In terms of scope I can agree that there is a bit of a difference, in that we're talking about whether you can copyright an entire library design versus a specific limited set of interconnection points. But even that would be problematic and contrary to existing practice; for example, Wine's reimplementation of the Win32 API would be a copyright violation if the Win32 API (and not just the code implementing it) is copyrightable. It contradicts the assumptions behind clean-room reverse engineering as well, which are that how-something-works documentation doesn't taint copyright of a new implementation that follows it as a spec.