Experimentation is always good, but for a few reasons I'm not sure this is desirable in the long-run. Firstly, while they are very different, there is overlap in what Freenet and Tor are doing, and so combining them is rather inefficient. The ideal solution would be to incorporate something akin to a mixnet into the first few hops of Freenet's routing (which are optional for the requestor). Secondly, while Freenet is true peer-to-peer, meaning that every client is also a server, this is not the case for Tor, where most users run clients, and a relatively small minority run servers. I remember years ago people were running BitTorrent over Tor, and the Tor guys considered this a form of denial-of-service attack on the Tor network, because it is so bandwidth intensive and bandwidth is a scarce resource given Tor's limited number of server nodes. I think they would have a similar issue with a lot of people running Freenet over Tor. I don't mean to be discouraging, this is an interesting experiment and worth trying. Ian.
On Wed, Jun 22, 2016 5:56 AM, Chris Double [email protected] wrote: On Wed, Jun 22, 2016 at 12:46 PM, Chris Double <[email protected]> wrote:
On Sun, Jun 19, 2016 at 10:56 PM, <[email protected]> wrote:
Yes. Freenet in darknet mode among other stuff.
This is great, thanks. I've been interested in finding a way to run
over Tor. I'll give it a try and see what performance is like.
I tried this out tonight. I configured two nodes, one binding to all network interfaces, which includes the onioncat address (node A), and the other only to the onioncat address (node B). I set node B as a darknet node only and traded node references. The nodes connected over tor (both node showed the onioncat address as the node address in the friends list). I was able to download an index site very slowly on node B. Given that it only has one darknet connection it's expected that it was slow. The connection between the two nodes would oscillate between Connected, Backed off, and disconnected. The times shown in the address area in the friend connection data were "(1701ms / 5361ms)". Upping the ping time out configuration settings reduced this oscillation but I got ForwardRejectedOverloads instead. An IRC discussion with nextgens and bertm suggested tweaking these settings was a very bad idea for performance and stability of the network. Some things to note while doing this: 1) I hit bug 6879 - https://bugs.freenetproject.org/view.php?id=6879 2) I hit bug 6862 because I used openjdk 9 - https://bugs.freenetproject.org/view.php?id=6862 3) Even though node B only used bindTo to a specific IPv6 address the physical.udp in the noderef enumerated all network interfaces. This exposed the nodes actual IP address if it had one. 4) I had to change wrapper.conf to change -Djava.net.preferIPv4Stack=true to -Djava.net.preferIPv4Stack=false otherwise I couldn't bind to IPv6 addresses. Some of the issues might be due to only being a 1 darknet connection - I didn't have the time to set up more nodes. What's a minimum number of darknet connections for reasonable performance? -- http://bluishcoder.co.nz _______________________________________________ Devl mailing list [email protected] https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl -- Ian Clarke Stacks - The AI CFO for your personal finances http://trystacks.com/ _______________________________________________ Devl mailing list [email protected] https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
