This, as the OP says, is a pretty basic approach -- but it's a start. It's probably best used for longer passages, where the amount of text in the viewport is pretty consistent.
What I'm really looking forward to is when eye-tracking apps get good enough and cheap enough that a browser will automatically scroll based on your eye movements. Add in a couple of other commands -- maybe wink to click, cross your eyes to switch tabs -- and you could do some decent browsing.
Then you'd just need a lip-reading app so you can mouth words and have them converted to text!