I've been considering a kickstarter for a new version of SMTP, while at least for the moment leaving IMAP alone. Specifically, the way headers are appended to mail in transit is unsupportable in a secure environment. The things I'm considering is that there doesn't have to be a flag day, you just need the vendors of several heavily used MTA's to support it as an option, then once 99% (or whatever number your company deems appropriate) of your email uses the new format you turn off the old.
This was poopoo'd in the past because there were 10s if not hundreds of thousands of email servers. Now people have pretty much stopped hosting most email and turned it over to google, yahoo, microsoft or one of the other major players. Therefore you're no longer faced with trying to get everyone to change things. You only need 5 major companies to change, and hopefully they're interested in the new protocol as well (nobody likes SMTP as it is, the question is can you get everyone to agree to some consensus of next generation email then move forward with it)
DJB's pull based email thing could be a part of this, maybe not the exact idea but something along those lines:
DJB's IM2000 (http://cr.yp.to/im2000.html). While I don't think all mail should be stored on the originating server, I think a mix could be used to provide more flexibility. Mailing lists could leave all the mail on the server, since a bunch of readers never read every message there isn't a point of exploding it out to thousands of mailboxes (except for reliability, and that could be gained by mail->nntp for public mailing lists)
Requiring domain keys could also be useful, since headers wouldn't be modified, just appended and signed.
If people are interested in crypto/privacy aspects, emails that aren't delivered but instead picked up by the recipients don't leak metadata like To, From.
It's probably best to approach this through the IETF, despite failures to make broad sweeping changes in the past, a new working group might be the best choice to get the interested parties involved.
Tangent here:
I also think that email clients need to be brought back and worked on. Thunderbird died because of two reasons: 1. Mozilla couldn't find a way to monitize it, and 2. Their biggest email competitor (gmail) and biggest contributor (google search) had already found a way to monetize email and thunderbird wasn't seeing significant updates at that point.
Other stuff I'd like to see in thunderbird:
Contact pictures on email (not something I think I would use, but nice for people used to facebook/twitter/etc). Integrated IM/Skype/Phone so you can effortlessly change the medium you're communicating through. Also the ability to send calendar events through IM or SMS would be nice.
Real synchronization. That includes plugins and every setting via a service like weave that is secure. This would also sync your passwords and gpg keys. Actually a generic weave-like framework that could be integrated with pidgin, thunderbird and other open source apps to sync across machines would be great. That would also fix major issues with pidgin's OTR.
So the reason I never kickstarted it is the same reason Mozilla doesn't work on thunderbird anymore. I have no idea how to monetize it in a way that would be long term sustainable. Users hate adds, they hate paying for software. Maybe an addon store, but that just means you're subbing the good development work to other people and then making the users pay to fix the things wrong with your app.