I have retired this approach in favour of a much neater solution that can be found here.
This tricky topic had me puzzled for some time and in earlier posts I went the way of using CSS classes to identify DOM elements; but was always a touch dissatisfied with the solution. Not only is it less efficient – valid XHTML pages should only have one element with any Id, although CSS classes can be shared by many elements – but it also feels all hacky ‘n stuff. I’m a bit older now, a bit more experienced and I RTFM. Without further ado, here is why it’s tricky, and how to fix it.