I came across this article on LuLa about optimum exposure. Basically what it says is to expose so that the brightest point of the scene is 99.5% white in the raw file and then correct in post production by bringing down exposure. Thus, if you photograph a black cat on a black carpet you would severely overexpose to have a raw file where everything is white (just short of blowing highlights) and then bring it back to the correct exposure in post.
The idea behind is that you capture more information. (Basically it seems to be taking ETTR to the next level.)
Intuitively it seems right and I have been able to myself to bring overexposed but not blown out shots to the right exposure. But I am curious what folks who know more than me about digital capture think about this.