On Tuesday 04 April 2006 01:59, Matthew Toseland wrote: > On Tue, Apr 04, 2006 at 01:48:33AM +0200, Jean Krohn wrote: > > On Tuesday 04 April 2006 01:00, you wrote: > > > Cool. Do you expect Galet to scale? Do you have anything in the way of > > > routing algorithms or do you rely on broadcast search? We could really > > > do with some local and semi-local stuff e.g. instant messaging (as an > > > extra incentive for people to add their friends, for example). How do > > > you do tunnelling? How do you do search? > > > > > > The big thing we have is the routing algorithm IMHO. Maybe our > > > architectures are complementary. > > > > Yes, fortunately it seems that, though our aim is the same, we have > > started at opposite ends of the problem :-). > > ATM, Galet is mostly local (the horizon is 1), so there is no public > > network on top of it yet, and no routing algorithm or search beyond the > > immediate neighbours, nor scaling problems. > > That's rather what I had hoped. > > > It has an invitation system based on cryptographic tokens to be > > transmitted through offline means (can be e-mail or whatever). The token > > consists of the hash of the nodes public key (which serves as the nodes > > identity) and a shared secret (which serves for the first connection > > only). > > Right. We are going to need to implement some sort of invites system > soon. It would be very useful to compare notes, although the low level > details may be somewhat different. How do you deal with this at a high > level? You just send them a file?
ATM, yes, it is a small XML file with base64-encoded data (oops I forgot, the token also contains the nodes address). Though a system of presentation is planned (e.g. Bob acts as a match maker for Alice and Charlie). This would be very usefull, as people tend to know each other in groups (if Alice knows Bob, there is a good chance she also knows Charlie). > > The key exchange for normal connections is station-to-station, which > > provides perfect forward secrecy. > > Right. We have an ephemeral Diffie-Hellman at present, but we do have > pubkeys, and will soon implement STS. Eventually we may use JFK, but > that's a bit of a monster. (It's VERY nice though). > > > Galet (in SVN) has semi-local IM, as in, if Alice knows Bob who knows > > Charlie, the three of them can chat together, Bob acting as a relay for > > the Alice and Charlie's messages. It also has pretty standard local > > file-sharing (resuming, priorities...). > > So 2 hops maximum? I tried to code something like this a while ago... it > was a pig... all the error cases and so on... But that was with routing, > and assuming a big network... No, it works recursively, so Charlie can invite Daniel to join in, and so forth... > Local filesharing has been suggested for Freenet, I'm not sure exactly > how it would work, how to integrate it in a clean way, etc. Sharing > bookmarks (known freenet web sites) is planned, at least. I do not know for Freenet, but with Galet it was pretty straight forward (index files, send index, send file chunks when asked, search on the indexes received). > > For the tunneling, Alice configures a tunnel in the GUI, by specifying > > a local port on her machine and a remote host and port to which Bobs > > machine will forward the data to (like the ssh -L option). Obviously, > > tunneling is disabled by default (Bob must enable it). > > Okay so one hop tunnels then? Yes. > > > Oh and Freenet is in Java, and always will be in Java, at least until > > > 1.0. :| (the language is fine, it's the runtime that's such as > > > *******). Swing or SWT aren't that far from QT, but anyway, client apps > > > can be in any language... > > > > I do not mind Java at all (a preliminary version of Galet was in Java), > > but compilation under Windows is not that bad when you get used to it :) > > and you do not need any extra package installed (JVM), while the > > application feels more 'native' to the user. > > And unfortunately I am not familiar with Swing nor SWT... > > > > > What core stuff do you think we are missing? > > > > I am not familiar with Freenet 0.7, so I cannot tell. > > Okay. You should probably try it sometime. I will :) > > > Have you read the papers on our routing algorithm? > > > > If the routing is the same as of old (1-2 years back) then yes. > > Am I right to think that 0.7 is the same routing algorithm than before, > > but on top of a 'darknet' ? > > No. Completely new routing algorithm, somewhat DHT-like. Ok, I will read it.
