This is one of the reasons I’m excited about “image mode”. I posted some thoughts kind of randomly in Base fedora image for constructing containers - #3 by mattdm. In the ideal world, we have a Base Runtime (*cough*) shared by (in the ideal) all our blocking deliverables. I think this layered structure can really help simplify the Quality team burden:
- Only one base to test for everything.
- More rigor and resources can be focused on ensuring the inner levels are solid, and explicitly less from Quality directly on the higher levels.
- Unlike prior minimization efforts, there is a tangible, immediate, and lasting benefit to moving packages out and up.
- A team working on a Spin or Edition starts with
from:fedora/ring3(the "Consensus Common Core[1]). They can be confident that that everything up until that line should just work, and be able to focus on their specific changes.[2]
With this, we can push more release-validation work to the Edition and Spin teams. We could even realize my ancient dream of pushing the release call for their own deliverable to that team, and not block on anything above the Consensus Common Core
We can (future-tense “can”) even build package mode (traditional rpm) deliverables using the image mode build pipeline, “flattened” as a last step. (As I understand it, @jligon is working on this.)
see the other message for more, but the rule of this layer is: everything common between all of the Editions, with implicit veto for every team that wants it smaller. Nothing to argue about — just move that thing up a layer. ↩︎
On the inner side of the line, I imagine a renewed “Base System Working Group” with ownership of the lower layers. I expect since the Edition Working groups do care about the end-to-end, they’d want to participate here as well (but wouldn’t have to). ↩︎