Many authors and publishers have a fundamental concern over Google's scanning of library books without getting permission from the rights holder. Those rights holders can opt out of the agreement on the Google Book Settlement website. My current understanding is that the settlement only applies to books published in the U.S. if the copyright was registered with the Library of Congress. I'm guessing this is either because the registration records are being used in the chain of custody to establish ownership, or because the owners of books who don't register their copyrights can't take Google to court until they do register them, at which point they would automatically become part of the class subject to the agreement. A cynical person might claim that rights holders without registered copyrights don't matter because they don't qualify for statutory damages or award of legal costs. If you think legal costs are minor, the amended agreement sets aside up to $30 million for the attorneys of the plaintiffs in the current settlement. Note that the U.K., Australia, and treaty countries are also included one way or another, but I didn't pay much attention to those parts.
The main sticking point for me was the definition of "commercial availability" and Google's absolute control over pricing for POD versions of the books, which they will most likely sell through third parties such as Lightning Source or Amazon. The agreement refers to a 60 day buffer period for the Rights Registry to contact the publisher or rights owner, after Google determines the book is not commercially available. This initially struck me as high handed, as there is no absolute requirement for Google to contact the publisher, even if the publisher is easily located.
The determination of commercial availability is to be made from online sources and electronic databases, and in cases where those data sources don't agree, Google has the right to make their best guess. An obvious example might be the Books-In-Print database showing books as being available because the publisher never officially declared them out-of-print on the Bowkerlink website, but as a self publisher, I'm not happy about online catalogs and databases being accepted as the gospel. I've known many self publishers who published without ISBN numbers (including myself) and registered the copyrights, making those books subject to the amended settlement agreement. If those books are only sold direct by the publisher or through specialty outlets, they will fail the test of commercial availability. However, buried deep in the settlement language was a paragraph stating that Google may not challenge the commercial availability of a title that the rights owner states is commercially available.
Another worry was that the agreement failed to address the issue of new editions. I initially believed that previous editions of a book that is commercially available may fall into the fair-game bucket, and publishers of multi-edition books might find new POD or eBook versions of previous editions being sold in competition against the newest edition, and without the publisher having any say in the pricing. But it turns out that the agreement covers this very explicitly:
"In-Copyright Principal Work. If a Book’s Principal Work is not in the public domain under the Copyright Act in the United States and that Book is Commercially Available, then any other Book that has the same Principal Work (such as a previous edition) is also deemed to be Commercially Available, whether or not such other Book is at the time in question also Commercially Available."
One area in which Google's long battle against web spam showed through was in their insistence of controlling outgoing links from author landing pages. It wasn't clear to me if Google intends to create an author landing page for all authors whose works fall under the agreement and who sign up through the registry, or how those authors may be given the ability to add a link to the author landing page in the first place. But Google explicitly reserves the right to remove links from an author landing page to an author website if the website has nothing to do with the books or is otherwise an inappropriate site.
I remain fairly confused over the meaning of the various deadlines and dates that pepper the agreement. While I think it's always possible to opt out in the future, it looks like the rights holder loses the basic per/book scanned payment by waiting too long, and may have trouble stuffing some of the other cats back into the bag once they're on the streets. The agreement struck me as a reasonable compromise between authors and publishers need to protect copyright, and Google's dual drive to disseminate the world's knowledge as broadly as possible, and make a profit doing it. But it is a compromise, and I believe copyright protection has been weakened by Google's ability to unilaterally force the issue, if nothing else.