Think about Google's autofill functionality. Currently, the way we've implemented it is that after every letter a request is sent to the server to find all valid possibilities of what the word could be. So, for example, if we're trying to search "hello", after we type the letter "h", then a request is sent. Then we type "e" and another request is sent. The number of valid possibilities becomes smaller and smaller as we type more and more of the word out. The current model of sending a request to a server after each keystroke is pretty expensive so what sort of algorithm would you use to limit the number of server requests sent while still maintaining the autofill functionality (meaning that it appears rapid to the user and doesn't stall).
