I have yet to see anyone use a URL loader in serious code beyond people just using it as an example. What I do see is lots of custom zipfile importers and thus for the majority if people this will continue to make sense. And even with a URL loader, get_filename would return a URL who probably has something like http://example.com/pkg/__init__.py which would still work as expected. If someone chooses to break from expectations they can just overload is_package().