I was thinking the same thing. In the past, when people have advocated replacing SMTP with something else to combat spam, that idea is always shot down with "too many clients, too much infrastructure" but if you have a new application and real demand (and it seems like there is) then you perhaps you could get enough traction to make that work.
Perhaps. Not sure I like the 'vaporizes in a week' thing. Point to point is good, and clearly something in the middle has to be designed in such a way that traffic analysis of the middle doesn't help. And like Lavabits you have to be careful that if your server infrastructure could be modified to break security, it can be compelled to be modified with an NSL and that is why I expect anything that is solved will only be solved with a server you "own" in your residence (at least in the US) to require the whole warrant process to get access to it.
Yeah, a few people have been looking at how to do things like this too (I've been talking to them); the idea being mix-net vs. onion routing, which is the big win of being async, and sensible defaults. It's a lot easier to do confidentiality/integrity/etc. on messages vs. traffic analysis resistance vs. strong adversaries, though.
