As time goes on, and more and more books are printed with their LC numbers, that code would end up coming out as even better.
Isn't it true that every book in the Library of Congress, i.e. pretty much every book in the English language, has an LC number, even if it isn't printed inside the book? If so, then LC numbers are close to ideal book identifiers.
However, an ISBN -> {all equivalent ISBNs} converter would also be really nice. Personally, I think there's nothing wrong with the traditional method of identifying a book by title and author though. Spelling differences can always be solved by appeal to authority: use the spelling of the Library of Congress.
Axel