On Fri, Nov 21, 2003 at 02:25:08PM -0600, Tom Kaitchuck wrote: > On Tuesday 18 November 2003 04:36 pm, Tracy R Reed wrote: > > On Tue, Nov 18, 2003 at 09:18:51PM +0000, Jonathan Howard spake thusly: > > > I think freenet is suffering because it doesn't have key space > > > specialisation. NGR is trying to route quickly, all un-overload are > > > > Yep. > > > > > speed. Don't know how to kick start it out of the current state though. > > > The only other way I can think of at the moment is nodes having > > > (semi)strict self key space specialisation. > > > > I did this to my ds and it has slowly but surely been un-specializing. It > > does no good if other nodes don't learn your nodes specialization and > > route to it. I suggest we stop routing data to nodes in desperation. If > > there isn't a pretty good probability that the node will have it or if the > > data is not clearly in the nodes specialization don't route to it. I would > > rather see a DNF than force some other node to de-specialize by forcing it > > to look for my data. > > How are you measuring specialization? If you are looking at your data store > than I would say you never were specialized. However ether way, you aren't > now. So, yes, we should do whatever we can to encourage this. IE: > non-alchemistic caching. non-alchemistic Query Rejection etc. > > I'm going to go out on a limb and claim that Freenet is not suffering from > Load balancing problems. It is suffering for Load problems. If all nodes are > overloaded, then no amount of QRing or backing off will solve the problem. It > is entirely possible that clients are using more bandwidth than the network > has. In order to fix this we need to address the client software. It is
Yeah, because nobody will ever use Freenet except in the ways we want, and the suggestion that somebody might release a killer app after freenet 1.0 that just happened to abuse the network slightly is ridiculous. > commonly stated that most of the bandwidth is used by apps like Frost, and it > seems reasonable. The problem with these apps is that they try to request a > key at periodically regardless of how long the previous attempt took. What's > more mos of their requests are not in the network to begin with. So to fix Theoretically, the failure table should sort this out. > this short term, let's place a limit of local pending requests. Mid term Fix > Frost so it knows when enough is enough. Then implement TUKs so they don't Why locally? If anywhere, surely it should be done at the node level - people will just tell their nodes to accept everything they throw at it. > need to request useless keys. Long term, implement pre-mix routing so that we I still don't believe that TUKs are the answer to Frost. Passive requests, maybe. But either TUKs or passive requests are a long way off - and they both require WORKING routing. > can have Manifests that have a node ref that you can connect to, so we can > have a limited point-to-point communication over Freenet. This will eliminate > the need for things like attempts at broadcast messages to a person. (IE: NIM > and FMB). > > Sound Good? -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so.
signature.asc
Description: Digital signature
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
