systemd 239 went out a few days ago and, predictably, my feeds caught fire. Half the reaction was breathless excitement about a couple of new bits, the other half was the usual chorus declaring the end of Unix as we know it. Somewhere in the middle was the actual changelog, which almost nobody seemed to have read.
The headline change people kept pointing at is the move to start treating certain failures more strictly, plus some tidying around how units get pulled in. There is also the ongoing work on the resolver and on portable services, which is the bit I actually care about. None of it is dramatic. It is the slow, unglamorous accretion of features that a project gets when it is genuinely used everywhere, and that is exactly what makes people cross.
Here is the thing I keep coming back to. I have been running systemd on every box I own for years now, and the overwhelming majority of the time I do not think about it at all. It starts my services, it restarts them when they fall over, the journal tells me why, and systemctl status gives me a sensible answer at three in the morning when I am too tired to grep. That is not nothing. That is most of what I want from an init system.
The criticism that lands for me is scope. systemd does a lot, and the surface area is large enough that when something does go wrong it can be genuinely baffling, with the relevant detail buried three layers down in a unit you did not write. The criticism that does not land is the philosophical one, the "this violates the Unix way" argument, because I have spent enough evenings duct-taping shell scripts and PID files together to know that the old way was not some lost golden age. It was fragile, and we just got used to the fragility.
So what did I actually do with the release? Almost nothing. I read the changelog, noted the couple of behaviours that might bite a service that relies on sloppy ordering, and made a mental note to test those on a spare box before they reach anything I care about. That is the correct amount of excitement for a point release of an init system. The internet would have you believe every one of these is either a revolution or a betrayal. Mostly it is neither. It is software getting slightly better and slightly larger, the way software does, and the right response is to read it, test it, and get on with your day.
I will say the resolver work is the part I am quietly pleased about, because local name resolution has been a low-grade nuisance on my network for ages and anything that makes it more predictable is welcome. But I am not going to pretend that is front-page news. It is just a good Tuesday.