On Tue, Apr 18, 2000 at 02:31:51PM -0400, Tom Ritchford wrote: > You have put your finger on a big problem, that of "absurd searches" which > return unreasonably large numbers of hits. > > However, I don't believe that you will be able to come up with a rule that > will exclude a given search as "absurd" before the fact. > > This means that we have to handle absurd searches in a graceful way. > > > I don't have a solution to this though the direction to go might be > that nodes that match a search are less likely to pass the search on.
My suggestion on that would be to limit the number of found items returned, and possibly mark it as such to encourage the user to narrow their query. Also if we have nodes designed to strip overly common keywords (esp. content types) then we can go a long way towards making searches reasonable. Your idea on passing on messages only if nothing or very little is found is also a good idea. Brandon's ideas on making Broadcast searches have less of an impact on the system as a whole should prevent the search system from damaging anything but itself. Floods will be dropped at the node being flooded, DOSing only that node, and caching will put a stop to non-malacious flooding / demand for a document. It shouldn't be possible to exert much leverage on anything but the searching subsystem if it is implemented properly. And even that attack would require a signifigant and sustained amount of bandwidth for not much result. --Adam Lydick -- Freenet -- Re-Wiring the Internet http://freenet.sourceforge.net My Node: tcp/rivendell.yi.org:19114 _______________________________________________ Freenet-dev mailing list Freenet-dev at lists.sourceforge.net http://lists.sourceforge.net/mailman/listinfo/freenet-dev
