>>31The problem with this is twofold. First, these supposedly optional components don't expose any standard interface (where "standard" means "having more than one implementation"). Second, said components are maintained together with the init system so there are far fewer practical impediments to their interfaces being unilaterally changed. Thus there is significant risk of the whole Linux user space becoming dependent on components with unstable interfaces. One can argue the same is already true for sysvinit scripts, but at least with sysvinit it's easy for non-developers to change the scripts.