I admire that … making a stand is about all us users can do
Yes, indeed. Sometimes it’s required to put up a sign.
Edit: Even if nobody cares. 
I suppose one could disassemble a flatpack, extract the package essentials, then install it without the container or the library overhead.
It might be easier to start from source code.
Yes, but then I’d partially accept the delivery of FP-only software. A complete refusal is for me a better approach.
Building from source is the last resort.
EDIT: Any FOSS software shouldn’t try to force its users into something.
I agree in principle
You want plain software without any packaging? … closest would be BSD ports.
It is not all that difficult. Bigest problem programs are monsters like firefox with heaps of dependencies.
But I see “us users” as regression testers also, “eating our own dog food” as MS was fond of saying in the Ballmer/Nadella era. Strange/esoteric use cases and bloat caused by the pervasive (invasive?) telemetry of the AI epoch “us users” are experiencing now is what can cause RedHat to declare 90% of its users employ Flatpak, right?
No, I want them to follow the system’s native packaging. That’s good enough; no need to reinvent the wheel.
I’m sure you aren’t the only one that prefers that. Simplifies a lot! But then we are heading for a future where rival tribes of developers insist their platforms are the only true path to Open Source nirvana. I love that Nix runs on nearly every OS and Hardware, even Mac, like Windows used to. The discussion about OSS vscode not running on the same machine as the vscode binary distribution is an interesting “edge case” but it brings up two questions:
- a portability question: can one share data locally between distros, if it is cloud-replicated?
- are declarative systems (yes Nix is my favorite) that do compilation and sandboxing natively, and can be atomically rolled back to when app x just worked, “the future?”
You are saying that the argument for containerised software is security?
Yes containers improde security … but you can do the same thing with firejail or VM’s.
If you care a lot about security use a specialised distro like Whonix.
Things like docker containers are easily distributed … you can just distribute the Dockerfile. But flatlak and snap did not do that … they invented their own container model… with bloat.
I. an unsperstand someone like Red Hat wanting to containerize everything… they are control freaks… but there is far less reason to impose containers on general Linux users.
My only concern with Flatpak was the size of some of the upgrades. One update was almost 1 GB in size! I don’t understand why an upgrade could be almost 1/4 of the size of a distro.
I have never liked Flatpaks or Snaps. AppImages I do like, as you have a bit of update control with them, plus they do not take up so much room, as a Flatpak and Snap packages do. It is the sandboxing of these Flatpaks and Snaps that take up all the room on your SSD. Plus they are rather slow to open and not always fully there, like the original Deb, RPM, or whatever package system. Call me old school, but I’d rather have the original Deb package installed. Why would a Linux user want to use Microsoft Edge? It is full of telemetry even here in Linux, it sends stuff back to Microsoft, plus it uses Bing, which is also sending stuff off to Microsoft. Microsoft Edge Linux release baffles the hell outta me.
Er, last time I looked, the total ecosphere of Docker was around 2+ GB, not including containers/images, which can get a bit bulky. The security model is, as you noted, superior to Flatpak and Snaps. With AI escaping its sandboxes recently, any Linux or other sandbox smells distinctly, um, urinary? Point being, any thing in a sandbox with an internet connection will have its own “threat model” to deal with. I used containers/VMs maybe 15 years ago to test antivirus software, a well-written virus can turn on network connections in its sandbox.
Yes that is the support. An image for one simple container has to include an OS minus kernel … plus the app it containerises…minimum is about 5Mb.
Running containers can share some of their layer structure… so 2 x 5Mb containers dont use 10Mb ram while running , but they each use 5Mb of disk.
Docker containers cant be executed like cli commands … you need a docker command to run them. I guess you could fix that with aliases or something?
You say sandboxes are not secure. I have to believe you. So what is secure, if anything?
I have never used AppImages so I don’t know how to use it. I will need to see if I can find some info on how to use it.
It couldn’t be easier.
- Download the AppImage. I prefer having them at /opt.
- Make it executable. (
chmod +x ...) - Run it
Usually the DE integration (launcher on desktop and menu) is done during the first run.
I have three versions installed on Linux: Stable, Beta, Dev.
I like the way it works. I have a bias against Google. I do have Firefox installed and use it occasionally.