On Wed, Aug 15, 2012 at 10:02:57PM +0200, fredbezies wrote:
Last threads on systemd was useless.
I disagree. In the last thread, I had to really dig for outside information to understand both sides of the argument. My research and tinkering has lead me to the following valuable conclusions: Init scripts are wrong. Sure, you can say they're adequate, that things start-up as you want them to. But that's not the same as them being right. They're gratuitously stateful, they leave you with a lot of uncertain state (can has idempotence?), they are large amounts of delicate code that only provide the illusion of stability because they've been so prolific and so heartily tested by a bounteous supply of users. systemd is also wrong. Pretty much, all anybody can say about systemd is that it isn't init scripts. This much is certainly true. It still does unnecessary parsing (place for bugs to lurk), it is highly coupled (having dependency on outside software including the kernel), and it goes to unnecessary lengths to nurture sloppy daemon developers. The arch dev's are making the right call. They can't maintain a fork of all the software that's going to be coupled to systemd. We're going to have to accept it sooner or later. systemd isn't really any better but I'm unsure if it's any worse. It's okay if we move the bugs into a place that upstream is more inclined to look at and fix. I know you guys can't be convinced to use daemontools and I'm not sure if you should at this point. Making the right decision is therefore not an option. You should just go with the wrong decision that's easiest for the movers-and-shakers to live with. I got all this from reading the several monster threads. Arch-general seems to be working for me.