Every time I tried Flatpak it did not work, always some other problem. Maybe this got fixed nowadays. I did not find it user friendly either, but this could be addressed by the distro.
AppImage actually works and is simple, so I use that.
Snap is trash because autoupdate cannot be disabled without code changes (what?!) last time I checked.
The thing with being around Linux since its early days (Slackware 2.0 in 1996), is that this kind of content regarding Linux Desktop fragmentation and fixes just around the corner gets tiresome.
KDE and GNOME in the 2000's already had everything in place, for vertical application development, frameworks and co, the only part missing was having OEM support for selling ready made systems, instead of reverse engineering hardware all the time.
Instead rebooting the experience, incompatible Gtk+ versions, multiple sound systems, a X Windows replacement that is yet to be widely available,....
Eventually the peace of mind is to use it the headless, time sharing approach that UNIX was developed for, and leave it at it, ssh and Web instead of telnet and X Windows of yore, into servers where OEMs actually care to support their hardware.
Possibly because the Gnome project has announced a Flatpak-based distribution (if I got that right anyway), which would have brought Flatpak prominently to people's minds again.
Snaps are centralized, provided by one company. Sorry, but no. It can be 'better' for some reasons, but free software should not be limited to one or few app stores.
I actually prefer flatpak over snaps if it has been made by original authors of software. Snaps get unusable really fast and have a lot of unexpected problems and inconveniences around the seams.
The only case where I do find Flatpak useful is to install proprietary electron based applications that do not interact with the rest of the system so much, but I try my best to avoid using those in the first place.
I don’t think the issue is really that api aren’t mature enough or that some developers refuse to use flatpak sandboxing correctly and instead use static permissions.
Linux, Unix and most computers were designed from the ground up as open system where everything can access everything.
It’s absolutely not secure, but that’s not what flatpak is supposed to fix.
Flatpak is supposed to make installations of apps easier by sharing common dependencies across all distributions. That’s all. And in a way, it works.
It has downside even on that, people who have slow internet connection will require hours to download a simple calculator app whereas it took a few seconds before, but it works.
Now on the sandboxing part, you can’t fix a sinking ship with a chewing gum in the hole. The whole structure is not designed to work this way.
It has to come from yet another distribution where permission are enforced at the operating system level, like on macOS. It’s a lot of work, and then you got to get developers on board, so it will never happen and it’s fine. Linux is not macOS, it doesn’t have a trillion dollar company backing it up and it’s not unified.
You got to play with your strength, not your weaknesses.
Actually, most operating systems were designed as multi-user systems where users cannot access each other's data. Trusted administrators installed software. This made sense in a world of expensive computers that needed to be shared.
In the modern world we do not share our computers with other users, and allow untrusted software to run on our machines from the internet. Without further sandboxing, it would execute as the user with access to everything for that user - which in practice means mostly everything.
I wish I could care less about Flatpak. But they inject themselves into every protocol discussion on Wayland and loudly insist only the flatpak solution is acceptable. That we have to keep using their dbus or whatever they do thinh. They drove off a xfce maintainer in record time who was just trying to discuss how to help their users, with terrible behavior. https://gitlab.freedesktop.org/wayland/wayland-protocols/-/m...
Everyone else is getting off dbus, is working to let the compositor have power, have authority. Dbus is this other system that has some security options but that rarely is well locked down.
And it's just wild to me that there's such ongoing protest against using Wayland protocol development that's so insistent, and it's especially annoying and frustrating seeing it from biased folk like Flatpak who want to make sure there is no discrete source of authority, who want to make sure they can own a chunk of the UX. DBus and their protocols make sure Wayland doesn't become the one way to run a good desktop, that you have to keep having multiple totally different messaging systems going. Fine, I don't begrudge you that as an option, but don't keep road blocking those who want to develop Wayland, please!
KDE and GNOME in the 2000's already had everything in place, for vertical application development, frameworks and co, the only part missing was having OEM support for selling ready made systems, instead of reverse engineering hardware all the time.
Instead rebooting the experience, incompatible Gtk+ versions, multiple sound systems, a X Windows replacement that is yet to be widely available,....
Eventually the peace of mind is to use it the headless, time sharing approach that UNIX was developed for, and leave it at it, ssh and Web instead of telnet and X Windows of yore, into servers where OEMs actually care to support their hardware.
The only case where I do find Flatpak useful is to install proprietary electron based applications that do not interact with the rest of the system so much, but I try my best to avoid using those in the first place.
Linux, Unix and most computers were designed from the ground up as open system where everything can access everything.
It’s absolutely not secure, but that’s not what flatpak is supposed to fix.
Flatpak is supposed to make installations of apps easier by sharing common dependencies across all distributions. That’s all. And in a way, it works.
It has downside even on that, people who have slow internet connection will require hours to download a simple calculator app whereas it took a few seconds before, but it works.
Now on the sandboxing part, you can’t fix a sinking ship with a chewing gum in the hole. The whole structure is not designed to work this way.
It has to come from yet another distribution where permission are enforced at the operating system level, like on macOS. It’s a lot of work, and then you got to get developers on board, so it will never happen and it’s fine. Linux is not macOS, it doesn’t have a trillion dollar company backing it up and it’s not unified.
You got to play with your strength, not your weaknesses.
In the modern world we do not share our computers with other users, and allow untrusted software to run on our machines from the internet. Without further sandboxing, it would execute as the user with access to everything for that user - which in practice means mostly everything.
Which is fine and the reason behind qubes OS decision to have passwordless sudo enabled on qubes.
Everyone else is getting off dbus, is working to let the compositor have power, have authority. Dbus is this other system that has some security options but that rarely is well locked down.
And it's just wild to me that there's such ongoing protest against using Wayland protocol development that's so insistent, and it's especially annoying and frustrating seeing it from biased folk like Flatpak who want to make sure there is no discrete source of authority, who want to make sure they can own a chunk of the UX. DBus and their protocols make sure Wayland doesn't become the one way to run a good desktop, that you have to keep having multiple totally different messaging systems going. Fine, I don't begrudge you that as an option, but don't keep road blocking those who want to develop Wayland, please!