>if it had the "public" flag enabled, would have its >routing table exported at a fixed location on the distrib servlet port
If this is done, then no changes to the gwebcache code are necessary - we'll be ready to go in less than an hour: Few addresses of caches will come bundled with installation and the installer script will query random one of them for the address of a public node. After that the installer will just download the refs from that public node. Or if we want we could ask for two or more public nodes, then compile seednodes.ref from theirs, etc. All kinds of permutations are possible. >The problem is any scheme like this trades some >centralization for some attackability - do we really want to make it so >easy to find a large fraction of nodes on the network? The caches as they are now are designed to store only the 20 most recent submissions, the number could be decreased to 5 for example. Also, they can run only in the few days after announcing a new release in order to help new nodes. After things stabilise, they can be shut down, assuming that we disable the cache annoucnement code in order to retain control over who runs them. If a new node can't connect to any of its bundled caches, it will just download the references from hawk as before. _______________________________________________ devl mailing list devl at freenetproject.org http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
