----- Original Message ----- From: "Nick Tarleton" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Tuesday, August 05, 2003 1:16 AM Subject: Re: [Tech] freenet not suited for sharing large data
> On Sunday 03 August 2003 05:49 pm, Gabriel K wrote: > > Well, I think a protocol should guarantee to find the data, much like > > ACHORD. > AFAIK, THIS IS NOT POSSIBLE without sending a request through the entire > network and placing ridiculous total load on it. Well, in freeNet maybe. If you have followed this thread you might have seen I have answered this question already. Also, I don't really see the point of making a network so large that you retrieve all it's data because of HTL. Better to just make smaller networks. Nodes may be part of several networks of course, so that data can spread among them. I think it's better to send the N nodes the request. But not directly, because that would not be good for anonymity. See my PDF as for how my suggestion is. http://www.student.nada.kth.se/~m98_khl/O-net.pdf > > So it seems to me that the original FreeNet lacks features that they are > > trying to put there now...? > > Such as shorter proxy paths, and optional upload of data? :) > There is something called 'path shortening' that may be what you want - I'm > not sure. Optional upload is already there: just insert with HTL 0, and it'll > stay on your node until someone requests it. Of course, there will be some > trouble _finding_ it. Yes, someone already told me about path shortening. > I do have an idea for a better optional-upload system that, combined with > NGRouting, would work well. It's really a specialization of Frost's system. > The uploader inserts a small announcement for the file, containing basic > information and its CHK, at a high HTL. The requesting client gets that, and > then tries to retrieve the file a couple times. If that fails, it then > inserts a small KSK (with a name containing the CHK of the file). The > uploader is constantly requesting this KSK. When it's recieved, the file is > inserted. > NGRouting's high-probability-of-finding would be required for it to work well, > obviously. Hmm.. I think someone has the same idea.. see previous posts in the thread.. it wasn't you was it? In that case you are repeating yourself :P Anyway, I think a good approach for a protocol would be to sort keys into the network, like chord perhaps since it's O(log n) to do a search. These keys point to the data. In my model (in the PDF) these keys would point to the ring where the data is, and the public key of the holder. That way, no request needs to be sent to all nodes. However, each node will know of more IPs in the network. The pdf does not describe this idea, because I haven't had time to think about it, if I can use ACHORD for sorting keys in MY topology. But I think it's possible because even if my topology is several rings, they can also be seen as one twisted big ring. I will have to work on that on my spare time :) /Gabriel _______________________________________________ Tech mailing list [EMAIL PROTECTED] http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/tech
