Systemd-homed: Systemd Now Working To Improve Home Directory Handling

[intelfx]

Is this sort of thing possible to use with Flatpak apps?

As it stands now, all desktop applications get hit by exactly the limitation I described. Long story short, you would have to checkpoint the app together with the X server (or Wayland compositor) — and, by extension, every other app that's connected to the same X server/Wayland compositor (same principle, but in reverse). DBus doesn't exactly help, either.

For this to work, CRIU would need to learn to extract X11/Wayland state from the display server/compositor (and inject it back upon process restoration), which is much easier said than done.

[starshipeleven]

Gnome
Linux
and
systemD
Operating
System

It sounds like a neat concept but we all know why it's a fatal combination.

(Explanation and spoiler for the game Portal: GLaDOS is a sentient AI which got so obsessed with carrying out its tasks that it forgot about the scientists it was running them for, killing them all in the process.)

Glados was always trying to kill them "within one picosecond of its activation" as said in the game, killed off a bunch of people in a few incidents, and then learned to lie and the scientists believed its lies and gave it the neurotoxin it used to kill them all and take control of the facility. https://half-life.fandom.com/wiki/GLaDOS

I would call your comparison invalid, as such.

[Djhg2000]

Glados was always trying to kill them "within one picosecond of its activation" as said in the game, killed off a bunch of people in a few incidents, and then learned to lie and the scientists believed its lies and gave it the neurotoxin it used to kill them all and take control of the facility. https://half-life.fandom.com/wiki/GLaDOS

I would call your comparison invalid, as such.

One picosecond is plenty of time for an AI at the scale of GLaDOS to forget why it's doing what it's doing.

[starshipeleven]

One picosecond is plenty of time for an AI at the scale of GLaDOS to forget why it's doing what it's doing.

Dunno, I sometimes also forget why I'm doing what I'm doing, but I don't descend into homicidal maniac mode and start killing off everyone in range.

Also, the ability to lie to secure neurotoxin to then use to kill off the scientists does establish causality links. That's a plan it set in motion, it's not a mistake, it's not random.

[AsuMagic]

I've got to admit I didn't expect a systemd thread to derail into a Portal shitpost.

[Alliancemd]

Sounds like another nice feature that people will hate on...
Go systemd, give me relocatable home directories, make backup simple

[F.Ultra]

One picosecond is plenty of time for an AI at the scale of GLaDOS to forget why it's doing what it's doing.

Actually the larger scale of the AI the higher the latency since C is and will always be a constant. In one picosecond you can only transmit data at a distance of 299792nm and that would be a very small scale AI :-)

[Danielsan]

I've never though that I would prefer be a homedless in my life...

However it is clear that no one of these changes introduced by Poettering serve to make GNU or Linux better, it is just the way how, since ever, Red Hat holds its dominance on GNU and Linux. The only thing is changed since before the introduction of Systemd is the fact that today there are more players on the scenario and thus Red Hat became more aggressive with its usual behavior...

[oiaohm]

I've never though that I would prefer be a homedless in my life...

However it is clear that no one of these changes introduced by Poettering serve to make GNU or Linux better, it is just the way how, since ever, Red Hat holds its dominance on GNU and Linux. The only thing is changed since before the introduction of Systemd is the fact that today there are more players on the scenario and thus Red Hat became more aggressive with its usual behavior...

Really would pay to look closer. Poettering is good for Linux as in the Linux kernel and most likely not good for freebsd or gnu long term. I will detail why.

Lets take pulseaudio. Lot of people don't remember or issue logs from before pulseaudio. People who say they use Alsa instead of pulseaudio today have a lot to thank pulseaudio for.

Before pulseaudio we had multi sound servers basically going down the route of a sound server per DE(desktop environment i am going to use this short a few times) . Just like we have people distro hopping because X feature does not work we use to have people DE hope because their audio did not work. We end up with fragmented solution and a lot of works for me DE outcomes. It was like 1 person tries DE 1 there audio is not working right they would switch to DE 2 maybe after multi DE switches then they would find a "Works for Me" now they would recommend DE 2 to person 2. Person 2 has the same kind of issue and after a lot of DE switching they would find a "Works for Me" on audio DE1 now think person 1 was a idiot not catching up they had just found another "Works for Me" solution.

The cause of this problem was found by pulseaudio. Pulseaudio resulting in fixing a lot of low level kernel sound drivers. Why pulseaudio documented that when X audio driver go X instructions it did X yes Y audio driver got X instructions it did Y when it should do X. This was a long list of what caused the "Works for Me" problem in the Audio stack. This was work done under Poettering lead.

Now lets look at systemd effects on lower down Linux kernel. Systemd is why we have cgroupv2 because once systemd started using cgroupv1 heavy again it demoed that it was a "works for me" if I am lucky solution by design. Kdbus a lot think as a failure if you look closer is not there were a lot of options people said for IPC that could give the performance that systemd developers were after for dbus that were also in a "works for me" if I am lucky state.

Then we have the big blow up over how the LInux command-line should be used over systemd. Techically systemd did nothing wrong as there was no rules so it was a "works for me" solution.

I could keep on going. Something Poettering is good at is finding a "works for me" solutions and providing pressure to that weak point to get a "works for everyone" solution in the Linux kernel. This homed would be another one to apply pressure all the faults homed has happened to everyone who attempts to run a business network with shared home directories on Linux if the machines are not all under a single central management. Heck how even if it is under a central management it can still go bell up.

Poettering is really like the real world bull in china shop for the Linux kernel. A real world bull can in fact walk though a china shop without breaking a single thing and that is the most common outcome Mythbusters and a few other parties have proven this. But looks as scary as hell of course so causes those who are not aware how a real world bull will behave to take extra protective measures. So Poettering being Poettering does the kernel a lot of good.

I am not going to say he is all good. For freebsd/"other operating systems", "other distributions" and gnu Poettering really does not give a rats about either. His wage is paid by Redhat like it or not he has to keep his employers happy. Redhat only support Linux kernel devices so Poettering cannot be expected to work on the other stuff. This means he has a Linux platform first policy. Some ways this is providing active pressure to those other solutions to have people from their camp to step up to counter Poettering.

[tildearrow]

In one picosecond you can only transmit data at a distance of 299792nm and that would be a very small scale AI :-)

Are you sure? Doesn't current slow down depending on the material?