@ncoghlan: The comparison I'm worried about is specifically this one: IIUC, right now it's safe to run 'pip --version' in an arbitrary directory, but it's not safe to run 'python -m pip --version' in an arbitrary directory. Am I wrong? (I actually couldn't convince either version to execute arbitrary code in 2 minutes of trying, but that's my understanding of the discussion so far.)

If that's correct, then I don't think this is like... the hugest security bug ever, but... I also think that it's irresponsible for e.g. packaging.python.org to be recommending people run 'python -m pip' the way it does now, and we need to find some way to change things so our beginner docs aren't triggering arbitrary code execution in a rare and subtle case.

We could add a bunch of warnings to packaging.python.org, explaining about how the code execution can be triggered, but that seems unsatisfactory given how those docs are targeted at beginners, plus there are so many places around the internet that recommend 'python -m pip' we'd never find them all.

We could update all those docs to instead recommend 'python -Im pip', but that has the same problems: no-one will understand, and people won't do it.

We could stop recommending 'python -m pip' entirely, but that runs into all the problems that have motivated this in the first place.

So I think we should find a way to make it so 'python -m pip' *never* executes code from the current directory (modulo the usual caveats, like the user explicitly setting PYTHONPATH to an insecure value etc.).

If 'python -m mypkg.myscript' is important, maybe we can make it 'PYTHONPATH=. python -m mypkg.myscript', or 'python -M mypkg.myscript', or making 'python mypkg/myscript.py' DTRT, or... something?