One more detail: HTL should be reset, or decremented, when we receive the message, NOT when we send the request to another node.
It might be argued that RNF should contain a nearestLoc as well as an HTL. I don't think this is profitable, on the grounds that if it does find a closer node, it is probably a false minima - a node which is not in the right place and randomly ended up with a location close to this request's target. Loop protection means that not including a nearest location on RNFs does not necessarily cause lots of extra hops. If we get a loop, we decrement the HTL. If we get an RNF, the HTL will be min(HTL-1, RNF.HTL). On Fri, Sep 23, 2005 at 02:58:51PM +0100, Matthew Toseland wrote: > Ian and I have agreed to a further change to the HTL mechanism in 0.7, > to solve a few problems: > > HTL starts off at the maximum value. > A request carries the value of the closest location so far seen to the key. > When we find a node which is closer than this, we update the location on > the request, *and we reset the HTL to the maximum value*. > If we go through a node which is not closer than the closest-so-far, we > decrement the HTL. > If the HTL reaches 0, we send back a DataNotFound. > If we cannot find any more nodes, we send back a RouteNotFound. > > Max HTL can then be as low as 5. > > There remain probabilistic decrement at maximum and minimum values of > HTL. > > For a subscription request, if we get a DNF and we have reset the > closest location, we know we are the closest location to the key, within > max HTL hops, therefore we are the root. > -- > Matthew J Toseland - toad at amphibian.dyndns.org > Freenet Project Official Codemonkey - http://freenetproject.org/ > ICTHUS - Nothing is impossible. Our Boss says so. > _______________________________________________ > Tech mailing list > Tech at freenetproject.org > http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/tech -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20050927/8f245b13/attachment.pgp>
