[ghshephard]

Mercurial is a canary - if they see so much pain involved in moving from Py2 to Py3, and little reward, then that's representative of the overall ecosystem.

Having people on older versions of the interpreter is a problem, because it divides the community in half when it comes to knowledge, skillset, capability, etc.. It also confuses outsiders who are looking at the situation, and don't understand which version they should move forward with, because it's not clear which one will by supported by the people they want to hire.

There is a real cost to having the community split - which is why so much effort is put into trying to pull it back together.

[cookiecaper]

Mercurial is a large, complicated project, so it makes sense that they'd delay a backend upgrade like Py2->Py3 as long as they can get away with it. It's not representative of the overall ecosystem. That's like arguing that as long as "canaries" like Facebook don't see any reason to move to Ruby or Python for web dev, the rest of us might as well stay on PHP too. There's a large investment in the existing infrastructure that often can't be ignored no matter how compelling the features of the new platform are.

The differences between Py2 and Py3 are not substantial enough that it "divides the community in half" in any tangible fashion. Yes, it's possible new software may not take Py2 compatibility into account and you'll need to switch to Py3 to get new stuff, but that's been the plan all along, right?

183 of the 200 most-used libraries are Py3 compatible now, including major systems like SciPy and Django. It's not 2009 anymore, and we should stop talking like it is. Py3 is the way forward and the Python Foundation needs to quickly and firmly deconstruct any sign of flapping on that front if they hope to convert the remaining Py2 holdouts.

I believe it would be very damaging for the community if Python backpedaled and said "OK, we know you just spent 7 years investing in the Py3 platform and converting your apps and libs to work on it, but visible projects like Mercurial still don't like it so we decided it wasn't worth it anymore." People don't take time out of their day to make their voices heard unless they're already discontented about something. We shouldn't assume that everyone hates Py3 just because Py3 users are going about their business quietly.

[ghshephard]

Don't get me wrong - I'm totally on board with Py3, I'm just saying for newcomers to the language, they hear, "Python 2, if you want to be compatible" and "Python 3, it's the future" - and sometimes, they just throw up their hands and say - "I have no idea what version I should use."

I think things will get better when operating systems start to default to Python3.

[saurik]

More of that effort needs to go into compromises back to Python 2.x as opposed to figuring out how to beat people into going up to Python 3.x; it should not have taken until Python 3.3, for example, for the u'' syntax to return. They need to sprinkle a few well-placed "this is a way to get compatibility with Python 2 and 3 at the same time, without crazy tools, at least as an interim state". When I shifted from Ruby 1.8 to 1.9, there was nowhere near as much of a flag day of pain, even though they were attempting to solve the same overall problem (Unicode), and even though Ruby 1.8 was actually abysmal at Unicode (unlike Python 2, which contrary to many of the people who like to try to carrot people to Python 3, does not have any issues with Unicode; in fact, if anything, Python 2.x is better than Python 3.0 was, as Python 3 decided to "fix" a bunch of things, like filename encodings, which actually should not be assumed to have an encoding :/).

[bemused]

my experience with python3 unicode has been just the other way round - there are some minor troubles with python3 in regard to unicode too, but when doing anything user facing (starting with i18n/l10n) I'd take python3 without thinking twice. its 2015 and bugs like the unicode usernames problem in GTA V lately are just embarrassing for developers

I'm not saying you can't do this with python2 but you have to pay attention to unicode all the time. this gets worse A LOT when trying to serve python 2 and 3 from the same codebase, something I wouldn't recommend at all if you want to do eg. application development. of course this is a bit frustrating if you learned to properly do unicode in python2 the hard way, but for new developers not having to worry about unicode issues really is a blessing