Linux, politics, and other interesting things
This post isn’t really about technology, I’ll cover the technology briefly skip to the next section if you aren’t interested in Linux programming or system administration.
I’ve been using the Systemd init system for a long time, I first tested it in 2010 . I use Systemd on most of my systems that run Debian/Wheezy (which means most of the Linux systems I run which aren’t embedded systems). Currently the only systems where I’m not running Systemd are some systems on which I don’t have console access, while Systemd works reasonably well it wasn’t a standard init system for Debian/Wheezy so I don’t run it everywhere. That said I haven’t had any problems with Systemd in Wheezy, so I might have been too paranoid.
I recently wrote a blog post about systemd, just some basic information on how to use it and why it’s not a big deal . I’ve been playing with Systemd for almost 5 years and using it in production for almost 2 years and it’s performed well. The most serious bug I’ve found in systemd is Bug #774153 which causes a Wheezy->Jessie upgrade to hang until you run “systemctl daemon-reexec” .
I know that some people have had problems with systemd, but any piece of significant software will cause problems for some people, there are bugs in all software that is complex enough to be useful. However the fact that it has worked so well for me on so many systems suggests that it’s not going to cause huge problems, it should be covered in the routine testing that is needed for a significant deployment of any new version of a distribution.
I’ve been using Debian for a long time. The transitions from libc4 to libc5 and then libc6 were complex but didn’t break much. The use of devfs in Debian caused some issues and then the removal of devfs caused other issues. The introduction of udev probably caused problems for some people too. Doing major updates to Debian systems isn’t something that is new or which will necessarily cause significant problems, I don’t think that the change to systemd by default compares to changing from a.out binaries to ELF binaries (which required replacing all shared objects and executables).
Recently the Debian technical committee determined that Systemd was the best choice for the default init system in Debian/Jessie (the next release of Debian which will come out soon). Decisions about which programs should be in the default install are made periodically and it’s usually not a big deal. Even when the choice is between options that directly involve the user (such as the KDE and GNOME desktop environments) it’s not really a big deal because you can just install a non-default option.
One of the strengths of Debian has always been the fact that any Debian Developer (DD) can just add any new package to the archive if they maintain it to a suitable technical standard and if copyright and all other relevant laws are respected. Any DD who doesn’t like any of the current init systems can just package a new one and upload it. Obviously the default option will get more testing, so the non-default options will need more testing by the maintainer. This is particularly difficult for programs that have significant interaction with other parts of the system, I’ve had difficulties with this over the course of 14 years of SE Linux development but I’ve also found that it’s not an impossible problem to solve.
It’s generally accepted that making demands of other people’s volunteer work is a bad thing, which to some extent is a reasonable position. There is a problem when this is taken to extremes, Debian has over 1000 developers who have to work together so sometimes it’s a question of who gets to do the extra work to make the parts of the distribution fit together. The issue of who gets to do the work is often based on what parts are the defaults or most commonly used options. For my work on SE Linux I often have to do a lot of extra work because it’s not part of the default install and I have to make my requests for changes to other packages be as small and simple as possible.
So part of the decision to make Systemd be the default init is essentially a decision to impose slightly more development effort on the people who maintain SysVInit if they are to provide the same level of support – of course given the lack of overall development on SysVInit the level of support provided may decrease. It also means slightly less development effort for the people who maintain Systemd as developers of daemon packages MUST make them work with it. Another part of this issue is the fact that DDs who maintain daemon packages need to maintain init.d scripts (for SysVInit) and systemd scripts, presumably most DDs will have a preference for one init system and do less testing for the other one. Therefore the choice of systemd as the default means that slightly less developer effort will go into init.d scripts. On average this will slightly increase the amount of sysadmin effort that will be required to run systems with SysVInit as the scripts will on average be less well tested. This isn’t going to be a problem in the short term as the current scripts are working reasonably well, but over the course of years bugs may creep in and a proposed solution to this is to have SysVInit scripts generated from systemd config files.
We did have a long debate within Debian about the issue of default init systems and many Debian Developers disagree about this. But there is a big difference between volunteers debating about their work and external people who don’t contribute but believe that they are entitled to tell us what to do. Especially when the non-contributors abuse the people who do the work.
In a world filled with reasonable people who aren’t assholes there wouldn’t be any more reaction to this than there has been to decisions such as which desktop environment should be the default (which has caused some debate but nothing serious). The issue of which desktop environment (or which version of a desktop environment) to support has a significant affect on users that can’t be avoided, I could understand people being a little upset about that. But the init system isn’t something that most users will notice – apart from the boot time.
For some reason the men in the Linux community who hate women the most seem to have taken a dislike to systemd. I understand that being “conservative” might mean not wanting changes to software as well as not wanting changes to inequality in society but even so this surprised me. My last blog post about systemd has probably set a personal record for the amount of misogynistic and homophobic abuse I received in the comments. More gender and sexuality related abuse than I usually receive when posting about the issues of gender and sexuality in the context of the FOSS community! For the record this doesn’t bother me, when I get such abuse I’m just going to write more about the topic in question.
While the issue of which init system to use by default in Debian was being discussed we had a lot of hostility from unimportant people who for some reason thought that they might get their way by being abusive and threatening people. As expected that didn’t give the result they desired, but it did result in a small trend towards people who are less concerned about the reactions of users taking on development work related to init systems.
The next thing that they did was to announce a “fork” of Debian. Forking software means maintaining a separate version due to a serious disagreement about how it should be maintained. Doing that requires a significant amount of work in compiling all the source code and testing the results. The sensible option would be to just maintain a separate repository of modified packages as has been done many times before. One of the most well known repositories was the Debian Multimedia repository, it was controversial due to flouting legal issues (the developer produced code that was legal where they lived) and due to confusion among users. But it demonstrated that you can make a repository containing many modified packages. In my work on SE Linux I’ve always had a repository of packages containing changes that haven’t been accepted into Debian, which included changes to SysVInit in about 2001.
The latest news on the fork-Debian front seems to be the call for donations . Apparently most of the money that was spent went to accounting fees and buying a laptop for a developer. The amount of money involved is fairly small, Forbes has an article about how awful people can use “controversy” to get crowd-funding windfalls .
MikeeUSA is an evil person who hates systemd . This isn’t any sort of evidence that systemd is great (I’m sure that evil people make reasonable choices about software on occasion). But it is a significant factor in support for non-systemd variants of Debian (and other Linux distributions). Decent people don’t want to be associated with people like MikeeUSA, the fact that the anti-systemd people seem happy to associate with him isn’t going to help their cause.
Forking Debian is not the correct technical solution to any problem you might have with a few packages. Filing bug reports and possibly forking those packages in an external repository is the right thing to do.
Sending homophobic and sexist abuse is going to make you as popular as the GamerGate and GodHatesAmerica.com people. It’s not going to convince anyone to change their mind about technical decisions.
Abusing volunteers who might consider donating some of their time to projects that you like is generally a bad idea. If you abuse them enough you might get them to volunteer less of their time, but the most likely result is that they just don’t volunteer on anything associated with you.
Abusing people who write technical blog posts isn’t going to convince them that they made an error. Abuse is evidence of the absence of technical errors.