On 19/10/15 15:07, Florent Daigniere wrote: > On Mon, 2015-10-19 at 14:11 +0100, Matthew Toseland wrote: >> I still think you should separate optimising routing (trees of all >> locations etc, long's instead of doubles) from changing behaviour >> (ignoring FOAF peers if they were also the peers of a node we've >> already >> visited). > That's what I'm trying to do here... I haven't done the last level of > optimization where we combine all of them into a single structure > (instead of iterating on each peer). It's still O(n)+ > > Here's how it's wired-in > https://github.com/nextgens/fred/blob/optimize-closerpeer/src/freenet/n > ode/PeerManager.java#L1054 > > As you can see, the logic is preserved, with the exception of how we > handle what's in the exclude list... What I'm wondering is whether that > new behaviour is fine or not. Do we need to compare what's in the > exclude list modulo Double.MIN_VALUE (old behaviour) or is an exact > match fine (new behaviour)? Double.MIN_VALUE is only really needed when we're doing arithmetic on doubles to get new locations. I don't think we are here. It should be fine?
Granted we do pass them around - in messages as doubles, in node references as text, so maybe???? > If it's not I just need more code to make it mirror the old one. >> Does our FOAF locations-of-peers announcement include nodes >> which are backed off for long periods? Nodes which are disconnected? > It's complicated. There are several checks before and after; I'm not > sure what the behaviour should be. The new code needs to mirror that - that's a bigger deal IIRC.
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Devl mailing list Devl@freenetproject.org https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl