A software API is like a novel's plot.
It is the implementation of the API that makes the novel, but the plot is intrinsic. However how many unique plots are there in stories? I'd argue a plot that included some of the implementation (character names, places, detailed plot breakdowns) would be copyrightable as it would uniquely identify the novel.
APIs have names, class names, interface names, etc. For example it might include an enum of Http Statuses by name. But they don't include the detail of the method of implementing getHttpStatus() on a HttpResponse object (or how that response object is created). Package names might be analogous to 'places' if you were high on opiates.
The interoperability argument does seem like a very strong path for Google to take. You can't change API names and places because that would break interoperability. Much like you can't rename Mordor and where it is between the LOTR books and the LOTR films - two different implementations of the same (detailed) plot. Argh.
Sorry, rambling insanity is now prevailing and the above analogy is full of (plot) holes. And it doesn't come out on one side or the other. I think the real issue is that the concept of someone copyrighting AND enforcing that copyright to the API to a programming language's core libraries in this day and age is pretty ludicrous.
Biting the hand that feeds IT
