asciilifeform: hm i must be thick.. what was the orig complaint about
asciilifeform: mircea_popescu: i meant re the orig stolen trilema piece
asciilifeform: meanwhile, in 'holy fuck, i thought we hit rock bottom but then heard a knock from below' lulz, http://www.loper-os.org/pub/mcleod/index.html << 'professor of engineering' tries to argue in favour of old-fashioned , carmichaelizable primality litmus ~instead~ of m-r, on acct of the 'rarity of carmichael #s' .☟︎
asciilifeform: ( how does copyrasty work in europistan anyway... does it actually make for a claim, or do you gotta be disney, like in usa..? )
asciilifeform: mircea_popescu: did the fraud complaint result in anyffin but the pictured lulgem?
asciilifeform: well-documented (e.g. jefferson lost his shirt )
asciilifeform: mircea_popescu: these folx so stupid they dun even know how to mit-phuctor ??
asciilifeform: this is almost as hilarious as the 'mathematical' conferences that happily ate perl-generated 'paper' etc
asciilifeform: not seen, but iirc mircea_popescu actually mentioned this item elsewhere
asciilifeform tuned out of subj when realized that can't point to a distinguishing algo between gavin, hearn, wright, et al -- they're idempotent , like electron..
asciilifeform: tru. who, incidentally, is the current gavin? i admit, lost track, dun follow subj actively☟︎
asciilifeform: iirc they already tried ~that~ ( gavin ? )
asciilifeform: recall that huawei was usg.fatwa'd for konsoomer pnoje also ( presumably logic is, 'ohnoez, even tards no longer want to buy ipnoje with 0 buttons and $2k price -- wat do? let's try banning all competition' )
asciilifeform: ^ ftr, the added cost of the piecewise-defined (correct for all possible inputs of N and Witness) , vs. yesterday's 'naked' variant, is <0.001s (on std test iron.)
asciilifeform: incidentally, m-r liars seem to be closed under multiplication, but i have not proven.
asciilifeform: ( would, if riemann were proven, conserve rng bits, tho , could use fewer of'em )
asciilifeform: in so far as anyone can tell, it's troo, and then 64bit witnesses suffice for 4096bit candidates. but not only relies on riemann, but dun win anyffin in ffa, where very small and very large number eat same cpu.
asciilifeform: mircea_popescu: it's that bach conjecture thing.
asciilifeform: afaik density of 'liars' is uniform across numberline.
asciilifeform: mircea_popescu: in that can check result with pen ? or canya think of some other reason..?
asciilifeform: this type of test is impossible on systems where m-r eats witness straight from rng, without possibility to override by hand.
asciilifeform: ( without having to use a microscope or whatnot )
asciilifeform: when we sit the thing down on a microcontroller with mask rom, or some other similar iron, it will be important to be able to spot-check the m-r and determine that for some input known only to you, it actually behaves as m-r.
asciilifeform: closest thing there is to a 'jesus bolt' in the thing, arguably.
asciilifeform: mircea_popescu et al : plox to lemme know if ~any~ part of this is unclear, cuz this is rather important moving part
asciilifeform: one oughta be able to feed any valid, per the theorem, witness, up to n - 2 , and get the expected output.
asciilifeform: per the proof, it is seemingly harmless ( a carmichael number has 1/4 of the integers as 'liars' ) but what it does is to prevent simple manual test with small numbers , which is imho quite typically kochian
asciilifeform inclined to reject koch's optimization ( which diana_coman retained ) where witness consists of rng(bitness_of_n - 2) , and actually make witness equal to rng(width) mod (n - 2) for full range
asciilifeform: this means that the gcd pre-litmus is an unambiguous win ( and you also dun leak anyffin if you terminate after X flunks the pre-litmus and you go an' get a new X)
asciilifeform: ( as discussed in detail in '17 thrd )
asciilifeform: naturally each new X is pumped from FG, rather than kochian increment or any such thing
asciilifeform: ( for n00bz, m-r is montecarlistic for probable-primality but deterministic for compositeness, i.e. if outputs 'composite' then you have in fact a composite )
asciilifeform: must point out tho that you dun leak anyffin if you terminate the m-rism for X after first failed shot.
asciilifeform: there's precedent -- e.g. the fella who tried to plug brezhnev, actually got out when su ran out -- after doing 30yrs in phonebooth-sized oubliette
asciilifeform: the mamies win their appeals cuz there's 100mil of'em and system aint set up to send ~all~ of'em downriver, there aint enuff clinks to house'em all. whereas 'terrorists' in the 'supermax' there's a coupla dozen of, and i've yet to heard of 1 being cut loose on appeal
