I highly suggest you look at this (https://bitcointalk.org/index.php?topic=307922.0) , which solves all Fiat transfer problems except fraud (anonymity, reversal, suspicious transactions, bank closure, reporting, etc.) by actually extending the scope of the project to allow spending of Bitcoin on anything.

In the past I was an economic consultant to a very similar business idea. Someone on my team was adamant that the idea eventually be made from website into software (and thus decentralized). We planned to allow the exchange rate to float, and charge 1% of each transaction to pay for dispute-resolution (and prevent Sybil on the reputation system).

Please post/contact me with any questions you have as I was disappointed to see this idea die. Hopefully it will see renewed interest.

Also, I have independently developed something (Truthcoin) which could be adapted for decentralized arbitration (ie, deciding who is right in the event of breach-of-contract). This would make the process completely decentralized.



This coinffeine idea is stronger from a game-theoretic point of view, because you'll notice that it is essentially a repeated version of what has been proposed with the collateral OVER the transaction amount. Blackmailing is therefore riskier (but still possible). Counter-intuitively, it might be best if the parties could never contact each other at all.