Uh, yes the idea is that NGR does work, but not while
the network is overloaded.  To reduce load, your
options are limited to: adding more resources,
improving routing and reducing queries.  My proposal
addresses reducing queries; which is what exponential
backing off is supposed to do but does not seem to be
working.  I think the current backoff method hurts
routing because it totaly closes down more optimal
paths (because optimal paths get more traffic) while
my proposal would throttle queries based on HTL (some
could still get through and so the node is not totaly
blacked out and routing can still gather data from
this route).  Lower HTL queries do take less time to
complete, IF they started out lower.  The assumption
with MinHTL is that the longer a key is on the network
the less likely you are to find it, all other things
being equal.  Also, MinHTL has the effect of reducing
the HTL of all queries without being biased based on
the original HTL.  Doing something like lowering
MaxHTL instead would prematurely kill new queries
while giving priority to older queries that are less
likely to be found (this would probably hose the
network even more, if that's possible).  I'd be very
surprised you could provide some evidence that the
failure table isn't working.  It's hard to imagine how
it could NOT be helping.  I'm sure the developers
disagree with you about this, why else would Toad
advocate expanding the failure table?

--- Edgar Friendly <[EMAIL PROTECTED]> wrote:
> Ed Tomlinson <[EMAIL PROTECTED]> writes:
> 
> > Hi,
> > 
> > This is an interesting idea...
> > 
> > Ed
> > 
> The problem with MinHTL is that if everyone is
> overloaded (as the case
> seems to be), the request will only be able to
> travel (MaxHTL-MinHTL)
> hops into the network (once its HTL hits the average
> loaded MinHTL of
> the network, it won't be able to go any farther.) 
> If anything, one
> wants to prefer lower HTL requests because those are
> going to take
> less time to complete (one way or the other). 
> 
> As far as higher HTL requests being good for feeding
> the failure
> table, there's not much evidence of the failure
> table being that
> effective, for some reason.
> 
> Thelema
> -- 
> E-mail: [EMAIL PROTECTED]                          
>  Raabu and Piisu
> GPG 1024D/36352AAB fpr:756D F615 B4F3 BFFC 02C7 
> 84B7 D8D7 6ECE 3635 2AAB
> _______________________________________________
> Devl mailing list
> [EMAIL PROTECTED]
>
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl


__________________________________
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to