New init system
 

Well, I didn't see this one coming.

http://www.joachim-breitner.de/heisse-news/news_27.xml

Google translate seems to do a pretty good job for those who don't read German.

Evo2.

... and still nothing official for Jessie.

Most agree that those old .rc scripts are simply outdated, stupid, chaotic and unmaintainable (sysv-init is dead).

Picking a new one (as default) is difficult for various reasons.
Whather they pick, it can't please everyone (resulting in articles as the above) ;)

Current options:
https://wiki.debian.org/Debate/initsystem/

Original bug:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=727708

A decision needs to be made, the sooner the better.

Is that your opinion? I'm interested to hear what exactly is wrong with SystemV. It works, and is rock solid. Why change something that works fine?

I'm really interested how it will work out.

Regardless the technical questions, Systemd is linux-only so they will have to come up with some sort of compatibility workarounds, Upstart requires you to sign the CLA to contribute upstream and i'm not sure how the Debian policy rules apply to this. OpenRC doesn't seem to be a serious advancement from sysVinit.
If Debian chooses Upstart it might be serious setback for systemd.

No.
It's from the majority (not everyone) of all DD/DM's and really about Debian and Debian only ...

http://lists.debian.org/debian-devel.../msg00651.html

Options:
*sysvinit (status quo)
*systemd
*upstart
*openrc (not available in sid - ITP: 684396)
*One system on Linux, something else on non-linux
*multiple: commit to supporting two or more specific initsystems, such that choosing a default is less important, and at least one of them is available to every arch

https://wiki.debian.org/Debate/initsystem/

Traditional SysV is lacking features for some user-cases/environments in Debian (blocking release goals for Jessie).
Supporting them all the same would be a mess.
OpenRC isn't even available in Sid.

So ... it's mostly about "upstart vs systemd" (+ a solution for kFreeBSD an Hurd if possible).

Hi,

actually the article is about a completely new solution based around suspending to disk.

Evo2.

And I can understand why there has been push-back. I always thought Linux was about choice - I'll start what I want when I want.
Not accept an image pushed down my throat by the devs.

Hi,

What push-back are you referring to? I've not seen any discussion of this new idea for an init system. I'd really like to see some though: do you have any links?

Thanks,

Evo2.

I was extrapolating from the (translation of) the article

Hi,
Ok. I don't find that last paragraph to be particularly surprising, and I don't really think of it as push-back against this new idea in particular. From what I've seen the upstart backers and systemd backers have already made up their minds and are unlikely to be swayed easily.

Evo2.

This link was in the first post;
https://wiki.debian.org/Debate/initsystem/

Which actually sets out the proposed possibilities quite well and I found quite interesting.

It really is not about "starting things when you want" as init already starts thing automatically when you boot. Yes you can change the order if you want but I don't see anything stopping you from doing that with the top three choices.

Upstart I doubt will be used. It is not that great anyway and the copyright policy of Canonical is offensive to too many people.

Systemd is used quite well by some distros already. I know it works fine in Mageia 3.

The only real objection to changing is that it is change or so it seems to me. There is some sense to that argument but there was the same sense in arguing for sticking with many systems that have now disappeared from use.

The objection of the systemd devs to porting it to other kernels than the Linux kernel seems like the only logical objection but that could be handled by a fork or by their becoming a bit more flexible in their thinking.

Be interesting to see what Debian does do with this. I suspect more integration of systemd, for which there is considerable support already in Debian, as sticking with init for Jessie as with Wheezy. Then a shift to systemd for Debian 9.

I think that is just more in line with the traditional way of change in Debian.

Hi,

I don't see any mention of Vorlon's proposal on that page. It's just the same old sysvinit, upstart, systemd, et al stuff.

Evo2.

It's called "satire".

... Ian “Vorlon” Bart hat im Interview mit heisse news das Ergebnis ...

Ian > Ian Jackson
Vorlon > Steve Langasek
Bart(h)> Andreas Barth(?)

http://www.debian.org/intro/organization

Does it? (... and that's called "sarcasm")
Their's obviously some"irony" in this ;)

PS: It's still under discussion:
*http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=727708
*debian-ctte: http://lists.debian.org/debian-ctte/

While I honestly respect all ctte members as individuals, it's difficult to overlook how 1/3 are paid Canonical employees as well ...

HI,
Damn, I should have spotted that.

I guess one needs to remember that subtly is usually lost with machine translation.

Anyway, thanks, this explains a lot.

Cheers,

Evo2.

Some Updates
 

Positions forming in the Debian init system discussion
LWN article: https://lwn.net/Articles/578208/

**Russ Allbery** (systemd)
https://lwn.net/Articles/578210/

**Ian Jackson** (upstart)
https://lwn.net/Articles/578209/

OpenRC is now in Experimental (...):
http://thomas.goirand.fr/blog/?p=153

A decision is expected this month...
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=727708

Quoting myself:

... sadly this could also turn in 4:4 scheme, plaguing Debian's democracy as happened a zillion times before.

http://www.heise.de/newsticker/meldu...g-2088684.html

Outdated, stupid and chaotic is the human brain. The spaghetti code you see in
most linux distributions (and BSDs) is not sysv-init fault. That's why users
with a poor knowledge of shell scripting and system administration see in
systemd a need.

I encourage anyone, indifferently of his knowledge about shell scripting and
Unix system administration to give Crux Linux a try. Read its rc files and see
how it boots. Do it and you won't want to hear about systemd/upstart tales
anymore.

Wasn't Arch using a similar method as Crux?

Yes, and still BSDs and Slackware.

I think eloi wasn't so much referring to the init system its self, but more so the users management of the init system. The BSD init system which is where Crux, Slackware and formally Arch adopted it from, is very simple and easy to manage.
I can't speak for everyone, so take this with a grain of salt, but i believe that some distributions (namely Arch) have migrated to System d because of Gnomes reliance on it, plus Systems d's consumption of other often used packages which were previously separate.

I am against System d because of this. This is a very aggressive method to almost force most distributions to use it, which gives Redhat an immense amount of power/influence on all the distros out there.
Slackware is a big thorn in the foot of this plan since Slackware abandoned Gnome a long time ago, so the pressure to conform is a lot less significant.

The thing I liked about Arch was their use of the BSD style init system.

So this must put Debian in a tough spot if Gnome is still the default desktop, and they're trying to keep kFreeBSD on the same page as their Linux releases, since I've read BSD doesn't have systemd.

Adding 200 lines to your network rc script just because you want network-manager in KDE justifies systemd existence.

That's mostly the result of a bigger question:
Is it still possible/useful to support everything inside a single distribution?

IMHO Debian would greatly benefit from a less modular design.

Upstream sysv-init wasn't compatible with freebsd and hurd either (the whole /proc stuff).

Fark GNOME, GNOME 3 is the ugliest, least flexible, most anti-intuitive, and resource hogging piece of crap GUI I've seen, and this includes Windows 8 and a good number of phone UIs.Also the developers are apparently only interested in bling and making it tablet-ready.
The last straw that borke the GNOME cammel's back for me personally came long before the whole systemd dependence.
Also MATE.
Fork YEAH!

What makes it so tablet ready? Why do you think they would design for a tablet? I like gnome because it is so perfectly configured for keyboard control. I wouldn't think that would be if it was designed for a tablet.

Maybe it is for advanced users?

A lot of people like Gnome Shell. I am not one of them. If you like it that is great.

You really should read some on the Gnome website however.

One of the main concerns in the design of GS is to be touch screen friendly. Yes I know there are laptops and monitors for desktops that are touch screens.

Most aren't. Probably won't be for some time to come. This leaves the target for design as phone and tablet use.

The Gnome folks and the Canonical Unity devs like to say this is for "advanced" users. Bullshit. They are both designed for touch screen use.

Tablets are the touch screens targeted. These are not the chosen computing devices selected by most "advanced" users. These are the most popular devices in use by people that want little from the device except the ease of messaging on social media and watching video.

I used GS from late 09 when it became available for testing. Used it for a couple years. For me it simply got worse. I now use Xfce.

For folks that like GS it is probably great. It is more flexible than Unity. I have no idea about W8 as the only place I have seen it is in adds on TV.

It is significant, however, that HP is now offering a deal on new desktop/laptops with W7 instead of W8 because they seemed to run into some consumer problems with W8.

Judging from the ranting and desperate bragging being done about the fine qualities of GS and Unity I would judge that I am far from alone in abandoning Gnome as my desktop. Both Lxde, particularly in the form of Lubuntu, and Xfce seem to be reaping a lot of users from ex-gnome users.

Mate is becoming a very, very competitive DE. While I prefer Xfce I do maintain Mate (Wheezy netinstall with the Mate repo added) for my wife as she really wanted to stick with the old style panel system.

Her reaction to Unity and Gnome Shell was "I don't want that crap on my box." I installed both on an external drive so she could try them out. She is not an advance Linux user. She is an advanced GnuCash user. According to her, after running one and then the other for a week apiece, they simply slow everything down making her do more key strokes and mouse movements.

I found this same thing to be true if trying to edit several images in Gimp. Or creating an image using parts of several other images.

They were so bad for my work flow that I now find that I vastly prefer KDE to Gnome. I have always loathed KDE and still do. You can, however, still have a sane way of dealing with several different windows of one application that does not require movements all over the screen.

I will say, in defense of both Unity and GS, that if using a touch screen I think they would work fine. I use a desktop. I use a vertical monitor. I am not going to be using, for hours, a touch screen with the attendent holding of my arms out in front of me.

Nor am I going to spend thousands of dollars to replace my hardware to use either of them when there are fine DEs that will do the same job on my existing hardware.

Hey think some of you guys should take a look at this : https://felipec.wordpress.com/2013/11/04/init/.

Thanks for that link. I have my problems with the Ruby syntax, but it is actually quite easy to get the concept. Interesting article indeed.

What he said. Thanks again. Very interesting.

Oh, you mean besides the paradigm shift from something that is perfectly suitable for desktop users to something that benefits tablets and other similar devices the most?

Who doesn't, also great job on the part of the original author for showing that there are things other than bash, which can be used to write an init script, enough to inspire me to try writing a few in Python;)
But really I personally have never seen much need for socket activation, I'd rather waste a few mb of ram than a few more seconds of CPU and i/o time if I wish to print something or to use ssh, and as pointed out xinetd could handle that.
It is even less relevant on servers, where you could have hundreds of GB or even a few TB of ram, and in any case they should be dedicated for one application only, not running the whole enterprise, ad what is true for physical systems is also true for VMs IMO, even considering the extra penalty most forms of virtualization incur.

It think a lot of people get put-off by SystemV due to the fact it requires so much scripting work, and few people want to take the small amount of time to learn scripting.

SystemV is far from being broken, messy, mangled, or outdated. SystemV works plain and simple, and learning scripting properly is only limited by one's ability to do the most basic of tasks in GNU/Linux that people still don't seem to grasp...

Read the documentation!

OpenRC, BSDInit, RunIt, s6, daemontools, and other replacement init systems all use scripting on some level, same as SysVinit.

The only issue is getting lazy admins to actually open up a document and read the bloody thing. Not to kick a dead horse but read Lennart's entire manifesto on why he feels SysVInit is outdated... It all stems from laziness and an unwillingness to properly read documentation.

I can't quite follow that logic. Due to being too lazy to write scripts or even read documentation this man does not only write a whole new init system from scratch, he also documents that new init system extensively and later extends it to the goal of becoming what he calls the "CoreOS"?

Sorry, but that doesn't make much sense. If there is one thing that you can't blame Poettering for it is laziness.

Laziness can come in varied forms though, so technically, Lennart might be a good software developer, but in terms of writing scripts and other configuration files he might be lazy in those aspects only. Remember this guy does have an ego so he could be one of those software developers that thinks he's too good and above the need to learn to write scripts.

However, then again Lennart's logic doesn't make too much sense either into why we need what has already been done in modular form, that he wants to do in monolithic form.

Ah, this again. systemd is not monolithic, as you can see for example in bartgymnasts effort to port it to Slackware, where he uses only the minimum he deems to be acceptable of systemd, not using the other modules.

This is relative. It could also mean "We are all so deep buried in the windowzy desktop paradigm and used to it that we can't possibly move on to something better". Gnome3 is actually neat, go try it for some time.
It's similar to arguing that KDE4 is oh-so tablet oriented because it allows you to put silly widgets on your desktop, just like Android does.

I hope you plan to join the Debian mailing lists soon, to refute all the arguments about "sysV is crap, there are much better alternatives". You obviously know something their Technical Comittee (such a band of lazy wankers, too lazy to write bash scripts) don't.

Actually systemd is monolithic by design even though it has various subsystems that act modular. Don't let the design fool you. Even though Bart was able to isolate out various components for Slackware and LFS was able to work out where udev and gudev subsystems can be extracted out individually, the problem is the core of systemd, when built, takes over and creates a foundation that everything then must compile into, and then it becomes a core dependency within the system for various other systems. Once this is built, systemd is no longer truly modular. It becomes a single system set with various components all working off a core foundation. Monolithic doesn't just have to mean it compiles into a single package file like the Linux kernel. Monolithic can extend to individual programs that work off a single core entity.

Even FreeBSD could be considered monolithic in terms of an OS. FreeBSD software is designed to where it will only work with FreeBSD, not any other system.

The whole Debian init Debate for dummies (with sound, pictures, interaction and other geeky fun stuff):
http://aceattorney.sparklin.org/jeu.php?id_proces=57684

... please stop derailing this topic in yet an other anti-systemd rant.

Haha, this is great! Pretty long too, must have been a lot of work.

Debating an init system like systemd is going to spark controversy. It's happened with EVERY topic on it. Just face the fact that hardly any Linux admin with common sense towards properly doing things is going to say "systemd is the best thing since the wheel was invented".

Edit:

The best solution we can hope for is one where other init systems are entirely optional, or we are given choices of init systems that can be supportive to sysvinit rather than destructive.

Example: RunIt can be ran alongside sysvinit as a daemon service supervisor as shown here from it's documentation page, without replacing sysvinit, upstart, bsdinit, or even launchd:

http://smarden.org/runit/useinit.html

That other software uses the APIs systemd is exposing does not change the status of systemd being modular. If you read bartgymnast's page about systemd without PAM on Slackware you will see that exactly one package, dbus, has to be recompiled, nothing else. Also, as you can see in the thread about that topic and acknowledged by you, bartgymnast uses only hostnamed, machined, timedated and journald, but none of the other modules. That should be prove enough that systemd is modular.

Learn to read, I clearly stated that I did try that stupid bloat and did not like it, and FYI I rather like FluxBox, so there is your retarded argument about Windows dependant mentallity.

Ah, a clear attempt to get out of actually answering a valid question/not debating by talking smack, begone troll.
Also I propose that anyone on any Debian technical panel that has been affiliated with either Fedora or Ubuntu be told to shove off and stop forcing Debian to copy other distros' solutions.

Funny, neither Fedora (or FWIW systemd) developers nor Upstart developers started that discussion, it was actually caused by Gnome 3 depending on logind, which in versions >204 needs systemd as PID1. By the way, you forgot the Gentoo developers with their OpenRC in that list of people trying to force Debian to use their init system.

Oh, and how many OpenRC developers are actually on the panel, or on any Debian technical panel for that matter?
In addition, I really don't give a rat's ass about Gnome 3's prerequisites, IMO the only prerequisite a stupid GUI should have is for X(Or Wailand, or Mir, when they are actually ready for wide use) and the appropriate GTK libs to be installed, it shouldn't even even have to come with crap like D-Bus, udisks, or the various *kits as dependencies.

As far as I am concerned the GUI is nothing more than nice window-dressing for lazy people a lot of the time.
I need it for actual administrative tasks, writing code, building various essential applications or managing files, users, etc. as much as a monkey needs a golf club, moreover a lot of those tasks are actually considerably faster of done via CLI.
Hell there are even a number of pretty nice music players for the console, I only need to use runlevel 5 if I am watching a movie , surfing the web, testing html/css/js crap, using some IMs, or viewing or editing some types documents.
netBeans, Bluefish, Geany and Eclipse are all nice, but vi with some plugins(or EMACS if you are a pervert) works pretty damned well, thank you.

Also, GNOME3 is still a slow, ugly, uncomfortabe piece of trash written by retards, if I had to choose between it and Windows 8 I would just use Windows 8.

The exact same number as systemd developers.
Ah, I get it, anyone who does not use the computer like you do or prefers a different visual style or type of DE is just a lazy retard. If I could use my ignore list you would have earned a place on it with post.

Haha lol, good advice from someone who obviously didn't get what i have written at all! Keep it going!
If you think fluxbox somehow magically doesn't use the window paradigm, well, think twice next time.
I personally prefer i3 over fluxbox, as it's less noobish (i can be elitist too) and windowzy and more keyboard oriented and tiling (hint hint, maybe now you'll get it), yeah who cares.

Far better than the tablet-moron paradigm, windows are far more flexible than the idiocy pushed by iOS, android and now GNOME are trying to push.Also from the screenshots I saw of i3 it still uses windowing, so if you are all hot and bothered about a "non-windows" paradigm, why not use BloatGNOME instead troll-face.

Excellent, lets get rid of the Upstart Ubuntu asshats now, so that we can have an objective discussion conducted by people that are not trying to reinvent hot water for the sake of being kool.

As to the idiocy you just spouted about me. well you are either trying to annoy me or you are a total moron, I don't give a rat's ass what other people use as their UI, just as long as I am not forced to use an init system because the dumbocracy decided that their bling must be dafault, and the idiotic developers need to force stupid dependencies, it is basically the same as if a car manufacturer forces you to buy the automatic gearbox model of car to get the leather upholstery.
What if tomorrow KDE decides that it should work with Wailand and OpenRC only?
And what about all the other Environments and window managers that do not have such stupid dependencies?

As someone that has been accused of being an elitist guru for suggesting that Unity was developed with illiterates working at fast food outlets, I really am embarrassed to advise you to try and be a bit more civil.

This is not a Windows forum. This is a Linux forum. The difference is that we try to treat each other with a bit of respect. We all have our ideas of how we like our work flow.

Mine is obviously superior to all others. For me.

I am sure that yours is for you. Even if it is outdated and moronic.

That last sentence is an example of what I am saying. I know nothing of what you do on your box or how. Therefore that sentence is completely and utterly stupid on my part.

Let's try to simply leave that kind of comment out of our discussions.

I enjoy table pounding discussions. I like to imagine sitting in a cozy room with the other parties, perhaps sharing a beer or three, stating, firmly, our opinions but having respect for the opinions of others enough to avoid table pounding from turning into head pounding.

Gnome3 isn't totally trash. A lot of good applications have come from the Gnome developers that have inter-worked with other desktops like KDE, Xfce, LXDE, etc. for a long time now.

Now I won't say it's not a pain in the butt to use, as it is and very clunky DE by design, but the applications that have come out of Gnome have been very beneficial to many UNIX and UNIX-like distributions, desktops, and platforms equally.

And lastly GUIs can be very beneficial to use and they aren't by any means just for lazy people. A lot of us use desktops for media and such that require GUIs.