On Sunday 28 Jul 2013 04:51:22 Robert Hailey wrote: > > On 2013/07/27 (Jul), at 2:03 PM, Matthew Toseland wrote: > > > You've just reinvented NGRouting. :-) > > I've heard that before. I had no idea that there were so many fun ways to > implement NGRouting!!! ;-) > > Seriously, though... at worst this is can be examined as two of "the > same"/"current algorithm" networks with half the nodes, and I happen to know > that: currentAlgorithim/2 != NGRouting > > Hmm... I guess that's not accurate, actually..... because a slow peer (e.g. > on a modem) might see a peer as fast that another peer (say, on an OCH3 line) > sees as slow... so there is certainly a non-trivial routing effect, and not a > sharp "two sets of nodes" distinction. > > > what determines performance is really... other things like... whether they > > find the data at all > > So that answers my first question... I see that my overall remote CHK success > rate is at 15.6%; I presume this should (theoretically) approach 100%, so do > you think that is closer to the root cause of the performance issue?
No, it shouldn't. Inevitably some content is lost, and the overall stats are skewed by being a mix of high htl and low htl answers (which go more hops so count on more nodes, skewing the success rates). The per-HTL stats are much more interesting, and at high HTL are much higher too.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Devl mailing list [email protected] https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
