Hmm, I will have to go back and review what I said to see if I misspoke. The basic issue is that open source licenses are copyright licenses, and so their enforceability is tied to the scope of copyright. The scope of copyright as applied to software is not at all clear, despite what anyone will tell you. And, probably not surprisingly, a lot of the ambiguity around the scope of copyright comes from the use of interfaces, APIs, and methods of linking. Python, like any dynamic language, is fuzzy around the boundaries here.

There are so many square pegs and round holes that anytime someone says something is "clearly" a derivative work, that person is almost certainly incorrect. The concepts of copyright were never meant to apply directly to software, and the law doesn't work the way people frequently intuitively think it does.

The problem is that applying copyright (and patent, but thats another story) restrictions to software has broad implications for what you can do with the software. This isn't just a practical issue; it is a freedom issue in the Free Software sense. I see so many bogus claims each day about IP that I have become very wary of its scope.

For example, a couple years ago Blizzard was able to get an injunction against another company because the other company sold a program that scripted Blizzard's software. No modification, just GUI automation. I was horrified, and disagreed, but they were able to put together a winning argument.

I unabashedly advocate boundaries on copyright. I think authors should say where they consider their code to end, and other code to begin. I think those boundaries should be reasonable within the scope of the law. I think this principle applies even (or especially) when it is free or open source software.