[xyzzy_plugh]

I miss writing sysvinit scripts exactly never.

Even if you were building a system for which init stages would be immutable, forever sealed and unchanging, determining the correct ordering means understanding every daemon and all their associated quirks and their dependencies and the quirks associated with their dependencies and so on...

I will never get the hours spent being a human SAT solver trying to safely order hundreds of lexicographically ordered start and stop scripts that may or may not be present in an unbelievable number of permutations.

The reason sysvinit is extinct is neither fashion or politics, but evolution.

[throwanem]

Given my experience writing systemd units and then trying to figure out why they don't behave as the documentation says that they should, this suggests that being miserable to work with must be an adaptively neutral trait in init systems, thus selected neither for nor against.

[hulitu]

> human SAT solver trying to safely order hundreds of lexicographically ordered start and stop scripts

Your doing it wrong. You don't want too many services on a single server.

[xyzzy_plugh]

Where did I mention anything about servers? The scenario I was referring to involved maintaining a large set of packages for a distribution, each of which may or may not be installed.

[oohffyvfg]

except writing systemd unit file hacks are just as bad. but with the added offense that things don't work as promised.

eg. if your sshd is set after initRandom, but unit random is set after timesync, and timesync depends on network because ntp... if you only have lan and your ntp servers aren't reachable (or offline) you don't get any log out user visible hint at all of why start sshd just hangs there forever.

no matter how bad init hacks were, nothing was this shitty. zero logs. zero hints. just wait there forever.

if you like systemd, it just means you're sheltered and too green.

[justsomehnguy]

> if you like systemd, it just means you're sheltered and too green.

Says the green name.

People can like whatever, but most of the time people like sysvinit less than systemd init.

[xyzzy_plugh]

I shouldn't be responding to flamebait but alas.

No, they are not just as bad. The scenario you describe is obviously broken, but the entire system can be inspected and easily debugged. With a flick of the wrist you can at your leisure summon a graph of the entire known universe. If you only have LAN then, typically, your host is physically accessible so this should be easy to debug.

Nothing about the scenario you describe is inherently worse because of systemd. The same scenario with sysvinit is literally a nightmare. This is literally and practically infinitely less of a head scratcher with systemd than anything else in existence.

I'm not sure how anyone can defend sysvinit -- you're certainly not even trying -- but shitting on systemd because you had a bad time is about the least productive way to participate here. It's no one's fault but your own if you don't understand your tools or the ramifications of letting yourself use them in ignorance.

Besides, at least at the time of this comment, your username is the one that is literally green.

[oohffyvfg]

what part of no logs and dependencies of dependencies not showing up you missed?