The future with Systemd

[confuseling]

I'm really not trying to suggest that companies or government agencies don't have self-serving motives, and aren't often corrupt.

Just that it doesn't (in this narrow sphere) matter. If the most horrible autocratic regimes in the world write good GPL code, we could still use it. The key is trusting the people overseeing the projects incorporating the code - and since there's a powerful political movement to keep corporate control of open source projects dilute, even if a project is taken over another would take its place (whether forked or not).

It's much like the scientific process in many respects. Money corrupts scientific publications, and that's a problem. But the scientific method still works - either results are repeatable, or they aren't, and theories eventually stand or fall on their merit. No company can falsify scientific results forever, as long as independent scientists exist.

What would happen if Canonical hired every Debian developer (or even a majority, and voted the rest out)? Do you really believe everyone would just roll over and give up?

It just seems like an incredibly fatalistic way of looking at things.

[llivv]

What would happen if Canonical hired every Debian developer (or even a majority, and voted the rest out)? Do you really believe everyone would just roll over and give up?

Why would they, DOJ did the majority of that dirty work for them. Now all they and other corp with a stake need to do is wait for everyone to forget what they've done.

It just seems like an incredibly fatalistic way of looking at things.

indeed.

[Randicus]

I did not post this, because it might be considered a person attack:
Your view can be summed up as blindly trust in faith.
But now that you have posted this:

It just seems like an incredibly fatalistic way of looking at things.

it is time to be honest. Your view is naivete bordering on bury the head in sand.

Corporations cannot take control.
They already have.

It must have advantages, otherwise systems would not be adopting it.
Extreme naivete. Politics and money could not be factors. Oh no. Have faith. They would never force something on us if it was not in our best interest.

If a few people do not like it, they will create a fork.
When GUIs and many other applications are systemd-dependent, there will be nowhere to fork to. This is not simply replacing the init system. It is a fundamental change to the core of the system that integrates several processes. A "fork" would be a new operating system. Hopefully UNIX-based.

What would happen if Canonical hired every Debian developer (or even a majority, and voted the rest out)? Do you really believe everyone would just roll over and give up?

Another example of naivety. Who are everyone? The developers who were kicked out or the users? The ex-developers would have no say in future development of the distribution they are no longer part of. The users would only have two options: continue using the system and be quiet, or switch to a different system.

My assessment is that this thread has long since reached the end of its usefulness as a venue for discussion. I shall be waiting for the next rare discussion thread with something interesting.

[confuseling]

I did not post this, because it might be considered a person attack:
Your view can be summed up as blindly trust in faith.

Nope, not exactly. You've presented zero historical examples of companies destroying GPL software, while in common use, without it forking. You've given no theoretical explanation of how they would.

My position is based on reason and experience, yours on... Well, what exactly?

But now that you have posted this:

It just seems like an incredibly fatalistic way of looking at things.

it is time to be honest. Your view is naivete bordering on bury the head in sand.

Corporations cannot take control.
They already have.

You're calling me naive? Good luck finding an operating system (never mind computers to run it on) with no corporate involvement. I said corporate control is kept dilute - it is, companies balance and check each other - which is presumably why you straw-manned me against the clear intent of my post instead of using a direct quote.

It must have advantages, otherwise systems would not be adopting it.
Extreme naivete. Politics and money could not be factors.

What exactly are you suggesting? That RedHat bribed the TC to make their competitor adopt the same terrible software they're using?

I didn't say there's no politics nor money in Free software, just that it doesn't make a blind bit of difference, in the long run, to the code. That's basically the entire point of the GPL.

Oh no. Have faith. They would never force something on us if it was not in our best interest.

If a few people do not like it, they will create a fork.
When GUIs and many other applications are systemd-dependent, there will be nowhere to fork to. This is not simply replacing the init system. It is a fundamental change to the core of the system that integrates several processes. A "fork" would be a new operating system. Hopefully UNIX-based.

This seems to be your central point - problem is it's unintelligible. GNOME depends on the external interfaces of logind. What's stopping other people writing their own version, or forking it? (free clue: people are writing their own versions, and forking it. The answer's "nothing")

What would happen if Canonical hired every Debian developer (or even a majority, and voted the rest out)? Do you really believe everyone would just roll over and give up?

Another example of naivety. Who are everyone? The developers who were kicked out or the users? The ex-developers would have no say in future development of the distribution they are no longer part of. The users would only have two options: continue using the system and be quiet, or switch to a different system.

A spectacular example of question dodging. If Debian announced they'd been 'bought out' by Ubuntu, would people give up, or keep developing a fork?

In short, what can corporate involvement in a distro / project actually threaten, aside from poaching individual devs (their choice) and the name (we'll cope)? Historical examples only please, no paranoid handwaving.

My assessment is that this thread has long since reached the end of its usefulness as a venue for discussion. I shall be waiting for the next rare discussion thread with something interesting.

You don't even use Debian any more, right? Since we're doing "home truths", what are you actually here for, apart from trolling?

[mardybear]

Being relatively new to the Linux world, <10 years, and with no exposure to old school Unix i found this article informative, especially the background information regarding the Unix philosophy. Much of this may have already been mentioned in this convoluted thread but if anyone's interested:

Choose Your Side on the Linux Divide
http://www.infoworld.com/d/data-center/ ... ide-248950

Edit: Sorry this article was also linked on the same page:
http://www.infoworld.com/d/data-center/ ... pse-248436

[buntunub]

Here we go again with this, "we are switching to XYZ monolithic app and if you don't like it, don't use Debian!!!" kick. The Debian leaders seem to taking a liking to authoritarian measures now. "You will do it or not use Debian!"...

And then these threads start. And then some in the community actually are clueless as to why people are upset. And then the flame wars start....

Have a look into the history of UNIX, Linux, and the human aging process and maybe some light can be shed.

GNOME 3, and Systemd are/were being forced upon this community, hell or high water. That's just a fact. The users have no say whatsoever. The decision being made from upon high dictatorial style, and then those who make the decision wonder why people in the DEBIAN world get upset about that. In my view, these types of decisions should be with community consenus, majority vote, to stay within the spirit of the social contract, which again in my view, was meant to bind us together as a community with transparency and open dialogue.

While Systemd does have some advantages over SysV, there should never ever be one monolithic system level app that controls everything, and to which everything else is dependant. This violates everything that makes Linux what it is, and the reason for it's widespread adoption. My suggestion would be to use SystemD to create a new OS, while retaining the bits of it that can be integrated into SysV scripts for Debian.

[mardybear]

buntunub wrote:
Here we go again with this, "we are switching to XYZ monolithic app and if you don't like it, don't use Debian!!!" kick. The Debian leaders seem to taking a liking to authoritarian measures now. "You will do it or not use Debian!"...

Beginning to see this myself even though i haven't been using Debian too long, similar to recent changes with Firefox and gnome3.

I know Debian pretty much lets you install it anyway you want but gnome3 is still the 'default'. As a universal OS me don't think current gnome should have ever been chosen as default. Many of the problem threads on this forum wouldn't even exist had Debian choosen a leaner default WM that does not require hardware acceleration and exposes new users to metapackage confusion. Just my 2 cents.

When i found Debian i was convinced i would never change - this is my OS choice for life. Maybe time to try slackware or a BSD derivative. Wonder if the old saying 'once you go slack you never go back' is still true?

[koanhead]

Well, I'm not entirely certain any more what the 'default' status means. I guess that it's the one that the main installer installs if you don't tell it to install something else.

If people don't want systemd forced upon them, then I think what's needed is for a subset of such people to put together a debian-installer that installs a version of Debian without the software to which they object, whether that's GNOME3, systemd, or other. This doesn't seem all that hard to me, but I admit I don't know much about the debian-installer and its customization. If others are willing to help, then I'll look into such customization and we can get together and make an installer or installers that doesn't install things the user doesn't want. If someone is interested in helping with this, please PM me.

I have other things to do, so this is not a project I will undertake on my own.

[keithpeter]

When i found Debian i was convinced i would never change - this is my OS choice for life. Maybe time to try slackware or a BSD derivative. Wonder if the old saying 'once you go slack you never go back' is still true?

I have Slackware 14.1 (the stable release) with KDE on this Dell i5 laptop with 3gb of ram and an atheros based pce-i wifi card. Seems fine and stable. Slackbuilds easy enough to do. Updates applied easily. I'm keeping Debian on the X60 (the one I carry round, half the weight of the Dell which I use like a desktop). Both of these are recycled laptops, the X60 is 8 years old and the Dell dates from 2011.

Remember that the slackers get round the absence of automatic dependency resolution to some extent by having a fairly huge default install with tonnes of libraries. 8.5Gb is the default, rising to 14Gb with a dozen or so 'slackbuilds'. You can leave out package sets at install time (e.g. kde) but then there could be issues with updates expecting libraries. There is a choice of DE/WM including light ones. No Gnome though.

The slackers are grappling with the issues raised in this thread. Have a look at threads in the slackware sub-forum on linuxquestions.org about systemd. Lots of ideas being kicked around and tried out (i.e. people are compiling and installing modified udev/eudev on 'current' installations and reporting back).

The Salix Linux distribution is based on Slackware but dispenses with the 'kitchen sink' approach of Slackware itself by using a small core and having a package manager with dependency resolution. They do xfce4 and Openbox variants. I tried the xfce4 one to see if I was going to be able to cope with Slackware and it worked fine. Just thinking about machine specs here.

This whole situation looks to me like Jugaad vs Waterfall as a development model. I'm sort of agnostic but I am confident the jugaadu in the Linux world will fork code to satisfy their preferences. We'll have choices I'm sure.

[edbarx]

If people don't want systemd forced upon them, then I think what's needed is for a subset of such people to put together a debian-installer that installs a version of Debian without the software to which they object, whether that's GNOME3, systemd, or other. This doesn't seem all that hard to me, but I admit I don't know much about the debian-installer and its customization. If others are willing to help, then I'll look into such customization and we can get together and make an installer or installers that doesn't install things the user doesn't want.

Some points are worthy of being made about what you claim.

First, systemd has nothing to do with OS installers whichever they happen to be. Second, getting rid of systemd is difficult as systemd is not some paint program for infants BUT a complete overhaul of the system initialisation base. It also replaces, INIT, the manager of the whole system. So, replacing systemd, necessarily means, modifying anything that has to do with systemd. There is a workaround by providing an interface layers of .so files, so that, processes still see systemd with init, but this is still difficult to achieve, as it requires many many hours of development and it is not a one time job.

Note, that installers don't control what I install on my system. At least, this has been the case with init. I use debootstrap and chroot to install my system, as I find that a beautiful experience of exercising power over my system. Obviously, no graphical interfaces are used, although, ironically, I can do that from within ANY desktop/window manager I deem fit.

systemd is a shackle even for CLI wizards. That is the reason behind the outcry of so many. Don't imagine ripe GNU/Linux users are afraid of rigid installers, as that is a non-issue for them!

[koanhead]

If people don't want systemd forced upon them, then I think what's needed is for a subset of such people to put together a debian-installer that installs a version of Debian without the software to which they object...

Some points are worthy of being made about what you claim.

First, systemd has nothing to do with OS installers whichever they happen to be. There is a workaround by providing an interface layers of .so files, so that, processes still see systemd with init, but this is still difficult to achieve, as it requires many many hours of development and it is not a one time job.

Are you claiming that d-i jessie does not install systemd?
That is the specific problem I propose to address. Anyone who can wield a package manager can deal in their own way with dependencies that want to pull in systemd. Not who can use synaptic also knows how to use debootstrap. I'm offering my assistance to those who want it and are willing to help build a partial solution.

Note, that installers don't control what I install on my system. At least, this has been the case with init. I use debootstrap and chroot to install my system, as I find that a beautiful experience of exercising power over my system. Obviously, no graphical interfaces are used, although, ironically, I can do that from within ANY desktop/window manager I deem fit.

That's fine for you, but this suggestion wasn't directed to you personally. Even if you were on board, I'd want more interested folks before rearranging my schedule to work on such a thing. There are others who do, in fact, use d-i to install Debian. I presumed that at least some of those people are both reading this thread and interested in a version of d-i which doesn't install systemd.

systemd is a shackle even for CLI wizards. That is the reason behind the outcry of so many. Don't imagine ripe GNU/Linux users are afraid of rigid installers, as that is a non-issue for them!

Believe it or not, I've somehow managed to learn a few simple commands in the last 18 years ☺. I have read the various threads on debian-user and -devel, and I think I'm pretty conversant with the various reasons for opposition to the systemd-as-default-init decision.
Systemd is not a big problem for me. It's not installed on the systems where I don't want it. I had to expend a small amount of effort to make that happen. I'm willing to expend a little more to enable people to use d-i to install jessie without systemd, if enough people want that. So far it looks like there is no interest.

Second, getting rid of systemd is difficult as systemd is not some paint program for infants BUT a complete overhaul of the system initialisation base. It also replaces, INIT, the manager of the whole system. So, replacing systemd, necessarily means, modifying anything that has to do with systemd.

Getting rid of systemd is as easy as not installing it in the first place. Sysvinit still works. As far as I know, runit, daemontools, and OpenRC all do also. Sure, you can't have GNOME without some systemd components. There's lots of depends on systemd stuff. I figure if anyone knows enough to object to systemd, they ought to know enough to use their package manager to keep unwanted packages off their systems.
I'm not a DD, I'm not on tech-ctte. I don't have any real influence on Debian decision making. All I can do is try to make a productive suggestion, and I did that. If I come up with something better then I'll suggest that instead.

[buntunub]

Well I think the interest is definitely there. Thus all the threads about it and the constant daily email list spam about it. From a thread above, it also appears there is a serious effort underway in the Slack community, so good news there. My concern though is with those folks like myself who use Debian for idiological reasons but do not have the coding experience to help with your suggestion. You definitely have my support though. I really would rather stick with Debian. However what I have seen from the Debian leadership over the last few years has not been encouraging.

Just curious, but what license does Systemd operate under? Does it meet the standards of The Debian Free Software Guidelines (DFSG) and the social contract https://www.debian.org/social_contract?

Edit.. I see its under the GPL, http://www.freedesktop.org/wiki/Software/systemd/
Searching through Wikipedia though it appears there are some fairly significant opponents to Systemd, such as Linus Torvalds and Patrick Volkerding. That is enough for me.

Adoption of systemd has been somewhat controversial. Linus Torvalds has expressed reservations about the attitude of a key developer of systemd toward users and bug reports.[26] Theodore Ts'o has also expressed reservations about the systemd philosophy and the attitude of the same key developer.[27]

In a 2012 interview, Slackware's founder Patrick Volkerding also expressed reservations about the systemd architecture:[28]

Concerning systemd, I do like the idea of a faster boot time (obviously), but I also like controlling the startup of the system with shell scripts that are readable, and I'm guessing that's what most Slackware users prefer too. I don't spend all day rebooting my machine, and having looked at systemd config files it seems to me a very foreign way of controlling a system to me, and attempting to control services, sockets, devices, mounts, etc., all within one daemon flies in the face of the UNIX concept of doing one thing and doing it well.

In January 2013, Lennart Poettering attempted to address concerns about systemd in a blog post called The Biggest Myths.[29]

Eric S. Raymond declined to comment on systemd at first, but stated, "I'm aware there’s a controversy."[30] Then in a March 2014 interview on Slashdot, he expressed some concerns about the goals and architecture of systemd:[31]

I want to study it carefully because I'm a bit troubled by what I hear about the feature set and the goals. From that, I fear it may be one of those projects that is teetering right at the edge of manageable complexity – OK as long as an architect with a strong sense of design discipline is running things, but very prone to mission creep and bloat and likely to turn into a nasty hairball over the longer term.

So now the question arises. Why does Debian feel the need to fly in the face of what appears to be some VERY knowledgable leaders within the Linux world? What is the justification for switching to this INIT system over what has worked oh so well up until Jessie? Does having a marginally better boot time merit abandoning the tried and true modular startup architecture of SysV?.. For that matter, who really and truly cares about boot times as a core part of their OS experience enough to want this switch? Lastly, WTF is up with log files in binary format!?

[keithpeter]

So now the question arises. Why does Debian feel the need to fly in the face of what appears to be some VERY knowledgeable leaders within the Linux world

It isn't *Debian* as such. This shift in a core technology has occurred in most of the major distributions, Debian being one of the last to adopt the systemd as a *default* component (remember the technical committee's decision was about defaults). This has happened (I think) because upstream projects are beginning to depend on the api/functions that systemd provides.

Debian packagers make (minor) changes to upstream. How much modification do you want to do and how much developer time have you got?

https://lists.debian.org/debian-devel/2 ... 00603.html

The debian-devel thread above illustrates the detailed issues that we face moving forward (read/skim the whole thread to get the full narrative). You have a developer asking how best to write an updated config file for a daemon. Should be an hour's work.

https://lists.debian.org/debian-devel/2 ... 00952.html

Read this thread and the associated bug. A suggestion for an alternative installer script and some *code* for a systemv init Debian 'blend' to be available in tasksel.

[mardybear]

keithpeter wrote:
The Salix Linux distribution is based on Slackware but dispenses with the 'kitchen sink' approach of Slackware itself by using a small core and having a package manager with dependency resolution. They do xfce4 and Openbox variants. I tried the xfce4 one to see if I was going to be able to cope with Slackware and it worked fine. Just thinking about machine specs here.

Just wanted to thank-you keithpeter for the slackware information - very informative. The whole kitchen sink approach does not interest me but does explain how they get away without some type of dependency resolution...makes sense now. Salix appears to be more to my liking and i will definitely check it out. Thanks again.

[keithpeter]

Salix appears to be more to my liking and i will definitely check it out.

Another year (at least) with Wheezy and I'll bet beer that when the systemd version for Jessie is chosen, there will be a systemd-shim package to match, so there will be a choice. The slackers are a pragmatic bunch - if it is just too hard to work round systemd they will go with it.

[edbarx]

Systemd is not a big problem for me. It's not installed on the systems where I don't want it. I had to expend a small amount of effort to make that happen. I'm willing to expend a little more to enable people to use d-i to install jessie without systemd, if enough people want that. So far it looks like there is no interest.

There is definitely interest in enabling users to get rid of systemd. If you found how it can be done, I suggest you to expend some more effort to write a howto. Users will be grateful for that.

[keithpeter]

Hello edbarx and all

http://www.linuxquestions.org/questions ... 175504377/

I'll try this recipe tomorrow Tuesday on my spare hard drive and see what happens.
(spare drive in a USB caddy, I'd install a base Debian Jessie* using debootstrap if that can work from a wheezy install which I think it does, add a kernel and then use qemu to boot off the USB and see what happens. If it boots OK, I'll add X and a window manager and pop the hard drive into a computer and boot that)

I'm actually agnostic in this particular discussion but I think choice is important.

Owing to a communications glitch at work I have some reading time this afternoon

[edbarx]

Thanks, keithpeter. The post of koanhead made me think there was a long term solution to avoid systemd, however according to TobiSGD there is no such guarantee after Jessie. systemd will inevitably introduce its API which will shape the other software that has to communicate with it. The latter is the reason why it is difficult to get rid of systemd in the long term. It is far from a matter of using a package manager like synaptic or equivalent. A package manager does not write code, but only passes commands to its backends which do the dirty work of package management. To avoid systemd one must find a way to emulate systemd through another initialisation program like init. Yes, you can go the dangerous way and mix repositories, but that, besides being highly discouraged for obvious reasons, will not replace systemd for ever. The only feasible way I can see, is to emulate systemd through an additional layer upon init or equivalent.

However, I am open to discussion and I am eager for more illuminating ideas.

[keithpeter]

To avoid systemd one must find a way to emulate systemd through another initialisation program like init. Yes, you can go the dangerous way and mix repositories, but that, besides being highly discouraged for obvious reasons, will not replace systemd for ever. The only feasible way I can see, is to emulate systemd through an additional layer upon init or equivalent.

No library mixing, just Jessie.

Yes, rolling forwards from Jessie there are no guarantees

Those who really want to keep systemv style init will need to emulate the api that systemd provides as you say.

I tend to look at 'how do I keep this going next year' and then worry about the long term later. Seems to work.

[mardybear]

I tend to look at 'how do I keep this going next year' and then worry about the long term later. Seems to work.

Agreed. Fret, fret, fret is all we ever tend to do...the modern human condition. Would imagine Wheezy will receive longer support, similar to Squeeze, especially due to the systemd controversy. So as long as Wheezy works well on any given system then it will be possible to avoid systemd for quite some time....without expending any effort. As keithpeter mentioned earlier, by then there will likely be readily available alternatives.