> a lookup chain has the chance to merge with a table chain (8-Rs) * 2^12 times, > where Rs is the starting round. When Rs == 0, then the probability to merge > is 2^32.5 (table height) * 2^12 * 2^3 = 2^47.5. > The 2^47.5 is not the expected 60% of 2^52 and the factor of 2^3.5 can be > attributed > to the affinity of A5/1 to produce merges when being clocked with the > majority rule. > the 2^3.5 fits well with the keyspace reduction of the 100 extra clocking > optimization. > > Each end point on disk is the thin end of a funnel and you can only stack > them on their > wide end (which represents the rest of the chain), thats why the density of > the end points is > at 60% of the maximum when only 2^32.5 of 2^52 possible end points are stored > in the table. >
Ok, the calculation of the merge probability is mathematics from the realm of fantasy and was actually an attempt to see if everyone is awake (in retrospect). But the main idea still holds. the density of the end values must be much lower than the merge ratio, because each end point represents 2^15 values that are opportunity for a merge. maybe someone can do the math correctly. _______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
