When I started this program, I had hoped to make something fully interactive so that I could post it on the web and surfers could make their own mosaics. This obviously adds a few more complications to the project. Perl and cgi script would be necessary for either loading an already established library or surfing google's image search for a substantial library associated with the user's chosen keyword. Then one would have to consider the shape of these library images. If the program continued to subdivide into square tiles, should the images be cropped or scanned for the "best square region" for that particular target? If the latter, another significant block of memory would have to be allocated (I think this is the method used by Robert Silvers' algorithm, but you might want to read up on that yourself).