Title: REQUEST & RESPONSE

Request & Response

A main difference of GNUnet compared to Crowds or Onion Routing is the relation between request and response routes. In the other two systems, the routes are basically the same (just inverse and, thus, differ only in their direction). The difference of GNUnet is that both routes can be independently determined, to a certain degree (see Figure 2), by the user. In case of rerouting requests, a user A may decide, if she wants to set her address as originator of the message (indirecting) or if she wants to keep the address B of the original sender (forwarding). In the latter case, the response will be delivered straight to B without being rerouted over A.

Figure 2: Anonymous data transmission in GNUnet. User A sends a request to user B who is deciding to forward the request to user E. The response passes from user F to user A in only four steps. This is achieved by means of optional shortcuts.

It is up to each user to decide how many and which other users will receive a request which is going to be rerouted. The amount of forwarded messages is determined by the current CPU and network loads, the reputation of the sender, and a random value. In that case, reputation of a user is determined by the amount of her requests and valid answers, in short her behaviour. Additionally, each request provides a field, which determines the time-to-live that is particularly useful to avoid that requests circulate forever in the GNUnet.  

Messages can be sent to the users in direct connection with the sender. However, for each message, which is going to be forwarded, the choice of recipients is not uniformly distributed, but rather depends on the hash value of their public keys. Recipients are the more preferred the more the public key’s hash value is “close” to the one of the request.

In general, requests (and responses) are not instantly delivered. It is rather the case that they will be buffered until either the buffer runs full or a random period of time exceeds. 

It is then up to each user which receives responses to decide whether she wants to store the response or not. That way, requested data is distributed more and more within the GNUnet by time. Moreover, one user may ask another to store data. However, whether the other user complies depends on the reputation of the requester and local resources, particularly storage space.