It's hard to tell just from the titles, but I have the vague impression that there's rather less computational geometry than I'm used to seeing at past SODAs.
One very important change (perhaps a consequence of going to electronic proceedings?): “your conference paper can be up to twenty (20) pages”!!
ETA: See also 3d pancakes, algorithmic games, kd-PhD, 3d pancakes again, biased coin, polylogblog, geomblog, motocicleta