On Fri, Jun 30, 2006 at 01:47:01PM +0200, Oskar Sandberg wrote: > Ian Clarke wrote: > >-----BEGIN PGP SIGNED MESSAGE----- > >Hash: SHA1 > > > >I don't think we necessarily have to prevent location swapping on > >opennet nodes, the destination sampling approach seems pretty robust, > >and as the network stabilizes, the number of location swaps should > >decrease. > > I don't think this matters either. A much bigger concern is that the > network could end up largely split into two - very few "open" nodes > talking to dark ones, and vice versa. For it to work, people who are > open would also have to want to authenticate people who don't directly.
In other words we need to figure out a system of incentives to make it extremely attractive, as well as easy, to add darknet peers. There is absolutely nothing wrong with incentivising the behaviours which will ensure the network's survival. We have to do this to some degree in e.g. load balancing, this is no different. Here's my thoughts: 1. Opennet takes ages to bootstrap. It has constant connection churn. While this can be a strength, it can also be a weakness. Darknet offers some level of stability. 2. We can provide some level of local "sharing". We can share bookmarks, and possibly file indexes, with our direct peers. We can send text messages to them, or files; we can integrate with Jabber perhaps. 3. Significantly increased security. We can have a "trust levels" system. If you have enough true-darknet connections then locally generated requests can be limited to true-darknet connections. 4. More security: I believe it will be extremely difficult to implement premix routing in any meaningful and safe way on opennet. Certainly it will require completely different structures. Both premix routing and swap enforcement *require* darknet AFAICS. 5. Preferential treatment. True darknet nodes will tend to have fewer connections and therefore more traffic can be handled from each connection. But we can go beyond this: While we should not misroute requests we have accepted to our darknet peers, there is nothing wrong with accepting more requests from them, if they want to send more requests. Load balancing will then adjust the input load accordingly (more darknet requests allowed, less opennet ones). Any other ways in which darknet is better, or means by which we can favour it without breaking opennet? > > A problem, in general, with this whole thing is that the incentives for > connecting to people are too small. It is hard to convince people that > they ought to go through the trouble of adding more then a neighbor or > two, if the only reason is that it is healthy for the network (when they > may not notice much difference themselves). Yes. > > When I first envisioned an applications of this type of Darknet, I > thought of it much more in the context of a IM/file sharing application > then Freenet. In such a system, people would have have motivation to add > "buddies" (presense, being able to surf their share directly, etc) which > they don't in Freenet... Why can we not have Thaw share its index files with the adjacent nodes? We could provide FCP support for local messaging. > > // oskar -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20060630/fe2049d5/attachment.pgp>
