Re: Request for comments: onion routing design paper

[Marc Rennhard <rennhard@tik.ee.ethz.ch>]

On Sun, 02 Nov 2003, Roger Dingledine wrote:

> We've finished the first (rough) draft of our design paper for Tor
> (The Second-Generation Onion Router). Please read it and comment.

In general, I like the paper, although it contains very much information
and I'm unsure if those that do not understand mix networks really well get
much out of the paper. Should be OK for the event you're submitting it
to, tough.

I've read the 17 3/4 page version I downloaded it at about 8.00 Monday
CET.

Sec 1: Using a proxy instead of SOCKS would still work with privoxy as
you can tell privoxy the following proxy to use. So you should not
use this argument to favour SOCKS instead of proxies.

Sec 1: We believe it to be unencumbered by patents. Try to get rid 
of the *believe* here.

Sec 2: The system you are referencing [39] is "Anonymity Network". AnonNet
is another project still under development without much information
being available.

Sec 2: Use TCP *segments* instead of TCP frames.

Sec 4: At the end: tell them also what will follow in Sections 4.4 and 4.5

Sec 4.2.2: I don't understand how the leaky pipe works. When a node gets
a cell (downstream), it removes a layer of encryption and gets the stream
ID (if the cell is indeed directed to this node). 
But how does the node recognise this is a real header and not another
layer of encryption? Do you just assume that another layer would seem
random enough to not mistake if for a header? Or do you check the checksum
at every node?

Sec 5.2: What do you mean with administrators seem to be used of port 80
abuse? Do you talk about generic attacks on web servers in Code Red style?
Or generic DoS attacks with spoofed source addresses?

Hope this helps,
Marc