[SOLVED] Is 2019 Still Too Soon For Intelligent Assessment of SystemD?
SlackwareThis Forum is for the discussion of Slackware Linux.
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
If you have any problems with the registration process or your account login, please contact us. If you need to reset your password, click here.
Having a problem logging in? Please visit this page to clear all LQ-related cookies.
Get a virtual cloud desktop with the Linux distro that you want in less than five minutes with Shells! With over 10 pre-installed distros to choose from, the worry-free installation life is here! Whether you are a digital nomad or just looking for flexibility, Shells can put your Linux machine on the device that you want to use.
Exclusive for LQ members, get up to 45% off per month. Click here for more info.
Seriously, who cares if that init binary have few dozens of library dependencies or just several? And why would care someone about this?
I care, because it makes it more fragile. The init program should be the smallest, tightest, static, most reliable program possible. If init dies (or won't run on reboot) the kernel panics. I probably don't have to tell you how important init is in the process tree.
I already stated one example... during updates. Systemd wouldn't restart, which broke the system, with no possibility of resuming. That's what prompted me to say this in the first place.
Last edited by TheRealGrogan; 08-12-2019 at 03:20 PM.
I already stated one example... during updates. Systemd wouldn't restart, which broke the system, with no possibility of resuming. That's what prompted me to say this in the first place.
From what I know, those system wide dependencies should not change the ABI in a stable release, even when there are updates.
Anyways, even Slackware uses a backup of the essential libraries, which is that aaa_elflibs, so the system probably will survive well when the update is properly done.
And for the users who tend to mess the system (like me), there's always the Swiss army knife: a bootable USB hard drive containing a full installation and the reference installation kit, which I learned that's a must to have for any honorable Linux user.
Yes, I learned from Darth Vader about USB drives, if that matters.
Last edited by ZhaoLin1457; 08-12-2019 at 03:36 PM.
A component's reliability decreases per the square of the number of things that can go wrong (for *any* reason, including software bugs or operator error).
Adding a dependency to a component increases the probability of that component failing which is exactly a factor of the probability of the added dependency failing.
Illustration: If component X has dependencies on components Y and Z, then X will function correctly if Y functions correctly -and- Z functions correctly.
If Y has a 98% chance of functioning correctly and Z has a 97% chance of functioning correctly, then X has a 0.98 * 0.97 = 0.9506 or 95.06% chance of functioning correctly together.
Adding more dependencies only decreases reliability. Adding dependencies on U and V, each with a 99% chance of functioning correctly, decreases X's reliability to 0.98 * 0.97 * 0.99 * 0.99 = 0.93168 or 93.168% chance of functioning correctly.
I made up these numbers for the sake of illustration, but the underlying math is true for all systems (and not just software). Since nothing is 100% perfectly reliable, adding dependencies will always decrease overall reliability.
If you want a more real-world assessment of the risks incurred, look at the changelogs of each of systemd's dependencies and see how frequently bugs are introduced to those dependencies. Combined, they represent bugs which might impact systemd's operation.
Given this, if a system needs to be highly reliable, it is good practice to avoid increased complexity as much as possible.
Higher complexity is tolerable in ancillary components, but not in primary components (such as an init system).
What I said about init, is an example of the best practices that make Slackware more reliable than other distros. Not that everyone is going to have a broken system.
I grudgingly put up with it on other distros, and I know how to work with it, but systemd is a shit show.
While I agree that adding complexity can, and usually does, add opportunity for disaster, and I very strongly trust the KISS concept, I am also aware that complexity always occurs with any progress or evolution. If this wasn't so we'd all be using DOS and driving cars from the 50s. Adding complexity is an investment, a cost that is hoped will increase benefit. In the above two examples, DOS and Cars, despite a few failures, we have PCs that do far more than what was possible in a single user, terminal only environment and cars that not only do more but last for hundreds of thousands of miles, instead of 30K to 50K.
This is what I want to know. I am aware of the costs of systemd. Has it turned out that it also has benefits and are those benefits anywhere near a net gain? So far, it appears not.
NOTE: I have started a nearly identical thread in Linux-General today that links to this thread. I also quoted the essential purpose and rules of the thread so nobody has to come here from other distros but some might. Let's please be adult and respectful of any that do. By all means be direct about negative software issues we can see from the outside looking in, but also remember that we are blind to some of the experiences they have and that both poles can benefit from honest discourse. Thanks.
I am ashamed to realize that even the wisest minds around refuse to realize that systemd is not a init system. Or, better: it is not just an init system, but a software collection.
That init system? This is just the upstart developed by Ubuntu. I guess nobody should explain what is udev. Or ConsoleKit - the first version. And so on.
In my opinion, those RedHat guys just grabbed different software, putting everything together and making it to work better together. And continued the development of those software togheter.
The sum of those dependencies already existed, from the separate components, then I do not think that those dependencies count increased dramatically for the final software collection.
I do not defend systemd, in fact I do not care about it.
But I prefer to look sober to this "issue" and even a modest user like me realize that this software collection is an idea of the RedHat higher ups, and they hired a team of programmers to make it, for their own RedHat use. If Mr. Poettering refused to take this job, another programmer should have been certainly hired.
I am on opinion that there are no devil plans or malefic guys, but just business from their part.
And from their POV, they arrived to conclusion is better to make and use this software collection. And other businesses considered that this is good, that's why they adopted this systemd too.
Last edited by ZhaoLin1457; 08-12-2019 at 04:12 PM.
I am on opinion that there are no devil plans or malefic guys, but just business from their part.
And from their POV, they arrived to conclusion is better to make and use this software collection. And other businesses considered that this is good, that's why adopted this systemd too.
And here, dear sirs, is a man that have never read a software license. Praise its innocent naivete.
I'm not ashamed, and I don't feel the need to mansplain my points here. Of course it's more than an init system (that's part of the problem), but it IS the init system. That's just arguing semantics.
I have first hand experience, and second hand observations to draw my opinions from. I don't give a toss about Redhat and their commercial motivations.
Nobody mentioned upstart, Ubuntu doesn't use that anymore, but I had a bad experience with that too during updates when telinit was run by the update scripts to rehup, and init was segfaulting. It left me dead in the water and I had to (temporarily) symlink telinit to /bin/true so dpkg --configure could continue. That's the same brand of fragility I'm protesting here.
even the wisest minds around refuse to realize that systemd is not a init system.
Critics have said it is not an init system for years-- that's also part of the problem. It should be an init system. It isn't.
Quote:
In my opinion, those RedHat guys just grabbed different software, putting everything together
That is the problem.
Quote:
and making it to work better together.
What they did is make it less suited to being maintained, overall-- now instead of smaller groups (communities) maintaining just one part, it requires a corporation to maintain the whole thing.
People talk about "fixing" systemd simply by getting it back into individual components.
If that's the fix, the problem should be very clear.
Quote:
The sum of those dependencies already existed, from the separate components, then I do not think that those dependencies count increased dramatically for the final software collection.
Lumping them together was indeed the problem.
Quote:
But I prefer to look sober to this "issue"
Subtle way to imply that anybody who finds more problem with it than you do is hysterical, rather than sober. Bit of a circular argument.
Quote:
even a modest user like me realize that this software collection is an idea of the RedHat higher ups
Lennart denies this outright, but it is clear to systemd critics or at least difficult to believe otherwise.
Quote:
If Mr. Poettering refused to take this job, another programmer should have been certainly hired.
That's an argument used to justify atrocities-- this is no different. "If you don't slaughter these civilians, they will just call in someone else to do it."
It's a fallacy, for what that's worth.
Quote:
I am on opinion that there are no devil plans or malefic guys, but just business from their part.
Are you implying that "just business" can't be malefic? I would say it's both.
Quote:
they arrived to conclusion is better to make and use this software collection
Yeah-- as one of the obvious effects of this design is package-deal-itis, which corporate software monopolies have used for years to reduce choice for customers. And lo and behold, it has reduced choice, which (as usual) the companies creating lock-in point to as "success" and "proof that this is good software."
A lot of people are around long enough to not fall for that old trick. Lennart and other proponents have tackled this by saying they're just greybeards who don't like change.
Quote:
And other businesses considered that this is good, that's why they adopted this systemd too.
Windows is best-- it's what most people use. I'd go a step further and say that BSD is probably "better" (in some very important ways) than GNU/Linux, but I am not a BSD user, I simply respect what it does best.
GNU/Linux is a compromise between broader appeal and broader application, and what BSD does better-- that compromise is why I use GNU/Linux. But it's not as much of a compromise as other well-known platforms, which is why I don't use them.
Either way, people who prefer something that isn't the most popular platform should already be inoculated against the fallacy that marketshare is a reliable measure of quality.
Last edited by freemedia2018; 08-12-2019 at 05:13 PM.
Either way, people who prefer something that isn't the most popular platform should already be inoculated against the fallacy that marketshare is a reliable measure of quality.
How is this not a contradiction in terms? and it also seems in direct conflict with your windows bashing rants. What's real?
Also, I'm still not seeing any systemd specific complaints based on actual use from you. This seems same ol same ol philosophical rant to me and useful to nobody. Please do step up your game with substance or stop posting here. I think I have been both polite and clear about what I'm looking for in this thread. Please give me the same respect I've given you. Just because In marked it "Solved" doesn't mean it morphed into a free-for-all.
How is this not a contradiction in terms? and it also seems in direct conflict with your windows bashing rants. What's real?
Also, I'm still not seeing any systemd specific complaints based on actual use from you. This seems same ol same ol philosophical rant to me and useful to nobody. Please do step up your game with substance or stop posting here. I think I have been both polite and clear about what I'm looking for in this thread. Please give me the same respect I've given you. Just because In marked it "Solved" doesn't mean it morphed into a free-for-all.
First, I'll mention that your sarcasm detector is turned off. Consider activating it.
Second, I'll point out that when you jump off of a very tall building's roof, you'll be quite fine for most of the journey.
Third, I'll mention that this S6 page describes what is required to bring up and continue to monitor those services required for a (presumably) Unix-ish system to continue to run. A good indicator would be that requiring a reboot for anything other than a kernel upgrade is a tell that your init system has pulled in too much stuff from userland.
Fourth, well, that's in the future. IMO, systemd may very well work more-or-less well at the moment, but it could have had a different architecture to correctly separate the different requirements of the various boot stages.
While you may poo-poo philosophical rants, getting the software architecture correct normally requires either:
Big bucks
or someone who wants (1) but is willing to set the stage for (1) after getting buy-in.
God knows that I've been around long enough to see various companies make very horrible decisions and pour shitloads of money into implementing those very horrible decisions. (For those who may wonder, I wasn't involved in all of those bad decisions.)
Don't get me wrong. After reading @freemedia2018's comments, I thought of this. To a non-trivial degree, that's an unfair thought. Engineers can be aimed by management, after all.
@enorbet -- I'm responding here to the other thread you started (I see it has been closed; how unfortunate). I've been following all of this stuff for years, especially the threads about it here.
I guess I use my computer for some fairly mild stuff compared to some other folks. I'm not the most knowledgeable Linux user but I've been running it for some time now. And systemd hasn't seemed to have caused any problems for this user.
I figured that the best thing I could do regarding systemd would be to just keep on using the distros I like using, let the experts hash things out, and see what happens. I've got Debian, Arch, and Kubuntu installed these days.
Maybe I'm just blind and systemd really is the horrible thing that some people say it is. Oh, well, none of my computers have burst into flames because of it yet, lol. I hope the Linux folks will figure something out that'll make everyone happy. As if.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.