Ever really destroyed your server because the it needed were available? I have. It was so much worse than a boot process that froze.
If Systemd was pausing due to a network share being down, it’s only because I (or you) told it to do exactly that. There are lots of good reasons to delay the boot process until all drives the system expects to be there are actually there or the network is up. Cleaning up the mess that happens when the system does not check these kinds of things at boot is so much worse. It’s never really some nebulous thing. Like it or not, intentional or not, the machine is doing exactly what you asked it to do and a delayed boot or a boot halted until you can solve the real problem is almost always better (or at least safer) than the alternatives. I’ve experienced all the things you’ve mentioned, dealt with each of those issues, and it was so much more of a hassle to diagnose before Systemd.






I’d like to see ideas like this make a comeback, hopefully with some modifications this time around to protect our privacy and resist corporate exploitation.
We used to use del.icio.us and other variants to do exactly this before browsers had profiles. Back then, its primary draw was that you could take your bookmarks with you anywhere to any machine (this being before that function was baked into browsers and before web browsers could be carried in your pocket). The secondary effect was that you’d share and tag those websites with your own categories/descriptors, thus crowdsourcing a new version of the old web’s link directories using Web 2.0. You could browse through symantic tag clouds to discover new things. Del.icio.us was for websites, but people were tagging and logging all of their favorite stuff and sharing it online so that like minded strangers could filled the gaps in their cultural awareness. We tagged our books with librarything. We tagged recipes with recipe thing. Audioscrobbler (later known as last.fm) logged our music listening to automate the tagging, not by direct symantic tagging, but by relational/temporal coincidence. If other people that listened to a lot of the stuff you listened to and they also listened to some other stuff you didn’t, those became recommendations for you. That kind of relational algorithm would survive the slow death of Web2.0 to become the backbone of recommendation services like Spotify and probably even TikTok.