That was my initial thought, but I was aiming for a compromise approach with SHOULD
.
I’ll note that we already require collections to document requirements within plugin DOCUMENTATION
, but that’s
Those three usecases are definitely valid, even if there’s other ways to address them. Thank you for bringing them up! I’m mainly hung up on the lack of direct benefit to users of the ansible package.
Regarding the last two points, we already strugle with collection maintainers running the required tests, so I’m not inclined to add a requirement for how they have to run them. Also, there’s the issue with runtime vs. testing dependencies. There should be a separation of concerns there, and asking maintainers to put a requirements.txt for test dependencies in the repo root doesn’t address that and doesn’t help with EEs which only need runtime dependencies.