I see this as a bit of a pox on both their houses. Oracle has every right to assert ownership of the *implementation* of the methods that they write and choose to licence it however they wish. I agree with the author on that point; that it equally protects copyleft code. But they cannot copyright the API itself, that is, Google can use the interfaces, structures, data classes, method signatures necessary to deliver the functionality specified by the API but must write their own implementation of those or licence it appropriately.
Oracle are being a bit tricky by omission. If they were being honest about it they would asset Google's right to the interface "code" and reiterate that their complaint is about the implementation code only. But I suspect that would drop the lines of code violation quite handsomely if they don't count those. If I was in oracle's line of business with some other global 3 letter megacorp that could claim ownership on a rather significant API and would therefore be making that distinction at every opportunity.
Google are being tricky here by pretending that some of the items weren't copied. Notwithstanding that for trivial methods, the same code can quite easily be independently written and that with the advent of refactoring tools that just renaming variables to make it look different might only take a few seconds, it certainly looks suspicious to me.
Oracle are also asking for what seems to me to be a non proportionate share of revenue here. I work with two pieces of business software weighing in at give or take 3 million+ and 500 thousand+ LOC, and that is nowhere near the complexity of a modern operating system. It's got me thinking about the status of snippets provided on stack overflow too. I can well imagine a number of methods that are heavily inspired by answers in similar forums. 11 thousand, whilst significant, is likely to include many fair use elements and even o methods that Oracle may find that someone else invented.
Biting the hand that feeds IT
