> David Sowder (Zothar) wrote: > > Is there more than one metric for which we are trying to achieve "small > > world"? If so, could that be confusing things for others as well?
I don't claim to be an authority on this, so I will leave any kind of detailed explanation to the people who really know about it, but I may be able to help on a couple of points: Swapping can't achieve the triangles, ie. if A knows B and B knows C, it's likley that A knows C. This is really the definition of a small-world network (looked at from one perspective), and the swapping algorithm will find it's job much more difficult if this property doesn't hold. Arguably this property does emerge with #freenet-refs, since people who join at similar times will end up peered with each other. The other property is geographical closeness. If all my friends suddenly started using Freenet, nearly all my peers would be in the UK, and thus have reasonably good ping times to my node. To some extent, people in the same time zone will be on #freenet-refs at the same time, and so it will have this property too. Much less so when you factor in ref exchange bots, of course. I believe this is what's already been said, but I'll add it if only for the sake if archive completeness: the swapping algorithm is merely a mechanism of assigning values to nodes so that we have a concept of location for routing in the network. It doesn't set out to achieve any kind of topology, merely represent the topology that it's given. The only way we can change the topology is by changing the connections. As for Colin's reply: > I've thought about this before, but it's difficult to do that in a way > that's both difficult to harvest lists of opennet users from Fundamentally, if a new Freenet user can automatically find new peers, so can the FBI. I'm sorry, but there's no getting around this, full-stop, simply because we cannot distinguish the two. Yes, you can have things like refbot.py that collect 10 refs and disappear, meaning that you have to watch #freenet-refs constantly, but it's still possible. Granted, doing it over freenet keys means that it's less blockable (you can still spam the key, which is a minor annoyance), but that's really equivalent to opennet via requests, so we're just getting down into implementation details, which is not what this discussion is about. Slightly long post, I know, but I hope it's helpful. On Monday 05 March 2007 20:37:22 Colin Davis wrote: > I've thought about this before, but it's difficult to do that in a way > that's both difficult to harvest lists of opennet users from, and not > having external requirements. > > You could arrange and find parters through IRC, but then it's trivial > for someone to /join that IRC channel and watch who's connecting, and > could be blocked by blocking IRC.. > > The best answer that I could come up with, was to post noderefs to > Freenet, NIM-style.. That would by in-band, so it wouldn't be any more > blockable than the rest of freenet.. It would still be harvestable, > though, and it would require you to connect with at least darknet user, > before Open-net could start.. > > Personally, I think that's a good requirement ;) > > -Colin > > > David Sowder (Zothar) wrote: > > I'm trying to resolve something in my mind about the small world model > > and how it relates to Freenet. My understanding has been that the > > relation was in Freenet node location distances and my assumption was > > that the swapping algorithm was intended to optimized the "small world > > model" of an arbitrary set of connections such that, in my mind, it > > would theoretically settle on all nodes having a small world > > distribution of peers: increasing numbers of peers as shorter distances > > from a given node. > > > > Toad has informed me on IRC a bit ago that the swapping algorithm does > > not make arbitrary interconnections achieve "small world", which leaves > > me with these questions: > > > > Is there more than one metric for which we are trying to achieve "small > > world"? If so, could that be confusing things for others as well? > > > > Can a given node and a list of potential peers be used to create a small > > world model, at least from the perspective of the given node? I assume > > this is somehow possible as I understand it that opennet will be doing this, > > > > Some of you may already know where I'm likely going with this. What can > > opennet built into fred do that a program like refbot.py couldn't do? > > Could refbot.py potentially say, add 50 peers and then remove (in an > > orderly fashion) all but 15 based on a small world location/distance > > distribution to achieve a small world model if say, all/most nodes were > > using this same algorithm? > > _______________________________________________ > > Tech mailing list > > Tech at freenetproject.org > > http://emu.freenetproject.org/cgi-bin/mailman/listinfo/tech > > _______________________________________________ > Tech mailing list > Tech at freenetproject.org > http://emu.freenetproject.org/cgi-bin/mailman/listinfo/tech >
