On Friday 19 September 2003 12:21 pm, Ian Clarke wrote: > Standardization is almost always easier said than done, and this really > overlooks major differences between the theory behind these network's > operations.
Mnet's, Enthropy's, and Freenet's goals are essentially the same. Gunnet and Grapevine differ from Freenet's goals only in that they think it is too hard to do, and so they are willing to make compromises to get something done now. > Freenet uses a heuristic routing model - if you want a standardized way > to build a heuristic P2P network with Freenet's qualities then > congratulations, you already have one, its called Freenet! If you > expect to find a common standard which encompasses heuristic and > non-heuristic P2P networks with a variety of different goals then > congratulations too, it already exists, its called TCP/IP. I was thinking more along the lines of Freenet there. IE: produce a document that is like 'here is how Freenet will ideally operate' (years form now). > > There are a lot of things that are easy to reach a general consensis on. > > (Encryption / routing etc.) > > Erm, no, encryption and routing are far from easy to reach a general > consensus on since the encryption requirements will vary depending on > the system's goals, and if everyone used the same routing algorithm then > everyone would basically be building the same P2P network. The routing > algorithm is the most distinguishing feature of any P2P network. I think that all the projects that have the same ideals as Freenet can agree that the best way to route in the way that gets the data as quickly as possible while remaining anonymous to all parties. > > However it might be nice to have files that did not need to be part of > > the data store, like in GnuNet. > > You clearly have no clue about how Freenet's routing algorithm works if > you don't understand why this wouldn't work for us. These P2P networks > work in different ways because they have different goals and > requirements. JXTA started with the vague goal of standardizing P2P > networks and look what became of it. I understand EXACTLY how Freenet works. I wrote a proposal on how this can be done and submitted it to this list SEVERAL times. (As recently as 2 days ago) http://www.mail-archive.com/[EMAIL PROTECTED]/msg01235.html http://www.mail-archive.com/[EMAIL PROTECTED]/msg01239.html Nobody has ever responded to any of my proposals, or given any indication that they would not work. So I'll continue to assume that it is possible, but nobody is willing to volunteer to write it. Please tell me if you have any thoughts on this. I know nothing of JXTA. > > However the hard part would be comming up with a standard for > > communications and metadata. Ideally the protocol would not care about > > firewalls / nats and utilize the underlying network as best as possible. > > I understand Jrand0m has some ideas in this aria. One would probably want > > to include some sort of simplistic search and flooding resistance, > > borrowed form GnuNet, and TUKs included in the metadata ala Grapevine. > > Dont forget the perpectual motion machine, it would be great if we had a > perpectual motion machine! It's pretty simple to start with; Data should be stored in CHKs. (I assume using Freenet's/Enthropy's algorithm) Meta Data is stored in an SSK which has a manifest and a TUK (which is how updates are done.) Routing is done biased on the hash. The best routing mechanism is the fastest one that is anonymous for all parties. Communication should be done in the Fastest way that allows all people to connect to the network, and it as difficult as possible to block. That's easy to agree upon, we don't need to come up with the best possible way to do all these things, just a specification that all the projects can work toward. > > If someone could even produce a preliminary outline of what this should > > look like and submit it to the other projects, I'm sure they would > > cooperate and try to come up with improvements. The important thing to > > remember is this is not just 'Here's how we do it', but rather an ideal > > network, even if it contains elements that you don't know how to > > implement yet. > > Keep dreaming :-) > > Ian. _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
