Two articles in one, which have to do with systemd

A news in phoronix comment that it continues the debian debate about what to do with your initialization system. For a long time there have been voices calling for renewal and to get rid of the old sysvinit. And within those voices the rivalry between those who support systemd, those who support upstart and (very, very few) those who support openrc ……… ..and are not willing to support more than one.

The discussion is fierce and is like putting together a book of several volumes (they go through the 2500 messages, and this bug was opened only 2 months ago!!!). systemd is supported by several distros that migrated successfully (Fedora, Arch, OpenSuse, etc.), but its followers regret that Debian has to maintain versions for the FreeBSD kernel, where systemd is not ported (nor does Lennart intend to port it) . What is ported to FreeBSD is OpenRC (actually port to Debian KFreeBSD was achieved), but only Gentoo and its derivatives use it (except Sabayon which uses systemd). And Upstart, has the advantage of coming from the downstream (Ubuntu and its derivatives and also Chrome OS), but it falls short compared to systemd. And if to this we add the discussion outside the Debian lists, among which is Lennart's opinion y patrick lauer answering (to Lennart), any flamewar is small in comparison.

It was also news in phoronix that there are already opinions within the Debian technical committee. On one side is Ian Jackson (Debian daemon maintainer) who is in favor of Upstart. He considers it for its minimalism, for being better integrated into a demon's code, for its ease of packaging, have a less arrogant community (according to him) and for being more ready to be chosen for Jessie (OpenRC is not yet). It also indicates that disadvantages such as lack of IPv6 and UDP socket activation or multiple socket activation do not require difficult structural decisions and can therefore be more easily resolved.

And on the other hand is Russ allbery who is in favor of systemd: First you think that OpenRC is the most conservative alternative and that you don't even want to bother solving bugs like the lack of integration with events at the kernel level or its dependency more on shell scripts than on declarative syntax. In terms of service management, the activation of sockets stands out (not only initialize them but do it in parallel), the integration of the status of the daemon (more complete than in upstart) and in-depth security. Also remember that Debian already uses systemd (especially logind) for certain applications like udev and gnome (whose version 3.8 is already in testing) and you already have the migration plan in mind.

And as for the question of portability, systemd fans on they say «There is no portable software, there is only software that was ported.I mean, either the Debian carriers for kFreeBSD and Hurd make it work, or they're going to shit. And this second option weighs heavily since (according to popcon) only 0,09% of Debian users have the FreeBSD kernel installed.

Meanwhile, KWin developer Martin Gräßlin is following the discussion on Debian that I had told them about, and he loves Russ Allbery's comparison between systemd and upstart and comments on his google + account that he intends to integrate systemd to Plasma, and incidentally that any environment that uses Wayland is passed to systemd. In particular you want to use socket activation to start your KWin session.

Christian Loosli asks that KDE not have a high dependency. Martin replies that KDE has a high dependency especially on QT, but fuck it, they only want it for features that are neither in OpenRC nor in Upstart, but more importantly, because they want KDE to depend on kdbus (your own d-bus service explorer a project that seeks to integrate d-bus to the kernel) which already depends on systemd. It also says don't worry about the initialization system because this will be independent of whether you use OpenRC or SysVInit (In fact, Gentoo uses systemd even though its init is OpenRC. So "there should be no problem with Debian"). Then it is Eric Hameleers (member of the Slackware coreteam) who complains that they want to choose technologies that are only for Linux (again the issue of portability). Martin asks you to read the post of the false myths that Lennart wrote. That he trusts Martin.

What do you think of the panorama? The next news I have to do is an article on systemd, I'm going to do it as a football game story.

systemd fans homer simpson

The content of the article adheres to our principles of editorial ethics. To report an error click here!.

31 comments, leave yours

Leave a Comment

Your email address will not be published. Required fields are marked with *



  1. Responsible for the data: Miguel Ángel Gatón
  2. Purpose of the data: Control SPAM, comment management.
  3. Legitimation: Your consent
  4. Communication of the data: The data will not be communicated to third parties except by legal obligation.
  5. Data storage: Database hosted by Occentus Networks (EU)
  6. Rights: At any time you can limit, recover and delete your information.

  1.   vicky said

    Not counting that three of those who vote in favor of upstart, two are canonical employees and one is former employee.

    Between Upstar / Systemd and Mir / Waylad there seems to be a controversial competition between software created by Canonical and that promoted by Red Hat (among others)

  2.   Rolo said

    I think it's fine that the implementation of systemd or upstart or OpenRC in Debian is discussed. It is evident that sysvinit is reaching the end of a great cycle and in short I think there will be no problem to continue using sysvinit in KFreeBSD and Hurd while one of these other alternatives is implemented in Linux.
    At the end of the day Hurd has to first manage to support sata disks, usb, other partitions that are not ext2, sound support, 64-bit architectures, among other things. so supporting systemd or upstart is at the bottom of a long list of priorities. I think KFreeBSD is going to have less trouble supporting it.

    on the subject of systemd or upstart it seems that systemd has a certain advantage
    technique and that upstart has the luck to look tied to ubuntu and nobody else.

  3.   cr0t0 said

    Very good the article diazepan about the current status of systemd in Debian and to find out a bit, since there are several distros that use this initializer (Siduction, which is DEBIAN, is already implementing it). As a debianite one sees him from outside, from time to time gossiping in the forums of the brave ARCH.
    The implementation and even the concept seems complex to me: Is it a framework? Among several things, it allows a faster load of the system when executing processes in parallel?
    Being the mother of several distro and being associated with STABILITY, they should take all the time in the world to discuss it (or at least before Jessie freezes)

    PS: If your name isn't IAN, won't you work on the DEBIAN project? xd

  4.   poor taku said

    Being debian (and derivatives) the most widely used distribution, portability and independence is crucial, but I'm just a simple GNU-programming learner so I can't be part of these technical discussions yet.
    The next ep is expected, reading this was entertaining

  5.   Fernando said

    Interesting article.
    I just wanted to comment on two things. The first is that if systemd is released under the terms of the LGPL license, Poettering's order not to port it to non-Linux systems is worthless. Anyone can do it because the license allows it.
    Regarding KDBUS, it is not a KDE project but rather the implementation of DBUS in the Linux kernel.

    I believe that, having been an Ubuntu user from 6.06 to 10.10 and today being an Arch user since late December 2010, systemd is superior to Upstart. The transition from sysvinit is easy and learning to handle systemd is quite easy.

    1.    pandev92 said

      What you should do is stop doing the webón! I still remember the great rubbish of pulseaudio and the suffering it caused us linuxers the first years, and that it was finally dedicated to porting systemd to other unixes.

    2.    mirage said

      The problem with porting systemd to other systems is that the infrastructure does not exist in those kernes since or there are not the necessary components or similar components that can replace them in practice. Creating systemd to kfreebsd would require porting other components as well, mainly cgroups. In other words, according to Lenart, it is like trying to land an airplane in a country where there are no airports. in my opinion they should use openCR for the other kernels and leave systemd on linux, it cannot be that 99.1% of users are condemned to use a 0.9% inferior solution. Not to mention that kfreebsd and hurd already use different configurations than the linux version

  6.   zipr said

    […] The FreeBSD kernel, where systemd is not ported (nor will it be ported, by express order of Lennart Poettering) […]

    Do you have any source / link for that order? Because it seems to me that Systemd is GNU, and as long as it remains free it is nobody to order anything. I think what he said is that he wouldn't do that job himself, that he only works for GNU / Linux, that's why what you've written in the news story sounds really bad, like Poettering was a monster or something.

    1.    diazepam said

      I've already corrected it, but yes. Lennart says that it is not feasible to port systemd to BSD and that they will not accept patches to make it portable on either BSD or Hurd (it's in the comments).

  7.   Christopher said

    Well, as long as it is transparent to the user, I don't care what bothers me is that flames involve everyone, it seems that everyone marries a technology and does not see which one is better

  8.   Christopher said

    Too bad the fight waited I ended up being technical when choosing.

  9.   AdrianArroyoStreet said

    In my opinion you should depend as little as possible on one or the other implementation. I say this for KDE. They should keep dependencies to a minimum. And regarding Debian, maybe Upstart is easier to implement since Ubuntu already has it and the number of potential bugs would be reduced; and if necessary, systemd could always be executed as indicated in the article.

    1.    mirage said

      what happens is that what is wanted is not "dependency" as such. what you want is to make use of some characteristics that by grace or misfortune. only systemd provides and upstart for design reasons it is difficult to implement the same in a similar way (for example it already has sockets activation, but very limited and does not allow parallel activation of processes that is supposed to be the reason for being of the sockets) so it is not that it depends, it is that you want to make the best piece of softare possible and to this day or there are alternatives or even projects that provide the same. for example gnome. gnome does not officially depend on logind. gnome relies on certain dbus interfaces that are only provided by logind or console kit at present. consolekit is deprecated and abandoned and loggind depends on systemd. but nobody prevents a third party from developing a daemon or mechanism to provide the same kdbus interfaces to use gnome, this is how in openBSD they have nome 3.10 even though no bsd has dbus or systemd.

  10.   Alex said

    Personally, after the move from Arch to systemd, I noticed a great improvement in terms of startup speed

  11.   Tesla said

    I think these discussions are for developers. The truth is that I have no idea what differences there are between one and the other, and I think that for ordinary users it is not too relevant. I have come across systemd on Manjaro and do not find a performance improvement over Debian or a worse performance. So I do not know…

    Anyway, let's hope the best is done, I don't know what option is. LOL

    A greeting!

  12.   peterczech said

    I am in favor of systemd since maintaining kfreebsd that is practically not used by Debian users seems illogical to me.

    1.    peterczech said

      Also because of how easy systemd is to use and the general improvements it brings: D.

    2.    pandev92 said

      With that way of thinking, no company should support Linux xD

  13.   geronimo said

    I do not use Debian but I hope they will opt for systemd, more than anything for its ease of use "besides knowing something about it" ^^

  14.   toyerd24 said

    What led Ubuntu to choose Upstart and not take the step with systemd, which in the opinion of many is considered better? Regards.

    1.    vicky said

      Upstart is canonical technology (they love to use their own software) and predates systemd by 4 years I think.

  15.   atlas7jean said

    Systemd to death xD

  16.   gallux said

    Debian has two essential focuses: stability and universality, in fact, that's where its support for the most varied architectures and hurd and freebsd projects comes from. My opinion is that they should defer the decision to the subsequent stable release and dedicate themselves to OpenRC, which would avoid these discussions.

  17.   roader said

    From my file I say, systemd seems to be the strongest alternative, anyway, with the resources that they should have, because they do not consider forking it

    "Being a fan of OpenRc is not explained, he regrets"

  18.   khourt said

    I vote for the chronicle for the next one!
    I am not a user who knows the terms used, but even for people we know little it is interesting and important to see how a discussion like this develops.

    What will predominate, democracy, the interests of some company, or the fact of choosing the one that at the moment is best suited for Debian's objectives?

  19.   Blue skull said

    From what I have been able to read, and as a programmer, I can say that systemd is much more advanced than upstart.

    It manages the initialization of services only when they are needed (reducing the loading time of the system), it also tries to substitute scripts for definitions (goodbye to the .sh slowdowns) and also there is the advantage of cgroups, with which the administrator The system has full control of everything that is released.

    If there is something that I hate very much it is mixing politics with technology ..., if there are technical reasons to use something, it should not be allowed in any way that the debates are manipulated by issues of commercial interests or simple selfishness, only reasons should prevail techniques, and in that systemd is way ahead of upstart in my view.

  20.   thorzan said

    I don't understand much of these advanced topics, but the discussion is passionate and well told. We want more!

  21.   elav said

    If you ask me: Systemd. Arch has shown me that it is worth using and that it is much faster than its counterparts.

  22.   rainbow_fly said

    Well .. limiting it to speed (since I don't have much technical knowledge)
    Ubuntu is supposed to use Upstart right? Ubuntu's startup was always very slow, it reminded me of windows sometimes, with the disk in the middle and with a half-old computer, both of them took to start long enough for me to go pour me soda and come back ... instead using Archlinux with Systemd the computer turns on faster than anyone I've seen in the rest of my life xD (I'm not exaggerating), every time someone wants to use it they don't expect it to start so fast hahahaha

  23.   mirage said

    kdebus is not from kde, it is a project of the freedesktop foundation whose objective is to integrate d-bus into the kernel to solve some deficiencies that linux has at the process management level. But what you want is to make several pieces of software designed to work together to improve security and sandboxing.

    1.    diazepam said

      it is already corrected.