Sort:  

Love a smell of holywar in the morning.

XD it's on

I guess that for some corner cases like scientific research you get to see stability issues but for 99% of the use cases it won't matter. And I think it really hurt the ecosystem that whenever you look for some resources you have to make sure that it's indeed for your python version. Like that moment when you find the perfect library but it's all python 2.7 so you have to rewrite it to make it python 3 compliant.

Or the fact that you have to support python 2.7 and 3 if you want to create a library that is used by everyone.

The point is that in our field, except numpy and scipy, the rest is mostly homemade, so that there is no strong necessity for python 3 at the moment. And the hardcore calculations are done in Fortran... We need speed :)