          • (Score: 2) by Magic Oddball on Thursday October 23 2014, @07:29AM

            by Magic Oddball (3847) on Thursday October 23 2014, @07:29AM (#109071) Journal

            Thanks for posting that -- Void Linux sounds like it might be just what I've been hoping to find as a road out of the systemd mess that doesn't require me to jump headfirst into an unfamiliar OS or compile binaries in my sleep.

      • (Score: 3, Informative) by Thexalon on Wednesday October 22 2014, @04:41PM

        by Thexalon (636) on Wednesday October 22 2014, @04:41PM (#108759)

        My understanding of the problems systemd were trying to solve:
        1. Linux used to flash a text screen showing a bunch of stuff starting up. (This was really really handy when something went wrong, because you'd see a "[ FAIL ]" in red where it started going wrong and a semi-informative message about what it was that failed to work properly.) This is usually a non-issue for geeks, but for marketroids and non-geeks. Systemd-based systems start with a graphical screen immediately and you have to Ctrl-Alt-F1 to see what's going on behind the scenes.

        2. Sysvinit and related systems run start scripts in a specific order, where later scripts may depend on the earlier scripts having already run. That means that you have to set up the script order correctly to ensure that you don't run into a situation where daemon A relies on daemon B, but B hasn't started by the time that A is trying to startup, so A's startup fails. This means that administrators and distro creators have to pay a lot of attention to dependencies and craft the order by hand.

        3. The problem described in (2) means that you can't start things up in parallel, which potentially unnecessarily slows booting.

        I have no problem with the concept of trying to solve these problems. I have lots of problems with the way that systemd goes about solving them: There are simpler approaches that would have provided the same benefits without coupling all userspace daemons to the init system.

        The simplest solution would involve a config file with a start command, a stop command, a check command to determine if the daemon is fully operational, a list of runtime dependencies for each daemon, and a list of whether this should be running at a particular runlevel. That would be enough to allow a smart init system to find all the daemons that should be started up for the specified runlevel, figure out what order they needed to be starting in, and then firing off commands in that order, waiting if necessary for earlier daemons to finish doing whatever they needed to do first.

        --
        The only thing that stops a bad guy with a compiler is a good guy with a compiler.
        • (Score: 2) by tangomargarine on Wednesday October 22 2014, @04:56PM

          by tangomargarine (667) on Wednesday October 22 2014, @04:56PM (#108774)

          Linux used to flash a text screen showing a bunch of stuff starting up.

          I miss that old text scroll. We had a Windows 95 machine when I was growing up that took 2 or 3 minutes to scroll through all that stuff, too. I've been sadly unable to reenable the bootup messages in Ubuntu/Mint these days.

          --
          "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
        • (Score: 1) by cesarb on Wednesday October 22 2014, @05:36PM

          by cesarb (1224) on Wednesday October 22 2014, @05:36PM (#108789) Journal

          > Systemd-based systems start with a graphical screen immediately

          That's plymouth, not systemd. It existed before systemd was created, and can be found for instance in upstart-using releases of Ubuntu.

          If you want to see the text messages, press Esc; the graphical screen will go away and show the messages underneath it. To get back to the graphical logo, press Esc again.

    • (Score: 3, Insightful) by tangomargarine on Wednesday October 22 2014, @04:04PM

      by tangomargarine (667) on Wednesday October 22 2014, @04:04PM (#108739)

      Reading the article through, I got the impression that what the writer was saying between the lines was really, "The opponents have some decent points but they should stop being dicks and just give in to the inevitable conquering of all by systemd."

      I'll at least give him credit for making it a lot harder to detect that bias than most things I read.

      I think comparisons to The Borg would not be too far off the mark. And in my book, "make it more Windows-like" makes me deeply suspicious and I want to see an actual list of concrete advantages other than "our retarded focus group likes the GUI."

      --
      "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
  • (Score: 3, Interesting) by Thexalon on Wednesday October 22 2014, @03:25PM

    by Thexalon (636) on Wednesday October 22 2014, @03:25PM (#108715)

    From what I understand, the fundamental problem is this: Unix philosophy states that each program should have one job to do, and do it well. Systemd's philosophy involves integrating as many things as possible into systemd to improve bootup speed and eliminate "glue" like the /etc/init.d scripts. Those two ideas are fundamentally incompatible.

    Personally, I'm firmly on the old-school Unix position: Whatever gains exist in the 3 seconds of boot time are lost in the additional complexity. For servers, I'm not going to reboot them often enough for the boot time to matter much, and for desktops the slowest part of the system is always between the chair and keyboard so the speed of the computer doesn't matter much. For me, it's also a lot easier to take apart and tinker with a shell script than to re-kejigger daemons when they're tightly coupled to the init system (e.g. adding an extra command-line argument to a database server startup). Old init systems do one thing well, namely figure out which scripts need to be run, in what order, and run them in that order.

    There is also a significant concern that stuff that can only start up under systemd locks them out of other Unixes like BSD that don't have the kernel features systemd relies on, splintering the userspace of Unix-like OS's. Anyone around in the 1980's knows how bad an idea [wikipedia.org] that is.

    --
    The only thing that stops a bad guy with a compiler is a good guy with a compiler.
    • (Score: 2) by carguy on Wednesday October 22 2014, @04:10PM

      by carguy (568) Subscriber Badge on Wednesday October 22 2014, @04:10PM (#108743)

      Here is a naive idea from an occasional Linux user (I know next to nothing about details of Linux internals):

      It seems that boot time is one of the major claims of the systemd proponents(?) I'm imagining a "compiler" (using the term loosely) that could compile all the normal startup scripts and other processes. Could this be the best of both worlds?

          * If you want a standard system, just use the compiled binary that comes with the distribution.
          * If you want to do something custom, tweak the various scripts as traditionally done with Unix until things are working, then you have the option to run the "compiler" if you need extra speed in booting.

    • (Score: 3, Interesting) by choose another one on Wednesday October 22 2014, @07:40PM

      by choose another one (515) Subscriber Badge on Wednesday October 22 2014, @07:40PM (#108840)

      The choice you crave - in this case the infinite choice and flexibility of configuration by script - leads to what you say is a bad idea. Choice inexorably leads to splintering and fragmentation, because without enforced standardisation, sooner or later people will disagree. http://xkcd.com/927/ [xkcd.com]

      Even before systemd, Unix init systems (let alone the entirety of userspace) were fragmented. SysV init vs. BSD, SMF on Solaris, OpenRC, launchd, busybox on embedded and a bunch of others. You can't even be sure what shell /bin/sh will be on Linux, let alone across all Unixen, or if init even uses shell scripts (even before systemd). All of which means you are going to have to rewrite init scripts for every platform. Systemd might even become standard enough on Linux distributions that it _reduces_ fragmentation.

      Old init systems do one thing well, namely figure out which scripts need to be run, in what order, and run them in that order.

      Doing one thing well is fine, but the issues come when it doesn't do something else. There is rarely only one requirement. Merely running scripts synchronously with no standardisation around how a service is determined to be running (or not) makes dependency management problematic, and responding to events.

      splintering the userspace of Unix-like OS's. Anyone around in the 1980's knows how bad an idea that is.

      Yes but somehow there always turns out to be a "good reason" to fork and fragment. The existence of Linux itself fragmented userspace at the GUI / window-system level because it didn't have Motif, which was the Unix-like OS standard widget library (even the FSF said it was GPL compatible, under the "normally ships with" system library clause).

      Efforts to find a free replacement for Motif on Linux were further fragmented, because one of them wasn't GPL-free enough for the FSF (despite being less restrictive than Motif, and fulfilling exactly the same role as a system library). Splintering and fragmentation were held, by some, to be more desirable than resolving that issue, and even though the issue was completely resolved some 10 or 15(?) years ago, the fragmentation of Linux desktop userspace persists.

      People can always find a good reason for a bad idea when they want to. And yes, Systemd may yet prove to be another example of that too...

      • (Score: 2) by Thexalon on Wednesday October 22 2014, @08:41PM

        by Thexalon (636) on Wednesday October 22 2014, @08:41PM (#108867)

        The existence of Linux itself fragmented userspace at the GUI / window-system level because it didn't have Motif

        This doesn't impose much of a cost on users, because you can run KDE/Qt-based applications in Gnome and Gnome/GTK+ applications in KDE, or either under another window manager. Where this has caused problems (e.g. clipboard management and sound systems), there are solutions now in place that make this basically a non-issue.

        Even before systemd, Unix init systems (let alone the entirety of userspace) were fragmented. SysV init vs. BSD, SMF on Solaris, OpenRC, launchd, busybox on embedded and a bunch of others. You can't even be sure what shell /bin/sh will be on Linux, let alone across all Unixen, or if init even uses shell scripts (even before systemd). All of which means you are going to have to rewrite init scripts for every platform.

        That's a point in my favor, not yours: Yes, you have to write a ~50-line init script to make a daemon run on new kind of init system, but that 50-line init script allows you to run the same daemon using different init systems on different kinds of machines. For example, on a desktop, I'd want my init system to get me up and running quickly while potentially still loading things in the background while the login screen is showing, whereas on a server I'd want to make it so that startup is completely unattended and a sysadmin will get notified if something goes wrong, whereas on an embedded system the priority is minimizing unnecessary RAM usage. Those different goals could well mean different init systems in use, which I don't see as a problem.

        The cost of writing that init script is low compared to, say, not being able to run two daemons concurrently on a box because daemon A expects to be started by init system X and daemon B expects to be started by init system Y. Also lowering the cost dramatically is the fact that init scripts are often quite similar across the board for a particular distribution: What you need to start or stop mysqld versus postmaster, for example, is not wildly different on the same platform. That also makes it easier to port an application to a non-Unix platform should you ever want to do that.

        The software design principle in play here is the idea of a simple glue or shim layer. By making it so that the pieces don't absolutely have to link together in a particular way, you allow the pieces to change and allow whoever is assembling them to do things you'd never even thought of. For example, if you're working with Legos, you have to stick the bricks together in a way that the bumps and the slots line up with right angles everywhere, whereas if you're working with actual bricks the mortar allows you to stick the bricks together to make round buildings.

        My concern isn't differences between systems, it's differences between systems that prevent them from running the same software. And coupling user-space applications to an init system that relies on a particular kernel is doing exactly that.

        --
        The only thing that stops a bad guy with a compiler is a good guy with a compiler.
  • (Score: 2) by halcyon1234 on Wednesday October 22 2014, @03:55PM

    by halcyon1234 (1082) on Wednesday October 22 2014, @03:55PM (#108732)

    From what I understand, it's a war on how to properly pronounce it. Half the people pronounce it like two words "System D", and the other half like a past-tense verb "Systemed". That's the disagreement, the cause of the whole war, right?

    I haven't seen this many bodies since serving in the Gif Wars (or should I say Compuserving)

    --
    Original Submission [thedailywtf.com]
  • (Score: 1, Flamebait) by Lagg on Wednesday October 22 2014, @05:00PM

    by Lagg (105) on Wednesday October 22 2014, @05:00PM (#108776) Homepage Journal

    My comment history pretty much promptly defeats that stupid overgeneralization as well as my video in which I walk you through how fucking simple the journal format is.

    and no I won't read on, being associated with something called "useless" is pretty apt for this author. As if he has any place to talk about people saying things that are enlightening.

    --
    http://lagg.me [lagg.me] 🗿
    • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @05:37PM

      by Anonymous Coward on Wednesday October 22 2014, @05:37PM (#108790)

      video in which I walk you through how fucking simple the journal format is.

      Link? "Fucking simple" is the guy who replaces a simple text file with a corruptible binary format requiring special tools to parse out.

      • (Score: 2) by Lagg on Wednesday October 22 2014, @06:43PM

        by Lagg (105) on Wednesday October 22 2014, @06:43PM (#108823) Homepage Journal

        Just because it's binary doesn't mean it's automatically worthless and bad. If that was the case then the "traditional unix" people are circle jerking furiously over would have been bad too. They conveniently forget things like the weird usage of dbm when it really shouldn't have been and also forget that the unix kernel itself was just about as far away as one gets from what they consider the unix philosophy these days (which should really be called the "plan9 philosophy"). So the whole "text is always better and binary is ebil" thing is not just ignorant it's getting pretty close to double standards.

        Plus, if there was a better way besides something that would have lead to horrific overhead like XML or JSON (and good luck deserializing larger logs going that route) to store grouped metadata fields I'm sure they'd have done it. And it's not as corruptible as people seem to think it is by simple virtue of being an append-only file format. Even my simple tool would probably handle and kind of "normal use" corruption (like an unfinished append, which is pretty hard to do in practice because as I recall journald buffers and writes out 64 bit aligned object blocks and only when they're ready to be written, I think it's an atomic operation too). The video is here [youtube.com]. I spend a minute or two glorifying the morons like the author by complaining about them but it's only at the beginning and I get to the code quickly.

        All that said, I am still not totally comfortable with not having plain text log files but the sheer usefulness and various ways of filtering log entries that grep simply couldn't replicate helps.

        --
        http://lagg.me [lagg.me] 🗿
        • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @07:59PM

          by Anonymous Coward on Wednesday October 22 2014, @07:59PM (#108848)

          systemd is not automatically worthless or bad.

          The fact that systemd's goal is to be a requirement for every linux system in the future is automatically worthless and bad. There are many who don't want to use your program. That's all they ask. To not be forced. Apparently that is just too unreasonable, what a bunch of moronic neckbeards they are for not agreeing with you.

          And that's why the debate goes nowhere.

          • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @11:12PM

            by Anonymous Coward on Wednesday October 22 2014, @11:12PM (#108947)

            There are many who don't want to use your program. That's all they ask. To not be forced.

            Nobody is forcing anybody to use [distro "forcing" systemd on them]. And this is why the concept of "forking" exists - people are unhappy with the "official" version, so they take the last version they thought was suitable and improve it how they think it should be improved. So why is that not happening here?

            • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @12:12AM

              by Anonymous Coward on Thursday October 23 2014, @12:12AM (#108963)

              You're not getting it. Several components of systemd are attemping to change (and by and large succeeding) the way that hardware and the kernel interact, thus making it very difficult to avoid systemd. (see udev, kdbus-udev integration)

              This is not why the concept of forking exists. This is why rm -rf exists. I don't want to write my own systemd-alike to handle overly complex interfaces, I just don't want to use it or any of its crap. Forking is not happening because no one (of the opponents) wants systemd, they are happy with what they have. And systemd is trying to make itself completely unavoidable as it bloats from an init system to some vauge collection of possibly related building blocks for an OS. The systemd opponents don't want a modified systemd, we want !(systemd).

              • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @03:31AM

                by Anonymous Coward on Thursday October 23 2014, @03:31AM (#109005)

                Are you serious? This article was posted on "uselessd.darknedgy.net", which is the homepage of uselessd, the fork of systemd...

              • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @04:58AM

                by Anonymous Coward on Thursday October 23 2014, @04:58AM (#109016)

                The systemd opponents don't want a modified systemd, we want !(systemd).

                So write and maintain one. Fork the distros that are "forcing" systemd on you. They're writing software for their own purposes, not for you specifically. If you don't like the direction they're taking, fork it. That is the whole reason forking is a thing.

                • (Score: 2) by tangomargarine on Friday October 24 2014, @03:52PM

                  by tangomargarine (667) on Friday October 24 2014, @03:52PM (#109616)

                  THEY should have forked in the first place instead of forcing it down distro's throats if they want their dependencies to keep working.

                  I don't control what decisions the distro maintainers make, so it's partially their fault, too.

                  --
                  "Is that really true?" "I just spent the last hour telling you to think for yourself! Didn't you hear anything I said?"
        • (Score: 2) by Arik on Wednesday October 22 2014, @08:08PM

          by Arik (4543) on Wednesday October 22 2014, @08:08PM (#108853) Journal
          "Just because it's binary doesn't mean it's automatically worthless and bad. "

          That may be the only true thing you have said yet. And it's only true enough for argument.

          The fact is that text logs are accessible using my choice of thousands of good programs, and anything I need to do with them there will be an easy tool to do it with, no matter how obscure the task. It also means that if they are open when the FS goes down I can still read them. Those are very substantial advantages that I rely on regularly.

          With a binary log you take both of those things away. Things I need and use. And what's the advantage in return? You use slightly less disk space? You really think I will give up being able to use my own tools for that?

          No way. You throw around the word 'moron' and I am trying to refrain from following suit, but you clearly must think your 'users' are morons for real.
          --
          If laughter is the best medicine, who are the best doctors?
          • (Score: 2) by digitalaudiorock on Wednesday October 22 2014, @09:46PM

            by digitalaudiorock (688) on Wednesday October 22 2014, @09:46PM (#108909) Journal

            The fact is that text logs are accessible using my choice of thousands of good programs, and anything I need to do with them there will be an easy tool to do it with, no matter how obscure the task. It also means that if they are open when the FS goes down I can still read them. Those are very substantial advantages that I rely on regularly.

            Absolutely. The "text isn't necessarily better" argument I keep hearing is BS, because logs are for when things are BROKEN and I don't need something making them hard to get to...so text is way way better for log files and always has been...period. Binary is great for a lot of things, like databases...but nobody with any sense wants critical logs in a database or anything requiring specific tools.

            Anyone who hasn't learned that from years of the Windows Event Log astonishes me.

            I want to be able to read my logs with nothing else required except the file system itself. Speaking of which, with systemd, if you boot to a system rescue CD, how the hell do you get to those things? I'm assuming you'd obviously need journalctl, but would that necessarily even work? I never plan on needing to know mind you. People can keep this bullshit as far as I'm concerned.

        • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @08:22PM

          by Anonymous Coward on Wednesday October 22 2014, @08:22PM (#108860)

          Thanks for the link and awesome beard :P

          I was going to post this comment in a base64 encoded binary format, you could easily write a parser to decipher it so any objections would clearly be bullshit. You wouldn't even have to write a parser, just copy from your browser, decode the base64 and run strings on the result. I think it's a winner of an idea. Also, when people talk to me in the street, rather than engage in verbal communication using a language we mutually understand I'm going to start beeping at them like artoo-fucking-dee-two. They can just record me and write a parser back to natural language that runs on their phone or something. Or not...

          All that said, I am still not totally comfortable with not having plain text log files but the sheer usefulness and various ways of filtering log entries that grep simply couldn't replicate helps.

          I never had any problem retrieving and reformatting relevant info from syslog using standard unix text tools and pipes.

          journalctl -o short --no-pager | grep [...]

          May as well duplicate the entire journal as a file and work on that, which makes the binary format useless. One of the arguments for binary logs was to prevent log fiddling but this is demonstrable nonsense. The format they created is not robust and file corruption EWONTFIX'd so log tampering is as easy as overwriting the LP string in the file header. Most tools would refuse to parse this file and most admins would simply assume the file was corrupt. Uh-huh?

          • (Score: 2) by Lagg on Wednesday October 22 2014, @08:44PM

            by Lagg (105) on Wednesday October 22 2014, @08:44PM (#108870) Homepage Journal

            You know what? Fair enough. I don't like things that can't be cat'd right into my terminal either. I don't like even a single level of redirection before I get to the raw lines but until we figure out a better way though this is pretty much what is going to happen but I'm sure that if anyone came up with a plain text and reliable format (and no, normal log lines aren't reliable. It needs reliably parse-able metadata) it'd be considered. It needs to be said though that these log files aren't in any way as corruption-prone as people think. See my note about atomic object writes and such. Deliberate corruption by overwriting the signature is kind of a moot point because if someone can do that they can also do rm -f /var/log/whatever.log

            Still, you have a fair point otherwise and that's the core problem here. This isn't some kind of moral "pro-life/anti-life" thing. It's an argument about the merits of systemd. It has nothing to do with either side being "enlightening". It's that everyone focuses on character attacks of Lennart and co. and never even bothers to make any kind of technical argument. This can even be seen in the responses to my parent post. No counter arguments, no technical responses (besides yours). Just whining about me calling people that deserve to be called morons morons. Ironic that the AC is the one who is giving some semblance of technical discourse.

            --
            http://lagg.me [lagg.me] 🗿
            • (Score: 1, Insightful) by Anonymous Coward on Wednesday October 22 2014, @09:07PM

              by Anonymous Coward on Wednesday October 22 2014, @09:07PM (#108886)

              Deliberate corruption by overwriting the signature is kind of a moot point because if someone can do that they can also do rm -f /var/log/whatever.log

              They can but then the administrator would know something was going on. More likely they edit the file but with journald they don't even need to go to that trouble.

              It's that everyone focuses on character attacks of Lennart and co. and never even bothers to make any kind of technical argument.

              I do get where you're coming from but also the personal attacks. Take the email app on my phone, I swipe to delete an email. That mail is then placed into the deleted items folder where I have to press and hold each email before pressing the trash icon at the top. There's no bulk delete or purge all option and I get hundreds of emails every day, each having to be individually deleted twice. I could probably find the source code and customise it, but then I'd have to maintain it. It's much easier for me to just say...

              "The authors of the stock android email app are morons with no friends who don't get email".

              Venting like that is a cheap shot but gives me some satisfaction. There's a lot of that going on in the systemd debate, people encounter a problem, decide they don't like systemd and vent. I fear attitude of the systemd developers and it's ever increasing scope makes the situation worse.

              • (Score: 2) by Lagg on Wednesday October 22 2014, @09:51PM

                by Lagg (105) on Wednesday October 22 2014, @09:51PM (#108916) Homepage Journal

                I understand that both from the perspective of a user and developer (in fact that is actually going to be the topic of another C3 video) but usually the two intertwine on some level when it comes to systemd so there's really no reason to be like that. You can both rant/vent and give technical arguments or critiques at the same time. Hell I do it all the time. Yes I agree that the maintainer attitude right now is terrible and aforementioned stupid attacks make them even less open to critiques, however I get the feeling that even if everyone was perfectly civil some of the maintainers would still be asshats. See for example the kernel command line thread. The arrogance and entitlement shown by them is disgusting. I feel that perhaps we'd get somewhere with this if most of them were ejected. Empathy needs to be shown on both sides of this.

                As far as the log files go that was just an example, they could also do a sed -i type deal and subtly change log lines to something unexpected. The point is that if they have write access you're already pretty much done for and the file format is irrelevant after that.

                --
                http://lagg.me [lagg.me] 🗿
        • (Score: 1) by tomek on Thursday October 23 2014, @10:53AM

          by tomek (3281) on Thursday October 23 2014, @10:53AM (#109108)

          I actually clicked on your youtube link. Its a half an hour video. It starts with a rant followed by a comment about lacking documentation and a an information about a macro you're unable to explain because you're crap at explaining. I stopped there.

          Questions:

          1. If you're crap at explaining why are you linking to your "explanation"? How is it helping?
          2. I never needed a half an hour introduction to understand syslog, How is the new format better?
          3. The log format is the easiest thing to fix (see passwd file format), what about the other issues?

  • (Score: 5, Insightful) by RedBear on Wednesday October 22 2014, @06:36PM

    by RedBear (1734) on Wednesday October 22 2014, @06:36PM (#108820)

    As a semi-outsider to the Linux world (I used to run various flavors of Linux as my main OS back near the turn of the millenium), I've been paying only peripheral attention to Linux and the BSDs for the last decade. Every year or so I'll download a couple of the major Linux distros or a BSD and play around with them for a while, just to see how things are progressing. My overall impression has always been that the BSD community has always placed more emphasis on correctness and conservative stability while the Linux crowd has always placed more emphasis on evolving quickly and trying new things. But either way, most of the people involved in the *nix world prize some things above all: Freedom, and stability, and respect for people who really know what they're doing. Systemd seems to fly directly in the face of all those things.

    The original Unix Philosophy, for instance, is a recognition by old highly experienced neckbeard software and hardware engineers back in early computing times that Murphy's Law is an absolute truth of the universe. The more complex a system is, the more likely it is to create problems instead of solving them. The more tightly integrated a system is, the more difficult it will be to adapt the system to different kinds of tasks. And of course, the more difficult it will be to maintain and troubleshoot. The systemd supporters don't seem to want to acknowledge or have any respect for any of these universal truths. This immediately alienates a good portion of the entire *nix community.

    Out of all the disagreements I've seen in a decade and a half of observing the Linux world, this is the first one I've seen that has a high probability of producing a real permanent split in the community, and possibly even a true forking of the Linux kernel if the situation gets bad enough. I've been very surprised at the lackadaisical attitude of Linus toward the rapid encroachment of systemd, and I think that's disappointed a lot of the community as well. One thing is for certain, in the entirety of the last decade I haven't seen nearly as many people talking about switching from Linux to BSD as I have just in the last several months. The schism that is being created by the systemd battle should be worrisome to both sides. I fall on the side of the anti-systemd crowd for the very simple reason that all they are asking for is the same freedom of choice that brought most of them into the *nix world in the first place, whereas the pro-systemd crowd just seems to want everyone to be forced to use their brand new toy. No one has yet said anything on this issue that has convinced me I need to know more than that.

    Choice is everything, or Linux becomes a joke to its most die-hard users and supporters. If systemd becomes an unremovable parasite on the kernel, even if it's a symbiotic parasite, those who do not wish to use it will simply flee the Linux world entirely. Can the Linux community survive losing a third or even a quarter of its developers over the next couple of years? It's an interesting question.

    --
    ¯\_ʕ◔.◔ʔ_/¯ LOL. I dunno. I'm just a bear.
    ... Peace out. Got bear stuff to do. 彡ʕ⌐■.■ʔ
    • (Score: 2) by choose another one on Wednesday October 22 2014, @08:56PM

      by choose another one (515) Subscriber Badge on Wednesday October 22 2014, @08:56PM (#108877)

      Can the Linux community survive losing a third or even a quarter of its developers over the next couple of years? It's an interesting question.

      Just a thought - what fraction of "Linux" developers are now actually Android or embedded developers ? Linux would survive and prosper in those spaces even if all the Desktop guys forked each other into oblivion. Maybe that is why Linus doesn't care ?

      • (Score: 2) by Magic Oddball on Thursday October 23 2014, @07:54AM

        by Magic Oddball (3847) on Thursday October 23 2014, @07:54AM (#109079) Journal

        I don't know about the percentages, but in "Just For Fun" (Linus' official biography, which he helped write) he makes a few rants that boil down to him wanting to be able to almost play a hobbyist developer's role in terms of working on the kernel while it's fun and not needing to pay attention to anything else, as he hates the pressure of feeling like his opinions or actions will affect others.

        IIRC at one point he talks about an earlier schism in the community where he refused to take sides or even investigate the argument because of that attitude, so I suspect that's why he's staying silent on the systemd issue.

    • (Score: 2) by digitalaudiorock on Wednesday October 22 2014, @10:46PM

      by digitalaudiorock (688) on Wednesday October 22 2014, @10:46PM (#108936) Journal

      As much as has been said on this, I'm not sure anything I've seen sums it up better than this. Thanks.

    • (Score: 1) by goody on Thursday October 23 2014, @02:29AM

      by goody (2135) on Thursday October 23 2014, @02:29AM (#108991)

      A saying that I heard years ago kind of supports what's going on right now: "BSD users love Unix, Linux users just hate Windows." BSD users love and embrace the Unix philosophy, the core tenets. Linux users in general don't have this same sort of love of the core tenets of Unix, they're more interested in killing Windows and getting Linux on the desktop so the proverbial grandmother can run it.

  • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @09:28PM

    by Anonymous Coward on Wednesday October 22 2014, @09:28PM (#108898)

    I wish I could even think of something that hasn't already been said before wrt all this nonsense, but it seems almost impossible. I'm kind of glad I never got involved in the first place.

    I don't understand the actual point of the uselessd author's post, but it is pretty well-written. I'll take it for a fact that he's right about this huge debate mostly being political, since his post also appears to be entirely political as well (unless I missed something, and I read the whole thing). I'm glad he's writing uselessd though.

    The thing that gets me fired up most of all is that I read the bug trackers for various things, and I see how willing a lot of software authors seem to be to just "go along with it". Even the famous raster of Enlightenment fame seems to be embracing the concept of switching to systemd without any fight, despite the fact that he noted that it's just a dependency that's required for getting input to work in Wayland (which I don't understand; I'm not into that).

    I don't like systemd because I don't need it. The thing I really hate is that otherwise pre-systemd portable software is now becoming unportable due to it. But that's why I'm rooting for the authors of these things like uselessd and the GSoC projects, etc, that hope to make it less viral and dependent on Linux than it is.

    I wonder if anyone is working on a pulseaudio shim, too? I've never thought to look.

  • (Score: 0) by Anonymous Coward on Wednesday October 22 2014, @11:55PM

    by Anonymous Coward on Wednesday October 22 2014, @11:55PM (#108956)

    Well, a few things...

    1. Nobody is going to flee to the BSDs. When the initially try, they'll find it uncomfortably different from GNU/Linux, and also the same reason they won't be able to swap out systemd with something else will affect their BSD usage: the software won't work without systemd, being integrated with it. GNOME is just a start.

    2. The highly experienced neckbeard software engineers back from the early computing times? That's kinda of a contradiction. See, more data is being processed, more functions called, more processes per CPU are being done in my HTC Android right now, than it ever did in ALL the machines of the yesteryear you speak of, COMBINED. Those times dealt with bytes and maybe kilobytes. That's the environment that was so constricted and limited that, naturally, sprang the UNIX philosophies.

    Systemd may be a badly written software. But it is a great idea. Either it will mature over time, get rid of the nonsense and become better, or something else will crop up to satisfy the ever increasing need for an operating system with all its elements unified in a single, working API, instead of "10 competing standards" ala the XKCD comic.

    • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @06:55AM

      by Anonymous Coward on Thursday October 23 2014, @06:55AM (#109057)

      The computational power and memory all have increased. The brain power of humans hasn't. But the latter is what is relevant for avoiding bugs. At the same time, the security risk has also massively increased, thanks to the internet. So the importance of being bug free has also increased.

      Unix philosophy is not about how to fit programs into small space. Unix philosophy is about writing software you can rely on, yet at the same time is flexible enough to allow you to do whatever you want to do.

      Indeed, all that flexibility in Unix costs memory. Not for the single program, but for the system as a whole. Fortunately memory is not a scarce resource these days (although some programmers use it in a so excessive way that it again gets scarce, but that's a different story; I don't think systemd is in that camp, C code rarely is).

      • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @11:04AM

        by Anonymous Coward on Thursday October 23 2014, @11:04AM (#109112)

        Then in that same context, systemd IS following the UNIX philosophy being a set of small units that work together. Sure, you can't swap out systemd-timed with ntpd because the former is integrated, so systemd is kinda monolithic, but it is highly modular which is very important in the context of what you're saying: small units, each doing its own job properly.

        Or you're objecting the fact that you can't swap out individual components with non-systemd ones? Because that has nothing to do with limited brainpower problem you speak of.

        • (Score: 1) by GeminiDomino on Thursday October 23 2014, @12:44PM

          by GeminiDomino (661) on Thursday October 23 2014, @12:44PM (#109135)

          small units, each doing its own job properly

          [Formal proof needed]

          --
          "We've been attacked by the intelligent, educated segment of our culture"
    • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @07:51AM

      by Anonymous Coward on Thursday October 23 2014, @07:51AM (#109078)

      The trouble with BSDs is not that they are different. They are not THAT different.

      The problem is hardware support. But for VMs this doesn't matter much. Just for that reason I plan on trying PC-BSD soon.

  • (Score: 0) by Anonymous Coward on Thursday October 23 2014, @05:41AM

    by Anonymous Coward on Thursday October 23 2014, @05:41AM (#109030)

    Redhat is doomed. They won't be around in five years.