On Thursday 11 September 2003 06:24 pm, Mr Zenaan Harkness wrote: > I couldn't find this concept/ meme out there yet, so here goes. > > It is just like a debian keysigning party, except those with big boxen > or at least enough disk space to spare and are inspired in some way by > freenet can get together, build a lan of all party-goers, and have their > freenet's cache each other. > > The implications should probably be thought out - as I don't know the > technicalities of freenet software/ protocol, I don't know all the > issues, but the usual ones like > - privacy (but presumably you're with a bunch of friends) > - should this local lan be offline from the rest of the Internet > - if so, for a specific time or until what condition is met > - anonymity of what in cache is being shared > > At such a party it should also be possible to easily inject significant > quantities of data, if that were desired (again assuming you trust those > at the party). > > The reason is that I know that some people devote literally 100's of > Gigs to their data storage needs. Efficient, immediate and accurate > freenet insertion - and very short-term dissemination (albeit to a > smallish (but greater than just one) initial group), given that a few > mates trust each other - might really be useful for freenet overall. > > And the point of raising it here - how would one go about achieving > this. At the moment all I can think of is everyone manually clicking on > a bunch of links and hoping that pulls stuff through, but that seems > terribly inefficient - you want a spider type thing - or some automatic > cache-copying or something. ???
Perhaps instead of transferring all that data there should be a feature so that you list specific nodes that you trust in your config file. Then periodically when you are idle, send a request to them to tell you the contents of their store. Then if they trust you and are idle they will send you a list of keys. While it would be easy to tell if they are lying, the real disadvantage of this, is that within a network of trust if any node is compromised, the attacker knows the contents of everyone's store. It would also take a while to transfer so many keys. Perhaps a better way to do it, would be to periodically request a node to give you it's own estimate for it's routing speed. (In the same format NGrouting will use.) Then you could average this with your own estimates of their performance. If the more the two are consistently different then the higher you should weight your estimates. (Or if they appear to really bad, disconnect form them totally.) The advantage to the second approach is that you don't need much bandwidth. IE: you don't have to physically move computers. But it also requires a much lower level of trust. For example using this approach, it would not be unreasonable to make it transitive. IE: If the node that introduced you to the new node is trusted by you, and says it it's experience the average deviation of it's requests from those claimed is X, (and X is low) then you can probably put some reasonable trust in the new node. _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
