When a node makes a request it is passed through a chain of nodes until it finds the file it wants which is returned along the same path. However sometimes on the return path nodes don't change the header to point to themselves as the source. This is done because it enables more efficient routing in the feature and also helps other nodes meet.
I propose this consept be adapted so that works with both the up and down stream traffic. So when a node makes a request along the request path sometimes the requesting node should not change the header to point to itself as the requester. Of course if the request failed that failour would be sent back to the node who most recently made the request so it could try another entry in it's routing table. As before, when a node does not change the header to point back to itself the nodes on ether side of it become introduced to one another. (Assuming they weren't already) However then if the data is found it is not returned through the middle node. It is returned directly to the previous node. So the middle node does not know or care whether the request succeeded or timed out. This is not a serious threat to anonymity for the same reason that it is not a threat in the current system. One cannot prove that the previous node was the originator, and it doesn't happen often enough for any particular node to gain any statistically usefull information. It would not increase this threat as you could rig it so that it happens just as often as it does now. This means that the node doing this would have to decide in advance wether or not it wants to cache the data if it is found. (And if it does, route through it.) Of course just because it doesn't want to cache the data doesn't mean it should activate this behavior as routing trough it both improves it's routing table and provides anonymity for both the sender and receiver. However if the node is particularly busy and there is a reasonable expectation that if the request were successful it would end up waiting in a queue for a while, or possibly even dropped then there is little point in it fordwarding the traffic. For security the likelihood of a node doing this should be lower for both extremely high and low HTLs. And for backwards compatability it should not do this if ether the sending or the receiving node uses too old of a version to support this feature. However I believe this could significantly improve routing times as it would reduse the number of hops by circomventing the slowest nodes in the return route. This would also enable nodes to reduce their load without dropping requests. (especially usefull for nodes with lower bandwidth) My biggest concern is the effects this might have on the routing table. If nodes are frequently routing traffic elsewhere they are temporarily deprived of the feedback provided by the success of finding a particular file. _______________________________________________ Tech mailing list [EMAIL PROTECTED] http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/tech
