On 23/08/11 2:37 PM, Tony Arcieri wrote: > The point I was trying to make on the thread I started about a global > distributed datastore is that I thought modeling the system in terms of > probabilities is a better approach than trying to enforce a systemwide > currency that gives exact tit-for-tat measurements.
Yes, I understood your point. But I would suggest it is correlation not causality. The tit-for-tat comes from human behaviour, not from using a currency or any other feature of a design. A currency just helps to allocate the tits & tats more accurately, it isn't the cause of the behaviour. So, if you average things out with some sort of sharing / probability approach, you just won't see the tit-for-tat or other human solutions so clearly. Hence, abuse or inefficiency in allocation won't be seen. Now, whether this assists or breaks your app is ... open to question. It works for e.g., Skype & TOR, which consumes resources from one person to allow a service to another, without any clear accounting. But it doesn't work to the extent that when I notice my laptop heating up or my bandwidth draining over phone, I turn Skype off! > When you don't have any guarantees of SLA when you "purchase" things > with some virtual currency, what is it actually worth? (SLA isn't an expression of value. It's just a feature demanded by some very large corps that can't cope with unreliability in non-core business. Reputation + best-efforts is far better than SLA+bigger price, IMHO.) > I think a better > approach is to cryptographically log the behavior of various > participants in the system in order to automatically make judgments > about them and how reliable you expect them to be. This works on paper, but most reputation systems so far have been worthless. Where, worthless means, didn't deliver enough value to justify the work put in to collect the info... iang _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
