## Subject: Re: Frothing remailers - an immodest proposal In-reply-to: <199502010520.VAA04884@largo.remailer.net> (eric@remailer.net)
Date: Tue, 31 Jan 1995 21:20:28 -0800 From: eric@remailer.net (Eric Hughes)
In article <9501312152.AA10208@toad.com>, <kevin@elvis.wicat.com> wrote:
It seems to me that the current remailer web suffers a fundamental flaw. It is simply too static.
Now, dynamic rerouting is good for better delivery, but is bad for the trust in silence. [...] The end users must be involved, either directly or through some (legal) agent, in the manipulation of these relationships.
Any solution which tries to do this independent of the end user is broken, by definition.
Eric
Well, pgp support multiple recipients of messages. Supose that the remailers would choose at random only one of the addresses the user (or their client program) requested in a header line like: Request-ND-Remailing-To: RM1@a.b.c, RM2@c.d.e, RM3@e.f.g and try to deliver. If the mail fails right away, then it tries another address. Etc. The very paranoid user would avoid this feature, and stick with the old fashioned system. The paranoid would list two remailers, and encrypt the folowing message to both of them, and probably add a few more levels to the chain, just to be sure. The compleatly trusting would only have two levels of remailing, but which listed every remailer as a posible recipient of the message they send to the first in the chain. In this way we get better reliability, but still have user control over selecting the remailers. In fact, the user can select arbitrary message reliability, and remailer trust parameters, and should be able to come up with a set of nd-hops to meet the parameters. Hey Wei, Hal: What is the cost of this in terms of likelyhood that the whole path of remailers actually selected is compromised? Is this about right? If 50% of the remailers are run by the enemy, then with only one remailer listed in each hop, the odds of the path being compromised is (.5)^h (where h is number of hops). The odds of successfull delivery are .90^h (asuming every remailer is 90% up). If at each step there were two remailers, and the evil remailers always selected other co-operating evil remailers, then the odds of the path being compromized is larger at ((1-.5^2)==.75)^(h). But the odds of sucessfull delivery are much better, (1-((1-.90)^2)==.99)^(h). To keep the same chance of the path being compromised, the user would need to have 'x' times more hops where x is such that (.75)^x == .5, or about 2.4 times as many. Hmmm... Noyb