pleroma.debian.social

pleroma.debian.social

i find it doubly painful and concerning that even with everything going on about llms and linux, they still choose to double down on "systemd or nothing"

specially since gnome overall has a anti-llm instance, it's mind boggling

@navi
I am always so suprised when people can't believe that other people like systemd.

Like, it works. It doesn't require 10hrs of setup. It's service file layout is good. Simplish command line.
Yea, Redhat sucks. The code works tho. When it doesn't, we'll see.
Why would you expect discourse to shift away from it when there are no notable blocker issues?

@bill88t i don't care if people like systemd

i do care about more and more pieces of software unrelated to service manager are now made to only work with systemd for no reason, and anyone else that doesn't like systemd has to either reimplement everything or leave

for no reason other than "oh we want to use systemd's apis", when most of those apis they talk about just wrap the kernel apis anyway

@navi rather sad that their are intent on building only for one interface. If I learned anything as aspiring engineer, then the fact that with two interfaces, there'll definitely be bugs. Which is good, since you can then fix them. Limiting themselves to systemd will miss those bugs.
replies
1
announces
0
likes
1

@navi the system dev equivalent of "this website only works in internet explorer / chrome / …"

@werdahias what annoys me here, is that they're building for an interface that mirrors another interface

using systemd for cgroups alone doesn't do much of anything at all; just do cgroups if you want to manage resources, delegation exists, it's not any more complex

@navi yeah, well, wouldn't be systemd if it they wouldn't reinvent their own version of things. Much of the systemd ecosystem is redundant anyway, we have existing solutions (cron etc.). Rather improving and maintaining them should be the rationale.

@navi
Excuse me, you expect "normal software" to instead of using a user facing API to go and do it themselves?!

It's already uncomfortable enough for a dev that only has systemd to support another init system's crankyness. Asking them to instead of using a tested and trustwortht API to go and do everything on their own is just delusional.

Your app needs a background task.
Do you:
1. Spin something out yourself, and add a CLI option nobody knows to kill it, have to handle user logout, have to handle autostart on cli login, IPC, logging.
2. Talk to systemd and ask for a service for yourself.
3. Talk to any possible init system, wasting so much time and code to maintain all that for the 2 people that use anything else.

@bill88t

> Excuse me, you expect "normal software" to instead of using a user facing API to go and do it themselves?!

the systemd api "write a value to an assignment in a unit file", the cgroups api "write a value to file in the cgroup"

so no, i'm not saying "do it yourself", it's, use the kernel api? that was made for userspace applications? and it's the literal same complexity to use; systemd just mirrors this same api for you

> Your app needs a background task.

you document that, and the distro takes care of it; shipping a systemd, or openrc, or s6, or dinit, service

all but s6 of those have user services as a feature, and s6 users can do it as well it's just not pre-setup

do less, document your stuff, and that's it