I'm Ok to drop 3.3 for 6.3; we just have to be careful, I don't think we've ever dropped a Python version a minor release though.

I think that we can provide guidance to pre-6.3 for anyone on Python 3.3.

I'm also fine with leaving it until we split 6.x from master, and targetting it for 7.0. I don't think there's any hurry to do this - I don't know of anything we're eagerly waiting to do that's only possible from 3.4.

I'm happy to wait a bit and see how much we are planning to keep the 6.x branch. We can fork master toward 7.0; though that might mean having a 5.x 6.x and master branch.

I wouldn't start 7.0 for this; it's a simple set of changes, it's not worth having to backport everything for. Let's just let it sit until we have another breaking change that gives us a reason to aim for a 7.0 release.

See #10911 for PR related to changing testing for 3.3. with docs

I think it's fine to stop testing with 3.3 in 6.x releases, and then make the changes that we know will break it for 7.0.

I think it's fine to stop testing with 3.3 in 6.x releases, and then make the changes that we know will break it for 7.0.

Ok, fair enough. This need rebase now. I think a nov/dec release would be welcome. Do we wait for Dec25 ?

I've rebased the changes and updated the version numbers in various docs and messages to say that 7.0 (not 6.3) drops support for Python 3.3.

@takluyver do you want to merge this, so that we start master being a 7.0-dev ?

That's my plan. I've already updated the version number in master.