Double spending isn't really about trust. It's about transactional atomicity.
After reading a bit, I've concluded that Ripple doesn't actually solve the consensus problem. Whether it works or not depends on the shape of the graph of the core validating nodes. If the graph is too centralized (as it is now, and probably will always be), it can be controlled by fiat.
Transactional atomicity is exactly the reason why ripple needs a native currency - the native currency is what provides byzantine fault tolerance[1], just as in bitcoin.
Supposedly, ripple reaches consensus under diverse graphs of validator node networks. But its a research topic still undergoing study (Ripple Labs say they have whitehats closely examining the protocol).
1. http://en.wikipedia.org/wiki/Byzantine_failure#Practical_Byz...
The idea that I'm interested in is that it seems to solve the double-spending issue that Bitcoin so eloquently solved, but without all the electricity wasting issue, and the 51% issue, that currencies with proof-of-work (POW aka mining) have. Currencies that use the POW system are designed to work well only if half of the computational power of the network is used by honest nodes. If by any means more than 50% of the network is in control of dishonest nodes, they could do all kinds of things and perhaps even kill the entire currency.
Replacing the proof-of-work system with something based on reputation seems very interesting in my opinion: https://bitcointalk.org/index.php?topic=10193.0