Systemd, ten years later: a historical and technical retrospective

Article note: This is a really good piece, it's thorough, very historically grounded, and makes cautious proposals about how things might improve with new infrastructure and ideas. It takes the same "Systemd does some things well, but it is neither UNIX-y, nor a perfect solution that should be considered permanent" stance I've been on for a decade (and people don't seem as mad about this as when I said it in 2012). The hideously stateful dependency-hell situation (3.5) is a big "Systemd is probably not the right abstraction" tell. As far as the closing, I'm still hopeful that if someone actually solved the problem in a consistent way it might gain traction on technical merit instead of simply as a matter of political circumstance and incremental advantages like systemd did, but I also wouldn't bet on it.