Erste Seite Zurück Weiter Letzte Seite Übersicht Grafik

Local reputation as P2P


Notizen:

For the mentioned technical and social scalability reasons, the Freeporters' reputations algorithm is implemented as a peer-to-peer network. Therefore every node may reside on a different machine. Every node only has local knowledge about the network. They only need to know about directly trusted nodes.
Instead of calculating all the reputation values beforehand, we let the news items travel along the connexions in the opposite direction of the ties. It is not necessary to send the whole content of a report, it is enough to send a notification along with a URI to locate the content on the World Wide Web. The notification must also comprise an interest value. This interest value is changed by the nodes while it travels through the network to reflect a particular node's point of view. Additionally, the way a notification travels can be recorded in the notification.
For an implementation, the multiplication can be used for the concatenation function and the maximum function for the aggregation. The scale of trust and interest values is between 0 and 1.
In this example, A composes a report and publishes it. Node A sets the interest value to 1. Node C receives the report. It automatically changes the interest value by multiplicating it with the trust value of the connexion and annotates, that it was received from node A.
Node B receives the report two times, from node A and node C. After calculating the new interest values, it only keeps the maximum which is 0.64. It also annotates, that the report was received from node C. Node C acts in this example as a recommender of A's report.
Node D does not receive the report at all, because it is not interested.
This distribution process can be stopped with threshold values. News items will only be transmitted, if its interestingness is above a certain value.
There is also the possibility for a maintainer of a node to manually increase the interestingness of the automatically calculated value. In this case, the report may also become more interesting for other nodes and may continue to reach additional nodes.