The interesting bits are in twitter_timeline.coffee:
In particular, the
didScroll method should be interesting. As we
scroll, we're checking to see if we should permalink the page. When we call
permalink, we use the History API
method to change the URL.
On the first page load the
TwitterTimeline constructor is called
and I parse out any
max_id in the URL to fetch tweets later than that point in time.
This makes the page "sticky" when you exit the page and come back to it via the back button.