Comment Re:HORNET vs Tor (Score 1) 61
As someone who ran a Tor exit node for years (and has the bright green t-shirt to prove it), I don't think the protocol has anything to do with the speeds of Tor. Not inherently, anyway. Tor is slow because the vast majority of its nodes run on asymmetric consumer links.
The Tor protocol *is* the problem.
Tor says "I will use one encrypted channel, and send your data out that one channel, regardless of its speed".
Not "I will borrow from IP. I will open many encrypted channels, and send packets of data out each channel. The end node will re-assemble data from those channels. We will use the IP protocol to retransmit lost packets over each channel.".
IP, even though it is physically 1 channel, pretends to be 8 channels that each can transmit one packet. Something very similar to this could be used, and then if you have a slow link, that one slow channel only sends a small number of packets.
Tor *relies* on high volume to mask individual traffic. The only way to get high traffic is to make it "fast enough". With the default behavior, that any node can be an intermediary, any channel may wind up with a slow node in the middle, and then suddenly speed is lost. When speed is lost, people stop using Tor.
That is problem number one with Tor.
Problem 2 with Tor is less obvious. It claims to know what kind of anonymity you want. If all I want is to hide who I am talking to / what I am saying from my direct eavesdropper (my ISP, the local wireless hotspot), then all I need is to talk to one intermediary. Tor forces 2. There's good reasons to require 2 if I want full privacy, but maybe I don't.
If I want to split my communication into many streams, and have it re-assembled, then there must be lots of one-intermediary-hop streams.
That's good enough to stop my ISP from spying on me.
The normal encryption is good enough to stop the wireless hotspot people from watching me.
For 80%+ of usage (probably closer to 95%), that's sufficient.
Requiring more means slower.
Design a protocol so that none of the node can tell if you are using a 2-hop or a 3-hop system.
The entry nodes cannot tell if they are talking to the originator or to another node. (I think Tor fails this -- the first hop uses a different protocol than the rest, as I understand it).
The intermediates cannot tell if they are talking to first and last hops, or another intermediate.
Many years ago, I tried explaining this on the Tor discussion lists, and did a bad job of explaining it.
The last time I checked, the Tor people wanted a security analysis of all proposals.
And the one thing I know for certain is that I am incapable/incompetent at that.