"The overwhelming majority of the nodes on the internet running bitcoind are not mining". I think this may well nullify the problem, but is it guaranteed? Is it possible for a malicious miner with many app instances (say, spread throughout a botnet) to get enough connections to a client to prevent the transaction being propagated as it should? I don't know how the client app decides who to connect to, and this could determine whether or not there's a vulnerability – Highly Irregular – 2011-11-16T21:01:21.087

Irregular, If you can prevent somebody from talking to the Internet, you can prevent them from using Bitcoin. However, this is a "problem" that is not unique to Bitcoin... – eldentyrell – 2011-11-17T07:09:16.320

That's not what I meant - what I mean is: is it possible to target a particular client to try to fill up all its connections with malicious nodes? – Highly Irregular – 2011-11-17T22:23:53.433

@HighlyIrregular You can fill up all its inbound connection slots, but then you're just preventing it from helping others. It will still make whatever number of outbound connections it is configured to make to nodes of its choosing. – David Schwartz – 2011-11-18T06:34:00.170

2Your paper analyzes a 'problem' that arises only due to the assumptions it makes. There is no problem unless you assume that nodes respond to the incentive not to send transactions but don't respond to the incentive to receive them. That is an irrational assumption, and in fact your proposed solution it merely to discard the assumption. Why would you assume people would hold onto transactions without also assuming they would sell them? It is strictly this assumption that creates the problem. Both incentives arise concurrently. Why assume only one is acted on? – David Schwartz – 2011-11-17T17:54:08.880

Aviv, your summary has a condescending tone and still willfully ignores the fact that only a tiny fraction of the nodes on the P2P network do any mining -- the rest of the nodes have no incentive at all to withhold transactions. – eldentyrell – 2011-11-18T11:12:44.397

1@eldentyrell I don't see the tone you describe. Besides, the tone shouldn't matter. The summary clearly states "So we suspect the problem is not currently occurring or is not severe at the moment", so the key thing to understand is whether it could potentially become a problem in the future (especially given that mining must at some stage change from hobbyist to commercial scale unless the fundamental architecture is changed). – Highly Irregular – 2011-11-19T20:42:14.927

1"All you have to do is broadcast your transaction to a few major mining nodes" - but this isn't currently the default behaviour, is it? That said, I don't actually know how the client does determine who to connect to! – Highly Irregular – 2011-11-16T20:57:05.333

1"To precisely the same extent nodes are incentivized not to share their transactions with others, nodes are incentivized to get transactions from others. It should cancel out perfectly." - this doesn't cancel out at all. It provides an incentive to receive but not send. – Highly Irregular – 2011-11-16T20:58:23.337

1@HighlyIrregular Same thing. The more incentive I have to buy coffee, the more incentive my local stores have to sell it. If you don't assume people respond to incentives, there's no problem. If you do, you have to assume people will respond to the incentive to receive transactions, likely by inducing others to send them transactions, those others will in turn respond to that incentive. – David Schwartz – 2011-11-16T23:58:40.393

1The coffee analogy doesn't work because there's an incentive for both parties in that case; the customer gets the coffee and the shop earns some money. With transactions, a miner has a monetary incentive to receive them, but no monetary incentive to pass them on as far as I can tell. – Highly Irregular – 2011-11-17T22:27:18.887

@HighlyIrregular You're missing the whole point. Forget about money, don't assume you know the answer. All you know is that the shop has coffee and that the coffee has value to the customer. The customer now has an incentive to try to get the shop to part with the coffee. The shop now has an incentive to try to get the coffee to the customer. In both cases, the incentive is to get some of the value of the coffee to the customer. The natural result of these incentives is the sale of the coffee (or trade or whatever). But all you need are the two matched incentives, then something will result. – David Schwartz – 2011-11-17T22:32:52.817

To get that there's a problem here, you'd have to assume that the two incentives somehow both go unsatisfied by all possible mechanisms. That's a completely idiotic assumption. In this case, the transaction's value drops to zero as soon as any miner includes it in a block, so they have a strong incentive to resolve the situation fast -- otherwise, they both lose out. (It's like the coffee is about to spoil. They'll work out some deal. The shop may even loan the customer money to buy the coffee!) – David Schwartz – 2011-11-17T22:33:21.133

"The shop now has an incentive to try to get the coffee to the customer" - I don't see how the shop has that incentive unless the shop is giving away free coffee for the good of society (ie it has some motive beyond the profit motive). To stretch the analogy a little further, a Bitcoin client has an incentive to notify as many coffee shops as possible that it wants a coffee asap (and is offering $x for it), but while coffee shop A is trying to race to make the cup of coffee it's in shop A's best interests to NOT tell shop B about the fact that the client is offering $x for coffee. – Highly Irregular – 2011-11-18T00:45:15.080

@HighlyIrregular If the coffee shop does not get the coffee to the customer, it cannot get any value from the fact that the customer wants the coffee. If the coffee shop does get the coffee to the customer, it can extract a fair share of the value of the coffee to the customer. (Again, forget about the money. It so happens that the way customers and coffee shops react to the incentives is by exchanging money. That doesn't matter. The point is, they react to the incentives.) – David Schwartz – 2011-11-18T00:55:55.300

let us continue this discussion in chat

Have added a comment to the question summarising my perspective of the results of our discussion. – Highly Irregular – 2011-11-21T08:50:10.410