Creating own standard distribution is a cool idea. It’s like the devices by the OS developers (Google Pixel, Microsoft Surface).
Purpose
Create Linux distribution with pure GNOME for regular usage.
Base
I think that GNOME OS should be based on Ubuntu.
Ubuntu is the most popular Linux distribution. If developers create software for Linux, then precisely for Ubuntu, because easier to port software to one distribution than to whole zoo of distributions.
So we won’t make Linux more fragmented.
Applications
On Google Pixel, only Google apps are pre-installed.
Let’s also deliver only GNOME core apps (Web instead of Firefox, Music instead of Rhythmbox). Users will install needed apps from Software Center.
Self-contained packages
Logical to use Flatpak, standard self-contained packages for GNOME.
It’s not for a lack of desire. I think you’re going to have a hard time finding broad support to build on a specific downstream distribution from those of us creating upstream. I think you’ll also find that few of us want to maintain an additional distro (since many contributors are employed to work on a distro already) with serious CVE updates when everyone is already spread so thin building the system components themselves. Not to mention, building upstream is very different from building an integrated distribution that is stable from release to release.
However, a number of people would like a boot-able, conformance testing, OS image that is updated daily (or close to it). If that were to happen, there are a few things that I think are important to consider.
GNOME Builder wants VM/simulator images so it would be wise to build VM bootable images with enough tooling to push/debug applications into the VM (flatpak could work here using ostree sync on the host w/ virtio-fuse)
We might need alternate architecture support for #1
It would be nice if we could replace/override components so that plumbing developers could use these VM images to develop the OS (Shell, Mutter, Gtk, GLib, etc)
Keeping the images up to date should be minimal bandwidth and safe
Interpreting the above makes a few technologies obvious from my perspective
OSTree for storage/sync/updates
BuildStream for components which can be replaced
Once we get into the phase of replacing OS components, things get much tougher. But from my view point, it’s the one thing that makes all this work compelling from a competitive tooling standpoint.
There’s no one distribution using GNOME - Endless, PureOS, Fedora, and Ubuntu are all examples of distributions that hire developers to work on GNOME or on things for their distribution.
If GNOME OS is to be a thing, I think it would need to line up closely with the work we’re already doing. Whether
that’s shipping things as flatpaks or having an ostree-based platform to build off of. This way integration is less of a chore.
In addition, I think a potential GNOME OS should first start off as development tooling for core system components like the Shell, rather than something consumer-facing.
