Sebastian Rittau

It's been my plan for a long time now to get some of the definition from `_typeshed` into `typing_extensions`. (Maybe under `typing_extensions.ext`?). But time is finite. :(

typeshed tries to follow the implementation, not the documentation. In this case the documentation is at least misleading.

Maybe re-exporting the "proper" classes from `multiprocessing` as `_EventType`, `_QueueType` etc. is the best solution.

We should not change those if they are already correct, and we should work towards fixing `Queue` as well. Lies in typeshed always cause more problems than they solve, in...

This is indeed problematic and the natural solution would be an overloads returning `NotImplemented`. In the case of `float` (where `FloatLike` would have to replaced with something appropriate): ```python @overload...

This is not just a numpy problem: ```python class Floaty: def __init__(self, f: float) -> None: self._f = f def __eq__(self, other: object): if not isinstance(other, float): return NotImplemented return...

Ideally, we'd put third-party tests into `stubs/*/@tests`, not in `test_cases`.

In theory, there is the [@type_check_only](https://docs.python.org/3/library/typing.html?highlight=type_check_only#typing.type_check_only) decorator, but that has never gained traction.

I'd like to see something like this formalized. But as it concerns stubs and type checkers in general, we should probably move this discussion to typing-sig.

I am not a big fan of this, after all tkinter *is* part of the stdlib. A similar rationale could be used for quite a few other stdlib modules, e.g....