The only bullshit part of the saying is that there will be a single year where this convergence will happen. When in truth it has already happened. Linux has great desktop choices for those who would like to use Linux. Those of us who do have moved over to the OS and probably haven't looked back, nor are they looking to bring the general public into it. If the general public wants the plastic toy OS experience then let them burn with that ship. I've spent a lot of time advocating for the move to Linux and more open source systems and voting with your wallet in the computer landscape that my general position is now to talk to someone about it and when they're upset with the status quo, I'll be on the other side saying "I told you so." Then I usually end up helping people when they come back to me.
Furthermore, the commandline is a core part of the computer experience in my book. Just because Apple, Google, Microsoft want you to live on the internet doesn't mean it has to go away or have a friendlier alternative. There will always be a divide between people who have an interest in computers and willingness to learn them and people who just want to like buy something off Amazon. The latter will probably never use Linux no matter how popular it is on their phone or desktop.
It's just another way of saying: it has it's niche and that we should recognize it for what it is, instead of trying to make it the globally recognized OS of the Year for desktop PCs, I guess.
I wanted to use Linux for a long time back in the early 2000s, but it just wasn't up to snuff for a wife variety of daily tasks. Was a great development machine, but as soon as you wanted to do things like audio, or video editing, or whatever, it just wasn't there.
So yeah, I wanted to, but it wasn't ready. Things are much, much better now.
I’ve been using Linux for better than 22 years. It was the only operating system in my home for more than 12 of those years and I pretty much live at the command line. I’ve voted with my wallet and backed open source initiatives over that time too, but there comes a point where you just want stuff to work and you want to be able to interact with the rest of the world with the same feature set they enjoy. This is where Linux falls short. It’s a last mile problem with window managers and their associated apps. It’s further divided between software ecosystems within Linux. Another commenter pointed out how web-based apps are increasingly powerful, and I think that’s helping but there’s still a gap.
I have an interest in computers and I know a whole lot about them, but I want my daily experience to be smooth and to just work and I want to be able to relate to the rest of the world in terms of software. So, I use Linux for work and Windows or OSX everything else.
My point is that your comment seems to say that if people had an interest in computers they would (of course) be using Linux, and that’s absolutely not true.
The article is about when the year of Linux on the desktop will happen. If you want your desktop to run Linux it will. If you want Windows and OSX do you can use those operating systems then you will move to those systems. And my point is that you make of it what you will. There will be no mass convergence of society to a Linux desktop. If Windows and OSX are important to you socially then that’s your value, not everyone’s prerogative.
Honest question as I'm not sure if it's just my own bubble - when was the last time you ever heard someone say "the Linux desktop will take off this year" or something similar? Not "I'm going to switch to Linux" or anything individual statement but something to the effect all of the "Year of the Linux Desktop" articles talk to. I.e. is there really this consistently significant portion of people that think Linux is going to take off in popularity in <current year> or is it just a bunch of people that like Linux try to get more people to try it out during these kinds of release cycle changes.
Sure, I can find many seriously saying just about anything on the internet if I go looking for it. I meant a literal you in "when was the last time _you_ ever heard someone say 'the Linux desktop will take off this year' or something similar?"
I've personally heard "year of the Linux Desktop" from people directly many times per year every year that I can remember but in terms of response/meme/joke/etc not actually taking the stance "it _is_ the year of the Linux desktop". Yet I can't even think of the last time I even naturally ran across someone proposing the concept online (sans trying to purposefully find such a person like Googling) let alone someone I know being serious about it.
Not sure if that is a normal experience and people just like to talk/joke about how it isn't all the time or if my bubble/filter is just abnormal.
It would bounce back and forth for a few years. And then people joked that 2017(6?) was "The year of the Minix desktop" when Intel added minix to their IME.
> especially since server developers switched to macs
doesn't make much sense. MacOS is certainly very relevant, but there has been no exodus.
Here there are some interesting stats:
The 2019 Stack Overflow developer survey provides no detail about particular versions of Windows. The desktop operating system share among those identifying as professional developers was:
Windows: 45.3%
macOS: 29.2%
Linux: 25.3%
BSD/Unix: 0.1%
I think that in the last decade, Windows has lost a very significant market share, which has been taken from both macOS and Linux (the latter having a surprisingly high prevalence).
I almost fell off my chair when I talked to a friend of mine (whose completely no tech ) and she told me she's using Ubuntu desktop. A friend of hers installed it and it ran completely fine.
I think it's because most everything we do now is web based. Also I think people tend to forget how big a deal google docs actually is.
I have a few non-tech friends who have installed Ubuntu themselves and never looked back.
It does everything they need to do, and it doesn't bit-rot like Windows. Anyone who just uses their computer for web browsing, playing music, and other simple stuff just doesn't need any more, and that's been true for at least a decade.
I've used Linux as my only desktop OS for 20+ years (since high school) and Ubuntu for the past 10 of those years (previously Gentoo, RedHat).
It's not yet ready for prime time. When I set up a new machine there are still tons of things I have to "fix" from the command line. Touchpad scrolling. Getting my Jabra to not auto-adjust volume to 0 every time I speak. Modprobe this, modprobe that. Setting up an L2TP VPN. Getting bluetooth audio to support AAC/AptX. Getting Chinese input to work properly. Configuring the extra buttons on my keyboard and mouse to do useful things. Disabling IPv6. Installing CUDA and CUDNN. Creating a virtual camera. Ignoring TCP reset packets (commonly needed when visiting China). Adding support for ExFAT SD cards, which almost all Canon cameras use. Writing files to USB thumb drives without root permissions. Mounting an SMB or S3 volume automatically at startup. Setting programs to start automatically at startup. Scheduling tasks to execute at certain times. I need to go to the command line to do all of these things. On Windows and Mac most of these things can be done graphically.
Oh and even Chrome doesn't install on a default Ubuntu 18.04 system. You have to do "sudo apt-get install libpango" first or "sudo apt-get -f install" after installing it.
I work with users who have never used a terminal in their life, but still need all 3 of those things. I can tell them "Go download X & run it" on Windows. Getting those things to work on Linux would be a monumental feat [for them].
Yeah, Windows is easier. I can provide a user a URL to a Windows installer to skip those steps. (they don't need to uninstall it)
For Linux, the software I described does not exist as a package with an accompanying GUI app (and even if it did, which package for which distro?). That's the thing about these discussions: when a person has a real complaint about something that is very difficult for a regular person in Linux, there is always a theoretical way that Linux is easier, but then in practice it doesn't work.
Well my "theoretical" Windows software doesn't have an MSI installer, and has to be extracted as a portable file where the binary is hidden deep in a release folder.
And also not to mention the Ubuntu GUI software store is slow as fuck. Zero attention was put into making it a snappy user experience. It has in every way been an afterthought. If I double-click a .deb file to open it, it takes a good 7-8 seconds. On an i7 8th gen. That's not acceptable. If it needs to process something it should show up in 0.1 s and show a progress bar.
I don't use it because it's faster to open a terminal and type `sudo dpkg -i foo<tab><enter>`. Most Linux users today probably do it that way, so nobody cared to optimize the GUI "alternative".
"Download this file, extract it (tarballs act just like zips in the GUI), double-click RUNME.sh, enter your password when prompted (gtksu)". Even the tarball is removable if you don't mind telling them how to make the file executable. Use zenity+gtksu and they never need a terminal.
Yeah I think this mentality is why Linux isn't a good desktop OS for many people. With Mac/Win the usual mentality is that almost everything can be done graphically, even if it involves digging into the registry that's still a graphical UI that you can operate with a mouse.
As a Linux user and command line aficionado I'm not saying that's bad, but until that mentality changes it's not going to be for the masses.
I believe the "Linux Sucks 2021" [0] video did a great job of explaining not just how the YoLD hasn't happened, but why it will never happen, now.
Linux is built with a bad architecture (spaghetti monolithic kernel with no stable APIs), which has a tremendously high cost of development, and also a high cost of maintenance. Large companies (e.g. Google) have realized this isn't sustainable, and are shifting away into better architectures (e.g. Fuchsia), which tend to be microkernel multiserver systems.
As money goes away, Linux maintenance cost can't be met, and it'll crumble on its own weight.
I do still hope a good desktop/workstation OS with a FOSS license will happen. It just won't be Linux, because its architecture was never suited for it.
I am not qualified enough to answer for NT and BSD (I only can note they appear very stable nowadays).
But XNU is not much better than Linux. It has a schizophrenic soul with its BSD and Mach duality for everything. It is bogged down by Apple's never-ending "security" lock down efforts. It has no equivalent of "jails/containers". And, again, it is practically unusable for anything that is not created by Apple and is not extendable now that they are phasing out kexts.
I think a very interesting recent effort in kernel architecture innovation is Fuchsia.
What do you mean by "kernel with no stable APIs"? Linux developers, as a policy, don't break user space.
And since when Google became the representative of stable APIs?
During 90s, the de facto desktop OS used a monolithic kernel, so clearly, having a monolithic kernel doesn't prevent common adoption on desktops. Fuchsia is a microkernel, which would play nice with phone vendors with binary blob drivers, but this is a licensing issue not a technical one.
There is an indirect relationship in that only out-of-tree kernel-space Linux drivers are inconvenienced by that, that lots of out-of-tree kernel-space Linux drivers are proprietary ones, and that most kernel-space Linux proprietary drivers are dubious on licensing grounds (to be clear, NVidia seems to be one of the most clean in that regard, but I'm not aware of tons of proprietary drivers as clean as they are).
The thing is: it makes little sens for a kernel under GPLv2 to maintain an internal kernel space ABI for drivers (and I'm not even sure Linus would qualify with the licensing, but maybe I'm more moderate than he is). Ok I'm aware there is a mess at least on the Android side, but maybe the majority of it is questionable on licensing grounds to begin with (but the Linux Foundation is particularly lenient on licence compliance, btw, not surprising when you see the list of corporate members, and for some of the main ones their past practices - hm VMware hm...)
And that being said, I would be curious to see tons of old Windows binary only kernel space drivers continuing to run on modern systems, or ironically even some user space drivers (IIRC, printer drivers are userspace but 32-bits ones do not work on 64-bits systems)... Ok there is more binary compat for a few major versions if you are lucky. But then MS can always randomly requires WDDM 2 for its next OS anyway.
In the end, some old and/or exotic hardware work better (or at all) on recent Windows while some old and/or exotic hardware work better (or at all) under recent GNU/Linux distros, but I'm quite unsure this has anything to do with the precise kernel space API & ABI policies. What does not usually work is attempting to project a model on something that has vastly different premises.
So yes, licensing can influence technical decisions.
If your ecosystem is vastly built around binary only vendors, a more stable ABI is vastly more valuable than if you ecosystem is vastly built around source only free software vendors, AND pushing for in-tree integration on top of that...
>What do you mean by "kernel with no stable APIs"? Linux developers, as a policy, don't break user space.
Sure, Linux tries hard for stable user<>kernel interface.
Within the kernel, however, there's no such thing and it's an utter mess.
>having a monolithic kernel doesn't prevent
Even within monolithic kernels, and even after restricting to unix-like, there's those that are better structured internally (such as openbsd or dragonfly), and then there's Linux, which is an unworkable mess.
None of this is particularly wrong, but none of it is particularly threatening, either. Google isn't going to fix API breakage by introducing "another competing standard"[0], and they're certainly not going to solve backwards-compatibility with it either: which is becoming an increasingly big question in the sector. Yes, Fuscia can promise a better world, but who's going to get onboard with Google? Certainly not the majority of FOSS developers, who have effectively nothing to gain by abandoning a well-supported, mature ecosystem. It might have potential to replace Android, but only because Google's repeated bastardization of the codebase has lead to a pretty unsustainable system.
> It just won't be Linux, because its architecture was never suited for it.
Linux was designed to be abstracted over, and extended for any number of use-cases. It certainly does a better job of this than Windows, which has a terrible habit of hiding things in system registries, creating directory structures that don't make any logical sense, and building completely un-automatable system utilities (ever heard of someone's Windows ricing script? exactly). MacOS is literally just BSD lugging along 50 tons of XNU zombie code that it will someday replace. Ask most developers what they'd add to MacOS, and 9/10 responses are going to be a feature that already exists on Linux (good package manager, performnant docker support, updated coreutils, you name it). So in a way, the current "leaders" are still following in the footsteps of giants. Unless you have any better suggestions for a desktop metaphor that nobody else has though of yet, I don't really see any way that the status quo could be improved.
Chromebooks sold 30M+ units last year. That's a bog standard Linux desktop accounting for about one tenth of 300M in total personal computer sales.
As everything gets more web based, the client OS gets less important, for better and for worse. Just as many people predicted 20 years ago. Radical changes takes time.
The Linux Desktop is like building a house by asking random craftspeople to each design, build, and install a single piece of the house. One craftsperson preparing one plank of wood; a second craftsperson preparing a second plank of wood, etc. Later, one craftsperson turning those planks of wood into a door frame. Then another craftsperson to add a handle & lock. And another adding the hinges. Yet another hanging it on the wall. On and on, for every single component in the house.
This does not lead to a well-functioning house. It will be creaky, leaky, drafty, and ugly, and more difficult/costly to maintain. Yes, you can live in it. It may even be lighter, or cheaper, or in some other way technically superior to other houses. But most people will get really annoyed by it, and just want a house built by a single contractor that "just works".
This is even ignoring the gigantic elephant in the room, which is hardware support. Getting hardware to work with a Linux desktop has only two options: buy from an OEM that ships a supported Linux OS, or play a combination of russian roulette & home auto mechanic.
I think this is overly pessimistic, although fundamentally correct.
The silliest yet most iconic (no pun intended) problem I can think of is the necessity to create bookmarks multiple times for each of the graphic libraries of the system, depending on the program run. Yikes.
On the other hand, I think it's crucial evaluate the audience and their expectations. Even with the small flaws, Linux has reached the point where it can be used by the vast majority of users, the ones without too many expectations. They will complain about bluetooth with any operating system anyway ;)
Hardcore users will fix the system at any cost (I do), so they don't really matter, and I think they're a few (in proportion).
I think linux is paradoxically not fit for middle ground - the power user, who expects a certain consistency when digging deeper than just regular user.
How much of the whole audience those three segments represents... it's hard to say.
All in all though, I'm certainly disappointed that even on laptops that officially support Linux, there are still minor issues.
If this isn't written with reference to the classic UNIX-HATERS Handbook [1], it certainly could have been - the metaphor of living in a building site is common to both, and no less accurate now than then regardless of whether this is reference or rediscovery.
You can add to this that some major carpenters (I point my finger at you, Gnome team) will one day come to your house and suddenly remove a few planks here and there that you've been happily using for several years and turn a deaf ear at all protests.
And it's not even just the UI. But yeah, that's basically for both systems, would be really hard to say "this one is overall more consistent than this one."
In my experience, Linux hardware support is significantly better than MacOS and Windows (particularly 11, but even trumps 10 and 7 in most regards), and I'd like to see proof otherwise.
How much time do you have? I could write an entire book on shitty Linux hardware support.
My current laptop:
- Every time I connect any Bluetooth headset it usually (but not always) defaults to HSP/HFP rather than A2DP. I have googled lots, and apparently it's some sort of known bug, though nobody knows why it happens. None of the publicly-known fixes (yes I tried all of them) fix it. Sometimes A2DP just disappears as an option entirely, and I have to unpair and repair multiple times until it comes back.
- My video sucks. This laptop has hybrid graphics, which is sort of supported, depending on the kernel, the userland, the graphics cards & drivers. There is no official documentation of how to set it up (edit there is of course one or two distros with a Wiki page on setting up one or two graphics cards on one or two laptops, but if you don't have that particular hardware, "good luck" because it works differently for different hardware), and of course the distro doesn't figure it out automatically, leading to a wonderful scavenger hunt across the internet, that I have never needed to do with Windows. A magical combination of kernel arguments, environment variables, and driver configuration, lead to at least the graphics no longer having random tearing artifacts. There's no apparent way to turn off the card, either, so the GPU just soaks up power & generates heat, which makes the CPU fan churn all the time.
- Speaking of CPU: In a previous, really old distro, after fucking around with CPU frequency scaling a bunch, I accidentally got the CPU fan stuck on full-speed for 6 months. Later I installed some userland program which magically reset the CPU fan. I was able to tweak /sys/ files to silence the fan or make it explode, based on CPU scaling. But on modern kernels? The fan churns no matter how much I turn down CPU freq. None of the userland tools for CPU or GPU freq scaling make any difference, presumably because Linux simply exposes different controls for different hardware, and of course it changes from kernel to kernel, and the userland apps either don't keep up, or the drivers' functionality just doesn't work due to regressions. It actually worked in the older distro, though. (That's another thing Linux literally does not have: regression testing for hardware support, much less software, or a single database of officially supported/tested hardware)
- I can finally use power saving with WiFi without causing either a kernel panic or packet loss. This was a very old driver defect in very popular wifi cards, and this laptop is 5 years old. The suggested fix was "get a different wifi card". And of course, Windows did not have this bug. But better late than never?
That's just this distro + laptop. I have tried previous laptops with different distros where hard drives couldn't be detected or used during the OS install, where hardware was simply not detected, buggy drivers with missing hardware support, not to mention trying to figure out UEFI's bullshit. I've had sound cards not work, keyboard buttons not work, trackpads' functionality not fully supported, touch screens not be detected/work, and most every built-in SD card reader I've tried to use doesn't work or is a pain. Then there's the often not-working USB/HDMI hardware (cameras, audio, video, docking stations, etc) and keyboard feature keys like on/off wifi, display switching, brightness, volume, microphone, camera.
Some will reply with "yeah you need to spend a week researching fixes, what did you expect?" And some will reply "something like one of those things happened to me on Windows!" But in almost every case, any PC hardware will be designed and tested by the manufacturer to work on Windows, before they even ship their product, and is officially supported on (usually only) Windows. They will test their hardware on a given Windows version, release driv...
You would be hard put to find a house that wasn't built exactly like the Linux Desktop is. Your general contractor (distribution) hires a variety of contractors to install a massive pile of different components very few designed particularly for your house. Your analogy only works on the surface because of the odd notion of putting the component boundary at the board somehow.
The difference isn't the unconnected nature of the work which is virtuous its the resources available. No general contractor does the work needed to integrate all the many any varied components into a pleasing design for free and none of the components themselves are free. This flow of resources is what funds both the creation of said components and their proper integration. Because software has zero marginal cost an acceptable Linux Desktop experience can be cobbled together from what efforts people are willing to donate and what can be gleaned from resources invested in more valuable endeavors like servers.
The elephant in the room isn't figuring out how to get people to donate enough labor to make a third party OS work on every computer in the world, an impossible Sisyphean labor, its either convincing enough users to buy Linux computers for Linux support or finding some other way of getting sufficient investment for a more polished experience.
In any case nobody wonders out loud how we can possibly get Mac OS to have better hardware support people just buy a mac. I submit that there is also a third option. Google "name of my prospective computer" and the word Linux.
Just need to say it’s impressive how so many replies to you are literal and pointing out flaws in the analogy. Rather than thinking about why product development on Linux can’t seem to ever catch a break. Kind of ironic, but I suspect you knew it’d happen.
I have never met anybody who has looking into Linux outside of a Computer Science degree or as a computer hobbyist. Those people as it stand are a marginal amount of the population. The kind of people that bawk when you open up control panel. Or can only open up windows settings and find a how to online. None of these people have ever touched Linux. If you told them they had to run just "sudo apt-get update" in terminal, you instantaneously keep them from wanting to use it further. The shell is so scary to most people that they think it's hacking. I mean jesus I saw on CNN several years ago that when they described hacking, they had a backdrop video of someone running Ubuntu doing something like "sudo chmod x+ 777" and then "./the_virus_cnn_is_covering" or something.
I’ve used Linux desktop for 3-4 years now. Love it. Only use windows when a program only runs in windows, which is maybe every other week? Usually graphics intensive programs
Short: Every single Desktop Environment, then and now, does an absolutely piss poor job of abstracting away the need for a casual user to ever whip out the terminal.
Yes. In the original MacOs, there was no command line. That was by intent. It was a decision Steve Jobs made. And it led to the popularity of the Macintosh. The Macintosh was not text-oriented internally. It was resource-oriented. You could tweak a lot of things using ResEdit. Where things were in directories was not too critical. The Finder would find applications anywhere, and they were self-contained in one file with resources.
This was elegant. It was a complete rejection of the DOS/Unix approach. Users loved it.
It suffered from some implementation problems that dated from the Macintosh's original floppy-based configuration. Writes were deferred too long and resource editing was very brittle. This gave resource forks a bad reputation, especially when applications wrote configuration and status into their own resource fork. Any abnormal program exit left the resource fork damaged.
When Apple took over NeXT, they got a system built on UNIX. So now they were stuck with the command line and the UNIX baggage again. They tried to keep the resource fork at first, but gradually gave up. It didn't play well with the UNIX file system model.
The Linux tradition is that GUI system tools suck. They exist, but try to do anything and you soon find you need to use a command line. The GUI tools are never complete, and they're often wallpaper over a command line tool. Often they don't really know what they're doing to the system; they're just creating commands and displaying the responses.
All software and by extension OS are text oriented regardless of appearance. If you take a wad of human readable text and use it to built a binary blob then the blob is a function of the original program text ergo the text is still there just liable to be inscrutable.
Classic Mac OS hot garbage. They switched because unix was a better basis for an OS than anything that could possibly have done with their existing OS short of throwing the hard drive containing it out the window and starting over.
You can install a user oriented distribution like Mint or Ubuntu and never see a cli if you don't want to. So far as I can tell the gui tools that come with these distros are complete for all standard tasks. Minimal or technical distributions like arch, void, gentoo are pretty obviously meant to managed with the cli although users can still use rich gui tools for many non system tasks. Few are wallpaper over cli tools. I don't know where you got that idea.
I'm not sure what you mean by this specific part; Windows is not "text-oriented internally"; neither is let's say Gnome. Actually if that's about how often e.g. DOS commands would be run by various GUI elements, it was likely even less a practice on Consumer Windows than on NT based systems, given the DOS command window was a VM with limited and/or weird interactions with the host OS.
There is nothing wrong with Linux that you can't fix by uninstalling X Windows and logging in with ssh.
I dream of bringing a video game character into our world with a mirror and doing a sketch comedy act with it somewhere, and Linux would be a good choice for that because I'm certain I could make it never pop up a dialog that says "Failed to download software updates".
(In the meantime I'd better learn how to work a deck of three-sided cards like a puppet.)
I read maybe 2 or 3 of these a month, and I get maybe a paragraph in before I feel that they're saying the exact same thing. If you can't use Linux, I'm not trying to deny that from you. I'm a musician, I understand how hard it is to get certain, specific creative tools working. But, here's the deal:
The current status of the Linux desktop is fine. There's a significant portion of the computer-buying market (probably 70-85%) that couldn't care less about the operating system they're running. MacOS, ChromeOS, Windows, Linux: it's all the same to many. You just need access to a browser and simple tools for the basics (text editing, email, calendar, etc.). Linux has you covered for those, and if you don't need anything else you'll likely be fine.
If you don't want to use Linux, then don't. But articles like these will neither sway current Linux users or stop the never-ending demand for it.
My main desktop is an ageing i72600k running Windows 7. It's still fast but I wanted something that could boot from nvme, have more modern USB 3.1/3.2 and not full of security holes. But thing is, I don't want to move to Win 10/11.
As a test I installed 64 bit Void Linux Musl on an old AMD A8 6500 using the (VERY CRAPPY) built in GPU and a PCIe to m.2 for a 500GB NVMe SSD. Installed flatpack and Steam and was playing Half Life 2, albeit at lower resolution and framerate. Cities: Skylines also just worked but had to drop graphics. I didn't have to tweak any config files or compile drivers. Then I installed some Proton flatpack and got some Windows games running like Two Point Hospital. Skyrim and amusingly but not unexpectedly, Crysis, fail to run. Though at least the skyrim loader window flashes for a brief second. Though everything else is solid otherwise.
This was all very painless and I thought some fringe setup like Void Musl sans systemd would be a no go for gaming but it was quite simple thanks to flatpack dragging in all the dependency and out of the box AMD GPU support.
Though it's not perfect. I still have issues with mounting ntfs disks and audio bugs: I lose audio in firefox and while browsing, anything which may trigger audio to play like a video generates a loud pop. Those are annoying and frustrating. Previously I was running Linux Mint which was always buggy in some way so every distro has its quirks.
Lastly, Game makers: Please support Linux! I no longer wish to surrender my computer to Microsoft (No to Win 11). I have a 16 core Thread Ripper coming in with 64GB of ECC RAM and a 1TB NVME plus my GTX 1060 which is likely getting Void Linux installed. As far as I am concerned, the year of the Linux desktop is here. Though it lags behind Windows in terms of "ease of use" but it's getting there.
Valid arguments all around.
Personally as a long time Apple user (20+ years) I have a plan to run some stable distribution with KDE on top.
RawTherapee is capable enough. Coding for web is not a problem. Resolve is functional under Linux. Vim/Markdown, etc.
I am utterly disappointed with Apple. Just colorful marketing tricks on top of rare innovation and telemetry. The whole move towards unification of Mobile and Desktop is a UX mess. The last softwares keeping me on mac os were Sketch and Logic.
Windows was never an option and never will be. Windows is just game host os.
So for me the year of Linux Desktop is here and now.
The 'Year of the Linux Desktop' is bullshit for the same reason that the 'Year of the Mercedes Highway' is bullshit. Not everybody drives a Mercedes, and not everybody wants to 'drive' a Linux Desktop.
The only people who scream about the 'Year of the Linux Desktop' are the proselytising fanbois. I stopped being one of those years ago. Who cares if Joe Sixpack wants Windows on his desktop. As long as I have Linux on my desktop, I don't give a rats about Joe.
Those of us that use Linux had our own 'Year of the Linux Desktop' at some time time in the past. My own 'Year of the Linux Desktop' was 2001, and which followed on from my 'Year of the UNIX Desktop' which was 1991.
Interestingly, I personally never had a 'Year of the Windows Desktop'. (I went directly from a MSDOS CLI to a UNIX CLI to a UNIX Desktop several years before Windows 3.0 even came out. When I finally bothered to have a look at Windows 3.1, I couldn't believe how ugly and primitive it was.)
And yes, I use lots and lots of xterms on my six virtual desktops. In fact, if the output of the program I am using does not depend on an image, I will be invoking it from an xterm. I eschew those new-fangled IDEs. They get in the way of my curses-based text-editor, make, and CLI to run that compiled program.
71 comments
[ 2.6 ms ] story [ 124 ms ] threadFurthermore, the commandline is a core part of the computer experience in my book. Just because Apple, Google, Microsoft want you to live on the internet doesn't mean it has to go away or have a friendlier alternative. There will always be a divide between people who have an interest in computers and willingness to learn them and people who just want to like buy something off Amazon. The latter will probably never use Linux no matter how popular it is on their phone or desktop.
So yeah, I wanted to, but it wasn't ready. Things are much, much better now.
I have an interest in computers and I know a whole lot about them, but I want my daily experience to be smooth and to just work and I want to be able to relate to the rest of the world in terms of software. So, I use Linux for work and Windows or OSX everything else.
My point is that your comment seems to say that if people had an interest in computers they would (of course) be using Linux, and that’s absolutely not true.
1) the idea of "the year of the Linux desktop"
2) articles containing 90% of the complaints that this one does.
I've personally heard "year of the Linux Desktop" from people directly many times per year every year that I can remember but in terms of response/meme/joke/etc not actually taking the stance "it _is_ the year of the Linux desktop". Yet I can't even think of the last time I even naturally ran across someone proposing the concept online (sans trying to purposefully find such a person like Googling) let alone someone I know being serious about it.
Not sure if that is a normal experience and people just like to talk/joke about how it isn't all the time or if my bubble/filter is just abnormal.
> especially since server developers switched to macs
doesn't make much sense. MacOS is certainly very relevant, but there has been no exodus.
Here there are some interesting stats:
I think that in the last decade, Windows has lost a very significant market share, which has been taken from both macOS and Linux (the latter having a surprisingly high prevalence).I think it's because most everything we do now is web based. Also I think people tend to forget how big a deal google docs actually is.
It does everything they need to do, and it doesn't bit-rot like Windows. Anyone who just uses their computer for web browsing, playing music, and other simple stuff just doesn't need any more, and that's been true for at least a decade.
It's not yet ready for prime time. When I set up a new machine there are still tons of things I have to "fix" from the command line. Touchpad scrolling. Getting my Jabra to not auto-adjust volume to 0 every time I speak. Modprobe this, modprobe that. Setting up an L2TP VPN. Getting bluetooth audio to support AAC/AptX. Getting Chinese input to work properly. Configuring the extra buttons on my keyboard and mouse to do useful things. Disabling IPv6. Installing CUDA and CUDNN. Creating a virtual camera. Ignoring TCP reset packets (commonly needed when visiting China). Adding support for ExFAT SD cards, which almost all Canon cameras use. Writing files to USB thumb drives without root permissions. Mounting an SMB or S3 volume automatically at startup. Setting programs to start automatically at startup. Scheduling tasks to execute at certain times. I need to go to the command line to do all of these things. On Windows and Mac most of these things can be done graphically.
Oh and even Chrome doesn't install on a default Ubuntu 18.04 system. You have to do "sudo apt-get install libpango" first or "sudo apt-get -f install" after installing it.
- search for software name online
- open first 3 links to find reputable download source
- decide on format (portable exe vs msi installer vs zipped binary) and download it
- invoke software installer specific to the software
- hope you never have to uninstall it
Experience for downloading software on Linux:
- open package manager
- find software
- click "install"
I'll let you draw your own conclusions, I guess.
For Linux, the software I described does not exist as a package with an accompanying GUI app (and even if it did, which package for which distro?). That's the thing about these discussions: when a person has a real complaint about something that is very difficult for a regular person in Linux, there is always a theoretical way that Linux is easier, but then in practice it doesn't work.
I don't use it because it's faster to open a terminal and type `sudo dpkg -i foo<tab><enter>`. Most Linux users today probably do it that way, so nobody cared to optimize the GUI "alternative".
As a Linux user and command line aficionado I'm not saying that's bad, but until that mentality changes it's not going to be for the masses.
Linux is built with a bad architecture (spaghetti monolithic kernel with no stable APIs), which has a tremendously high cost of development, and also a high cost of maintenance. Large companies (e.g. Google) have realized this isn't sustainable, and are shifting away into better architectures (e.g. Fuchsia), which tend to be microkernel multiserver systems.
As money goes away, Linux maintenance cost can't be met, and it'll crumble on its own weight.
I do still hope a good desktop/workstation OS with a FOSS license will happen. It just won't be Linux, because its architecture was never suited for it.
[0] https://www.youtube.com/watch?v=WtJ9T_IJOPE
But XNU is not much better than Linux. It has a schizophrenic soul with its BSD and Mach duality for everything. It is bogged down by Apple's never-ending "security" lock down efforts. It has no equivalent of "jails/containers". And, again, it is practically unusable for anything that is not created by Apple and is not extendable now that they are phasing out kexts.
I think a very interesting recent effort in kernel architecture innovation is Fuchsia.
And since when Google became the representative of stable APIs?
During 90s, the de facto desktop OS used a monolithic kernel, so clearly, having a monolithic kernel doesn't prevent common adoption on desktops. Fuchsia is a microkernel, which would play nice with phone vendors with binary blob drivers, but this is a licensing issue not a technical one.
The thing is: it makes little sens for a kernel under GPLv2 to maintain an internal kernel space ABI for drivers (and I'm not even sure Linus would qualify with the licensing, but maybe I'm more moderate than he is). Ok I'm aware there is a mess at least on the Android side, but maybe the majority of it is questionable on licensing grounds to begin with (but the Linux Foundation is particularly lenient on licence compliance, btw, not surprising when you see the list of corporate members, and for some of the main ones their past practices - hm VMware hm...)
And that being said, I would be curious to see tons of old Windows binary only kernel space drivers continuing to run on modern systems, or ironically even some user space drivers (IIRC, printer drivers are userspace but 32-bits ones do not work on 64-bits systems)... Ok there is more binary compat for a few major versions if you are lucky. But then MS can always randomly requires WDDM 2 for its next OS anyway.
In the end, some old and/or exotic hardware work better (or at all) on recent Windows while some old and/or exotic hardware work better (or at all) under recent GNU/Linux distros, but I'm quite unsure this has anything to do with the precise kernel space API & ABI policies. What does not usually work is attempting to project a model on something that has vastly different premises.
So yes, licensing can influence technical decisions.
If your ecosystem is vastly built around binary only vendors, a more stable ABI is vastly more valuable than if you ecosystem is vastly built around source only free software vendors, AND pushing for in-tree integration on top of that...
Sure, Linux tries hard for stable user<>kernel interface.
Within the kernel, however, there's no such thing and it's an utter mess.
>having a monolithic kernel doesn't prevent
Even within monolithic kernels, and even after restricting to unix-like, there's those that are better structured internally (such as openbsd or dragonfly), and then there's Linux, which is an unworkable mess.
> It just won't be Linux, because its architecture was never suited for it.
Linux was designed to be abstracted over, and extended for any number of use-cases. It certainly does a better job of this than Windows, which has a terrible habit of hiding things in system registries, creating directory structures that don't make any logical sense, and building completely un-automatable system utilities (ever heard of someone's Windows ricing script? exactly). MacOS is literally just BSD lugging along 50 tons of XNU zombie code that it will someday replace. Ask most developers what they'd add to MacOS, and 9/10 responses are going to be a feature that already exists on Linux (good package manager, performnant docker support, updated coreutils, you name it). So in a way, the current "leaders" are still following in the footsteps of giants. Unless you have any better suggestions for a desktop metaphor that nobody else has though of yet, I don't really see any way that the status quo could be improved.
[0] https://xkcd.com/927/
As everything gets more web based, the client OS gets less important, for better and for worse. Just as many people predicted 20 years ago. Radical changes takes time.
This does not lead to a well-functioning house. It will be creaky, leaky, drafty, and ugly, and more difficult/costly to maintain. Yes, you can live in it. It may even be lighter, or cheaper, or in some other way technically superior to other houses. But most people will get really annoyed by it, and just want a house built by a single contractor that "just works".
This is even ignoring the gigantic elephant in the room, which is hardware support. Getting hardware to work with a Linux desktop has only two options: buy from an OEM that ships a supported Linux OS, or play a combination of russian roulette & home auto mechanic.
(Linux desktop user for 20 years)
The silliest yet most iconic (no pun intended) problem I can think of is the necessity to create bookmarks multiple times for each of the graphic libraries of the system, depending on the program run. Yikes.
On the other hand, I think it's crucial evaluate the audience and their expectations. Even with the small flaws, Linux has reached the point where it can be used by the vast majority of users, the ones without too many expectations. They will complain about bluetooth with any operating system anyway ;)
Hardcore users will fix the system at any cost (I do), so they don't really matter, and I think they're a few (in proportion).
I think linux is paradoxically not fit for middle ground - the power user, who expects a certain consistency when digging deeper than just regular user.
How much of the whole audience those three segments represents... it's hard to say.
All in all though, I'm certainly disappointed that even on laptops that officially support Linux, there are still minor issues.
[1] https://web.mit.edu/~simsong/www/ugh.pdf
https://news.ycombinator.com/item?id=27556754
My current laptop:
- Every time I connect any Bluetooth headset it usually (but not always) defaults to HSP/HFP rather than A2DP. I have googled lots, and apparently it's some sort of known bug, though nobody knows why it happens. None of the publicly-known fixes (yes I tried all of them) fix it. Sometimes A2DP just disappears as an option entirely, and I have to unpair and repair multiple times until it comes back.
- My video sucks. This laptop has hybrid graphics, which is sort of supported, depending on the kernel, the userland, the graphics cards & drivers. There is no official documentation of how to set it up (edit there is of course one or two distros with a Wiki page on setting up one or two graphics cards on one or two laptops, but if you don't have that particular hardware, "good luck" because it works differently for different hardware), and of course the distro doesn't figure it out automatically, leading to a wonderful scavenger hunt across the internet, that I have never needed to do with Windows. A magical combination of kernel arguments, environment variables, and driver configuration, lead to at least the graphics no longer having random tearing artifacts. There's no apparent way to turn off the card, either, so the GPU just soaks up power & generates heat, which makes the CPU fan churn all the time.
- Speaking of CPU: In a previous, really old distro, after fucking around with CPU frequency scaling a bunch, I accidentally got the CPU fan stuck on full-speed for 6 months. Later I installed some userland program which magically reset the CPU fan. I was able to tweak /sys/ files to silence the fan or make it explode, based on CPU scaling. But on modern kernels? The fan churns no matter how much I turn down CPU freq. None of the userland tools for CPU or GPU freq scaling make any difference, presumably because Linux simply exposes different controls for different hardware, and of course it changes from kernel to kernel, and the userland apps either don't keep up, or the drivers' functionality just doesn't work due to regressions. It actually worked in the older distro, though. (That's another thing Linux literally does not have: regression testing for hardware support, much less software, or a single database of officially supported/tested hardware)
- I can finally use power saving with WiFi without causing either a kernel panic or packet loss. This was a very old driver defect in very popular wifi cards, and this laptop is 5 years old. The suggested fix was "get a different wifi card". And of course, Windows did not have this bug. But better late than never?
That's just this distro + laptop. I have tried previous laptops with different distros where hard drives couldn't be detected or used during the OS install, where hardware was simply not detected, buggy drivers with missing hardware support, not to mention trying to figure out UEFI's bullshit. I've had sound cards not work, keyboard buttons not work, trackpads' functionality not fully supported, touch screens not be detected/work, and most every built-in SD card reader I've tried to use doesn't work or is a pain. Then there's the often not-working USB/HDMI hardware (cameras, audio, video, docking stations, etc) and keyboard feature keys like on/off wifi, display switching, brightness, volume, microphone, camera.
Some will reply with "yeah you need to spend a week researching fixes, what did you expect?" And some will reply "something like one of those things happened to me on Windows!" But in almost every case, any PC hardware will be designed and tested by the manufacturer to work on Windows, before they even ship their product, and is officially supported on (usually only) Windows. They will test their hardware on a given Windows version, release driv...
The difference isn't the unconnected nature of the work which is virtuous its the resources available. No general contractor does the work needed to integrate all the many any varied components into a pleasing design for free and none of the components themselves are free. This flow of resources is what funds both the creation of said components and their proper integration. Because software has zero marginal cost an acceptable Linux Desktop experience can be cobbled together from what efforts people are willing to donate and what can be gleaned from resources invested in more valuable endeavors like servers.
The elephant in the room isn't figuring out how to get people to donate enough labor to make a third party OS work on every computer in the world, an impossible Sisyphean labor, its either convincing enough users to buy Linux computers for Linux support or finding some other way of getting sufficient investment for a more polished experience.
In any case nobody wonders out loud how we can possibly get Mac OS to have better hardware support people just buy a mac. I submit that there is also a third option. Google "name of my prospective computer" and the word Linux.
Yes. In the original MacOs, there was no command line. That was by intent. It was a decision Steve Jobs made. And it led to the popularity of the Macintosh. The Macintosh was not text-oriented internally. It was resource-oriented. You could tweak a lot of things using ResEdit. Where things were in directories was not too critical. The Finder would find applications anywhere, and they were self-contained in one file with resources.
This was elegant. It was a complete rejection of the DOS/Unix approach. Users loved it.
It suffered from some implementation problems that dated from the Macintosh's original floppy-based configuration. Writes were deferred too long and resource editing was very brittle. This gave resource forks a bad reputation, especially when applications wrote configuration and status into their own resource fork. Any abnormal program exit left the resource fork damaged.
When Apple took over NeXT, they got a system built on UNIX. So now they were stuck with the command line and the UNIX baggage again. They tried to keep the resource fork at first, but gradually gave up. It didn't play well with the UNIX file system model.
The Linux tradition is that GUI system tools suck. They exist, but try to do anything and you soon find you need to use a command line. The GUI tools are never complete, and they're often wallpaper over a command line tool. Often they don't really know what they're doing to the system; they're just creating commands and displaying the responses.
Classic Mac OS hot garbage. They switched because unix was a better basis for an OS than anything that could possibly have done with their existing OS short of throwing the hard drive containing it out the window and starting over.
You can install a user oriented distribution like Mint or Ubuntu and never see a cli if you don't want to. So far as I can tell the gui tools that come with these distros are complete for all standard tasks. Minimal or technical distributions like arch, void, gentoo are pretty obviously meant to managed with the cli although users can still use rich gui tools for many non system tasks. Few are wallpaper over cli tools. I don't know where you got that idea.
I'm not sure what you mean by this specific part; Windows is not "text-oriented internally"; neither is let's say Gnome. Actually if that's about how often e.g. DOS commands would be run by various GUI elements, it was likely even less a practice on Consumer Windows than on NT based systems, given the DOS command window was a VM with limited and/or weird interactions with the host OS.
(a) Well, Desktop Linux has been a reality for me for over a decade. I've even had it installed in my parents laptops.
(b) But Linux dominates the mobile space with Android.
I dream of bringing a video game character into our world with a mirror and doing a sketch comedy act with it somewhere, and Linux would be a good choice for that because I'm certain I could make it never pop up a dialog that says "Failed to download software updates".
(In the meantime I'd better learn how to work a deck of three-sided cards like a puppet.)
The current status of the Linux desktop is fine. There's a significant portion of the computer-buying market (probably 70-85%) that couldn't care less about the operating system they're running. MacOS, ChromeOS, Windows, Linux: it's all the same to many. You just need access to a browser and simple tools for the basics (text editing, email, calendar, etc.). Linux has you covered for those, and if you don't need anything else you'll likely be fine.
If you don't want to use Linux, then don't. But articles like these will neither sway current Linux users or stop the never-ending demand for it.
As a test I installed 64 bit Void Linux Musl on an old AMD A8 6500 using the (VERY CRAPPY) built in GPU and a PCIe to m.2 for a 500GB NVMe SSD. Installed flatpack and Steam and was playing Half Life 2, albeit at lower resolution and framerate. Cities: Skylines also just worked but had to drop graphics. I didn't have to tweak any config files or compile drivers. Then I installed some Proton flatpack and got some Windows games running like Two Point Hospital. Skyrim and amusingly but not unexpectedly, Crysis, fail to run. Though at least the skyrim loader window flashes for a brief second. Though everything else is solid otherwise.
This was all very painless and I thought some fringe setup like Void Musl sans systemd would be a no go for gaming but it was quite simple thanks to flatpack dragging in all the dependency and out of the box AMD GPU support.
Though it's not perfect. I still have issues with mounting ntfs disks and audio bugs: I lose audio in firefox and while browsing, anything which may trigger audio to play like a video generates a loud pop. Those are annoying and frustrating. Previously I was running Linux Mint which was always buggy in some way so every distro has its quirks.
Lastly, Game makers: Please support Linux! I no longer wish to surrender my computer to Microsoft (No to Win 11). I have a 16 core Thread Ripper coming in with 64GB of ECC RAM and a 1TB NVME plus my GTX 1060 which is likely getting Void Linux installed. As far as I am concerned, the year of the Linux desktop is here. Though it lags behind Windows in terms of "ease of use" but it's getting there.
RawTherapee is capable enough. Coding for web is not a problem. Resolve is functional under Linux. Vim/Markdown, etc.
I am utterly disappointed with Apple. Just colorful marketing tricks on top of rare innovation and telemetry. The whole move towards unification of Mobile and Desktop is a UX mess. The last softwares keeping me on mac os were Sketch and Logic.
Windows was never an option and never will be. Windows is just game host os.
So for me the year of Linux Desktop is here and now.
The only people who scream about the 'Year of the Linux Desktop' are the proselytising fanbois. I stopped being one of those years ago. Who cares if Joe Sixpack wants Windows on his desktop. As long as I have Linux on my desktop, I don't give a rats about Joe.
Those of us that use Linux had our own 'Year of the Linux Desktop' at some time time in the past. My own 'Year of the Linux Desktop' was 2001, and which followed on from my 'Year of the UNIX Desktop' which was 1991.
Interestingly, I personally never had a 'Year of the Windows Desktop'. (I went directly from a MSDOS CLI to a UNIX CLI to a UNIX Desktop several years before Windows 3.0 even came out. When I finally bothered to have a look at Windows 3.1, I couldn't believe how ugly and primitive it was.)
And yes, I use lots and lots of xterms on my six virtual desktops. In fact, if the output of the program I am using does not depend on an image, I will be invoking it from an xterm. I eschew those new-fangled IDEs. They get in the way of my curses-based text-editor, make, and CLI to run that compiled program.