Comment so... (Score 1) 347
How long before they have to replace anything beyond the GNU stuff with something out of the 90s to avoid dragging in the systemd shoggoth via some dependency or other?
How long before they have to replace anything beyond the GNU stuff with something out of the 90s to avoid dragging in the systemd shoggoth via some dependency or other?
The only behavior systemd "expects" is for daemons to talk to it either via dbus or libsystemd.
Everything else have some kind of problem attach that is just waiting for the admin to tun his back on the rack.
LAMP offered a way to rapidly spin up a
Kernel devs have in the past and will in the future (at least as long as Torvalds is in charge, and i fear the day he retires) keep sub-systems and such in the code if they know there are users of it out there.
Damn it, even though everyone says that you should use the ip command for anything network related, ifconfig still works. This because the interfaces between the program and the kernel is still there, after having been labeled "depreciated" for a decade or more.
A basic operating rule for Linux kernel developers is "do not break user space!".
Best i can tell, because it take into account the whole USB address tree when linking a device to a driver.
For me there is a line between cheating on a single player game for my own amusement (poking at ram values in a console emulator for instance) and automating online play against others.
In the first instance, the only one that is potentially hurt is me. This in that i may ruin my enjoyment of a game.
In the second i am ruining the fun of everyone that share the server with me.
Yeah, nofail. Thats up there with their abuse of debug in the kernel command line.
nofail is simply about supressing the error message on a failed mount, nothing more, nothing less.
Have the whole system go into panic mode because of a vestigial USB mount is missing is not sane by a long shot.
When the issue was raised to them enough for Poettering to take notice, their "solution" was a default timeout on every service for both bootup and shotdown.
This then blew up the following Fedora alpha, because they had a service that ran on first reboot after a system update that ended up taking long enough to tripped said timeout...
Their overall development methodology seem more at home with a website than a central OS component.
Iirc, systemd no longer optimize for "spinning rust" as the major devs have all moved to SSD...
Ah yes, SSD. Didn't Poettering recently yank the "spinning rust" optimizations from systemd because all the devs used SSDs?
This will the kernel devs decided to keep an old subsystem around because someone, somewhere, was still using it?
The difference in attitude between the kernel and userspace is staggering. I swear userspace devs are actively user hostile at times.
Heads up, nofail is only about sending an error. Not evacuate the ship because a busboy is missing!
Systemd's handing of nofail is up there with its handling of debug, and likely with the same level of arrogance from the devs.
Likely he has a network mount going, and because systemd can't tell NFS from EXT it yanks down the network before unmounting, as networking is earlier in the dependency graph.
So it will sit there and wait for the network mount to time out before moving on.
I guess having a network mount service alongside a mount service get systemd panties in a twist. Or perhaps Poettering only use dropbox...
It is also the one init that goes into crisis mode if that usb devices you have a entry for in fstab is not present on boot.
A short hint Poettering, when it comes to unix boot, / is paramount, everything else is "best effort". Because once / is up, the sysadmin has the resources to get anything else sorted and mounted.
Oh wait, systemd is part of the effort to push everything into
If you want to make PoetteringOS go do so, and stop shitting all over long established, and time tested, behavior!
The overarching problem is that systemd can decide to rearrange the boot sequence at any moment.
If a admin has set a sequence to be XYZ they have a very good reason for doing so. So why should the init suddenly decide that YZX is the way to go?
Sounds ironic given that systemd (and the rest of the fd.o cohort) seem hell bent on reproducing Windows (or at least Active Directory and Group Policy).
For God's sake, stop researching for a while and begin to think!