<html><head></head><body><div>On Thu, 2015-09-10 at 14:25 -0400, Jim Kinney wrote:</div><blockquote type="cite">It seems that once I dig past the issues of the personalities of the devs, the complaints are rather thin. It would be highly impractical to have the same command set used for two different outputs from two different inits. It's not like systemd suddenly appeared 6 months ago. It's been around more than 4 years. There's a global array of people with a good track record of picking new technologies to work into Linux distros. From my perspective, there's very little that's hit the big "FAIL" button and a tool as central and vital as the init process will have been beat up on operational theories before being accepted into use by the top 95+% of installed Linux distros. So, time to learn new tricks.</blockquote><div><br></div><div>Agreed.</div><div><br></div><div>As I mentioned previously, systemd (with its networkd, journald, and udev) use less memory post-boot than busybox (with its mdev, no syslog, and the Debian network management scripts). By itself, that's very valuable on an SBC if you need every last page of RAM.</div><div><br></div><div>The binary key-value store which the journal provides allows applications to use a universal interface to create rich log output. No longer do you even have to serialize your output for the system logger: you save a bunch of RAM by just putting your binary data in the log files, verbatim. Move the expense of stringifying to the developer's or operator's workstation, and not the embedded appliance that wants to be very power efficient and just get its work done.</div></body></html>