As someone who designs systems as part of his job, “it violates Linux philosophy” is a perfectly valid reason to hate something as critical and low level as an init system.
I think “hate” is a strong word, mind, but following consistent design patterns and constraining scope matters. The problem is that it’s one of those things that doesn’t tend to matter until shit hits the fan.
The scope creep exponentially raises the efforts needed to properly test, troubleshoot, secure, and just generally work with the project code. The unique design creates rough edges for integrations with other pre-existing systems and tools, while also raising the effort needed for troubleshooting when something goes wrong.
There’s also the whole “Chesterton’s Fence” aspect to a lot of it, where the project runner’s overall attitude has been dismissive of existing conventions and separation of functionality.
There’s a lot of perfectly valid reasons to dislike it. That said, I think full on hate is extreme for an aspect of the overall Linux “system” and “user experience” that most users never interact with. Like having an opinion on the order your BIOS handles things in.
Which part of the “Linux philosophy” does systemd “violate”.
The scope creep exponentially raises the efforts needed to properly test, troubleshoot, secure, and just generally work with the project code.
I keep seeing these words, “scope creep”, in this thread, and I don’t think most people understand what they mean.
The scope of systemd is an init and service manager, and any basic service required for a minimal, coherent system to work. If you think that it goes too far, then we should be talking about the Linux kernel and all the stuff it has that could run on userland and doesn’t, or coreutils, which essentially is the shell equivalent to systemd without the integration.
Sure, and that’s just as valid points against them as well. For my part at least, none of this is meant as some sort of death knell, or direct damning evidence that some software is some sort of terrible piece of garbage.
I just refuse to believe that someone would literally be entirely incapable of comprehending any of the reasons people dislike systems to only be capable of engaging with the criticisms in bad faith.
That only has impact on the security risk surface exposed for your own install and for troubleshooting on your own machine. The maintainers and other coders on the project still have far more space, complexity, and features to cover.
It’s not a death knell by any means, but still a valid cause for concern.
Predictable, more-or-less identical systemd, with its stance on limiting eveything and being secure by default is more of a security risk than old-timey inits that were different for every distribution with init scripts made by different people with different abilities and customizations?
When I started my ops journey systemd wasn’t there, and it was far harder to make compatible software. Also if you are designing a system how exactly general use systems come into picture at all?
By every reasonable measure systemd is a win, it started to appear even in automotive industry, which is regulated as hell, by 2022 (when I left).
As someone who designs systems as part of his job, “it violates Linux philosophy” is a perfectly valid reason to hate something as critical and low level as an init system.
I think “hate” is a strong word, mind, but following consistent design patterns and constraining scope matters. The problem is that it’s one of those things that doesn’t tend to matter until shit hits the fan.
The scope creep exponentially raises the efforts needed to properly test, troubleshoot, secure, and just generally work with the project code. The unique design creates rough edges for integrations with other pre-existing systems and tools, while also raising the effort needed for troubleshooting when something goes wrong.
There’s also the whole “Chesterton’s Fence” aspect to a lot of it, where the project runner’s overall attitude has been dismissive of existing conventions and separation of functionality.
There’s a lot of perfectly valid reasons to dislike it. That said, I think full on hate is extreme for an aspect of the overall Linux “system” and “user experience” that most users never interact with. Like having an opinion on the order your BIOS handles things in.
Which part of the “Linux philosophy” does systemd “violate”.
I keep seeing these words, “scope creep”, in this thread, and I don’t think most people understand what they mean.
The scope of systemd is an init and service manager, and any basic service required for a minimal, coherent system to work. If you think that it goes too far, then we should be talking about the Linux kernel and all the stuff it has that could run on userland and doesn’t, or coreutils, which essentially is the shell equivalent to systemd without the integration.
Which could apply to GNOME or KDE given they do more than just desktop environments.
Sure, and that’s just as valid points against them as well. For my part at least, none of this is meant as some sort of death knell, or direct damning evidence that some software is some sort of terrible piece of garbage.
I just refuse to believe that someone would literally be entirely incapable of comprehending any of the reasons people dislike systems to only be capable of engaging with the criticisms in bad faith.
You are allowed to build minimal systemd, which didn’t change that much in many years, no scope creep, consistent, simple.
That only has impact on the security risk surface exposed for your own install and for troubleshooting on your own machine. The maintainers and other coders on the project still have far more space, complexity, and features to cover.
It’s not a death knell by any means, but still a valid cause for concern.
Predictable, more-or-less identical systemd, with its stance on limiting eveything and being secure by default is more of a security risk than old-timey inits that were different for every distribution with init scripts made by different people with different abilities and customizations?
When I started my ops journey systemd wasn’t there, and it was far harder to make compatible software. Also if you are designing a system how exactly general use systems come into picture at all?
By every reasonable measure systemd is a win, it started to appear even in automotive industry, which is regulated as hell, by 2022 (when I left).