Investigate usage of fully portable app formats like FlatImage or Conty

[fiftydinar]

Links

FlatImage:
https://github.com/ruanformigoni/flatimage

Conty:
https://github.com/Kron4ek/Conty

List of experimental FlatImage applications:
https://github.com/flatimage

Explanation

FlatImage & Conty basically pack whole desktop environment, along with application into the compressed binary.

That can be tweaked, as FlatImage supports minimal Alpine base too, where app developer can tweak which packages they want to install, to maintain working app, while retaining acceptable file-size.

One improvement that could be made to these, is to have something akin to standardized flatpak runtimes.
F.e. you would have minimal alpine image as a base & then you would layer runtimes into it, like Mesa GPU runtime, proprietary Nvidia GPU runtime, Gnome (GTK) runtime, KDE (Qt) runtime etc.

That would standardize things much better & make it much easier for developers to test their FlatImage/Conty applications.

Comparison

Compared to AppImages, they would work everywhere, even with Linux distributions which ship musl instead of glibc.
It is known how much of a hassle is to pack complex AppImage applications to work properly.

2nd advantage of those is that they support sandbox & setting permissions using bubblewrap, like flatpaks.

3rd advantage is non-dependency of libfuse (for FlatImage). It requires fusermount only.

Honorable mention

Universal Blue's experimental Bext project:
https://github.com/ublue-os/bext

As far as I know, it is intended for CLI apps only & it requires systemd to be present, as it uses systemd-sysext functionality for mounting images.

This one is too technical for me to understand, so I will update this part of the post when I get more information.

[ivan-hc]

All projects above have what my ArchImage miss: hardware accelleration.

For all the rest, they are bloated, being them complete systems/containers into a sort of AppImage.

ArchImages can work everywhere too, in particular if bloated (i.e., without removing unneeded stuff).

So, the only thing to investigated on here is "how to integrate hardware accelleration in ArchImages?"

[fiftydinar]

Hmm, interesting take too.

I somehow didn't see this apart from your AM/AppMan work.

Your distribution method is a lot more efficient as I look into it compared to others.

It would be great if this becomes a standardized way of making AppImages.

[ivan-hc]

I don't limit myself developing a distribution method for AppImages, I've also my own AppImages and my own scripts/tools to create them:

• https://github.com/ivan-hc#creating-appimage-packages

• https://github.com/ivan-hc#my-appimage-packages

By the way, they are also frowned upon by the AppImage community. Please, don't ask my "why?"