I notice that the article says not to use urllib, but doesn't say what to use instead if you need an HTTP client. The correct answer, of course, is requests...w

I notice that the article says not to use urllib, but doesn't say what to use instead if you need an HTTP client. The correct answer, of course, is requests...which isn't in the standard library. So there's a tradeoff between a good API that you have to figure out how to install, and a lousy API that's universally available. (This tradeoff isn't as bad as it once was—the Python package management experience has improved a whole lot over the past decade—but getting packages installed is still a barrier for new users, especially on Windows.) It's not just a matter of saying "use the stdlib"; the stdlib is imperfect and its flaws are not limited to the concerns of power users. Also, it's worth separating the issue "this library has an overcomplicated API surface" from the issue "this library uses Python magic in ways that make its semantics different from the default expectations for a Python API". For instance, there's nothing remotely magical about urllib; it's just unwieldy to use. Most of the other listed packages, like pandas, have the opposite problem.