Quote from: r.willis on April 11, 2013, 01:04:14 PM

Quote

The client is. The server will be as soon as it's stable enough.

Do you know you are misleading people? Particularly when you use word "open" and Open Source Initiative logo describing your sistem?

I don't think we are. We've been quite clear about what's open now and what are future plans are.

Quote

Quote

Again, we sign every new ledger, about every 20 seconds.

Quote

A new ledger is closed approximately every 5 seconds.

Which one is true?

It depends on conditions. Right now, they're averaging about every 20 seconds. It can drop to every 5 if there's continuous transactions. The "idle interval" can change based on network timing accuracy as well. See https://ripple.com/wiki/Continuous_Ledger_Close

Quote

Quote

We designed the system so that people would have good reasons to trust it without needing a central authority to trust.

Quote

In practice, most people will use the default UNL supplied by their client.

Seems a little conflicting to me.

There's no conflict. For example, Bitcoin doesn't require you to trust any central authority, but most people either use online wallets (where they trust the wallet operator) or download client software (where they trust the software provider). The important thing is that Bitcoin doesn't *require* them to trust people in this way. They just do so because it's convenient -- making a considered judgment to balance risk and convenience.

Quote

Quote

I don't know what you mean.

I mean, how new coins of ripple currency are issued?

All XRP were created when the genesis ledger was created. No further XRP can ever be created. Other currencies are created when a pathway is created with a non-zero limit and then funds are pushed along that patway.

Quote

Quote

By consensus.

Do you have hard proof that 1000 nodes would converge to consensus in N iterations in presence of 10 malicious nodes? Do you believe they can execute this N iterations in 5 (20) seconds?

We've simulated precisely that condition. How fast it converges under those conditions depends upon the latency between the nodes. Under most realistic conditions, it can converge within 4 seconds, even with 1,000 nodes. Large numbers of conflicting transaction can increase the time. 5 iterations won't typically take more than 6 seconds. The network will take as long as it takes to reach consensus. There is no particular required time. Nodes that are particularly slow will "bow out" to improve the network's performance. There's a safety to ensure we don't shrink to a small set of "super nodes".

Quote

"Technical Description" of consensus on wiki doesn't seem very technical to me.
What about network split? Will it be permanent and not-repairable?

A worst case failure will result in a persistent network split not unlike the hard fork Bitcoin had recently. One difference is that every Ripple node will immediately know that there's a hard split rather than requiring manual intervention to stop processing transactions as Bitcoin did.

The resolution would depend on the cause of the split. If both halves were equally good and the split occurred due to some freak occurrence, a decision of which fork to take would have to be made as it was for Bitcoin. If the cause was malicious, honest people should choose the non-malicious fork, even if it's in the minority somehow.

I'd say such a split is basically inconceivable, but I would have said the same thing for Bitcoin, and it happened. So, yes, it could happen.