I'm discussing it with the author on bitcointalk.org - what he has presented so far doesn't address the sybil attack problem for me. His system prevents attackers from acquiring a lot of trust over a short period of time, but as far as I can see, the long con will still be effective.

http://blog.emunie.com/?p=53#more-53

TL;DR: Transactions carry meta-information that assigns trust to signing nodes. In case of conflicting transactions, nodes select the transaction signed by the signers with highest collective trust. Sybil attacks are countered with fees.

Sounds like a system that combines elements from TAPOS + DPOS with an unlimited number of signers.