I'm starting to look into Sovereign Keys, covered in more detail in  and .
That the current PKI system is brittle is accepted by many people. Brainstorming and prototyping new internet service authentication approaches is first-order important, and Sovereign Keys is worthy of further investigation and support.
Seems like Sovereign Keys does introduce a few new concepts that need security analysis; the timeline servers offer an interesting capability, I wonder about vulnerabilities. For that matter, I wonder about patents in this space. There is a minefield of granted timestamping patents and while the timeline servers may not specifically address timestamping, I wonder of some of those patents were written generally enough to impact Sovereign Keys.
Note that a proposal from Adam Langley and Ben Laurie of Google  also introduces the notion of a public append-only log, in some ways similar to timeline servers, but not domain-specific.