The insight I got from that exercise was that, to him, APIs represent a product (e.g. I put a lot of effort into making this API, it's nice and clean, why should someone else be allowed to copy it?) and the implementation is typically straightforward grunt work, with intrinsically less value (e.g. we all know how to square a number).
As a developer, it was easy for me to sympathize with that thinking, because who hasn't been amazed by how elegant some of the APIs are and how difficult it is to create a simple, yet powerful API?
However, I believe that in the general case, it is the implementation that's the difficult part, which is why I'm against copyrighting APIs. Anyone can come up with an API for a Map, not everyone can create an efficient implementation of HashMap.
Doesn't this sound vaguely familiar? I think it's the same argument as "ideas vs execution", it's just that, as a community, we've discussed the pitfalls of ideas, NDAs and execution an order of magnitude more than copyrighting APIs vs implementation, and so there's less of a divide.
If you side with Oracle and you agree that execution > idea, then you should consider that implementation > API, and, similarly to how you can't copyright an idea, you shouldn't be able to copyright APIs.
In the case of Oracle vs Google, both the API and the implementation are trivial, which is a special case of the more general "API vs implementation", and it is unfortunate that a precedent will be set for the general case, based on a special case. That sounds like a recipe for future trouble.