Systemd 254 released

Posted Aug 4, 2023 16:49 UTC (Fri) by anselm (subscriber, #2796)
In reply to: Systemd 254 released by Cyberax
Parent article: Systemd 254 released

I still have fairly new software (less than 3 years old!) that only has SysV scripts. Yeah, I can easily writer a wrapper script, I know. However, having a tool to do it is nice.

You have that tool now. It's called systemd (or systemd-sysv-generator if you want to be picky; it even comes with its own man page). There's nothing wrong with copying its output against a rainy day.

So yep, removing it is a dick move.

I'd consider it a friendly suggestion to some people to get off their lazy fat behinds.

Systemd has been around as the de-facto standard for mainstream Linux distributions for more than a decade now. If in 2023 there are people who have the chutzpah to distribute commercial software that comes only with a SysV init script, they deserve a stern talking-to. (And if the software in question is FLOSS, here's your chance to do something nice for the community.)

Systemd 254 released

Posted Aug 6, 2023 3:34 UTC (Sun) by Cyberax (✭ supporter ✭, #52523) [Link] (3 responses)

> Systemd has been around as the de-facto standard for mainstream Linux distributions for more than a decade now.

Yup. Just like Linux should drop anything that is older than a decade. Those developers having chutzpah not to move all their API use to io_uring in 2023!

Yeah, I understand that it can be at least a couple of orders of magnitude more complicated task. But still, the thing that users like is not breaking their stuff. And this particular change will absolutely break some users' workflows.

Also consider this, supporting systemd-sysv-generator requires almost no maintenance. Yet now many distros will probably have to scramble and keep patches on top of systemd to support this. And it will absolutely require an order of magnitude more aggregate man-hours to do that.

That's why it's a dick move.

Systemd 254 released

Posted Aug 6, 2023 11:12 UTC (Sun) by bluca (subscriber, #118303) [Link]

In case you never noticed, Linux (as in the kernel) breaks compatibility all the time. Every new release something breaks and need changes somewhere, whether it's a netlink message, or a uevent, or an LSM, etc etc. The answer is always 'deal with it'.

Systemd 254 released

Posted Aug 6, 2023 12:21 UTC (Sun) by anselm (subscriber, #2796) [Link]

That's why it's a dick move.

I disagree. At some point the training wheels need to come off the bicycle.

Properly-written systemd unit files make it possible for system services to avail themselves of all sorts of convenient systemd features which improve stability, security, and general service handling. Most of this remains unused by services which just keep using SysV init files with autogenerated service unit definitions. Programmers who, in 2023, insist on providing only SysV init files do their users an obvious disservice, and there is no reason anymore for systemd to be complicit in this.

Note that there is nothing at all preventing programmers from providing both a systemd service unit and a SysV init file for their system services. Nobody talks about forcing anyone to give up SysV init support completely. It's just that nowadays most people can do a lot better with proper systemd support rather than legacy SysV init files – and even if software authors are too lazy and irresponsible to acknowledge this, they can still provide a minimal 10-line generator-style service unit that just calls the SysV init script, big deal.

Systemd 254 released

Posted Aug 6, 2023 17:21 UTC (Sun) by tao (subscriber, #17563) [Link]

Surely if supporting systemd-sysv-generator requires almost no maintenance, it'd be a great target for newcomers who want a good way to get their feet wet and to get major kudos from the apparently enormous community of people who want to use initscripts instead of systemd units.