So if Bob owes Alice $500, and Alice owes Bob the same amount for some reason - that makes sense it would cancel. But since Ripple allows you to make a 'chain' of debt-issuance, I don't see how a third party would take a passed along debt and cancel it out that way.
Think about the way banks settle. I write you a check for $50. You deposit it. Now, my bank owes me $50 less and your bank owes you $50 more. If you then write a $50 check to someone who uses my bank, our banks can just cancel them out and don't need to do any settlement. If there's a net imbalance, they can settle.I don't know. From what I have gathered, it is a distributed database. Users would only need to hold their own data, but available to everyone upon demand. It wouldn't be a 'blockchain' with every transaction.
Nodes that wish to process transactions must have all current state information. They don't have to have a historical record of past transactions. You don't need to run a node that processes transactions to use the system.
This is the bottleneck.
