Andy @ surfMind.com comments on the DHTML menus at the new USC site, and mouse trajectories, which we noticed, but unfortunately, by that time, it was rather late in the development process, and due to various delays we were unfortunately already far behind schedule on implementation.
That being said, as Andy Mentions, there are a number of post-fixes that can greatly increase usability. A delay at the very least (or for a more complete solution a sticky-bit triggered delay and perhaps even a vector mouse-tracking algorithm), and a wider mouseover area (right now one must traverse the mouse horizontally within a 25-pixel band – it’s positutely rsi-inducing) would help greatly.
That’s actually currently on my post-launch todo list. There are a few things above it right now, but the menu needs a rejiggering to be be more S.508 friendly anyway (currently, the inactive submenus will all show up at the end of a page on Lynx – if they’re written out w/ JS, that should fix it).
In the long run, my personal opinion (months ago, and now) is that if a dynamic menu is used, it should probably be a vertical one. The horizontal tracking problems disappear completely, and it’s a more familar paradigm anyway. Of course, the DHTML meus, while a usability concern, is actually the least of the website’s usability errors. When I get a chance to take a breather at work, I’ll be writing up an analysis, and will probably post some thoughts on the ol’ blog as well.
I’ve also run some preliminary traffic analysis on the first three days of logs (3.5GB). While one might not think so immediately, there’s actually quite a bit of usability data that can be mined from there.