I don’t understand what problem they are meant to solve. If you have a FOSS piece of software, you can install it via the package manager. Or the store, which is just a frontend for the package manager. I see that they are distribution-independent, but the distro maintainers likely already know what’s compatible and what your system needs to install the software. You enjoy that benefit only through the package manager.
If your distro ships broken software because of dependency problems, you don’t need a tool like Flatpak, you need a new distro.
Just a shot in the dark, but it sounds like you have never maintained software. If you have, then my apologies, but your experience must not have included much by way of dependency and version management. Dependencies and versions are a major source of headache for any software engineer and it’s a problem of our own making.
Having completely self-contained applications, while more bloated, makes maintenance and distribution exceedingly simpler and puts the burden of managing that solely on the developer of said application instead of burdening everyone down the line.
I remember when package managers didn’t exist. It was painful. This is the next step in the evolution of the Linux desktop that was mostly solved in Windows and effectively completely solved on macOS since some early 10.x version.
I would not call it a next step. Just another option. The big downsides include much larger on block storage size and worse yet much bigger memory foot print since your app cannot benefit from shared images. Worse system integration too.
In a world where block storage is huge and cheap and memory too maybe less important. I would not say it is without issues though. Maybe convenient but not optimal in a lot of ways.
From my understanding, Flatpak is built on top of OSTree, which will automatically deduplicate files across different packages. That said, I’m not sure if this extends to downloading packages. The site claims that it does do “delta updates,” which would hopefully mean that it doesn’t download files that are already on the system, even if they’re part of another package.
I’m just going off what I read in the docs. Someone with more understanding of the system can clarify.
Flatpak does try to account for storage size by using shared base images. The main problem is that some Flatpak apps don’t update to the latest base, and some use different base images altogether, meaning most of the time it needs to have several bases anyway.
If people want the Linux desktop to become more ubiquitous in homes, it better damn well be the next evolution. Someone’s grandmother isn’t going to get on the command line when apt inevitably decides to break.
The concept is not new, and Apple has had .app containers for a very long time that almost always just works. So clearly the concept has long been proven.
Perhaps flatpak, snap, appimage aren’t the final forms of this concept on Linux, but it’s a step toward making application packaging and distribution much more friendly for the common masses.
It takes work off both the distro maintainers and the developers of the software, because they only need to provide a single package that works anywhere instead of packages for every single distro.
One thing I like about about Flatpak in particular is it allows me to have newer applications on distributions with older package bases (for instance, Debian.) I don’t much care for rolling release distros, and I’m not a fan of having to hunt for a 3rd-party repository, so for that purpose I really love the option to just get a Flatpak.
Also Bottles. Bottles is great.
I’m on Debian because the software in the Debian repos is stable. So for mission-critical software, at least for my purposes, I’ll pick the version in the Debian repo, especially if it requires detailed integration with the operating system such as real-time audio. If the software does get updated, it is probably important and nearly guaranteed not to break. A great example has been KDE Plasma: I don’t get the bleeding-edge features, but it’s been a rock-solid, fast, still modern desktop environment on every computer I installed it on, including an old laptop that is so underpowered that Windows 10 is a Power-Point presentation upon a fresh restart. If Debian takes several months or longer to update it’s Plasma packages to Plasma 6 when it comes out next year, that would be fine for me because I don’t desperately need any new features from Plasma.
However, for software that really benefits from being up-to-date and isn’t a showstopper if it breaks, for example FreeTube, I prefer the Flatpak. I primarily use Discover for simple package management and upgrades, and it was trivial to install the Flatpak backend, so now my Flatpaks get updated like anything else. However, Librewolf (a browser, which I prefer to keep up-to-date) is installed from a non-Flatpak external repo because I had problems giving its Flatpak version webcam permissions (even if I enabled them in Flatseal).
AppImages have been great for working on new computers because I can (usually) just download them and go. Except for programs that I expect to be portable, I don’t typically use them in the long haul. Still, they’re super convenient to have around.
I don’t touch Snaps because of the closed-source backend and their role in Canonical’s transparent attempt to lock down Ubuntu, but if they open-source the backend I might consider trying them.
IMO part of why I’ve stuck with Linux is because there is (usually) a choice of how to compute. I.e., there are several ways to solve a problem where Windows or Mac would pigeonhole you into their workflow. Having multiple options is inherently a good thing as far as I’m concerned, even if I don’t use all of them.
I like Flatpaks and AppImages for application delivery and here’s why:
-
Software doesn’t just magically appear in various distros’ repositories. There is a considerable amount of work (time/effort/energy/thought) that goes into including and maintaining any given program in a single distro’s repo, and then very similar work must be done by the maintainers of other independent repositories. To make matters worse, some programs are not straight-forward to compile and/or may use customized dependencies. In those cases, package maintainers for each distro will have to do even more work and pay close attention to deliver the application as intended, or risk shipping a version that works differently in subtle ways and possibly with rare bugs. (Needing to ship custom versions of deps for a certain program also totally eliminates a lot of the benefits of shared libraries; namely reduced storage space and shared functionality or security.) That’s part of the reality of managing packages, and the fact is that there’s a lot of wasted effort and repeated work that goes into putting this or that application into a distro repository. I have a ton of respect for distro package maintainers, but I would prefer that their talents and energy could be used on making the user experience and polish of their distro better, or developing new/better software, than wrestling with every new version of every package over and over again multiple times per year.
-
As a developer it’s very nice to know exactly what is being “shipped” to your users, and that most of your users are running the same code in a very similar environment. In my opinion, it’s simply better for users and developers of a piece of software to have a more direct path, instead of running through a third party middle-man. Developers ship it, users use it, if there’s a bug the users report it, developers fix it and add features and then ship again. It’s simple, it’s effective, and there’s very little reason to add a bunch of extra steps to this process.
-
The more time I spend using immutable, atomic Linux distros like Silverblue, the more I value a strong separation between system and applications. I want my base system to be solid as a rock, and ideally pretty fucking hard to accidentally break (either on the user end or the distro end). At the same time I also want to be able to use the latest and greatest applications as soon as humanly possible. Well, Silverblue has shown that there’s a viable model to do that in the form of an immutable and atomic base system combined with containerized applications and dev environment. What Silverblue does may not be the only way of achieving a separation between system and applications, but I’ve never been more certain that it’s the right direction for creating a more stable and predictable Linux experience without many compromises. I don’t necessarily want to update my whole system to get the newest version of an application, and I certainly don’t want my system to break due to dependency hell in the process.
-
The advantages of the old way of distributing applications on Linux are way overblown compared to the advantages of Flatpak. Do flatpaks take up more drive space than traditionally packaged apps? Maybe, I don’t even know. But even if they do, who the hell cares? Linux systems and applications are mostly pretty tiny, and a 1TB nvme ssd is like $50 these days. Does using shared library create less potential for security flaws going unfixed? Possibly, but again, sometimes it just isn’t possible or practical for applications to share libraries, Flatpaks can technically share libraries too, and the containerized nature of Flatpaks mean that security vulnerabilities in specific applications are mitigated somewhat. I’m not a security guy, but I’d guess that Flatpaks are generally pretty safe.
Well, that’s all I can think of right now. I really like Flatpaks and to some extent AppImages too. I still think that most “system-level” stuff is fine to do with traditional packaging (or something like ostree), but for “application-level” stuff, I think Flatpaks are the current king. They’re very up-to-date, sandboxed, often packaged by the developers themselves, consistent across many distros, save distro maintainers effort that could be better used elsewhere, easy for users to update, integrate with software centers, are very very unlikely to cause your system to break, and so on.
It would be really hard for me to want to switch back to a traditional distro using only repo packages.
I disagree with so much of this.
You might not care about the extra disk space, network bandwidth, and install time required by having each application package up duplicate copies of all the libraries they depend on, but I sure do. Memory use is also higher, because having separate copies of common libraries means that each copy needs to be loaded into memory separately, and that memory can’t be shared across multiple processes. I also trust my distribution to be on top of security updates much more than I trust every random application developer shipping Flatpaks.
But tbh, even if you do want each application to bundle its own libraries, there was already a solution for that which has been around forever: static linking. I never understood why we’re now trying to create systems that look like static linking, but using dynamic linking to do it.
I think it’s convenient for developers to be able to know or control what gets shipped to users, but I think the freedom of users to decide what they will run on their own system is much more important.
I think the idea that it’s not practical for different software to share the same libraries is overblown. Most common libraries are generally very good about maintaining backwards compatibility within a major version, and different major versions can be installed side-by-side. I run gentoo on my machines, and with the configurability the package manager exposes, I’d wager that no two gentoo installations are alike, either in version of packages installed, or in the options those packages are built with. And for a lot of software that tries to vendor its own copies of libraries, gentoo packages often give the option of forcing them to use the system copy of the library instead. And you know what? It’s actually works almost all the time. If gentoo can make it work across the massive variability of their installs, a distribution which offers less configurability should have virtually no problem.
You are right that some applications are a pain to package, and that the traditional distribution model does have some duplication of effort. But I don’t think it’s as bad as it’s made out to be. Distributions push a lot of patches upstream, where other distributions will get that work for free. And even for things that aren’t ready to go upstream, there’s still a lot of sharing across distributions. My system runs musl for its C library, instead of the more common glibc. There aren’t that many musl-based distributions out there, and there’s some software that needs to be patched to work – though much less than used to be the case, thanks to the work of the distributions. But it’s pretty common for other musl-based distributions to look at what Alpine or Void have done when packaging software and use it as a starting point.
In fact, I’d say that the most important role distributions play is when they find and fix bugs and get those fixes upstreamed. Different distributions will be on different versions of libraries at different times, and so will run into different bugs. You could make the argument that by using the software author’s “blessed” version of each library, everybody can have a consistent experience with the software. I would argue that this means that bugs will be found and fixed more slowly. For example, a rolling release distro that’s packaging libraries on the bleeding edge might find and fix bugs that would eventually get hit in the Flatpak version, but might do so far sooner.
The one thing I’ve heard about Flatpak/Snap/etc that sounds remotely interesting to me is the sandboxing.
-
In that specific area, that’s where compared to each BSD operating system, Linux will forever be trash garbage.
In what way does BSD solve this issue? I would not consider limited number of distributions, small user base, and package count exactly a solution.
That said, if BSD had been released as FOSS a decade earlier I imagine we would all be using BSD not Linux. Would have been an interesting twist of fate.
BSD is FOSS, unless you are an idealogue.
BSD does not have distributions, those don’t exist.
BSD totally has distributions. Some versions of BSD are separate operating systems from each other, not distros, but things like GhostBSD or MidnightBSD are absolutely FreeBSD based distros.
Wrong, guess again. Read the websites.
You mean like where immediately on the front page of the GhostBSD website it says that it’s built on top of FreeBSD code? Just because they don’t use the term distro doesn’t mean they’re anything different.
You stuck in the cult of Linux and projecting your mentality onto other things without deciphering each on their technical marits. You look at all software in Linux terminology rather than making a distinction to articulate correct phrasing in a cohesive manner.
I call a spade a spade. If you can’t handle two binary compatible versions of BSD being called distros just because it’s a Linux term even though by every possible definition of that term that doesn’t include the word “Linux” they absolutely are distros, that’s your problem.
I think the comment refers to this:
Not sure. This is a container basically not a distribution format. Not sure how this is different from Linux containers though Linux has a bunch of options. Not sure which is most similar.
Yeah, I think the BSDs lead the way with some things, like jails. But they’re not distribution formats. But jailing is part of things like Flatpak. And we have chroots and systemd nspawn. I think I misread the comment and it was more shitposting than anything of substance.
That is the thing about Linux. Linux is so huge and has so many ways of doing things that the days of knowing everything are gone. Same with many of the important tools too. Python has gotten way vast too for example.
This is kind of old people behaviour. I’m still not 100% sure if I’m getting more conservative, having difficulty with things changing, or if things really used to be better… They’re different, that’s for sure. And I have some valid criticism for some things, too.
Frankly I loved my Commodore 64. My Linux box is better in every measurable way but there was something to simplicity and a time where just making a sound and drawing on the screen with a computer you could afford was quite a thing.
Same with Python. Started using it about 1998. It was simple enough that I learned the language in a day. Now there is so much more. Then add packages for everything these days… lot of the work is understanding packages, venvs, how to deploy, not just opening idle or pywin and writing stuff. Sure Spyder or one of the other IDEs can do static checking, have doc at your fingertips, integrate a debugger, and have a graphical shell where you can do all sorts of stuff. Changes the feel of programming though.
