I *think* how it works is when the spider is parsing the links it just records them, where they are coming from and where their going.