This is slightly less interesting when you consider the fact that there's no possible way that these 'shelves' are actually stored on servers. There aren't enough hard drives on the planet to store it all.
I'm not familiar with exactly how this algorithm works, but I guarantee it's similar to
Tupper's Self-Referential Formula, which is a function(edit: not technically a function but an inequality. yes I know it doesn't pass the vertical line test) that creates what's essentially an infinite list of 110x15 bitmap images arranged in a vertical column. If you want to find a specific bitmap image on that function, you treat the image as a binary number, multiply it by 17, and then graph the function with the range set from that number, to that number plus 15.
What the website is doing is taking your string and using math to find where, if this algorithm was run for eternity, your string would be in its infinite library. The fact of the matter is that this infinite library does not actually exist, and the hypothetical location and surrounding text of your string is created on-call when you enter in your message.
The take-away from this is that a 'library of babel' would be essentially impossible to generate. You'd have the world's best super-computers working on filling it for decades, and we would struggle to find even one coherent sentence in the piles and piles of incoherent gibberish. It takes a human being to provide it with an information-carrying string before it can ever find out where the meaningful sentences are, and that's why human brains are so much more fascinating than a computer program designed to spill out unique strings of nonsense.
