The @media conference is over and taught us mainly that there was a misconception of what the audience would be. The first day was a bit of preaching to the converted, whereas the second one seemed to realise that and start an expert discussion instead.
Excellent point, IMHO. We use object detection to check if the browser supports certain scripts, and punish Opera 6 by cheekily testing for createTextNode and getElementById.
This is a much cleaner way than browser sniffing and easy to check. However, when it comes to screen readers, a lot of our testing really is assuming what they should do or not. Especially new DOM developers assume that a screen reader works like a text browser, which is not true, as modern screen readers do detect and speak out DOM changes.
More over a pint later.