On 01/19/2015 10:42 AM, Lonnie Olson wrote: > Tod was correct. I was primarily objecting to the interview as > opposed to systemd in general. It didn't cover what I thought it > should, and it felt very much like a puff piece for Lennart.
Yes okay, point taken, though I disagree with the assessment of the interview. Maybe I'm colored by other interviews with Lennart where he articulated very well the reasoning behind his work. > The arguments against systemd that I find interesting enough to place > some seed of doubt are: > > * Re-inventing the wheel concerns like that of networkd. I haven't > had a system with networkd to try it, but it feels bad. Can't we try > the upstart idea of handing off network stuff to another system that's > "tried and true" and continue with dependencies from there? Except this isn't really true. Short version: most of us will never install or use networkd. networkd is really designed for a container environment, virtualized servers, or other specific needs. Certainly in this space, there are probably not a lot of tried and true mechanisms to deal with networking and container lifecycles. It's not meant to do general networking for a desktop or laptop. It cannot do WiFi like NetworkManager can, and I doubt it ever will, unless that because something necessary for a virtual server. > * journald replaces a bunch of functionality of a syslog tool, but > rarely enough to go without a real one. That means I have to run both > since journald is required. I recognize the reasons some logging > entity is required for early boot in systemd, but shouldn't be > required late boot, or it should implement enough functionality to > replace at least 80-90% of the functionality of a real syslog daemon. > (log to files, accept or relay UDP messages) journald can certainly work in conjunction with syslog. In fact any enterprise distro is probably required to have a normal syslog for audit purposes (RHEL7 does by default, and it also has journald). journald certainly does things that syslog does not, such as allowing fine-grained logging and debugging from the very start of init. I think the idea of journal is a good one; I just wish they'd use an on-disk format that wasn't an opaque binary dump; perhaps this will change in the future. But like I say, syslog is still available, even if it's not on by default in distros like Fedora. Some people compare journald with the Windows event log and viewer, but this is rather ignorant I think. > I guess that's about it so far. Pretty minor gripes overall. Also > probably just growing pain issues. Reminds me of PulseAudio, some > ugly growing pains very early, but can't live without now. Very true. I do tend to respond to people complaining about systemd because it seems like there is a lot of misinformation about. I myself came out rather hotly against systemd a few years ago on this very list! That was before I knew anything about it. Hence I was publicly and rightly castigated for that. /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
