> > There is nothing wrong with using the > > data source node, as long as it is not overloaded or physically farer > > away then the cache. > > How do you know whether the data source is overloaded or physically more > distant? Freenet's way of distributing load might not be optimal in all > cases, but deliberately concentrating load seems like a poor alternative. >
The requests to a data source node, that are caused by a redirect can be marked, to be low priority. The data source node may answer with the query reject message, that overloaded nodes use now. The load is thus just as much concentrated on the node as it can handle. An overloaded caching node can even unburden itself by returning a redirect instead of the data. You could use the ping time or the number of common leading bits in the ip address as a measure of distance. > > How do you keep spammers or flooders from using "hint" messages to make > > their data popular? > > How do you keep spammers from using automatically generated requests for the > same effect at even greater cost (with or without redirects), or inserting > tons of their garbage to push everything else out of everyone's cache and/or > clog up the network? The spammer will need much more bandwidth to do that at least for the inserts. And existing weak points in freenet are not an argument to open up even more. > Hit-count updates would only have as much credence or > effect as the recipient is willing to give them based on authentication, > past history, etc. If it smells funny, drop it. After all, it was only a > hint so, by definition, dropping it won't break anything. Dropping > redirected requests would be much more disruptive. You misunderstood me. The node receiving a redirect may choose not to follow it. It can always request the data again and specify, that it won't accept a redirect. This works after a dropped or failed request to the data source node, too. -- Thomas Leske _______________________________________________ freenet-tech mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/tech
