The always interesting Jon Udell has recounted an instance of rabid anti-spam filtering on Jon's Radio.
What he is talking about is 'registered email', where mail is spam by default and only known senders get through. This clearly removes spam, but as Jon points out If we rule out spontaneous association then we will not have defeated the spammers. They will have defeated us. A world in which no one can approach you is a sad world indeed.
We're back to that identity thing. When we're approached in the real world we know for sure that we are approached by a real person, and we can directly hold that person accountable for the approach. Sure, we all know the handout spammers on the city streets but they are a minor annoyance. Receiving email should be like that.
An identity header to a navigable 'sender' information source would keep email free, but upgrade the message with more reliable sender information. Obviously I wouldn't expect the person at the other end to share their life story - but something else will do. A photo or any other kind of sharing of trust will do.
I think it would be interesting to see if spammers would actively seed such an identity pool with 'fake people'. IF they do - at least there's now an infrastructure path for building trust chains for that kind of information.
This would open up a 'reputation economy' like the one in place for mailservers (open relay databases). A personal profile with a low reputation for spamming would be marked 'below threshold', sorta like slashdot, only for person to person email, not forum posting.