Why? There will be some clustering from the IRC, and it seems to be producing *roughly* the right link degree distribution. And IMHO it will be difficult to write the server software to produce a small world network...
On Thu, May 18, 2006 at 11:26:58PM -0400, Colin Davis wrote: > Freenet .7 is designed to work through a series of connections > between individuals- When a person first joins the Freenet darknet, > they are expected to join with at least three people they know and > trust. > > While this may be attainable once there are a high number of network > users, With the network in it's current small-size a new users is > unlikely to have any existing friends using the service. > > > While people might get their refs from their friends, I worry that > the vasty majority will join after having read about it in a > Magazine, or in Slashdot. They may add their friends afterword, but > they need a way to get started without them. > > > > The current solution to providing people with node references to > exchange is an IRC channel, in which people can exchange their > noderefs, but I don't feel that this is an adequate way to solve the > problem- > > People joining the channel are usually there seeking references, as > they are newcomers to the service. The majority of people they > exchange with are thus likely to ALSO be primarily newcomers, due to > the number of people looking at any given time. > > This problem is exacerbated when a news story or new release of > freenet is announced. A Slashdot article may bring hundreds of people > to #freenet-refs, flooding the channel. Since they will be the > primary people IN the channel, they will link incestuously with one > another, but little into the existing network. > > I think that one way to solve this in the short term may be to > implement a link-exchange website, which people can go to to post > that they are seeing references- The website can then hand out > references to visitors, tracking the connections it hands out, and > the ones that it has handed out in the past. > > By handing out these references, it can form a rough map of the > network- While it will miss the dark nodes which don't use the > service, it at least gives a rough idea. > By doing this /very rough/ mapping, the server can decide /where/ to > hand them out, based on the number of other connections it's given > out in certain areas. > > > Example: the node initially knows about (and has contact > information/references for) five nodes, A, B, C, D and E. > > A story goes life on Slashdot, and 20 people apply for noderefs. > F-Y. > With the existing IRC solution, they would primarily > link to one another > with one or two links from their network to the > existing base. > > The website solution would give each of the next 10 a > connection to two of the existing servers (between > A-E). > It would then try to balance the new 10 connections > similarly, using the entire set of 15 it now has. > > > I know it's a simplistic view- Please don't get hung up on my > semantics of HOW to try to balance.. Oskar is Smart, Matthew is > Smart, and Ian is smart. I'm sure a good algorithm could be designed > that balances better than nothing at all. > > > > I'll admit this is far outside my area of expertise, but it seems > that almost ANY solution, even an imperfect balance, which matched > refs via a server page would be preferable to the current IRC scenario. > > -Colin > > _______________________________________________ > Devl mailing list > Devl at freenetproject.org > http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl > -- 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/20060519/6a8a142d/attachment.pgp>
