On Saturday 11 September 2010 14:31:20 Dennis Nezic wrote: > On Sat, 11 Sep 2010 06:58:00 -0400, Uriel Carrasquilla wrote: > > I am thinking more along the lines of membership interests. I want > > to be in the yoga darknet group but also in the tennis darknet > > group. But if I join both, I have now bridged these two groups into > > a new yoga-tennis group. If I was to carry this operation of members > > joining other darknet groups, eventually I end up with one single > > group, what we have today, opennet. > > Yes, effectively. (Opennet behaves a little differently -- your > neighbouring peers are constantly being swapped and optimized to > approach a small-world topology.)
No. IMHO our key goal is to build a large global darknet. Which is completely different to a large global opennet. In that it actually has some level of meaningful security. It is possible to rubberhose and remote-root your way across the network, of course, if you are e.g. doing a mobile-attacker-source-tracing attack based on predictable keys (CHK inserts of predictable files or messaging posts e.g.), but this is *VASTLY* more expensive than the equivalent on opennet, which could probably be implemented on a domestic connection with no resources and a moderately determined geek. > > > What are the advantages of a darknet? > > The main advantage, I believe, is security -- opennet nodes are > relatively easy to monitor and traffic-analyze, given a strong opponent > like Big Brother, by compromising your (constantly changing) opennet > peers. In darknet, they would have to physically compromise each of your > friends. Also, since opennet nodes are ... open ... all opennet node ip > addresses can in theory be listed, and blacklisted. To do this in > Darknet would require physically traversing the entire network. Fundamentally the problem with opennet is you can find and connect to all opennet peers quickly and electronically without needing much in the way of resources. This means that ALL attacks are easy. Although "easy" varies from trivial to maintaining connections to 20,000 nodes (which is still "easy" in that it's probably feasible relatively cheaply with bandwidth being the main cost). > > > I take it faster routing (the number I heard is up to 20 members). > > Maybe, although I don't think it's necessarily the number of peers that > affects this -- the number of peers you are connected to is a > limitation of your bandwidth. Unfortunately darknet, especially with small numbers of peers, does not work particularly well in terms of load management. Fixing this is one of the goals of the new load management system. > > > It is a predefined same-interest group (as long as no random members > > are added). > > Again, I wouldn't think of things in terms of groups. There will be > cloudy clusters of common interests, but in general it's a open sea of > connections. Common interests are not all that important. They certainly help in terms of caching but it is not necessary. Freenet routing assigns locations and puts the same data on the same nodes, and the small world property of the underlying network enables it to find that data efficiently. The small world property results from people connecting to their friends i.e. people they already know. Opennet on the other hand manufactures it. > > > I thought version 0.7 was supposed to fix the problem of anonymity > > for nodes by creating darknets with the understanding that within a > > darknet anonymity is not necessary since in theory all the members > > know each other anyway. > > Darknet was implemented to fix the rather serious security issue of > opennets. (Opennet in 0.7 was only supposed to be a transitional thing, > for newbies and people not too concerned with scary opponents.) It still is!
Description: This is a digitally signed message part.
_______________________________________________ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe