I think removing 2.7 support is a fine idea, but I wouldn't remove the tests for 2.7 until the actual 2.7 functionality is removed and the major version is bumped.

I'm not personally relying on 2.7 support but I'd imagine many are, and removing tests for functionality people are still relying on sounds like a bad idea to me.

(Of course to play devil's advocate, using EOL software in the first place also seems like a bad idea to me.)

It's a tough call. If we assume AWS Lambda is the main target (do we have data for this?) the versions supported should follow what AWS Lambda currently supports:https://docs.aws.amazon.com/lambda/latest/dg/python-programming-model.html (2.7, 3.6+)

Python 2.7 isnot scheduled for deprecation on AWS.

I think, for clarity, it'd be best to release a new major version if we drop 2.7 support.

I'm assuming people still on 2.7 have old apps that aren't under active development and they just need to keep them running. In that scenario, it'd be clearest to set a constraint on the major version. After all, you're stuck on old versions of most libraries already...

Maybe we leave this open to hear if anyone hasactive 2.7 apps, but it's hard to see why the answer isn't "update to 3 already."

AWS is dropping support for 2.7in a few months. I think we've reached a consensus that a major version release should drop support.

Draft PR to drop 2.7.