On Tuesday 07 October 2008 03:40, Daniel Cheng wrote: > 2008/10/7 Matthew Toseland <toad at amphibian.dyndns.org>: > > On Sunday 05 October 2008 08:31, Daniel Cheng wrote: > >> 2008/10/4 Matthew Toseland <toad at amphibian.dyndns.org>: > >> > On Saturday 04 October 2008 02:23, Daniel Cheng wrote: > >> >> 2008/10/4 Matthew Toseland <toad at amphibian.dyndns.org>: > >> >> > On Friday 03 October 2008 17:27, Michael Rogers wrote: > >> >> >> Can't remember whether this has been raised before, but a random walk > >> >> >> terminates at a given node with probability proportional to the node's > >> >> >> degree; does this mean high-degree nodes are more likely to receive > > swap > >> >> >> requests than low-degree nodes? Seems like that could be disruptive in > >> >> >> two ways: > >> >> >> > >> >> >> 1) When a high-degree node changes its location, many other nodes are > >> >> >> affected. > >> >> > >> >> If you/vive/oskar are looking at the the degree thing... please review > >> >> this as well: > >> >> http://code.bulix.org/20bjpk-68537 > >> >> > >> >> This patch remove the opennet location from swapping -- > >> >> essentially seperating the darknet and openet. > >> > > >> > Why would that be beneficial? You're still treating them as the same for > >> > routing purposes? > >> > >> darknet links are stable, opennet links are not. > >> swapping should depends on (and only depends on) something stable, > >> or the location won't be stable. > > > > Possibly. But then we have to figure out how to route between independant > > networks, which we haven't solved yet. > > The good thing about opennet is: It don't need swapping to work. > Just give each node a random location, it would figure out the linkage > and routing itself. > > Darknet need swapping in order to work. > Disabling swapping on hybrid node just make the darknet links > completly unusable. > > I suggest *enable* swapping on hybrid node and *ignore* opennet location. > > *enable* swapping on hybrid node > ---> * darknet would work only if you swap > > * too few darknet-only node on the network, > swapping location between ~100 unevenly distributed > nodes is ..... useless. > > *ignore* opennet location > ---> * opennet location is unstable, no reason depending on it
The opennet location is fixed. So it is stable. > > * this does not hurt openet: > opennet is self-organizing, > as long as the location is stable, it would work. > > * bad (unused) opennet link would just fallout > there is no waste of connection slots . > > > >> >> >> 2) There might be some correlation between degree and other > > properties: > >> >> >> high-degree darknet nodes might belong to committed users with large > >> >> >> stores, in which case it's particularly disruptive if those nodes keep > >> >> >> moving. > >> >> >> > >> >> >> Just a thought. > >> >> > > >> >> > I don't know. This looks like a question for vive/oskar. > >> >> >> > >> >> >> Cheers, > >> >> >> Michael > > > > _______________________________________________ > > 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 > > -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 827 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20081022/48bfd416/attachment.pgp>
