[thristian]

In Python, it's both - the basic structured-programming advice to avoid global variables is always good, but it's a specific quirk of Python that makes global variables (including built-in functions) slower than local variables.

Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes. Because Python is a dynamic language, every time you refer to a variable, the Python interpreter looks for it in the local scope first, and then each enclosing scope until it hits the containing module. A local variable will always be found in the first iteration of that loop, a global variable will take at least two iterations.

[encukou]

Another CPython quirk is that global (module-level) variables are looked up in a dict, but a function's local variables normally get a reserved chunk of memory that's directly indexable.

[adgar]

> Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes.

You are describing lexical scope, not dynamic scope.