Shubham touched on it a bit, but I want to bring providers into the fold a
bit as well. I don't think there is enough focus on provider versions. Us
maintainers of Airflow have greatly benefited from being able to have
breaking provider changes, but I always get the impression the average user
just looks at the core version and ignores the provider versions almost
completely.

I'm not saying our position today is "wrong", but I think we (knowingly)
overlook how semver compliant "Airflow" is overall. Sure, core is backwards
compatible, but Airflow overall isn't, and the core version is the focal
point... Sucking more things out of core (like executors) just makes this
problem worse.

I don't have a fix here though, just want to express my thoughts that we
aren't as strict overall as the one version string implies. I'd be in favor
of allowing major version releases with breaking changes in core as well.

Now, one elephant in the room - the 5 year security patches thing Jarek
brought up. I freely admit I haven't tracked this at all, so please correct
me if I'm wrong. If that ends up panning out though, I think we will have
to reassess our strategy with providers too. As of today, we have 294 major
versions of providers, all released within the last few years. Being on the
hook for maintaining all of those for a 5 year period is a pretty good
reason to be much more picky on when we have major provider releases I'd
say.

Reply via email to