On Sun, 10 Feb 2008 01:47:01 -0600, Scott Bennett wrote: > But, Roger, will the 0.2.0.19-alpha release at least confirm during > the reachability tests that it is talking to itself and not to some other > server?
I am not sure that is what is happening. For example, it may be that the reachability checks are being performed on the current (and correct) IP before the update to the suggested new IP is put into effect. Anyway, I suggested making an explicit verification of the reachability of the current IP before updating to a suggest new IP as a means to help counter incorrectly suggested new IPs. E.g., if the old IP still works, ignore the suggested new IP; if the old IP does not work, update to the suggested IP. (This sort of thing doesn't solve the root issue of how best to establish trust in the IP address observations of directory servers, but maybe it helps a bit. Once you get an accurate IP suggestion, you are protected against incorrect suggestions until such time as you become unreachable again, when the game starts all over and you were back down anyway.) There are probably many ways to perform that check - one way I threw out there was to explicitly call the check reachability routines before updating; another (and perhaps better) way might be to only process suggested new IPs when Tor notices that it is no longer reachable. Unfortunately, neither did I supply a working patch to implement explicit verification of the old IP as an additional workaround for the root issue nor did I even confirm the cause of the successful reachability tests. Scott, you have been quite vocal in this area, so perhaps you might want to put in some legwork here. I am sure the Tor team (and the rest of us) would appreciate such efforts. -Andrew
