There are several problems with this approach. First one is that it doesn’t recognize notebook users connecting over a cellphone. It can’t really, since browser environment is literally the same, unless it would try to measure latency and bandwidth of page elements and guess from those results, which is neither easy nor reliable. 3G networks can be rather fast and have a better latency than a wired connection from somewhere like Tanzania.
There are two ways you could go about this. My first go was finding style nodes in DOM and looking at their disabled property, which is commonly used in style sheet switchers for turning sheets on and off. It doesn’t work, since ‘wrong’ media types in Firefox are ignored, not disabled. Their disabled value is still set to false.
A proper way of doing it would be using DOM Style Sheets methods. Basic idea is to compare actual values as set in the page with values read from style sheets and resolving which ones were used. While not exactly trivial, by little forethought it can be made to work fine in Opera and Firefox. It can also work in Explorer using its own methods, but it’s a pain to make it work in Safari, which in this case is not only an incompetent liar, it’s also lying in some weird accent. If you’d like to learn more about it, then there’s a great page describing current state but the gist of it is that you can’t rely on methods being there and even when they are, in some form or another, they might not work reliably (Safari) or might produce weird results (yup, Safari again).
The end result certainly isn’t perfect or pretty, but it keeps amount of bookkeeping to minimum and limited to CSS files. Even using a “safe” property like font size can lead to problems (e.g. 0.9em is sometimes interpreted as 0.90em), but nothing difficult to overcome.
It might not be of production quality, but it will work as good starting point for further exploration.