Isn't the Oracle case more about the definition of an API (what I would consider to be the header files, for example), versus this, which is more of an implementation of an API, though?
Agreed, I should have clarified that I meant hypothetically if the author started out with some example code that was supplied by the API owner under the GPL.