This hack can only tell you what sites a user has visited out of a list of sites you specify. A general ad platform based on this hack would presumably have a list of sites to query against, acting as a litmus test for determining a user's interests and socioeconomic status. I would argue that such a list would be difficult to compose (javascript/JSON size vs. generality vs. granularity) and no better than the cookie / flash cookie / geo IP lookup technologies currently being employed by doubleclick and others. Such a list would necessarily have to target sites in the top 500 whereas networks like doubleclick can mine niche sites that may serve as better indicators of a user's true behavioral profile.
It's not necessarily quite that bad. The list can be made modular.
One option would be to start of with a list of the top 500 sites and then determine from that information what demographic the user fits it. Then you could load a list of sites pertaining to that demographic and further narrow down the user's behavior.
Also, if you're talking about ads on a certain site, the publisher can make a list of the sites in their niche and then you could load that list and move on from there.
All in all, this has some very interesting potential. The question is, why hasn't Google done it already?
They have. I'm pretty sure Doubleclick does this. Without the CSS hack though(using the cookie method).
The CSS hack is only worthwhile if you don't have the kind of data sources that the big guys have. If this technique actually made money, you would probably want to sign up for something like Lookery or RightMedia because they'd get you better results.
