Computer science is changing: the amount of data available for processing is growing exponentially, and so must the emphasis towards its handling. Like the 19th century change in physics from mechanics to statistical mechanics, the new algorithms sacrifice the precision of a unique answer for the fast search of statistical properties. The following draft of a book by Hopcroft and Kannan breaks the path of what most future algorithms manuals may look like:
Heavy on proofs, many topics have been selected for their mathematical elegance, not their pragmatism. On the final version of this much anticipated book, I would love to see more content on hash algorithms, parallel algorithms, graph spanners or a more extensive discussion on Support Vector Machines.