With the sheer brokenness of Python packaging I wish that the Python 2->3 transition had been between two completely different languages, such that the entire t

With the sheer brokenness of Python packaging I wish that the Python 2->3 transition had been between two completely different languages, such that the entire thing could have been thrown away. Yet because there was breakage, but not enough breakage to amount to moving towards a better rewrite, nobody's ever going to attempt a true Python 4. They blew their chance. Now we have a never ending train of hotshot developers that think they can "solve" Python packaging with another tool, without realizing that you can't design a tool to fix a broken 15-year-old system. It's not just a "there are now X competing standards" issue, it's also the fact that anyone who makes an earnest attempt is doomed. They will never be able to get it past the point of universally good enough. I lament the fact that I'm dragged back into the Python ecosystem every time I need to do something with a PyTorch-based tool someone spent years of effort building. I would never recommend the language to a newcomer. Ruby pollutes the global namespace, and Gemfiles execute arbitrary code, but at least with Ruby there's Gem, and then there's Bundler, and nobody complains. Will we forever be beholden to the mistaken decisions of the original developers in the early 90's, just because Python is simply too widespread? Because even people like me who detest the packaging experience are forced to go back there because nobody's going to rewrite the hottest ML project in Elixir or something? Even the simple mistakes like putting venv files in Scripts/ instead of bin/ forcing one to work around it in cross-platform CI make no sense in hindsight. Nobody remembers why it was made that way after so long. Now it just floats over you like a spectre.