00:11 MasterDuke left 00:51 MasterDuke joined 02:01 MasterDuke left 02:17 MasterDuke joined 03:34 AlexDaniel left, AlexDaniel joined
samcv so i did some work over a few days. and i wasn't able to do a meet in the middle attack on our current hash implementation. due to how it has three state variables rather than just 1 in most hashing systems 04:06
the space of three uint32's is too great to be able to attack it easily this way it appears to me 04:08
i may still switch to siphash since it's been studied more though, but the need isn't very pressing
and i think they're about the same speed as each other in the testing i had previously done 04:10
though i may be able to speed up what we're doing now a bit 04:11
and i would like a 64 bit hash value 04:17
04:17 AlexDaniel left, AlexDaniel joined
samcv I got this with asan gist.github.com/samcv/4d12bb5e80a6...4dd6efb541 05:55
in t/concurrency/01-thread.t. it's a flapper 05:56
actually it only happened once :\ 06:03
even though i tried runnig it in a loop
06:23 AlexDaniel`` left, wictory[m] left 06:24 ilmari[m] left 06:26 benchable6 left, greppable6 left, squashable6 left, bisectable6 left, releasable6 left, releasable6 joined, squashable6 joined, greppable6 joined, bisectable6 joined, benchable6 joined 06:31 lizmat left 06:54 lizmat joined 07:05 MasterDuke left 07:35 stmuk_ joined 07:36 ilmari[m] joined 07:54 domidumont joined 08:01 domidumont left 08:02 domidumont joined 08:12 wictory[m] joined, AlexDaniel`` joined
lizmat do we have a writeup on MoarVM's garbage collection somewhere I can refer to in an article? 10:05
10:53 MasterDuke joined 11:50 domidumont left 14:59 lizmat left
japhb samcv: www.schneier.com/blog/archives/201...s_law.html seems relevant. :-) 15:10
15:11 Ven`` joined 15:16 MasterDuke left 15:25 Ven`` left 15:26 Ven`` joined 15:46 domidumont joined 15:54 domidumont left 16:10 Ven`` left 16:11 Ven`` joined 17:18 domidumont joined
samcv japhb: yep very true. switching seems good just because i haven't found any security research on it. though it doesn't appear Meet in the middle vulnerable. there may be other ways to exploit it :P 17:47
or it is, though i fairly suspect it's not given it has three but not two state variables 17:48
since you need all three 32 bit uints to match up to do it properly. just not likely enough to make it feasible and i think slower than a brute force attack 17:49
18:11 Ven`` left 18:15 Ven`` joined 18:17 domidumont left 18:27 Ven`` left 18:28 Ven`` joined 19:17 MasterDuke joined
samcv siphash is slower though 19:28
19:28 Ven`` left 19:29 zakharyas joined 21:07 MasterDuke left 21:10 lizmat joined 21:29 ggoebel left 21:45 ggoebel joined 22:14 dalek left 22:29 zakharyas left 22:57 Ven`` joined 23:09 Ven`` left 23:27 synopsebot left