Fred von Lohmann of Electronic Frontier Foundation has the best single-post set of observations to date on the GBS settlement proposal.  It notes that the settlement accomplishes a degree of access that litigation might have taken years to develop, but also observes critical areas of concern. 

Fair Use: How will this agreement impact future fair use cases involving book scanning? Others (like the Open Content Alliance) are scanning books, and they may not have Google's ability (or budget) to strike a deal with the world's publishers. UCLA Law's Prof. Neal Netanel has a few preliminary thoughts along this line at the Balkinization blog.

Innovation: It seems likely that the "nondisplay uses" of Google's scanned corpus of text will end up being far more important than anything else in the agreement. Imagine the kinds of things that data mining all the world's books might let Google's engineers build: automated translation, optical character recognition, voice recognition algorithms. And those are just the things we can think of today. Under the agreement, Google has unrestricted, royalty-free access to this corpus. The agreement gives libraries their own copy of the corpus, and allows them to make it available to "certified" researchers for "nonconsumptive" research, but will that be enough?

Competition: In the words of Prof. Michael Madison, "Has Google backed away from an interesting and socially constructive fair use fight in order to secure market power for itself?" Does this deal give Google an unfair head start against any second-comers to book scanning? The agreement creates an independent, nonprofit Book Rights Registry to dole out Google's royalties, and the parties clearly hope that the Registry will be able to license others on similar terms. But the Registry is empowered to cut a deal with Google on behalf of all rightsholders by virtue of the class action; in order to offer similar blanket licenses to others, it would have to independently acquire rights from each and every copyright owner individually. How long will that take? What about the Registry itself? It hopes to be a monopoly that fixes prices for the entire market of copyright owners -- precisely the kind of thing that landed ASCAP and BMI, which dole out blanket licenses for music, in antitrust trouble decades ago.

Access: This agreement promises unprecedented access to copyrighted books. But by settling for this amount of access, has Google made it effectively impossible to get more and better access? The agreement allows you to "purchase" digital access for out-of-print books, but does not include the right to download the book (unlike public domain books). So you can read the book, but only on Google's terms. Libraries get more access, but for an undisclosed price (OK, one computer for free) and still with a variety of restrictions. In the words of Harvard's head librarian, "As we understand it, the settlement contains too many potential limitations on access to and use of the books by members of the higher education community and by patrons of public libraries."

Public Domain: Early reports are that public domain materials are not regulated by the agreement. Moreover, Google has negotiated a "safe harbor" that protects it from liability for mistakes in evaluating the copyright status of a book. That should result in more willingness to forge ahead with the free PDF posting of books published between 1923-1963, where a public domain determination turns on checking government records to see whether the copyright had been renewed. But will Google impose restrictions on these "safe harbor" public domain works? Will the libraries that receive a digital copy of their own public domain holdings impose restrictions on those copies?

Privacy: The agreement apparently envisions a world where Google keeps all of the electronic books that you "purchase" on an "electronic shelf" for you. In other words, in order to read the books you've paid for, you have to log into Google. Google is also likely to keep track of which books you browse (at least if you're logged in). This is a huge change in the privacy we traditionally enjoy in libraries and bookstores, where nobody writes down "Fred von Lohmann entered the store at 19:42:08 and spent 2.2 minutes on page 28 of 0-486-66980-7, 3.1 minutes on page 29, and 2.8 minutes on page 30." If Google becomes the default place to search, browse, and buy books, it will be able to keep unprecedented track of what you read, how you read it, and collate that with all the other information it has about you. Does the agreement contain ironclad protections for user privacy?