My main criticism of systemd is that the project doesn't seem to have any bounds. They are actively trying to re-implement everything instead of having a clear

My main criticism of systemd is that the project doesn't seem to have any bounds. They are actively trying to re-implement everything instead of having a clear definition of what problems they're trying to fix. That tends to end badly, usually in a unportable web of dependencies and components being shoved in the wrong places (D-BUS inside the kernel just doesn't seem right to me). Having said that... It does have some nice things. It's very easy to create robust service startup definitions (units) without resorting to brittle shell scripts checking PID files and whatnot. Also, services start in their own resource groups for better resource allocation between them. These alone have changed my opinion about systemd. These solve real sysadmin issues.