Ed Tomlinson wrote:
Hi
On the otherhand, maybe we want to just eliminate QR and use something
like this. Nodes have 100ms to reply with a Query Accepted otherwise the
query is deemed to have been rejected. NG should be more that able to handle nodes that abuse the trust implied by the above scheme. Its major
advantage would be a massive reduction in traffic. Consider now where
we are getting 3-5 QRs per hop. This implies a request at HTL 25 will
generate 4x25 QRs. Even if we manage to reduce this to 1 QR per hop
we still get 25x the QRs vs DNFs or Datafound...
But NGR already sets a cost to QRs. This proposal does nothing more than add up 100ms to tSearchFailure when the failure is due to QR. Certainly this would cause NGR to avoid nodes which QR a bit more, but it's very arbitrary. How do you know whether 100ms will achieve the goal of load balancing ("to have a node receive as many requests as it can handle but no more")? Why not 50 or 200ms?
I think the solution has to be adding information to the QR so that requesting nodes *know* when they should back-off and when they can come back.
-Martin
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
