Stéphane, I don't know much about urllib to review this. The reason I said it's better document this is that there was an extended discussion on issue25439 along with a PR about type checking breaking some behavior and ambiguity in what is allowed currently. This PR seems to encode str to bytes implicitly and it needs to be documented as such that this str is encoded to bytes under the hood. I am not sure of backporting it as a fix to stable releases. I think it's better to raise an error instead of doing something implicitly.  I am adding Martin and Senthil to the issue who might have a better context.