Haha, point taken. Sorry about that. As a student, I think that one or the other would make a good project, as both would be too much work.
-Tyler On Mar 19, 4:21 pm, Michael George <mdgeo...@cs.cornell.edu> wrote: > Tyler Laing wrote: > > Hi, > > > That sounds remarkably like Bittorrent. If you check the internet > > archive or other sites, you should be able to find the Python code for > > Bittorrent before it was closed. > > I agree - it's very like bittorrent, but with different properties. > With BitTorrent every client wants a complete replica, whereas with user > rated content, we'd be happy if lower-rated content only resided on the > hosts that specifically cared about it. We can afford to be much > fuzzier here.> About server discovery, my concern is that most systems can be > easily > > polluted with malicious entries. Look at all the issues with DNS > > (which is pretty much exactly what you propose), and what needs to be > > done to prevent ARP poisoning attacks. Any such system would have to > > be designed to distribute the authentication in a secure way. And if > > attackers have control over one or more nodes, it can be difficult to > > ensure authority and correctness without a centralized server, hence > > the Root DNS nodes. > > The difference from DNS is that DNS answers the question of "find me the > host named x" whereas we would want "find me some nearby servers to pick > from". Malicious entries would be much less of a concern - the worst > damage you could do would be to either prevent a game server from > showing up or to always make your own server look good. Neither attack > is nearly as interesting as, for example, making yourself look like a bank.> > It sounds like some interesting ideas, but the problems are ones that > > still stump PhD's... > > As a PhD in training, maybe that's why I think these things would be cool :)> > -Tyler > > --Mike