And another point: Nobody wants to add latency. (There may be a few applications, but they are rare.) What people want is security and one way to get it (sometimes) is by adding latency to their messages.

Latency is essential to security, though high volume reduces the latency that's needed to get a given level of security. It's not enough - if your latency is a constant, it doesn't buy you anything, and if there's not enough traffic, the fact that the remailer waited 3 days before sending out the one message it received doesn't help much either. But if it's low, and relatively predictable, you lose rapidly to traffic analysis instead of potentially losing slowly. Lucky's point that traffic analysts can learn a lot because messages decrease in size in type-1 remailer networks is important; Mixmaster's constant-size message blocks are probably the best way to go (though you then need more random interblock latency), but the ability to have each remailer add padding is important if you don't use fixed-size blocks, and I'm not aware of anybody implementing that. (The obvious implementation would be for remailers that randomly pick another remailer to chain through to reach their destination, since they could put padding inside the encrypted package.) Thanks! Bill Bill Stewart, stewarts@ix.netcom.com Regular Key PGP Fingerprint D454 E202 CBC8 40BF 3C85 B884 0ABE 4639