ruanformigoni/flatimage

idea: Introduce concept of standardized runtimes

Opened this issue · 2 comments

fiftydinar commented

I find that using a standardized minimal image like Alpine & layering needed application dependencies on it would be the best, in order to retain acceptable FlatImage file-size.

To make this easier, standardized runtimes would be introduced, which are similar to flatpak runtimes:

Examples:

  • Mesa GPU runtime
  • Gnome (GTK) runtime
  • KDE (Qt) runtime
  • Fontconfig runtime
  • Some freedesktop runtimes
  • etc.

I guess that those runtimes can be implemented as a script of install commands with grouped dependencies for each base, which would be easily visible & nicely sorted in this repo, so others can improve & fix it when necessary too.

This implementation would make things easier for developers to make & test their FlatImages for reliability before releasing them.

I know that this is easier said than done, so it's just an idea for now.

ruanformigoni commented

Hi @fiftydinar , thanks for the suggestion! I got this on my plans for flatimage, something like:

  1. Distribute the runtimes as ready-to-use dwarfs files.
    • There would be scripts to generate each runtime.
    • The runtimes would be provided with each base, both generated from github-ci.
  2. Overlay dwarfs runtimes with overlayfs (easy to switch between them on the fly, and they will also have a smaller footprint).
  3. Allow users to create their own runtimes, similar to:
# Create runtime 
./alpine.flatimage fim-runtime-create --from=qt5-qtbase,qt5-qtwebengine --to=qt5runtime.dwarfs
# Include (If a pre-configure qt5 runtime is downloaded, only this step is required)
./alpine.flatimage fim-runtime-add qt5runtime.dwarfs # <- overlayed over /usr

I would still need to maintain a gnu base alongside alpine, this is due to applications like gameimage, wine is problematic on alpine and symlinking the host drivers to inside the container for nvidia (like it is currently done) would not work due to alpine using musl instead of glibc.

Let me know what you think 😄

  • 👍2
fiftydinar commented

@ruanformigoni I'm glad that you had that in mind, as your input sounds great!

I cannot pinpoint any issues with this approach, but I'm not an expert, so take this with a grain of salt.

  • 👍1