I’ve been running Linux for ~10 years. Again, I can get it working just building from source. My complaint isn’t that I can’t get Blender to work, it’s pointing out that Linux does break sometimes. I don’t make excuses for Linux breaking is all. I think it should strive to do better.
His complaints are solved by distro agnostic packaging, hence flatpack/snappy.
I never said the Linux desktop market share was large, but that a small number of distros has the vast majority of users in that market.
If end users wanted compatibility more than anything else then they would all choose Ubuntu which is already the most popular by far.
Things that attract one user to one distro instead of another is typically that which separates them, if they don’t care, they are going with the most popular one.
What exactly is this ‘very low-hanging fruit’ ?
What’s this about ?
Are you claiming that the Linux distros are changing things around all the time ? Please give me some examples. Biggest change we’ve seen the past 10 years would be the wide adoption of systemd.
Sounds like you should just choose a GNOME/KDE desktop based distro, and install through the bundled graphical application installer. It just ‘works out of the box’ in the way you described.
Then either back one of flatpak or snappy, or wait and hope one of them becomes a de facto standard. You can’t prevent competing formats from existing.
I mean, how do you handle the existence of both Steam and Microsoft Store on Windows ? Sounds like total chaos for a developer !
Please elaborate.
If dependencies are updated which break abi compability, the packages are updated as well automatically through the package manager update.
As long as the dependencies don’t break stability, the binaries do not need to be updated, Linux and glibc are very common dependencies with a stable userspace abi, I have binaries that are at least 5-6 years old and run perfectly, I could probably run binaries decades old, I just don’t have any.
I’ve never used AppImage as I found it to be a crippled solution.
I have tried flatpack, it supports the ‘Desktop Entry specification’ standard for integrating with a desktop environment (show up in menus, icon, information on what file types it supports/associates etc).
You install the container solution through your distro package manager, I listed some distros supporting flatpak (it’s still in development) in an earlier post, then you can click on package files or search for packages through repositories.
Since the format is not fully finalized (I think), there are only a small range of packages available for testing right now, you can find some here: http://flatpak.org/apps.html
Apart from being distro agnostic, there are other advantages like sandboxing.
Yes, but they have no control outside of Linux, and distro maintainers have no control outside of their own distro. No one can force anyone to adopt anything.
If you are not going to back up your statements with anything remotely concrete then please don’t bring them up at all. I have no interest in a ‘X SUCKS because I say so!’ exchange.
You don’t want to ‘install some application’ on your production system, you want to install applications which have been thoroughly tested and is officially supported on your system.
At home I run Arch, I would never consider running it in production.
Since we are in a 3d forum, a logical example would be Disney, Pixar, Dreamworks, etc who all run their production pipelines on STABLE Linux enterprise distros, it’s not as if their employees get to install some non-vetted software.
That’s your interpretation, Microsoft suddenly deciding to implement the BASH shell on Windows tells me another story.
Looks like Ubuntu is dropping Mir and Unity. I’m glad to see Mir go. I’d rather everyone just supports Wayland.
Well, I don’t consider it solved (for the reasons I mentioned). Linus (or rather his team) doesn’t ship flatpack/snappy, they ship AppImage for Subsurface.
I never said the Linux desktop market share was large, but that a small number of distros has the vast majority of users in that market.
Of which each are fractions of a percentage, like I said. I don’t even know how we’re disagreeing.
If end users wanted compatibility more than anything else then they would all choose Ubuntu which is already the most popular by far.
How is Ubuntu in particular so great at compatibility? Ubuntu breaks stuff all the time, as well. Packages disappear on upgrade. Binaries are not compatible across releases. Their new desktop compositor breaks many applications and it’s not compatible with Wayland (which also breaks stuff, but at least it’s has good reasons).
I’m not even sure if Ubuntu is really that popular on the Desktop, especially after introducing Unity. What’s your statistic?
What exactly is this ‘very low-hanging fruit’ ?
Being able to distribute software and have it work neatly (comparison: Windows/Mac OS) on the majority of systems, without user intervention.
Are you claiming that the Linux distros are changing things around all the time ? Please give me some examples. Biggest change we’ve seen the past 10 years would be the wide adoption of systemd.
Things are renamed, moved, packages are removed, dependencies in particular versions are removed, configuration file conventions change… those are not big changes, but they break things nonetheless.
The last time I’ve wasted an extraordinary amount of time was trying to get xrdp to work. I’ve stumbled upon this guy’s blog, who has half a dozen entries on getting it to work just on the latest *buntu.
Sounds like you should just choose a GNOME/KDE desktop based distro, and install through the bundled graphical application installer. It just ‘works out of the box’ in the way you described.
That only works for applications maintained by the OS itself. That’s a completely different topic.
Then either back one of flatpak or snappy, or wait and hope one of them becomes a de facto standard. You can’t prevent competing formats from existing.
That’s not my job. The distributions have to integrate this, like I said. I have no problem with them integrating all of them.
I mean, how do you handle the existence of both Steam and Microsoft Store on Windows ? Sounds like total chaos for a developer !
Those are storefronts. They’re providing value, visibility, i.e. they make it worth your while. Except the Windows Store actually doesn’t provide that much value (while bringing tons of restrictions), which is why there is mostly garbage on there.
Please elaborate.
If dependencies are updated which break abi compability, the packages are updated as well automatically through the package manager update.
Yes, but only the ones that are being maintained by the distribution. That’s the fucking problem! As a third party, you can’t ship a zillion packages for every combination of dependencies out there.
As long as the dependencies don’t break stability, the binaries do not need to be updated, Linux and glibc are very common dependencies with a stable userspace abi, I have binaries that are at least 5-6 years old and run perfectly, I could probably run binaries decades old, I just don’t have any.
A C library can maintain its ABI for a while. Anything using C++ (without a C interface) can’t. Your average C++ program will have maybe five dependencies (also C++) some of which probably depend on some version of Boost (try counting how many versions of that are in your repos…) and it all just goes downhill from there.
I’ve never used AppImage as I found it to be a crippled solution.
I have tried flatpack, it supports the ‘Desktop Entry specification’ standard for integrating with a desktop environment (show up in menus, icon, information on what file types it supports/associates etc).
I’ll check it out the next time I actually find an application using it.
Yes, but they have no control outside of Linux, and distro maintainers have no control outside of their own distro. No one can force anyone to adopt anything.
That’s all the control you need. Systemd got adopted because those people decided to go for it, even though lots of people whined about it (and they had good points too!). It’s the standard now. Adoption didn’t come from “below”.
For something like a container format (which doesn’t even bother anybody like systemd), the same thing can (and needs to) happen.
If you are not going to back up your statements with anything remotely concrete then please don’t bring them up at all. I have no interest in a ‘X SUCKS because I say so!’ exchange.
Ordinarily I would agree, but you’ve already been weaseling around so much in this endless quote battle, so there really isn’t a point. I happened to give the example of XRDP above, feel free to take that apart, I don’t care either way.
You don’t want to ‘install some application’ on your production system, you want to install applications which have been thoroughly tested and is officially supported on your system.
That’s kind of bullshit, because almost no software is really super-stable these days (nor does it need to be) and the distro maintainers aren’t helping, because they are not the developers. All they can do is withhold updates.
As an adult user, I’ll be the judge of which software is stable enough to run in my production.
Since we are in a 3d forum, a logical example would be Disney, Pixar, Dreamworks, etc who all run their production pipelines on STABLE Linux enterprise distros, it’s not as if their employees get to install some non-vetted software.
Nobody here is Pixar, Disney or Dreamworks…
That’s your interpretation, Microsoft suddenly deciding to implement the BASH shell on Windows tells me another story.
They now have a “Windows subsystem for Linux” (based on Ubuntu), which is the continuation of their much older “Windows Subsystem for UNIX” (which made Windows POSIX compliant!). This basically replaces Linux-VMs for a couple of people, making actual Desktop Linux even less relevant. Like I said, many developers need a Linux system for testing, one way or another, because on the server, Linux is ubiquitous.
Just so we are clear, Subsurface is probably available in the official repos of every Linux distro with more than 10 users.
As for them using AppImage, the improved alternatives are still in development (although I’m not sure how snappy’s future is now with Canonical going back to Gnome, perhaps they will deprecate snappy in favor of flatpack).
Every Linux desktop statistic I’ve seen over the years has Ubuntu firmly at the top, you’ll find some gathered here: https://en.wikipedia.org/wiki/Usage_share_of_operating_systems
How is that ‘low-hanging fruit’ when it comes to Linux distros ? Again a solution like flatpak/snappy is the closest thing.
The distro/package maintainers should give you ample warning if updates break existing setups, and descriptions on how to fix it, if they don’t, vote with your feet.
Of course it’s not as if Windows upgrades don’t break things, even leading to lawsuits.
It doesn’t seem to be packaged by many distros, on Arch it seems quite easy to get up and running despite not being officially supported: https://wiki.archlinux.org/index.php/xrdp
Perhaps a prime candidate for flatpak/snappy type packaging.
Given that your complaint was about not being able to target cross distro Linux support, using one or both of flatpak/snappy sounds worthwhile.
Which is why third-parties typically only support a select few, alternatively statically compile all dependencies (comes with it’s own can of worms, I know).
This is what flatpak/snappy should solve.
Heh, those libraries must have taken years off package maintainers lifespans.
I provided a link to some 30 apps, including Blender
Key point is ‘decided’, they decided on systemd because it was a huge improvement on existing solutions.
I guess the next big wide transition will be towards Wayland, particularly now that Unity is dead.
I hope either flatpak or snappy will turn out to be a superior solution, because then we will see the vast majority of distributions adopting one of them and thus minimizing fragmentation.
No argument here.
Production software which you depend on for your business needs to be stable, bugs and crashes cost money if you’re lucky, can be disastrous if you’re unlucky.
If we look at enterprise distros, they employ a lots of developers who can indeed fix bugs, Red Hat being the prime example.
No, they can pass bugs upstream and apply upstream fixes downstream.
Yes, in YOUR production, if you are the boss you can do whatever you want, I fail to see your point.
The same reasoning applies, if your livelihood depends on your software pipeline working correctly, introducing and depending on untested software during production is folly.
It has been that for ages, this is a recent reaction from Microsoft.
Of course Linux is also ubiquitous on ‘the cloud’, which is were a huge amount of software development is moving.
Yes, since there doesn’t seem to be any technical motivation for Mir I can’t say I’m sad to see it disappear, full focus on Wayland as the X successor !
That said, has there been any official word on Mir, or is that just implied as it was so closely tied to Unity ?
I’m not sure. The OMG Ubuntu article says Mir specifically, but the blog post doesn’t seem to. I can’t imagine why they’d keep it if they’re ditching phones, though.
Moderation: BeerBaron, PolyPincher… please lay off on the quote battle. You can continue your debate by PM. Future quote battle posts of yours will be removed from this thread.
Sure, but that’s besides the point. The people that really need this are smaller FOSS developers and non-FOSS developers.
As for them using AppImage, the improved alternatives are still in development (although I’m not sure how snappy’s future is now with Canonical going back to Gnome, perhaps they will deprecate snappy in favor of flatpack).
Again, I do not care which solution wins out. Either one of these sounds fine. As long as these aren’t integrated and work out of the box on all major distros, I do not consider this problem “solved”.
Every Linux desktop statistic I’ve seen over the years has Ubuntu firmly at the top, you’ll find some gathered here: https://en.wikipedia.org/wiki/Usage_share_of_operating_systems
I actually can’t find clear data there. I wouldn’t be surprised if Ubuntu was at the top, but anecdotally, it’s not actually that popular.
How is that ‘low-hanging fruit’ when it comes to Linux distros ? Again a solution like flatpak/snappy is the closest thing.
It’s a low-hanging fruit because it’s not a big technical effort, but apparently in the Linux world people can never agree which fruit to pick, so they’re all rotting away. This could’ve been solved fifteen years ago, and now I’m told I still need to wait for something to be done.
Then there’s people like you that are discussing away the problem indefinitely. As a result, whenever there’s an opportunity for Linux (such as the Windows 8 or the Windows 10 debacles), Desktop Linux isn’t ready.
The distro/package maintainers should give you ample warning if updates break existing setups, and descriptions on how to fix it, if they don’t, vote with your feet.
I am voting with my feet, I’m not using Linux for the Desktop. I’d also be very wary shipping software for the Linux Desktop. That whole “then don’t use it” attitude is really bad for adoption, though.
Of course it’s not as if Windows upgrades don’t break things, even leading to lawsuits.
… and when it does, people are rightfully outraged. Most of the time, breakage is by accident, not intentionally. Microsoft doesn’t tell you to vote with your feet, they take compatibility serious.
I provided a link to some 30 apps, including Blender
I’ve now checked this out on a Debian Stretch installation. Flatpak needed to be installed manually (fair enough). After installation, I downloaded a “flatpakref” for GNOME FeedReader. The mimetype wasn’t setup correctly, so “.flatpakref” opens as a text file. I had to manually install the file from the commandline. Installation was successful, but the program didn’t show up in the menus.
This is a near-complete failure, concerning the requirements I have defined. This is why integration is so important and why the problem isn’t anywhere near solved.
Also, this is not the solution I asked for, because this is not a container format, it’s a format for referencing packages from a repository. That package then needs to be downloaded through a program. That download may break in the future, repositories need to be maintained, installation requires an internet connection (or some other elaborate setup), etc.
Key point is ‘decided’, they decided on systemd because it was a huge improvement on existing solutions.
Arguably, arguably not. I’m not going there now though, I don’t care.
I guess the next big wide transition will be towards Wayland, particularly now that Unity is dead.
Good riddance…
I’m fine with Wayland.
Production software which you depend on for your business needs to be stable, bugs and crashes cost money if you’re lucky, can be disastrous if you’re unlucky.
I’m not gonna speak for every business in the world, but if a crash is disastrous to you, you’re probably doing something wrong. You need to be fault-tolerant, because there will be bugs in your software, guaranteed. The solution here is not having some maintenance gnomes wave their chickens for a year before shipping, it’s implementing failover.
You also need to be able to work around problems when you will be the first to find them. All the testing in the world will not prevent Murphy’s Law from taking effect.
If we look at enterprise distros, they employ a lots of developers who can indeed fix bugs, Red Hat being the prime example.
Fair enough, but that’s for a very specific selection of packages.
No, they can pass bugs upstream and apply upstream fixes downstream.
I’m not sure how this is different from “withholding updates” from a user perspective, that’s just an extra middleman in there. Unless you’re talking about backporting fixes, which probably should make you nervous as well if you care about stability so much.
The same reasoning applies, if your livelihood depends on your software pipeline working correctly, introducing and depending on untested software during production is folly.
The first party that should do (and usually does) the testing is the developers, not some maintainers that often barely understand the packages they are responsible for. “Tested” also doesn’t mean that software is “stable”. Just because some maintainer checks off something as “stable” doesn’t mean it is perfectly reliable. For Blender, at any given point in time, there’s bound to be 100+ unfixed bugs. Where do you draw the line?
Come on… this is a useless thread anyway. Your moderation adds nothing. Let us have our talk, it is meant to be public.
Looks like Canonical is laying people off and looking for investors. I haven’t used Ubuntu in a long time, but I always liked Shuttleworth. I think the last 5 years or so they lost focus and started throwing stuff at the wall to see what sticks. Either way, too bad things don’t seem to be panning out in Ubuntu-land.
I rather think they focused on something that nobody asked for. This quote from the article is a real gem:
“Unity 8 was axed after seven years’ investment because it’s been determined it could not make money.”
Who could’ve foreseen that, after watching Android take over and Microsoft failing with basically the same strategy? Here’s to the all the curmudgeons and naysayers who said this was gonna fail all along. They were completely right!
Would definitely consider very closely.The problem is all the other stuff like Marmoset, Topogun and Painter. That, and will they allow you to transfer your licences without any penalty?
Large studios use Linux because it can be integrated into their pipeline easily due to it being a wide-open platform. These studios have huge IT and Engineering resources that individual artists don’t have to support the customized environment that their Linux workstations exist in.
So no. It doesn’t make a whole lot of sense for individual artists to use Linux, and certainly not for stability/quality control - and furthermore, most desktop users are using Ubuntu, which isn’t even officially supported by the much of the commercial software that these studios use - so it’s a non-sequitur.
Again, it boils down to choosing the right platform for the purpose, and has absolutely nothing to do with the subjective “superiority” of one over another.