I find it paradoxical that people would question the need for a competitive marketplace of ideas in societies that tout competitive marketplaces for businesses. I also find it paradoxical that societies that tout individuality and freedom would question the need for others to express those very values, in preference of conformity and centralized decision making. We have seen the consequences when it comes to both commercial products and open source projects. Just consider the uproar of when KDE, Gnome, or Windows changes the user interface in controversial ways.
Yes, it does cost Linux in terms of user base. On the other hand, it is better (in my opinion) for the open source community to differentiate itself based upon its core values and to ensure its long term viability because of that.
The past is littered with failed companies that tried to take on the juggernauts while Linux has survived. A big part of the reason for Linux's survival is it's fluidity. When one project stalls, another can take over. When one group's needs are unmet, a new project can emerge.
> When one project stalls, another can take over.
Desktop environments are too big an effort to see this sort of dynamic. And it shows: the two major projects after 25 years are still the very first two. Around them we have a plethora of little projects who typically die as soon as their main sponsor dies, and contribute very little "R&D" to the main two - in fact, KDE and GNOME more often than not push experiments that the minnows then implement or reject.
I also don't see it as necessarily a zero sum game. The choice is not always "do thing A or contribute to a similar incumbent project B" but "do my own thing A or nothing at all, because B doesn't want my idea."
A new popular project has the potential to bring in more users.
what desktop linux is really lacking (especially with respect to hardware compatibility) though is really good QA.
It might have been coincidental but i can still remember when ubuntu threw it's weight behind unity after 10.04 (which i remember fondly coz it was absolutely rock solid) and quality really dropped off for a few years.
it's a perpetual problem with FOSS that people chase shiny baubles and underrate fundamental, boring stability and quality (or, for that matter, power usage).
This is somewhere where MS actually has a non-unfair advantage coz it takes stability, QA and power usage very seriously on windows.
Playing devil’s advocate herem by this logic we should also splinter the kernel development for multiple, competing, projects.
I think this is a better comparison since only a few desktop environments are forks of other desktop environments, and most of the viable ones are forks of Gnome (much as most of the viable forks of kernels are forks of BSD).
The word you're looking is "contradictory"...
A desktop on Linux that looks good and works, and is shipped pre-installed on desktops and laptops like.... Windows and MacOS?
The chances are pretty small, but if they want to write their own desktop, then good luck to them.
There are enough Rust enthusiasts, that if they choose to also open source it, they'll likely make a lot more progress too. This is the kind of thing that the 'rewrite it in Rust' crowd get excited about.
Nonetheless ive found it to be much more stable in the last 18 months. They have recognized their problem and improved.
My use case: a HiDPI muxed dual-GPU laptop with one or two also HiDPI monitors plugged in sometimes, but I try Wayland session once in a while to see if we’re there yet. This seems to konfuse KScreen enough that it sometimes treats 200% scale as if it was 400%.
> Doing their own thing just means duplicating 20 years of effort that those projects have done.
Not really, because part of that time is spent learning things, trying thigns, and part of that time may be unnecessary.
It is a much more difficult to maintain a different kernel than it is a different desktop environment. So there is a huge incentive to cooperate on one kernel, even among companies with wildly different concerns (e.g. embedded IoT devices vs. cloud servers).
The kernel is also developed with an entirely different support culture than, say, Gnome. Backward compatibility for all supported use cases is taken extremely seriously by kernel developers. If it works on today's Linux kernel you can be almost assured it will work in 5 or 10 years. So, kernel users (i.e. developers, distro maintainers, etc.) stay happy because their stuff doesn't break.
Contrast this with the more "opinionated" DEs and you see features yanked and changed at a comparably dizzying rate. So, DE users experience a rate of change and support dynamic that is actually much "worse" than kernel users do.
Maintaining a different desktop is a massive undertaking. The total amount of lines of code in GNOME or KDE is comparable to the lines of code in the kernel. It is no joke to make a new desktop environment, to the tune of hundreds of millions of dollars. It's not just a matter of take GNOME and replace the window manager and then you're done. But I suspect that is what System76 will start with and then they'll see how much money they can raise from that. (Probably not much if Purism is any indication)
Backwards compatibility to the kernel only matters to the syscall API. There are areas of the kernel that get quite a lot of churn. Or things such as obsolete/unmaintained drivers and architectures getting removed. But nobody complains about those because it only matters to driver developers or other low-level people like the systemd maintainers. And when something broken does make it into mainline, the few people who get affected by those drivers who do complain get told "too bad, there is no money/time to fix it, buy new hardware or pay a support vendor" or something like that. You know, the usual thing you hear from any open source project. I really don't see how the support culture is much different. I know it's common for people to call out GNOME but I never understood any of the complaints there, you can still grab GTK1 and GTK2 and compile them if you really want to use those, as bad an idea as it may be. Those are about as unmaintained and crufty as any of the older areas in the kernel, and they're still technically around.
If you ask me, it pays to just learn to be flexible and learn to appreciate and work with the strengths of a software even if you don't like some aspects of it.
The reason for the minimalist approach simple enough, I am accustomed to to it. Experiences with desktop environments tend to start with the idyllic romantic honeymoon and end with the realization that, after the initial setup, I am not actually using it for anything aside from launching programs and managing windows. The bulk of my time is being spent in a handful of actual applications, so why would I want to put up with all of the additional clutter?
> If you ask me, it pays to just learn to be flexible and learn to appreciate and work with the strengths of a software even if you don't like some aspects of it.
I want to be flexible in my work, but not with my tools. Having to be flexible with my tools means that I can never reach a level of mastery. I would be building my skills on shaky ground.
GNOME and KDE have needed help for the past 15 years. If it hasn’t happened by now, it’s not going to happen.
That's just not true. CDE was released before KDE, and the Motif widget set CDE uses is older still.
Stream discussion 1: https://youtu.be/mvk5tVMZQ_U?t=1247
Stream discussion 2: https://youtu.be/mN3QFw2BEcw?t=3587
Stream discussion 3: https://youtu.be/5MAlgKdsdvg?t=1820
Stream discussion 4: https://youtu.be/h0boUNc1JOg?t=3148
These are just their discussions (foreshadowing the upcoming videos) but there is some really insightful discussion about what is still hard for someone who hasn't used Linux and what isn't. Some of their first hurdles they had to overcome was issues with their DE, window manager, and other random things that sort of worked correctly but needed a lot of polish to be "just right" or competitive with Windows or MacOS for most users.
Linus is coming at it from a very realistic angle, but some people are very defensive about the criticisms. They're basically telling him he's "holding it wrong".
The Linux community needs to listen to criticisms and find ways to improve the user experience. The nature of choice when it comes to everything you run on Linux means that nothing needs to be compromised for the power user. So what's the harm in making the default experience more friendly and seamless?
There is a lot wrong with Linux usability. But Linux != desktop environments.
The main usability problems that I've encountered with my non-tech friends are app availability and hardware support. Neither of those will be affected by rewriting a desktop environment from scratch in another language.
Even more surprising was how calmly Luke handled everything. Contrary to Linus' disposition, Luke meets every problem with a thoughtful solution and a willingness to learn, which leaves me rather unsurprised that his Linux experience has gone a lot smoother. On the latest WAN show, Linus flies off on a six-minute tangent where he tries to argue that CLI tools shouldn't exist, only to have the chat dogpile him and remind him that piping between CLI programs is an essential component of Linux. Luke caught onto this pretty quickly, and played devil's advocate, while Linus continued spinning his wheels with the 'muh accessability' bit.
In other words, to nobodies surprise, Linus lets his ego stand in the way of correct technical solutions.
I didn’t find it that surprising, mainly because nobody getting into desktop Linux without prior knowledge is going to expect any connection whatsoever between a random third party app and the OS’ desktop environment. It’s like if clamping a smartphone mount to your car’s dash caused the car’s rear left wheel to fly off, it’s completely unintuitive and unexpected.
And realistically, the two shouldn’t have any influence on each other. As much as he should’ve read the package manager’s warning, it shouldn’t have been a possibility in the first place.
> On the latest WAN show, Linus flies off on a six-minute tangent where he tries to argue that CLI tools shouldn't exist, only to have the chat dogpile him and remind him that piping between CLI programs is an essential component of Linux. Luke caught onto this pretty quickly, and played devil's advocate, while Linus continued spinning his wheels with the 'muh accessability' bit.
This feels like a mischaracterization. His argument wasn’t a that CLI tools shouldn’t exist, but instead that for 99.9% of common use cases pulling them out shouldn’t be required.
I also noticed the difference between the two, and you highlight important aspects of it. But are you sure the cause is ego so much as age?
When I got into Linux, I was a kid. I had a lot of time, and that gave me a certain kind of patience and determination for tinkering, for customizing, for breaking and fixing things, and for learning to do things in new ways.
By comparison, nowadays I'm pretty much a grumpy old man when it comes to my desktop environment and software management tools: I like using my computer a handful of ways, and when I am in an environment that can't support a pretty similar workflow, I find it very irritating.
Linus also has a more bespoke setup, using a somewhat cutting edge hardware configuration, and he likely has more responsibilities than Luke at home (he has children!).
I do think there is an ego element: Linus thinks of himself as someone who's technically competent— who knows how to use computers— and being in an environment where working like he's used to doesn't work well is probably especially frustrating on that basis. But I don't think it's all of it.
I hope other readers don't dismiss all of your observations just because of your apparent personal animus for Linus.
This, though:
> In other words, to nobodies surprise, Linus lets his ego stand in the way of correct technical solutions. > correct technical solutions
We should be pretty careful about how we use the concept of ‘correctness’ when we're talking about desktop operating systems run on our personal computers for activities we undertake for pleasure. Otherwise we can end up saying something pretty silly.
In the 1980’s, computer users had to use a CLI, but very few of them could do more than start programs. Using output redirection or pipelines was extremely rare; understanding them (rather then copy-pasting magic incantations from computer magazines) even rarer.
I don’t see why that would be different today.
(1) unless the size of the desktop computer market shrinks significantly, of course. If we get to a situation where most desktop computer users are programmers, that may change.
I don't think this is uncommon behavior for the name of a product to effect who is willing to try it out. He likely thought Fedora was being suggested by his users because of the meme of fedoras. Google "fedora meme" for more information.
Now, is this the most rational behavior? No, not at all but it is what you need to expect from the average user. Branding is important. If I started an advanced system service company for IT and called my business "ASS Help" I'm going to get a lot of errant requests.
> with nearly every problem he has it seems to be exacerbated into a total disaster
Linus plays an energetic character. I wouldn't put much stock into how "irate" he seems. I would however expect anyone who hits similar roadblocks to him to actually act this way. If an average user plugs in another monitor and it doesn't instantly work they will be very upset.
> Even more surprising was how calmly Luke handled everything
Also important to note: Luke adjacent to software developers, has done some software development work, and has used Linux before for an extended period of time.
Luke also plays the straight faced silent type in their Penn and Teller.
> Linus flies off on a six-minute tangent where he tries to argue that CLI tools shouldn't exist, only to have the chat dogpile him and remind him that piping between CLI programs is an essential component of Linux
1. I don't think Linus thinks CLI tools should not exist (as he uses them often for advanced tasks) but he feels like they shouldn't be the only option. As someone who daily drives Linux on my desktop, laptop, and work computers since 201Xs I rarely use the command line for average user things.
2. The end consumer user of a product doesn't really care about the philosophy of the software they are using for. I'd even go so far as to say that the Linux community doesn't even feel this way. If we did, we would all be using BSD-style 100% POSIX compliant systems. I'm much faster with the CLI than I am with a GUI for most development related tasks (installing software, unzipping things, moving files, diagnosing performance things) but that is not representative of the average user. For example: ask the average Mac power user how to check how overworked their CPU is. Will they pull out `top`? Probably not.
> Linus continued spinning his wheels with the 'muh accessability' bit
Mockingly saying "muh accessibility" doesn't address the concern: if we want Linux adoption on the desktop we need to make it so that the average Windows or MacOS user can be productive in the system without any CLI involvement. Or, at least, no more CLI involvement than with Windows and Mac.
> Linus lets his ego stand in the way of correct technical solutions
I think, ironically, this is the ego of the Linux community. We cannot expect users to be rational. We need to make things simpler (not in terms of "purity" like POSIX adherence) for the end user. Linus brings up a great point: almost no distro has a full time team working to A/B test the OS to see if it's friendly for new users. I can think of only one instance of this happening sometime around 2015: Gnome.
Also, if this was pure ego, Linus' opinions would be very different. From this Friday's episode Luke was negative on the prospects of the Steam Deck being shipped with Linux while Linus thinks it will be great. I agree with Linus, and the reason he thinks it will work great. Steam Deck, and Steam OS, will tune all the nobs and installed packages in a standard way for game devs to target. This makes it simple! Linus and Luke both agreed that for non-gaming normal user tasks Linux OSs are currently at an A to an A+ in terms of usability.
Yeah, I totally understand his attitude on this. I'm not a daily Linux user myself and it's been about a year since I last tried it so please take what I say with a huge grain of salt, but all the times I've tried it out and set up my dev environment, I've run into issues which the forum or Stack usually tells me to resolve using the command line.
It's not that I personally mind using the command line, since I use it all day long as a programmer, but there have been plenty of times where I don't feel it's "warranted". People shouldn't be forced to touch the command line unless it's really necessary.
In general whenever I try out Linux I run into a lot of things in the UI that seem just plain strange and not quite right, even though I understand you can't expect things to work the way you're used to on a different system.
Substantiation for some idiots who will downvote me despite parent comment not providing substantiation: Slowness, slowness, bugs, bugs, more bugs, security vulnerabilities mostly due to insane NIH ways of doing basic IPC stuff.
You're not making the case for not using Linux here.
The only reason my wife doesn't use Linux is b/c it doesn't have the apps she wants, and hardware issues (Bluetooth headsets don't work without ruining the audio quality, sleep/hibernation is hit-or-miss, etc.). I just don't see any issues with DEs that are preventing people from using Linux...I would love to shown wrong, because I would love to see more Linux adoption.
But this just seems like a waste of time.
Funnily enough, since when Linux started and somehow some people thought it'd be the "Windows killer", Linux now powers pretty much the entire real world (servers / routers / phones / appliances). There are billions of devices running Linux.
And still... Because somehow Linux never got to more than 2.5% desktop market share, many see Linux as a failure.
This is both funny and mindboggling.
If you have any linux distribution with snapd installed, there is a wide range of commercial and OSS software that you can install. E.g. Slack, which is used in many companies. Chrome, Spotify, and many other things. Of course the experience is not always that good and there might be better ways to get the same software, depending on what distribution you are using. But it's a nice baseline way of getting what you need going without too much fuss.
And at least you are not dealing with n different packages for n distributions in various states of obsolescence, support, and brokenness, etc. anymore. Not quite as easy as installing stuff on a Mac but it definitely narrowed the gap a lot and makes it easier for developers to support their software on Linux as well.
The status quo for linux desktop environments has been for many years that there are many options to choose from and that most software tends to work across them without much issues. So, having another one is probably not that harmful. Though you could question whether it will move the needle in any way. It's a nice way for System 76 to show that they care at least and who knows they might actually do a decent job of it and gain some fans.
If I'd be buying a laptop to run linux, I'd probably want to install it myself and not be locked into vendor specific software choices. But that's just me. I'm sure they would not stop me from doing that.
It depends. If you're talking about one written in C/C++ with JavaScript bindings, there's plenty of those already. And they all suffer the same memory/type bugs that make them unreliable and resource-intensive.
I'm not saying rust is a silver bullet, but in my experience the compiler is pretty good at preventing footguns, and the strong type system makes it easier to fork and extend than C-style "send raw bytes to this memory region and pray your gods" APIs. Overall i'm curious to see where this goes!
> hurting their own sales in the long term
I don't think people buy a laptop for a specific desktop environment. In any case, switching desktop environment / distro should take under an hour.
[0] https://phoronix.com/scan.php?page=news_item&px=GNOME-Shell-...
Also, rust's "fearless refactoring" paradigm enables for incremental optimization which i've found much harder to deal with in other programming languages. I think that's a point worth considering for a big piece of engineering like a desktop environment.
EDIT: Also memory bugs are not just "memory leaks". There's plenty of subtle bugs introduced by off-by-one errors and undefined behavior. Just because memory usage doesn't rise unexpectedly doesn't mean software is working as intending. I don't know about you but i personally have had weird bugs with all desktops i've tried so far.
Arc<> does exist though, and is used quite extensively in the rust ecosystem.
Am i missing something?
Disclaimer: i'm not a good programmer, nor a low-level programmer; i use Rust as a faster/safer python-like high-level scripting language
MATE: https://github.com/search?q=org%3Amate-desktop+%22memory+lea... (265 results)
KDE: https://bugs.kde.org/buglist.cgi?quicksearch=%22memory%20lea... (97 results)
That's just for the ones we found. I would be curious to run a full DE through valgrind :)
EDIT: Anecdotal datapoint ahead: i just had to kill atril (MATE PDF viewer) on a machine where it was eating 6GB RAM after running for a few days.
The current desktop environment situation is more uniform than ever and the situation is quite good and easy for users. Ubuntu and Fedora + RHEL are all-in on Gnome with no changes to the situation in sight, and KDE is a high quality runner up with users enjoying good compatibility between these. Most specialized desktop distros like SteamOS, Clear Linux, etc are also Gnome.
No, and fragmenting the community even further only hurts the ecosystem as a whole.
Sometimes putting your ego on a side and contributing to existing projects does wonders. Unfortunately the RIIR crowd might be unable to do it.
A leading hardware vendor is a great starting point. It can force shame other DEs to support nonsexy but necessary features, it might spark other Linux hardware vendors to join / contribute.
If a viable high end Linux hardware market can be spawned with apple-like margins then things can progress.
But it might just be a other dead end. At least someone with resources is trying.
I think it's a recent trend that as established Linux distributions have largely gone beyond their phase of trying to differentiate through changing the desktop, now sticking much closer to an upstream default Plasma or Gnome, the idea of "we need our own HMI to differentiate our integrated product offering" is coming more from the folks doing hardware. A lot of these products are boutique offerings that need higher margins to compensate for lower volume, likely driving the thinking toward software as value-add.
e.g. Purism with Phosh for the Librem, System76 doing this, etc.
I don't think it's so much about differentiating. It's about funneling resources towards innovation. Some distros have become pretty conservative over time. Don't get me wrong, i like that my distro doesn't just apply thousands of random patches introducing subtle bugs.
But at the same time, System76's PopOS appears to be among the few spiritual descendants of Mandriva's user-friendliness: if these few corporations who have a few bucks to spare are not investing in bettering the desktop experience, who will?
I'm personally unconvinced by current desktop environments: they're either too complex/opinionated and ridden with bugs (GNOME/KDE) or too minimal and not packaged with user-friendly apps (i3/openbox). Those that aim for the sweet spot in the middle usually don't have enough resources to iron out all the bugs (MATE/Cinnamon/Budgie/LxQt...), and rustc compiler is pretty good at ironing out all the bugs you write before they end up in a "usable" binary.
I believe so. Rust is a much higher-level language than C so it's easier to hack on, but at the same time has much stronger guarantees.
Overall, its type system prevents many many bugs (it is my personal experience that when code compiles, it just works) and enables "fearless refactoring" which for something as complex as a desktop environment is a killer feature.
I don't know if counting on GTK as a toolkit is a good choice (though it has certain benefits eg for accessibility) but for sure writing a desktop environment in Rust is an interesting experiment.
I don't know if there's much advantage to give outside of a paradigm switch in how windowing works a la tiling. Even then tiling isn't for everyone or every task.
I do have many users in my day job on Linux laptops/VDI (Gnome 3/Mate) with no experience prior and they have adapted fantastically. So much easier to support than Windows.
There are so many desktop environments already, all causing grief for anyone trying to switch to Linux with their own stupid choices and incompatibilities. It's probably too early to say anything about System76's project, but I fear it'll only complicate Linux adoption more, hurting their own sales in the long term.