Slashdot Mirror
Is The Linux Desktop In Trouble? (zdnet.com)
"I believe that, as Microsoft keeps moving Windows to a Desktop-as-a-Service model, Linux will be the last traditional PC desktop operating system standing," writes ZDNet contributing editor Steven J. Vaughan-Nichols.
"But that doesn't mean I'm blind to its problems." First, even Linus Torvalds is tired of the fragmentation in the Linux desktop. In a recent [December 2018] TFiR interview with Swapnil Bhartiya, Torvalds said, "Chromebooks and Android are the path toward the desktop." Why? Because we don't have a standardized Linux desktop. For example, better Linux desktops, such as Linux Mint, provide an easy way to install applications, but under the surface, there are half-a-dozen different ways to install programs. That makes life harder for developers. Torvalds wishes "we were better at having a standardized desktop that goes across the distributions."
Torvalds thinks there's been some progress. For software installation, he likes Flatpak. This software program, like its rival Snap, lets you install and maintain programs across different Linux distros. At the same time, this rivalry between Red Hat (which supports Flatpak) and Canonical (which backs Snap) bugs Torvalds. He's annoyed at how the "fragmentation of the different vendors have held the desktop back." None of the major Linux distributors -- Canonical, Red Hat, SUSE -- are really all that interested in supporting the Linux desktop. They all have them, but they're focused on servers, containers, the cloud, and the Internet of Things (IoT). That's, after all, is where the money is.
Linux desktop distros "tend to last for five or six years and then real life gets in the way of what's almost always a volunteer effort..." the article argues. "It is not easy building and supporting a Linux desktop. It comes with a lot of wear and tear on its developers with far too little reward."
His solution? Having a foundation create a common desktop for all Linux distros, so the Linux world could finally reap the benefits of standardization. "This would mean that many more Linux desktop developers could make a living from their work. That would improve the Linux desktop overall quality.
"It's a virtuous cycle, which would help everyone."
"But that doesn't mean I'm blind to its problems." First, even Linus Torvalds is tired of the fragmentation in the Linux desktop. In a recent [December 2018] TFiR interview with Swapnil Bhartiya, Torvalds said, "Chromebooks and Android are the path toward the desktop." Why? Because we don't have a standardized Linux desktop. For example, better Linux desktops, such as Linux Mint, provide an easy way to install applications, but under the surface, there are half-a-dozen different ways to install programs. That makes life harder for developers. Torvalds wishes "we were better at having a standardized desktop that goes across the distributions."
Torvalds thinks there's been some progress. For software installation, he likes Flatpak. This software program, like its rival Snap, lets you install and maintain programs across different Linux distros. At the same time, this rivalry between Red Hat (which supports Flatpak) and Canonical (which backs Snap) bugs Torvalds. He's annoyed at how the "fragmentation of the different vendors have held the desktop back." None of the major Linux distributors -- Canonical, Red Hat, SUSE -- are really all that interested in supporting the Linux desktop. They all have them, but they're focused on servers, containers, the cloud, and the Internet of Things (IoT). That's, after all, is where the money is.
Linux desktop distros "tend to last for five or six years and then real life gets in the way of what's almost always a volunteer effort..." the article argues. "It is not easy building and supporting a Linux desktop. It comes with a lot of wear and tear on its developers with far too little reward."
His solution? Having a foundation create a common desktop for all Linux distros, so the Linux world could finally reap the benefits of standardization. "This would mean that many more Linux desktop developers could make a living from their work. That would improve the Linux desktop overall quality.
"It's a virtuous cycle, which would help everyone."
And we have that, in spades. Gnome, MATE, KDE. Sure, they have variations, but Windows had variations between releases. Concepts stay the same - menu button or ribbon with launchers in it or combination thereof. It isn't a software usability issue per se. Personally, I prefer MATE, and I prefer it the way Mint ships/configures it.
Not that it is my place to put words in Linus' mouth, etc. but it seems that what he is really complaining about is the package management landscape, the variations in libraries and versions and compile options, etc from distribution to distribution. Even starting with one of the Big Distros like Debian, you never know when/what Ubuntu (and therefore Mint, etc) will grab when they pull from -testing or -unstable to start their next release. The only real place you have cross-distro compatibility somewhat guaranteed is with true parent/child distros like Mint and its matching Ubuntu release that it shares package repos with.
Don't blame me, I voted for Kodos
XFCE are a pretty example of robust GUI standard (my desktop is practically the same for more than 10 years now...)
This thread perfectly illustrates the point by the OP. Everyone immediately turned on each other claiming that "THEIR WAY was the ONLY way" and "all other opinions were shit." Everyone immediately went to discussing the underlying technology of their preferences vs the point that THAT is the problem. Nobody gives a flying **** if you've been compiling your own desktop environment and workflow for the past 30 years. Nobody cares that YOU like x package manager over y. Its irrelevant. You aren't more or less linux than anyone else. The Linux community is virtually without equal in its ability to cannibalize itself with infighting and elitism. The major survivors, Ubuntu, RedHat among a few others quickly realized that trying to unify the rabid base into any cohesive strategy was pointless and worthless. Too much vitriol. I'm not the biggest fan of Linus at times but he is on point here. Of course the opposing point of view that Linux doesn't need a standard desktop is just as valid. There are plenty of "easy button" Linux desktop solutions in the marketplace and a little bit of research will show that basically everyone can get almost anything working on nearly any flavor. Rant over
The end user interface is the last thing that needs standardization. Desktops that look or act differently aren't the problem. What needs standardization is the back end API. There should only be one way for the installer to interface with the desktop manager for adding a new program icon. One way for a program to register its "settings". A single "control panel" where any program can add its configuration settings to. There should only be one form of IPC. One way for a printer to register a driver.
Once those issues are solved, once we have a rock-solid core set of standards there, then there can be a million distributions that look and feel different, that distinguish themselves by catering to X, Y, or Z. It won't matter. Any program will still be able to run on any of them, because they may look and feel different, but they will act and be configured the same.
Monoculture for UI is stifling. Monoculture for API is liberating.
I use XFCE on all my computers, since the early days of Gnome 3.
I love it. And bless it, for it has never changed and has continued to work well. Software perfection.
I don't use desktop icons. I use a full width bottom bar with a menu in the corner. Focus follows mouse. 10 virtual desktops. Many instances of xterm.
My wife uses it and she probably doesn't even know it. Or what OS is running. That's another part of software perfection; users who don't care don't even know about it, they only know about their applications.
Change is great when a tool doesn't work right, but it is often toxic when the tool already works.
OP says "we need a standardized user interface".
Your reply is "we have three of those standardized user interfaces". Looking in Wikipedia, I found this:
On desktop systems, the most popular user interfaces are the GUI shells, packaged together with extensive desktop environments, such as KDE Plasma, GNOME, MATE, Cinnamon, Unity, LXDE, Pantheon and Xfce, though a variety of additional user interfaces exist.
A standard is a standard. One single thing. Not eight. Certainly not eight over umpteen distros.
I have an old ASUS Eee PC 1005HA which came with Windows 7 Start or something like that. In time, that ugly-ass sticker with the license key has faded away so I installed a Linux Mint distro on it. We plan to use that Eee PC in the kitchen, to look up recipe instructions while cooking, and my girlfriend was asking me about its OS. She's a Windows user and so am I (most of the time). I was telling her it has Linux installed and if she doesn't like the interface, there are others around. She asked "so which is better?" - hell, I don't know.
Now, if you have a normal PC user, who knows just enough about an OS UI to be able to configure the OS and use it without asking for help, how would you present these user interfaces and the difference between them? When faced with a choice between multiple software solutions I tend to construct a table having the solutions in columns and their features in rows, with each cell marked on or off showing whether A certain solution has a certain feature, compared to the rest. In this case I realized I don't know what the difference is. I'm not sure I should care, either. So why, then, do these competing solutions even exist? They don't compete commercially, because they are free to use. They don't compete from a functionality perspective, because (and I make an assumption here) top 30 UI features for any modern interface are present in all of them. So why have all those solutions, if the top reason to use one over the other is personal preference? Which, by the way, needs to be developed, and a new user (or a converted one) doesn't have.
Last thing I need when switching to Linux Desktop is a consultant to help me decide which user interface better suits me. I would very much like to install a distro and have a way to choose between the eight user interfaces above, on the fly, by choosing from a menu or something, much like themes work on an Android phone. Then yes, it would indeed be a matter of preference.
I remember when Windows 8 was released, with their new Tile-based desktop and their horrible choice of redesigning Settings, a half-assed implementation which destroyed usability. Even today, with Windows 10 v.1809, Settings are a mess. Half of them are present in the "new" UI, and half are still in the classic UI (which was way more functional, if you ask me). I, the ever-desktop-click-and-OK user, had to rely on PowerShell or Command Prompt a lot more to change settings, because the UI way was more frustrating and slower. So, yes, there is ample opportunity for Linux-based desktop UIs to replace Windows-based UI from that regard, but fragmentation is one of the big hurdles.
...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
(And another annoyance - Torvalds sees Snaps and Flatpaks as the "solution" to the package management/distro issue? Really? Yeah, let's just replicate the userland for each application you install to deal with what was a non-issue.)
Yes really. It's the natural end game for an entire system where libraries are maintained and update completely individually and programmers are forced code against a moving target. This shouldn't be a surprise. The whole point of a distribution, and what makes the maintaining of a distribution so difficult is the endless juggling of new versions of software and libraries and the inevitable incompatibilities between them.
If you want the most up to date software where you can happily install without any affect on your system what the vendor provides on they day of release then your only safe solution is a packaging system like Snaps or Flatpaks. The alternative is screwing with your system in ways the distribution maintainer doesn't expect.
The only time I've ever given up trying to repair a Linux system and flat out reinstalled the OS (aside from obvious malicious damage like deleting root recursively) was when someone years ago tried to get the latest version of some CCTV software on their Debian system. The distro version didn't support some feature so they added a repo for the current version, installed it, force updated some libraries, and by the time he was finished X stopped working, and the entire apt database was so screwed up that it was basically impossible to revert to a working system thanks to the library structure of Linux.
Snaps didn't get created in a vacuum. They are a solution to a real problem.