On Sun, Jun 18, 2006 at 11:06:00AM +0100, Dave Baker wrote: > Firstly, I don't profess to be an expert on this subject, and I haven't been > following the discussion up to this point, but I've read the summary, and one > thing leapt out... > > > - Load is propagated back to the source. If node A is connected to node > > B to node C, and A floods, then part of the flood will go through B to > > C. C will then reject some requests pre-emptively and may have some > > timeouts. This will result in B reducing its send rate to C. That will > > result in more requests to B being rejected, which should result in A > > reducing its rate of sending requests. But even if A does not, B will > > refuse almost all of A's requests. > > Is it right that due to A's misbehavior, the connection between B and C is > treated as overloaded? Likewise, if A sends a request to B, who forwards it > to C, but it times out between B and C, the connection between B and C will > be considered overloaded (right?) But will the perfectly healthy connection > between A and B also be affected?
Only to the extent that the requests would have been routed to C. As I said, no solution is going in to the source that doesn't propagate load back to source. Otherwise we are worse off than we are now, and back into the 0.5 days of universal overload and data simply flowing to the least overloaded node. > > Dave -- 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/devl/attachments/20060620/be00ef09/attachment.pgp>
