I don't understand this too. If you set the mode to paranoid, this gives the NSA metadata that you're trying to hide something from them and then the NSA will archive all of the paranoid emails as they currently archive OTR messages, in the hopes that one day they can decrypt them. There needs to be just one mode, private by default - unless their specification is designed in a way to make the NSA unable to detect which mode a user has enabled?
From my brief reading of the spec, the three modes are mostly just a matter of where the keys are generated and stored. I think the messages themselves are indifferent to the mode (and therefore intermediaries can't determine between the three).
I haven't gone through the whole paper yet, but other than showing up on the site (which hasn't been updated for a year until two days ago), and in an initial talk a year ago, I don't see the Silent Circle guys mentioned anywhere else.
What have they actually contributed to this project? I get the feeling they wanted to do something together, but then they probably got busy with their own Silent Circle + Blackphone + moving to Switzerland, and they might've had some slight disagreements on what DIME should look like as well, and then it was all up to Ladar Levison and someone he hired to work it all out.
It's good that someone is looking into making the most important electronic communication mechanism secure. Making email secure without figuring out a way to make the architecture spam-proof would probably end up in a situation where people would be forced to open up their email contents to someone, thus adding a privacy loophole. Or it might end up with ton of spam similar to usenet, which essentially makes it unusable. Also, why would one want to name something "Dark Mail" which has some negative connotation to that.
