Systemd 209 Is A Massive Release, Readies Kdbus

[oleid]

Can you point to some? Hurd, Minix, and Mach are not, I have seen some research kernels though. They couldn't even run a server though, so of little practical usability, whereas sysvinit has been of practical usability for decades.

The microkernels work, however are lacking some features. As does sysvinit The most promising seems to be Genode at the moment. It's related to L4.

[Scimmia]

When reported as solved, they usually turned out as config issues or packaging bugs. Transition period.

Yes, there have been plenty of configuration issue, but there were atleast as many configuration issues with the previous sysvinit/initscripts.

That's not what you were arguing, though, you were arguing that systemd itself causes instability. Put up or shut up.

[gilboa]

I preach to keep or improve the quality of any critical component. This does not preclude new development, but raises the bar of it. Adding new points of failure to a critical component is unacceptable, as that would reduce the quality.

In this assumption the sysv number would be 0, and so the raise from 0 to 20 would be infinite % .

I agree with you that the kernel is a much larger source of bugs. But this brings me back to what I've said before, "shit sucks" is not a reason to "make it suck more". Adding points of failure to where previously were none is unacceptable. In other words, "kernel sucks" cannot be used as an argument to "let's make init suck too".

In *your* use case, sysv has zero bugs.
In *our* use case, sysv was anything bug free or even manageable.
I (or actually, we) develop a fairly complex software system that runs on embedded system that's connected to a large number of devices and has a fairly large set of kernel modules and user services.
All of this meant that it we had to patch the day-light out of sysv and had well-above 5000 LOC of bash and python scripts in an effort to get the initialization order in check. The lack any type of service dependency, misbehaved service detection, service logging, resource and/or user management or any type of graceful error handling meant we more or less had to develop our own systemd/upstart just to get the damn thing working.

... And than systemd came and we simply deleted our lousy proprietary service and kernel module load management and replaced it by ~500 LOC of systemd units and our problem was solved.

Back when I started using Linux for embedded projects (2000'ish), sysv was fine and dandy. Its now 2014, machine do come with 24 x 10GbE NICs w/ complex configurations (bridges, trunks, etc), multiple RAID systems with multiple encrypted FS and 100's of interdependent services. Its high time someone took sysv behind the shed and shot it.

- Gilboa

[curaga]

Either show your evidence that systemd is a detriment to stability, or shut up. Endless anecdotes about lines of code and complexity won't help you because they cannot be quantified or falsified.

That's the entire damn point! You cannot prove that systemd is stable, whereas you can for the competitor. If you fail to understand that, that's your loss.

You can continue to not use systemd, nobody cares, but please spear us the endless bullshit.

Nobody is forcing you to read.

@gilboa

If systemd works well for you, I'm happy for it. My issue was with some proponents thinking that systemd is a shining white shield with no issues at all, and everyone opposing it must be an emotional lunatic, let's throw some more insults!

If systemd works for you, use it, but spare the goddamn insults.

[kigurai]

That's the entire damn point! You cannot prove that systemd is stable, whereas you can for the competitor. If you fail to understand that, that's your loss.

I am sorry if this has laready been mentioned in this thread. But is this proof of sysvinit's stability available somewhere?

[tuubi]

But is this proof of sysvinit's stability available somewhere?

Curaga argues that sysvinit is so tiny that any bugs would be obvious and easily spotted in the code. No proof needed if you accept that argument. Personally I think he misses (or rather sees but ignores) the big picture.

[Skrapion]

My issue was with some proponents thinking that systemd is a shining white shield with no issues at all

No, systemd definitely has issues, but when you take the whole environment into account, it has less issues than sysvinit.

I would never claim systemd will never crash. I'd never make such unqualified, absolute statements as:

You cannot prove that systemd is stable, whereas you can for the competitor.

[Gusar]

You cannot prove that systemd is stable, whereas you can for the competitor.

Only if you ignore the entire framework of scripts you need to put on top of the competitor to make it do anything, because the competitor itself does so little.

Can you prove that the framework of scripts and helper applications on top of sysvinit is stable? Not to mention that every distro invented their own framework, with their own limitations and quirks and edge-cases and such. When the scripts act up (don't properly restart a service, don't start things in the correct order, etc), or when they can't provide something you want to do, it doesn't help any that pid1 is provably stable. On the other hand, systemd's pid1 may not be provably stable, but you have yet to provide reports of actual instability happening in practice.

Bottom line: You may have a valid point, but it doesn't have much merit in practice.

[psychoticmeow]

If systemd works well for you, I'm happy for it. My issue was with some proponents thinking that systemd is a shining white shield with no issues at all, and everyone opposing it must be an emotional lunatic, let's throw some more insults!

If systemd works for you, use it, but spare the goddamn insults.

Isn't anything emotional about it, you're making claims and then not backing it up with any actual information to support your position, which is about as strong an argument as to make any creationist proud. If you're not going to base your position on facts, statistics or testable claims, and instead choose "intuition" or wherever you're pulling this shit from, I'm going to continue calling you out as a complete and utter fucking idiot.

[gilboa]

My issue was with some proponents thinking that systemd is a shining white shield with no issues at all

I doubt that anyone above the age of 5, thinks that systemd is anything beyond piece of software that may/will make life easier for admins in the short/middle/long term.

and everyone opposing it must be an emotional lunatic, let's throw some more insults!
If systemd works for you, use it, but spare the goddamn insults.

I must admit that the amount of hate spewed at systemd or, for that matter, Lennart Poettering himself, does exceed what I consider acceptable.
Especially given the fact that both systemd and pulseaudio can be avoided fairly easily.

However, as others (and I) pointed out, your claims do require actual number to substantiate your claims.

- Gilboa