As a non-blind user, the title expresses my feelings too. And I feel like it's getting worse over time, not better.
From little things to kernel lockdown breaking hibernate on a fully encrypted system just because you should be happy to get your laptop battery killed by s2idle or disable secure boot. Yay, security.
I can only imagine the pain of all the accessibility issues on top of what I experience.
I agree, all the modern technical additions over the years to operating systems have entirely been 2 steps forward, 3 steps back.
It's always, "Oh, well, you can no longer run two or three monitors any more, but your primary display is higher resolution now!" Except DPI adjustments make it irrelevant and now my (i)GPU has a higher minimum load.
Or, "Oh, well, we only give you 2 ports now, but they're all <port>!" Great, but those larger bandwidth ports don't offset the fact that I can't plug in as much any more, and USB hubs are not a solution, they're a hack, wildly variable in operation, and some devices are not compatible with them.
Under Linux, you can still use the -dpi setting in xrandr (or in your x.conf). It sets the font bigger without blurring all the icons.
I prefer it over the replacement approach that modern desktop environments (and wayland) use. I've been exclusively using high-DPI displays for much longer than Mac OS or Windows have supported them, and the old approach was much better.
There's some argument that you need to blur everything badly (instead of setting a session-wide DPI) if the user is simultaneously using two displays with wildly different DPI's. That user is going to have a bad experience no matter what, so I've never understood that argument.
I noticed the one bright spot in the article is debian, though even that's broken for me thanks to systemd.
I switched to devaun, and things are much better, for now. It's unclear how long new software will keep reliably working under X11 without systemd.
Anyway, as a sighted user, my experience almost exactly matches the article, toned down about 10x.
(Concretely, on the systemd side: I hit the same issues with pulseaudio, and the new session stack regularly perma-blanked by screen until I rebooted. I can't reliably share machines with family members because elogin is so bad.)
GTK were talking about dropping X11 support at some point, so X11 folks using GTK apps would probably need to some sort of Wayland to X11 proxy, or to migrate from Xorg to some sort of multi-protocol display server that supports both X11 and Wayland, like Mir from Canonical/Ubuntu, or Arcan.
Yeah, I got much peaceful experience after I started running it on VMs, as modern PCs got good enough virtualization hardware, and I have used enough distributions since that 1995's Summer, starting with Slackware 2.0.
In the 2000s, a video of Nelson Mandela explaining "Ubuntu" was included in every iso[1]:
"A traveler through a country would stop at a village, and he didn't have to ask for food or for water. Once he stops, the people give him food, entertain him. That is one aspect of Ubuntu, but it will have various aspects.
Ubuntu does not mean that people should not enrich themselves. The question therefore is: Are you going to do so in order to enable the community around you to be able to improve? These are the important things in life, and if one can do that, we have done something very important which can be appreciated."
Since then, Canonical's customers have shifted with its mission; Ubuntu Desktop is mostly a promotional vehicle for paid services, and desktop changes on the whole over the past decade (post-systemd) weren't made to better represent the needs of our "travelers", but corporate customers or Canonical themselves.
Userland has always been a frontier space; as the author notes, you get out of it what you put in. But we can point to discrete events where shifts of priority and reductions of service have occurred for certain groups.
Is there a way to understand why they now have to ask for what they were once given in the spirit of ubuntu?
The big commercial distros hit a common failure mode for open source projects. There's a lot of money in cloud, which means there are a lot of vendors that want to extend things in ways that make the vendors' products relevant. Those vendors end up being the ones that fund future devlopment work, so they control the direction of the ecosystem for the projects they rely on (instead of users / hobbyists / hardware vendors / etc).
So, we get stuff like Ubuntu's fleet management service and flatpak/snap instead of things that actually improve the user experience.
The whole ecosystem is this way. Look at the stuff RedHat has been shovelling. The article mentions gtk3/4, wayland and pulseaudio. I'd add systemd to the list. All of those introduced multi-decade regressions into the desktop experience. The only reason they were funded was because they force churn in smaller open source projects, killing competition from lower-resource developers, and consolidating the ecosystem in the big shops.
Heck, RedHat Enterprise Linux is a giant GPL violation, but IBM throws a lot of money around, so most people just look the other way:
RedHat ships source code of GPL binaries for RHEL, but only if you pay them. The source code comes with the following restriction: If you redistribute it, they'll refuse to renew your support contract. This clearly violates the "no additional restrictions" clause of the GPL.
I think RedHat's theory is that the GPL doesn't entitle you to updates or support, so restricting them can't be considered an additional GPL restriction.
You could say that without the updates, RHEL is not fit for purpose, and therefore you should be entitled to them. Sounds like a complicated legal argument to have though.
I agree that RedHat are certainly not ethical actors in the FOSS community though.
Money talks, I guess. Two of those companies are bad actors in this space (google, with the android rug pull, and user abusing software built on open source, and aws, the repeat open source business killer):
Considering who Conservancy are, that is very unlikely. They are the main organisation doing GPL compliance work these days. Everyone else either opposes enforcement, or doesn't have the resources to do it, or does it but doesn't talk about it.
I find it ridiculous how "writing scripts to keep things working the way I want" can be a source of frustration.
To me, that's a source of pleasure.
I've never expected Windows or Mac to work exactly how I want them to. In fac, they can't. So given that, how can Linux be a poorer experience/
The truth is that for virtually everyone, Linux will be the absolute closest experience to having everything work exactly the way you want things to, because it is that open to being modified.
For you, it's writing scripts so that things work the way you want. For the author, it's writing scripts to be able to use the computer at all. And how would that work if you have to write the scripts in a state the computer doesn't work for you? That's the difference between being fun for you and being agonizing for someone who has to rely on it un order to use their computer.
Unrelated, but playing around with Orca I can't help but notice how behind the times the speech synthesis is. The voice is incredibly distorted and robotic. I know the goal isn't to have a natural voice with human cadence, but there's little reason to have it sound like your speakers are dying. Maybe it's just the defaults that are bad?
As always, framing the problem as a "Linux" problem really makes it difficult to properly discuss it. If I understand the article correctly, Linux actually is accessible enough. The problem is that Fedora Workstation, Ubuntu etc are not. But if the actual problem is that Fedora WS is not accessible enough, then you can actually start iterating the things that do not work and maybe even think about how that can actually be fixed.
But the "Linux desktop ecosystem" is a too fluffy thing to even be able to discuss in any coherent way.
To be fair, I've never once read or heard someone describe the Linux ecosystem as "it just works." The entire time I've grown up with computing, Linux was something you tinkered with as a nerd.
Debian and Ubuntu were like the distros you used if you wanted a semblance of a life as a GNU/Linux user.
For a non blind user and on a machine that is well supported (rule of thumb: a thinkpad of at 3years ago or more), it does just work. At least as well as on win/mac where frustrations and issues also do exist.
I haven't used anything but Linux on a laptop in over a decade, and for me it really does "just work" (unless you're including one-and-done BIOS setup issues and the like).
It really depends the kind of laptops. Gaming laptops with nvidua gpus and exotic stuff? maybe. Business laptops with intel everything from integrated graphics to lan/wifi are usually painless. That has been my experience on Fedora in the last 15 years.
Caveat being to do some research before purchase about support of third party devices like printers or scanner.
As a long-term strategy, such focused distros aren't the best idea. Derivative distros often have single developers, or smaller teams, making them vulnerable to disappearing. Its usually better to get all the changes needed into upstream projects, so that all desktops/distros are more accessible by default, and then contribute to testing/fixing accessibility in independent/major distros, so that all the downstream ones are more accessible by default.
The author prefers NixOS as a solution though, so each individual controls exactly what is on their system so they can roll back if something breaks and add their own tweaks. Eventually that just results in the one-person distro scenario though.
After reading this post I've just installed Orca (I already had speech-dipatcher installed and fully working in my vts (ttys)) on this Linux Debian cutting-edge rolling release that I use for years now...
And it is working greatly here (In Emacs, browser and everything else);
My ricing setup is: Linux Debian Siduction Distro plus StumpWM... Could you test this setup in your PC/Lap?
NOTE: I do not use Gnome/Kde DEs (at maximum, I maintain XFCE4 here for others who do not know how to use StumpWM.
I tried setting up some text to speech stuff years ago, and I ended just giving up. Everything kind of worked, except when it didn't.
Linux generally works fine for everything else for me, but this is definitely not its strongest point. My general impression was that many tools were very much developed by people in their spare time, and just not having enough of it. In theory it could all work because all the bits are there, but there's no "chief of accessibility" that can patch things over in the various projects.
The flexibility and decentralisation of the Linux ecosystem is great for some less common use cases; I wrote my own WM and some other X11 tools that work a bit quirky, but works really well for me. This would be very hard to do on Windows or macOS. But for some other less common use cases ... yeah, not so great.
Considering the amount of issues I've had over the years with audio in Linux, I completely empathize. I wonder what the author's experience with Windows / Mac is?
As a non-blind user, I am struggling to understand why blind users would want a desktop session in the first place.
Based on intuition I would think the pure text experience of using cli tools would be superior for blind users. And since there is usually a cli/tui version of every single gui app / use case, it is not like one would be missing out on availability.
Or is it because terminal multiplexers like tmux and screen aren't accessible enough?
It might help to know that blindness takes many forms. Having no sight at all is what many people assume but is not what many blind people experience. So some blind people can use accessibility tools together with the sight they do have which may for example only be in a small part of the area most people can see, or may be blurry, or otherwise impaired.
Put another way, as a user that prefers CLI + text mode (and stuck with DOS whenever possible for many years after installing Windows 3.0), I'd rather not start a desktop environment.
These days, there's basically no choice. Most software (even the text editor at work: VS Code) has abandoned text mode.
There are many days I would say it isn't a bad thing, until I'd realize I can't fill some important form online so that I don't have to queue during office hours at a desk.
exactly! red hat is a huge “contributor” to the “linux ecosystem”, but those contributions are mostly regressions. as the article states, gnome circa 2010 was stable and complete. gnome today is one breakage after the next—even if your vision is 20/20. don’t even get me started on wayland! or pulse. or systemd.
I was misfortunate enough to be born in 1979 and was also just old enough to be cognizant and tangentially involved during major computing/communications epochs. I was around for when personal computing proliferated down to family homes, and no longer just in the homes of computing enthusiasts building their own 8-bit computers. I was pre-teen/teen when the World Wide Web got huge, I was around for the jumps from 8-bit computing, to 16-bit, to 32-bit and, of course, 64-bit. As a kid we had an Apple ][, and a Mac XL (as well as a Commodore 64 and a Nintendo Entertainment System). Everything but the NES made it into my bedroom (my stepfather loved that NES!). I read all the manuals and books I could, I saw War Games, bought a modulator/demodulator with my own money, and discovered BBSes.
Man, what a preamble! What does this have to do with Linux? I first got into Linux when I was using OS/2 as my main OS. A good friend of mine (who got me into OS/2) also got me into Linux, specifically Slackware, in the early 1990s. He was 10 or so years older than me and worked at a computer store and taught me everything about computing that I had not learned myself.
Linux back then was HARD. The prerequisite being you had to be a massive computer nerd to even read the documentation! Linux was really not user friendly at all. The process of "installing" Linux wasn't done in an hour, an afternoon, or even a day. It was a continual process over weeks really. Editing this config file here, compiling this program there (we had no package management and fat binaries were rare). Once all that was done, you had all your hardware working, and the programs you wanted running and life was good. Until the install got borked by the result of a power outage in your apartment building. Fun times. It was a massive undertaking and you learned while you installed it. Your only support was, in many cases, confusing and poorly written documentation. You could go to the Newsgroup or mailing list and get some suggestions though. Many of the people offering "support" were very rude and unsocial neck beards who loved to gate keep and were very good at discouraging people from joining the Linux community. Luckily I had my ThinkPad running OS/2 Warp 3 (later 4) to email, use FTP, surf the web and read my Usenet newsgroups while I was learning how to use Slackware.
Linux wasn't about it being easy, it was about it NOT being a Microsoft Product. Microsoft's predatory, unethical and flat out illegal business practices (in some cases) was consuming the entire computing world and stifling innovation: Linux was the answer, it was simply NOT MICROSOFT.
Now here we are in 2025. Last night I installed Debian 12 on a recently acquired ThinkPad Yoga S1. It is a 2-in-1 Laptop/Tablet hybrid with touch and pen screen. Everything worked out-of-the-box. No issues. The tablet features, everything works. The entire process took about 15 minutes.
There are more Linux distributions out there than we have teeth in our heads. If you need support there are variety of options out there for you, but you probably won't need it if all you want to do is word processing or watching YouTube.
If you don't like one distro, try another until you find one you like, or change up from the default software in the chosen distro and pick different software to install. Linux is incredibly user friendly now. You're not dropped at bash prompt and told "good luck" anymore.
Are you blind or using Orca to interact with your computer? If the answer is no, then your recent experience with Debian is not particularly relevant to what the article is talking about, since you obviously wouldn't have run in to any of the issues discussed.
Ignoring the comment you replied to, I am curious about the claim that pipewire is incompatible with the audio card.
I get that pulseaudio cannot default to using an output with speakers. This happened to me too on every single login, so I switched to pipewire, which is capable of saving settings between seessions (pulseaudio claims to support this, but it's been broken for over 10 years based on my internet searches).
I guess I got lucky that pipewire can talk to my card?
I'd expect it to be far enough up the stack where hardware compatibility wouldn't matter.
I distinctly remember having to install a package before audio would work on my machine. Needless to say, that isn't the best option if you can't rely on your eyes to find out what's missing.
Unfortunately, I'm not knowledgeable enough about how audio is done in interaction with hardware to help satisfy your curiosity.
From little things to kernel lockdown breaking hibernate on a fully encrypted system just because you should be happy to get your laptop battery killed by s2idle or disable secure boot. Yay, security.
I can only imagine the pain of all the accessibility issues on top of what I experience.