I think routing could be improved by having nodes deliberately cache data that
they think will likely be requested from them in the future, rather than a
random sample of whatever comes their way.
To do this I propose the probability of storing any given piece of data should
be equal to:
The predicted probability that OUR node has the data, as defined by the
NGrouting function that is used to predict the probability that other nodes
have a piece of data, but applied to our own node using the entries in our
routing table in the same way that previous successful requests from the
other nodes are used in the function normally.
Divided by:
The average predicted probability for all possible keys.
And multiplied by:
The presentage of data that we are caching as defined in the config file.
I hope this is sufficiently clear.
_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
