On 2012/08/02 (Aug), at 3:58 AM, Matthew Toseland wrote:
> On Wednesday 01 Aug 2012 17:44:10 Robert Hailey wrote:
>>
>> On 2012/08/01 (Aug), at 11:35 AM, Matthew Toseland wrote:
>>
>>> Basically they let us gather statistics on the network so we can find out
>>> why routing etc aren't working as well as they should be.
>>
>> Perhaps now would be a good time to remind those interested that the current
>> announcement algorithm (shallow-first rather deep-first [linearly, not as it
>> relates to graph theory]) is theorized to have a negative effect on routing
>> (as seed nodes are pulled away from the network by the bootstrapping nodes).
>
> Is that theory testable? If not it's not much use.
It is testable, and can likely be simulated too. Not sure what to make of the
"not much use" part....
For example:
(1) start with the seed node simulator (already written)
(2) turn off caching of fetch data & re-bootstrapping (& path folding?)
(3) wait for network to stabilize
(4) insert a CHK
(5) fetch CHK, should succeed
while (true) {
(6) select a location (either randomly or in a sweeping pattern 0.0-1.0)
(7) create a new test node using the above location
(8) bootstrap the test node onto the network using the test seed node
(9) sleep for a few seconds?
(10) fetch CHK (eventually will fail)
}
In the "best" case, I suspect that one could tune the simulator to fail having
bootstrapped even number of peers that the seed node has.... or actually, even
"better" would be inserting leafs (nodes with a peer limit of one), which could
even be considered a direct attack on network topology.
--
Robert Hailey
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20120802/68a3567c/attachment.html>
