I don't think I'd try Wayland desktop on a low-power box like a Pi, I would probably go command-line, especially for Apache and file-sharing. I'm not really familiar with Wayland and all its talking points, I decided to try it after running x for many years, and I can't say I've been disappointed. Maybe it's too bad that corporate interests (IBM, née RedHat) have taken over great swaths of Linux development, and bent it to their needs. OTOH there seems to be a paucity of developers who will work for free.

First commit Sep 2008, first release Feb 2012, API stability Jul 2013.

:/

Long before Wayland reaches a satisfactory level, it'll have been replaced by SystemD's display server..

I mainly run everything headless anyway. The default for Raspbian (Debian) is now Wayland, although you can choose to switch back to x11 with the raspi-config tool. I'm running x on both machines at present but neither of them are on the latest distro. I suppose I should update my Nextcloud box as this is pointing at the web.

I have no idea how IBM run their Redhat business: probably every bit as heartlessly as any other vast corporation. I used to have a lot of time for the IBM people I worked with who all seemed pretty decent people. But then again, that was mainly on DB2, and DB2 is a world within a world at IBM, and a huge number of the staff are enthusiasts who love working with DB2.

Linux gaming is amazing currently.

There are some Linux native games but game devs really don't know what they're doing with Linux so Proton (Valve's Wine fork) is usually preferable.

Virtually everything on Steam runs in Proton. The only exceptions tend to be stuff with very invasive anti-cheat. There is support for that in Proton but the dev has to toggle it on.

Everything else runs great. Generally within singe digits fps of Windows performance, sometimes better than Windows, and improving constantly.

I game a lot and I've not touched Windows for well over ten years.

The Steam Deck has helped too. Virtually everything was running in Proton anyway, but now devs care about Proton performance.
 

Haha. I'd kinda welcome that. Much as systemd can be annoying at times, their shit *works*.

Wow. I must try that!

Except when it deletes your entire home directory, refuses to shutdown, falls-back to Google DNS servers, misuses private Google time servers, etc...

The google DNS servers was a weird choice, yeah. The rest I've never encountered.

In what situation does it delete your home directory? :')

Do you have to be using homed?

Call me insane but, I have this hunch Peter is not your average Linux user.

I don't believe it was dependent on homed (there's no mention of that being the case).

It's caused in v256 by trying to remove temporary files with a tool called "systemd-tmpfiles" which is described as "systemd-tmpfiles, systemd-tmpfiles-setup.service, systemd-tmpfiles-setup-dev.service, systemd-tmpfiles-clean.service, systemd-tmpfiles-clean.timer - Creates, deletes and cleans up volatile and temporary files and directories"

If a user decided to use --purge to clean up those files, but doesn't know that this tool also manages the /home directory by default, that's their fault for not reading the entire man page and every config file it mentions to discover this. (Of course, there was no mention of it managing /home in the man page, and even after the issue was raised and closed it still doesn't explicitly mention that but dances around with things like "Historically, it was designed to manage volatile and temporary files, as the name suggests, but it provides generic file management functionality and can be used to manage any kind of files".)

There was also another bug in an earlier release, if using systemd-tmpfiles in the intended way, which destroys the entire root partition, because the incompetent programmers allowed "/path/.*" to not just match dotfiles, but also match "/path/.."

In both cases, there wasn't just the bug, there was a shitty reaction, claiming there's no issue and blaming the person raising the issue, rather than just shutting up and fixing it.

> Call me insane but, I have this hunch Peter is not your average Linux user.

*shrug* I try to pay at least superficial attention to what major parts of my system are doing - not entirely through choice but because it seems to be the only way to keep it being somewhat my system.

Three decades ago paying attention to what things do wouldn't have been atypical amongst people using Linux systems?

Again, I'm very out of the loop with all things Linux, but it didn't take me long to discover a very abrasive relationship between some in the Linux world and the systemd devs. It also seemed likely that this is in part (at least) to the social skills/attitudes/agenda of Lennart Poettinger. 

I accidentally nuked my downloads directory (of ~4-5 years worth of assorted downloads) a few months back. It had stuff I wish I could recover, but rm -rf is pretty fucking final. I was trying to do a backup of it but lost track of which directory I was working in. Gone.

I think systemd exposed a tension that was emerging in Linux between wanting things to be simple to use and simple to understand. And you can't really have both.

Systemd's init is really simple to use. Give it a unit file and it basically handles the rest. Don't have to worry about dependencies or ordering or what happens if the thing crashes, it just handles all of that and works well.

But if you want to know what your init is up to - what it's doing when and why, it's a nightmare. It's virtually a black box.

It really represents the professionalisation of Linux. It's less something to play with and understand and more something you can stick on a 'workstation' and not worry about.

I use a lot of systemd stuff, some of it is very good. I like its networking stack and its timers for example. But that professionalisation does rub me up the wrong way. I don't want corporatised Linux.

If it gets too bad, there's always BSD.

Speaking of which, this is an interesting and balanced talk about Systemd by a BSD dev.

(I know you basically said you're not interested but someone might be!)

That was *kind of* the feeling I was getting in my meandering around the topic. Not so much that I'm not interested as I don't understand what systemd is doing as it does it - or a lot of Linux these days. 

Yeah, that's an interesting talk. I was coming round to thinking what he had to say about Lennart Poettinger, ie death threats are not cool. But at the same time, Poettinger seems like a guy who wants to corporatise the start up and management of linux, but can't abide the corporate controls (such as thoroughly logging and investigating bugs) that go with that. He also seems every bit as resistant to change that he isn't making as any change-resistant linux dev, eg changing from using Google DNS and Time servers as a fallback. I saw an exchange with somebody who raised that as an issue, pointing out that Google doesn't necessarily share the aims of every linux user, where he closed the bug report with: 
 

Christ, what's next? You accuse us of controlling people's minds with vaccinations we get directly from Bill Gates? And that systemd uses 5G to spread CoV-2? I don't think we need the input from the script kiddie peanut gallery here.

But, whatever, a nice run through of the history and how systemd got where it is. I suspect it's there to stay judging from the pies it has its fingers in.

I'm not a big fan of the "if you don't like it write your own" argument. It's a bit disingenuous. Not many people have the resources, time, financial support to do that, and certainly not the clout to get their version incorporated into multiple distros. 

Not sure how I feel about containers, and I say that as somebody who runs nextcloud from a snap which is next door to the same thing. I thought he rather side-stepped the question from the guy who mentioned dll-hell.

 

Yeah, generally I like that the people behind FOSS projects are actual humans with human virtues and failings. Some of them are lovely, some of them are dickheads. Rather than stepford PR pleasantness of corporations, like.

I don't like it when FOSS (or adjacent stuff) gets corporate or startuppy. The Mastodon dude leans that way.

It's nice when devs are attentive to their users but I also don't mind it when they're like: I made this, it works for me, you can use it if you like but it is what it is, I'm making the decisions.

The problem with Pottering is that he's doing that latter *within Redhat*, which, as you've said, is just kinda weird.

"Write your own" is bullshit, yeah. At least when aimed at individuals. And he *totally* sidestepped the DLL thing, yeah.

I think the interesting part of the talk was just identifying the necessity(?) for a system layer which takes care of all that semi-persistent state stuff on modern computers. I buy it *up to a point*. As soon as he gives examples he kinda loses me because he's not describing anything like how I use my computer. I *think* he's describing something like a phone.

I think flatpaks and snaps, and also wayland and systemd to an extent, are really about making Linux more friendly to corporations and proprietary software vendors. Flatpaks and snaps make no sense to me whatsoever in a package-managed system. Same with docker on servers. I use docker images when that's what's supplied but I hate doing so. It's a very unlinuxy (and inefficient and inflexible) way to distribute software.

I like containers in principle. They're a better way to do a lot of what was previously done in VMs. I know Steam makes a lot of use of containers on Linux (no idea what it does on Windows), to essentially give each game all the stuff it needs, and that makes sense - that's the neatest way to do that (but again, it's really working around the flaws in proprietary software. Games in this case).

And I like distrobox. It's a little program that lets you install and run another distro on your running kernel. so if I want to try out the Debian version of some program, or use a program that's not provided by my distro, it's an easy and pretty light way to do that.

Linux is getting more coporatey over time. And I fear that as it gets more mainstream (aided by stuff like Steam Deck), what happened to the web will happen to Linux. Or it'll become Android. As soon as I can run my games on BSD I'll probably hop over.




 

Also, re: dockers etc.

It's not even an especially new of novel idea. I mean, I struggle to see what's different in principal between say a snap, and an old windows or DOS program that installs all it's gubbins, and maybe a runtime or whatever into its own program directory. I'm not saying there aren't differences, but... 

When I started running nextcloud as a snap, of course I was impressed. I'd run owncloud for years and migrated it across boxes and reinstalled it when upgrading distros. then I moved to nextcloud because I'd hit an owncloud error which I couldn't resolve and I was at the limit of free help from the company. Then I found I needed to rebuild the box yet again and the prospect of reinstalling apache or nginx and configuring letsencrypt and all the bollox all over again seemed such a drag, so yes, one command "sudo snap install nextcloud" which did 90% of the work seemed pretty good to me. It has it's irritations, like logs being in a different place and you have to use snap commands to configure things (which I always forget), but I'm not complaining. But I imagine if you're very used to a system managed by apt or rpm etc. then it could be annoying in the extreme.

Back on the topic of systemd, yes, exactly that. I can't see how a layer to manage the launch of things that persist is anything other than a good idea. Then I think, well hang on, so how did stuff work before? Things didn't just fall over or run amok all the time (well yes, some did, but they still do). And when you look, of course there were other solutions. It's just that systemd is sweeping all away. The interesting question is "why is that happening?".
  Or write your own kernel, operating system suite, systemd enquivalent? Call it DinLin, Drew is not linux.

Yeah it's not a million miles away from just bundling DLLs in a directory. We'll go to great lengths to avoid the thing Linux (and BSD) is really good at and everything else is bad at: Dynnamic linking.

Yeah if I were installing nextcloud I'd go with a docker or whatever too. Anything that requires a database and a web server is a pain in the arse. It *shouldn't* be a pain in the arse, I think we fucked up somewhere and containers are a dreadful solution to the problem. But for now, it's easy and it works.

I think the need for a kinda explicit system layer comes with USB, wifi and having multiple devices. Hardware was less transient in the past so things didn't need handling in a dynamic way, really. And you didn't want to take your "account" or whatever elsewhere.

My OS will be written in 100% bash. And it'll be called Dim. Dim isn't Minix.


 

Yes, it's a different world now, and especially with the web as he said in the talk. 

Database, web server, maybe a dash of webdav and that's a recipe for something complicated that shouldn't be. I suppose it's because there are so many dependent components. I lose count of the number of times I've followed a script to install something and broken down because PHP is out of date, or 'command not found' or a folder isn't where it should be because it all changed 6 months ago. Or something. 

Anyway, Dim. That's a properly recursive acronym.