Datenwolf looks at the problems from the sysadmin perspective and all he is seeing and complaining about is complexity that introduces hard to debug problems and bugs. Sure his criticism is derided by Lennart as he has far more intimate knowledge about the software. However 'Did you file a bug?' and 'Do you hate disabled users?' actually are not helping here either.
It's complicated and modern Desktop Linux userland (ConsoleKit/PolKit/dbus/pulseaudio/GNOME...) feels like a sub-optimal solution from a sysadmin perspective. Maybe systemd will fix some of these issues but some are afraid it will introduce others.
Having dealt with "modern Desktop Linux" I can totally relate to Datenwolf. I understand that he lacks knowledge but if you are not a developer of these systems it's impossible to deal with this mess. And yes - I think it's largely a undocumented, poorly implemented and overly complex mess.
And that seems to be the eternal issue. Poettering seems to come at this not from an admin perspective but from a developer perspective. And with RH shifting their focus more and more towards cloud computing, so is systemd.
And in cloud computing the admins are not kings, but peons.
Peons tilling the cattle (server) farm of 100s of servers, each running any number of VM instances. If one or more of those instances go down, new ones are spun back up.
And you see this attitude within systemd. If a daemons crashes you don't leave it down and try to figure out why it crashes. Instead systemd will simply restart any daemon it finds not to be running when it should.
Basically this is not uptime by way of applying carefully measured and maintained administration. This is uptime by machinegun.
It is what allows anyone with some grasp of php to rent server time on a amorphous blob like the Amazon EC2 and spin up the next Twitter or Facebook.
No need to optimize for or maintain the hardware. If the current load is too much, wave your credit card and have 1000 fresh instances behind the the load balancer, courtesy of Amazon or RH.
Datenwolf is demonstrating the kind of exasperation that have in the past driven people to look for alternatives to Windows. They thought they had found the promised land in Linux, but now the blight is coming over the walls and taking up residence even here.
Because no one knows what they are talking about. There is a difference between a bug that's inside a program and a bug that's caused by the unexpected interaction of half a dozen programs. The only people who will see the later are those who actually use programs in the wild, e.g. system admins and not developers.
Which is exactly his point. "Have you submitted a patch" followed by "it's not our bug" doesn't help, it's down right idiotic when it comes to emergent bugs. Everyone involved can honestly say "not my problem" while you're stuck with a dead system "because you have the disabled".
Mr.Poettering et al seem not to realize that bugs don't just come from within programs but from how programs interact with their environment. When you hide all the details behind a single monolithic process it is either "this works" or "this doesn't" and there's nothing you can do about it. When it's a shell script trying together 20 programs it's "this kind mostly works apart from when x happens, which I can have a check for".
> When you hide all the details behind a single monolithic process it is either "this works" or "this doesn't" and there's nothing you can do about it.
The exact kind of behavior that has driven many from Windows to Linux.
