@lmedinar @2ck I think Python is the example of how not to manage a breaking release on the “too cautious” end of the spectrum (Obviously the other extreme also exists). Telling everyone “You ought to upgrade your scripts” for over a decade caused a lot of confusion. I vastly prefer something like a two year window or simply renaming, and not branding it as an upgrade.

@aexiruch I think it's difficult to do in any case. I've heard the story that there weren't enough good things in earlier Python 3 releases and there were too many breaking changes at once. For me though, at least on the surface level, they were good changes: the bytes/str split alone was welcome and necessary -- I didn't even realize how forked my mental model was with strings until I had to really *think* about string encoding.
@lmedinar

@2ck @lmedinar I agree Python3 is the better language, but the way the process was managed strikes me as ridiculously timid. A couple years ago I scoured the mailing lists for the rationale and my impression was a mixture of not actually giving a fuck (“that’s up to the distros”) and not wanting to create a bad experience, which, to me, is exactly what that approach did.

@aexiruch hmm. yeah, I can see that. I became "serious" as a Python programmer when Py 3 had already been around for a few years, but Python 2 being the default everywhere I went was definitely a head-scratcher. Communication about the changes and getting key training resources, tools, and packages on board is essential for avoiding the kinds of fractures we've seen (and even still see, in some quarters). From the outside, it seems that wasn't well appreciated given how packages like numpy didn't support Py 3 for like 3 years and projects like https://github.com/brettcannon/caniusepython3 had to be made @lmedinar