Outside of tiny niches the world is on Python 3 now, so I don't think older versions are particularly relevant when evaluating how complex Python is. For that r

Outside of tiny niches the world is on Python 3 now, so I don't think older versions are particularly relevant when evaluating how complex Python is. For that reason old vs new style classes don't cause issues. Nor do the existence of niche implementations like Stackless, unless we want to also claim such implementations as sources of complexity for C++ or (especially) Common Lisp, at which point it's a wash. Having trained many newly-graduated data scientists and engineers to a level of being reasonably productive in Python, my assessment is that it is simply not that difficult to get a handle on. But if I'm expected to take non-career-programmers to comparable C++ fluency, that would be a considerably longer process. Things they'd have to worry about in C++ that they don't have to worry about in Python: memory leaks, memory debugging, complex templates, libraries written in the lowest-level way possible, move semantics, ...