Re: Re: Drop Dynamic Querying
Arne Babenhauserheide <[email protected]>
| Newsgroups | gmane.network.gnutella.devel |
|---|---|
| Message-ID | <[email protected]> |
Am Montag, 26 de März de 2007 17:10 schrieb Bill Pringlemeir: > Credence addresses both cases, but requires that users vote wisely (and > often); I assume that it would suffice, if users could just vote down by banning files. Excluding bad results is done quite routinely, and if tehre's a nice friendly Trash Button, I assume that most people will vote often, if only to get rid of spam results for themselves and normal use-patterns would result in a good voting base; maybe not of many good votes, but at least of many bad results, which would solve the spam-problem. And if people share files they want to support, they are likely to vote them up. As I remember from my useage of Credence, the main problem was finding other Credence users, and finding them again. But this shouldn't be a problem anymore, as soon as we can use teh DHT to locate unique hosts. As long as the network consists mainly of good hosts (which we have to assume when we want to have any chance; if tehre are more spammers than regular nodes, there won't be any possible protection, but in the current network, they'd need millions of _active_ spam-nodes for that), Credence should solve the spam-problem. > DQ should be altered to exclude low correlated nodes from searches. And to exclude the results in the count, so the leaf keeps DQ active until it has enough uncorrelated and high correlated results. > GUESS shifts searching from the network to the node, so it *seems* more > difficult for SPAMers to monopolize queries with their responses. GUESS > can also exclude any nodes with SPAM responses from future searches. But a node can still send me spam - and the network loses te ability to protect itself (for example, every node could deny forwarding to low correlated results, which gives no advantage to spammers, because they can already do this, and it doesn't require a central blacklist (which is doomed to be outdated at all times), because it solely needs decentral voting). So sending results out-of-band can become a probem instead of an advantage. Or did already become a problem. Wishes, Arne [Non-text portions of this message have been removed]