On Mon, Aug 20, 2007 at 02:49:05PM +0100, Michael Rogers wrote: > vive wrote: > > This scenario corresponds roughly to many nodes that are alive in the > > network and a small flow through of positions by nodes joining but decide > > to leave for good. Some results soon... > > OK, I guess we have different ideas about churn - in your model it's > always the new nodes that leave, whereas in my model every node leaves > eventually (though not necessarily after an equal amount of time).
It's a matter of selecting something interesting and simple enough so we can simulate the principle isolated and see if it appears.. of course simulations could be done by trying to simulate how the actual networks grows.. If we see it appearing in simple models, we may then refine them. > But even if it's not the same, churn will eventually make the network > navigable anyway, as long as the new edges have the right length > distribution. Hopefully so :) If it grows well enough reflecting a (e.g. real-world) navigable graph or if we also use opennet. > But if you're constantly adding *and removing* nodes, then it's not > necessarily important for the new links to be navigable using the old > locations: what's important is for the new links to be navigable using > the new locations, because eventually all the old links that were > navigable using the old locations will be removed, and only the new > locations will be relevant. In other words, make the node findable in > the network that's being created, not in the network that's being destroyed. I agree, and this is what seems to happen to most networks, but we don't know really how it works or how to do it. > But again I guess this comes down to a difference of opinion about the > nature of churn: does churn replace old nodes, or just add and remove > new nodes? Both effects should be there if we should be realistic... I've started with simulating the second part to see how sensitive a network with the swapping algorithm is to an inflow/outflow. regards, Vilhelm -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 187 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20070820/6ac5d1ff/attachment.pgp>
