|
"I *think* he's
saying that if we fake-route it to another node, then the datasource of that node would have fake-routed it to another node" How do you figger that?
I may have misunderstood the workings of the
estimators/nodes, but certainly it can't be impossible (toads boss' says so ;-)
to make something that only deals with the estimate, WITHOUT actually relying it
(the request) further to another node?
I'm viewing this from the point of the node that requested
it: how can that node know, whether or not the node it contacted did have the
data and is acting if it retrieved it from another node, or *actually* retrieved
it from another node?
It is true that, if it hadn't had it in it's store,
and it had to search it on another node, that node, not knowing it's the second
in line (or can it, with the htl?), would also fake it, if it had it in it's
store.
It's worth noting, however, that this is not an
exponential proces, but a linear one; where as it would normally take one hop
(if it was in the datastore), a node has to take two hops now. Since it's
already established most things in the network will be found within (the time
of) 6 hops, this would now be (the time of) 12. Not an unreasonable amount,
given the benefit that one could never say with any certainty that it was or
wasn't in the store of the first node contacted.
One could not say, however, as Tom claims with the fixed
time-delays, that you know it came from the first nodes' store, by looking at
the total time elapsed. Because, if you have a node with very good connections
to other exellent nodes, the time to retrieve it, even if it had to go through 4
nodes could be less then a node that had slow nodes connected to it, even if it
only had to hop two nodes further (due to the estimators). The number of
possibilties of elapsed-timeframes would be almost infinite, and even taking
averages wouldn't help much, since no1 knows what the estimators are of another
(specific) node at one given time.
|
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
