On Mon, Sep 01, 2003 at 09:01:26AM -0500, Edgar Friendly wrote: > Toad <[EMAIL PROTECTED]> writes: > > > What is the purpose of QueryRejected.hopsToLive ? > > > It's an attempt to have the network recover from part of it having bad > routing in a pretty optimal way. I.e. using exactly the HTL indicated > in the initial request (modulo the fuzziness in HTL decrementing at > the end of the request). > > > Surely this sort of decision should be taken by the node sending the > > request? The current code reduces the HTL to > > min(hopsToLive-1, qr.hopsToLive). > > > I agree it seems like there's a breach of our usual tactic of not > trusting other nodes. OTOH, if the node were evil, it would be better > to DNF the request instead of QR it. Or even more evil, repeatedly > QueryRestarted it, and just stall forever.
DNFing the request, even after a long delay, will bias NGRouting against routing to the node. Repeatedly QueryRestarted'ing it... hrrm. Interesting idea. We don't reduce the timeout each time? Surely that allows a cancer node to do some pretty evil things... yes, I know inserts are vulnerable to cancer nodes, we will fix that, later. > > > Is there any reason not to get rid of this? If we need to reduce the HTL > > much, we should do it in the requester node, not the rejector node. And > > NGRouting should eliminate most load balancing problems anyway - this > > looks like it was added in the middle of the load-panic era. > > If it's any amount of work to get rid of it, I say that it's better to > spend elsewhere. If it's already done, I don't have any worries > either. It'll increase the amount of work a broken network will do if > the requesting node underestimates the proper HTL decrement, but > hopefully the network is going to be able to eliminate QRs anyway. It's a potential attack. NGRouting does not know about the reductions in HTL due to QRs, nor should it. If we need to reduce it significantly, that should be done in the node, not by the node we sent it to, which has demonstrated that it is of little value by QueryRejecting it. > > Thelema -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so.
pgp00000.pgp
Description: PGP signature
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
