Comment Re:My opinion on the matter. (Score 1) 826
Again, the scope of the systemd project isn't only to be init system; the scope is to provide a coherent base OS layer that controls all processes from boot to shutdown, enabling a stateless zero config boot if needed, including any OS containers booted from the main OS.
You can't have good service management without good logging, and you can't have good logging without an integrated logging systemd like systemd's "journald" that provides early boot logging info; syslog can't receive logging info until fairly late into the boot.
syslog has been outdated for decades; it doesn't provide really structured log files, anybody process writing to syslog can pretend to any other programs (journald gives kernel guarantee that the named services in the log files, are actually those that wrote the log entry), it works badly with multi language logging info, the logfiles are scattered all over the place (with systemd all logs, including utmp and xsession logs are stored in one central logfile). The worst part is, that since the logfiles have no real structure, the actually text output have become an API: change the daemon's output from "failure" to "error", and thousands of log watching scripts will break. systemd's logging facilities are a huge improvement over the old legacy syslog system.
Systemd is all about starting processes and services, including controlling whatever is needed for the system to boot, to ensure those processes are started correctly, at the right time, supervised and logged, and that everything needed is provided in the correct order and as automatically and fast as possible. Basic things like network and ntp control are part of this too. Mind you, you don't have to use the systemd version of sntp, or dhcpd, just use your own sntp or full ntp daemon.
That systemd provides such daemons is simply because, that when launching eg. 5000 OS containers at the same time, then it really matter whether it takes 50 or 5 seconds to get a dhcpd lease. And the systemd daemons are designed to be really lightweight, and really fast.
Again, most work systemd for a long time have been about OS containers; the long time goal is to launch any unmodified Linux OS as an OS container in a secure, sandboxed manner.
There is still a lot of work to be done (with safety), stuff like kdbus and cgroup is part of that effort.