I don't understand the whole collision thing. I mean a sha1 is 160bits so if you are hashing information longer then that collision is a fact, being able to forge a piece of information with constraints is the challenge and even that with enough power you end up being able to try all the combinations. What I understand from that collision reported is that they use PDF format which can have as many data inserted to it as comment /junk as you want so all you need is enough processing power to find the right padding/junk to insert to get the collision. Am I missing something here ?