On 1/15/07, Matthew Toseland <toad at amphibian.dyndns.org> wrote: > On Mon, Jan 15, 2007 at 05:25:58PM +0000, Michael Rogers wrote: > > Matthew Toseland wrote: > > > The other possibility is that this is happening naturally somehow. My > > > theory was that the node moves peripheral locations to peripheral nodes, > > > and then the peripheral nodes drop off the network. > > > > That would explain how there could be more than one cluster. If that's > > the cause, periodically resetting to a random location should help, right? > > Would it? Or would it drive the system to be more and more absurdly > specialized? Currently, a node is introduced; its location is well > within the core area of keyspace; we swap locations with a core node, > and the peripheral node gets a peripheral location. Then the peripheral > node disconnects, and the peripheral location is destroyed. So the > keyspace gets more and more clustered towards a single point (or > multiple points). This doesn't however explain why the cluster we see is > around 0.0/1.0. Anyway, would periodic resetting help? I dunno. A new > location if it was peripheral would still be moved to a peripheral node; > a new location which was close to the core would be moved to a core node. > However, these are no longer tied to the addition or removal of nodes, > so they have more time to hang around on the network before they are > removed... So I suppose it would help. Anyone feel like running some > simulations to confirm the theory and the fix? > > > > Cheers, > > Michael
Won't we be able to test this theory from the forced network reset? My node has a different location now (not 0/1 anymore). However, it seems my peers while distributed better than before are moving to 0/1. Four already have. -- I may disagree with what you have to say, but I shall defend, to the death, your right to say it. - Voltaire Those who would give up Liberty, to purchase temporary Safety, deserve neither Liberty nor Safety. - Ben Franklin
