On Tue, Jun 23, 2015 at 12:06:41AM -0400, grarpamp wrote: > Longer reply may come, but I think it is useful to again say > that it may be that you must disassociate the classical "tor > centric" idea of fill away from the idea of filling the "tor circuit". > Of course any circuit level fill from end to end will still be visible. > But if the network itself is doing its own node to node fill underneath, > independant from whatever circuits ride on top, there may be > your answer regard to the casual looker of GPA.
Even if, for the sake of argument, the padding node to node would work perfectly for the security you want there, it would be irrelevant to protecting correlation of client to destination. It is infeasible on the network links that matter: those between the client and the traffic-security network and those between the destination and the network. The system is intended to provide protection of communication with destinations that do not participate in its protocols. So even ignoring who would pay for this overhead, the vast majority of intended destinations won't participate. And the vast majority of clients could not afford to have always-on, full-length padding to connect to the network. Nor would they like the performance of the limit to that rate (e.g. no bursts above it). aloha, Paul -- tor-talk mailing list - [email protected] To unsubscribe or change other settings go to https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk
