[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: mixmaster delivery?

On Mon, 2004-02-02 at 20:30, Douglas F. Calvert wrote:
> On Mon, 2004-02-02 at 20:15, Nick Mathewson wrote:
> > On Mon, 2004-02-02 at 20:04, Douglas F. Calvert wrote:
> > > Hello,
> > >  How come no one has mixmaster delivery turned on? I turned it on for a
> > > second and then became nervous when no one else had it turned on.
> > 
> > Totoro1,2 and Moria1,2 have smtp-via-mixmaster delivery enabled.  The
> > reason that most people don't have it is that it isn't especially
> > useful: sending a Type-III message through a direct Type-III exit node
> > is likely to be more reliable, since it doesn't get the Type-II network
> > involved at all.  It doesn't _really_ create another exit node: it only
> > fakes it.
> > 
> How come it does not show up when I do a list-servers? 

"Direct smtp" and "smtp-via mixmaster" are both implementations of the
same abstract service, "smtp".  When you run "mixminion list-servers",
it shows them both as SMTP.

>                                                 When I had it
> turned on for a little while it said something that indicated it had mix
> capability I don't remember what it was exactly. Is this because it was
> limited because I am only set as a relay? 

I don't _think_ so.  If you turned on smtp-via-mixmaster, and your
server published, and you waited for the directory to include the
updated descriptor, it should list your as providing SMTP. 

> If I can not run an exit node do you think it is worth it for me to turn
> on mix and frag? My reasoning is that if I can not run a real exit node
> I might as well allow mixmaster delivery. 

Again, I don't think it's a good idea, especially in combination. (See
my last message for reasons why: basically, smtp-via-mixmaster
complicates the network and makes it harder to debug, and when combined
with fragment reassembly, it makes it unlikely for large messages to get
through successfully.  Remember: the Mixmaster payload size is only 10K,
and if any packets are lost, they cannot be reconstructed.)

> On another note I have a question about the drop functionality. Is their
> a reason why drop always picks chains of the same length? 

Picking a path length is (mostly) independent of picking your exit
address (drop, smtp:foo@bar.com, mbox:fred@nodename, etc.).  So your
real question is, "Why are my paths always the same length?"

I'll assume you're running version 0.0.6 or later.  If you don't specify
a path on the command line, Mixminion looks in your .mixminionrc to find
an option called "ForwardPath" (or 'ReplyPath' or 'SURBPath', depending
on what you're making).

So the reason you aren't getting random-length paths is that you're not
asking for them. :) 

You can specify a random-length path by using the ~N (approximately-N)
notation.  For example, to ask for a path of about 5 nodes, say:
        mixminion send -t drop -P '~5'.

To ask for a path starting with ccc, followed by a random node, followed
by moria2, followed by about 3 random nodes, followed by totoro1, say:
        mixminion send -t drop -P 'ccc,?,moria2,~3,totoro1'

>                                                         It seems that
> a random selection of chain length would add a touch more traffic
> randomness.
> And finally I am beginning to write a paper that has something to do
> with the economics/costs of anonymous mail/traffic. Exactly what I am
> not sure yet. What is the best forum to ask for input? Here? Mixmaster
> devel? Or some other list? I need to write/research a topic for a
> telecomm class and do a economic/cost/game theoretic modeling of some
> aspect of communication. I was interested in the recent traffic analysis
> papers and think I am going to do something along those lines. Are there
> any open questions right now that I might be able to look into and aid
> the community? 

Not sure what here; others may know more.  You should at least read
http://freehaven.net/anonbib/#econymics ("On the Economics of
Anonymity").  I'm not entirely sure how you would link traffic analysis
to economic methods, but I'd be keen to see it done. :)


Attachment: signature.asc
Description: This is a digitally signed message part