Show Idle (>14 d.) Chans


← 2017-11-12 | 2017-11-14 →
a111: Logged on 2017-11-13 04:46 mats: "And we are pushing robots to the limit in terms of the speed that they can operate at, and asking our suppliers to make robots go way faster, and they are shocked because nobody has ever asked them that question. It's like if you can see the robot move, it's too slow. We should be caring about air friction like things moving so fast. You should need a strobe light to see it. And that's incredibly critical to CapEx
asciilifeform: the 'tecan evo' i worked with ( small industrial gantry bot ) could make fully 100x faster motions than vendor recommended . BUT -- 1) the default winblowz turdware ( that afaik every installation other than mine, used , lock stock etc ) would result eventually in smashed equipment, dead bearings, and possibly injured meat and ..
asciilifeform: 2) vibration
asciilifeform: just because thing can 'move faster than you can see' dun mean that it stops on a dime, or that it won't oscillate ( ever see rifle barrel on high speed camera? ) , ring like a bell, destroy whatever illusion of accuracy.
asciilifeform: this is an entire field, i cannot begin to sum it up here. but even in 19th c machinists understood 'fast cut xor accurate cut'
asciilifeform: as for 'critical to capex', vey rarely is robot the chokepoint in the line.
asciilifeform: if actually care about capex -- speed up, or better yet, eliminate -- the ~meat~ in your line.
BingoBoingo: Or parallelize lines
BingoBoingo: If Hu cares to
asciilifeform: 'tesla' of course is not a commercial op in the usual sense, but rather an elaborate fraud prop, similar to a konsoomer-flavoured lockheedmartin
asciilifeform: for milking usg
asciilifeform: so it is quite in keeping with this, for it to , say, pioneer 'robot works in vacuum, and moves without hindrance of air resistance, ReallyFast!' etc
ag3nt_zer0: pete_dushenski http://www.contravex.com/2017/11/12/bigs-vancouver-house-the-last-pied-a-terre-youll-ever-need/#identifier_6_19987 <<< thanks for the introduction to Ingels. My family is from a small town near Wright's Taliesin and a good old family friend of mine owned a classic FLW student-designed house that sat on 40 acres above Lake Wisconsin - I practically lived in that house for 3 seasons one year and there is really nothin
ag3nt_zer0: Never been in a place that has equaled that in regards to my definition of a warm home... looking forward to learning more of Ingels
mircea_popescu: in other news : it was established in teh minigame torture rooms that in point of fact 4096 bit keys contain only 4090 bits of entropy at the very most (minus whatever koch-gpg manages to shave off in other ways).
mircea_popescu: the reason is that (in a translation of what koch-gpg does into sanity) you take 2045 bits of rng for each possible prime, stick 11 in front and 1 in the tail and THAT is your 2048 bit prime candidate.
mircea_popescu: the reason you stick the 1 in the tail is to ensure odd numbers -- large even numbers are never prime. this much is a math-forced reduction.
mircea_popescu: the reason you stick the first 1 in the front is, evidently, to not end up with sub-4096 Ns
mircea_popescu: the reason you stick the 2nd 1 in front is, not evidently, also to not end up with sub-4096 Ns : if you had the exceptional case of your primes being each 2^2047 + 1 your N would then be 2^4094+2^2048+1, which is shorter than 4096 bits.
mircea_popescu: this is evidently a "loss" of entropy, in the sense that what is advertised (4096) differs from what is actually delivered (no more than 4090). i am of a good mind to start calling them 4090 bit keys tbh.
diana_coman: p and q are different so there won't be exactly this limit case there, but obv same 4095 bits n instead of 4096 for other low-enough odd numbers that might be primes
mircea_popescu: diana_coman never, because 11 * 11 = 1001.
mircea_popescu: anyway, back to rsa discussion : there's about 6.5e612 primes in the interval 2^2045-1, 0 (by teh prime number theorem). every key needs a pair of these, and no number can EVER be repeated (if it is -- phuctor breaks both keys).
mircea_popescu: the chances of such repeating happening naturally'd be ~the inverse of ∏(1 - 2i/6.5e612), 1<i<n where n is the number of keys ever made. fortunately this evaluates to "never" on all extant iron.
mircea_popescu: this is not "true for very many keys like a billion trillion keys". this is true all the way up, by the time one's made 10^609 keys we're starting to get into five-nines assurance of unicity.
shinohai: Johnson, clear my schedule for the day.
shinohai: Go ahead and clear Tuesday too.
mircea_popescu: wait, your johnson makes your schedule ?
shinohai: of course, im the baron titsbare
mod6: haha
mod6: damn, nice loft
mod6: <+mircea_popescu> in other news : it was established in teh minigame torture rooms that in point of fact 4096 bit keys contain only 4090 bits of entropy at the very most (minus whatever koch-gpg manages to shave off in other ways). << uugh. every time we peel a layer back...
asciilifeform: http://btcbase.org/log/2017-11-13#1737169 << iirc we had a thread , re this spectacular lul
a111: Logged on 2017-11-13 11:40 mircea_popescu: the reason you stick the first 1 in the front is, evidently, to not end up with sub-4096 Ns
asciilifeform: ( picture extending this by induction : 'why not glue another 1 after that 1 , lest we end up with sub-4095...' )
asciilifeform: http://btcbase.org/log/2017-11-13#1737171 << the only properly forced loss is of the bottom bit
a111: Logged on 2017-11-13 11:43 mircea_popescu: this is evidently a "loss" of entropy, in the sense that what is advertised (4096) differs from what is actually delivered (no more than 4090). i am of a good mind to start calling them 4090 bit keys tbh.
asciilifeform: http://btcbase.org/log/2017-11-13#1737177 << nobody's ever running out of these
a111: Logged on 2017-11-13 12:44 mircea_popescu: this is not "true for very many keys like a billion trillion keys". this is true all the way up, by the time one's made 10^609 keys we're starting to get into five-nines assurance of unicity.
asciilifeform: 'Tesla’s CEO seems to be fully unaware of why industrial robots have limits, affecting actuators, speed and precision when handling heavy parts reliably and minimal downtime. Air friction is certainly no constraint, but moments, acceleration and deceleration. '
asciilifeform: ( in re yesterday's thread where http://btcbase.org/log/2017-11-13#1737159 )
a111: Logged on 2017-11-13 05:11 asciilifeform: so it is quite in keeping with this, for it to , say, pioneer 'robot works in vacuum, and moves without hindrance of air resistance, ReallyFast!' etc
asciilifeform: re the rsa key entropy, it is possible to trivially regain the lost bottom bits' worth of entropy -- you save the discarded bits and use them later as triggers for 'take nextprime(p) instead of p' and 'take nextprime(q) instead of q' . there may be other possible algos
asciilifeform: thinkaboutit.
asciilifeform: as i see it, this circle is satisfactorily squared nao.
mod6: im not sure about that
mod6: even /if/ doing some prime selection based on 'discarded bits' would net you anything what-so-ever, is it shown, presently that they even do this?
asciilifeform: mod6: noshit koch doesn't do this
asciilifeform: but you can trivially show that using the bottom bits in this way lets you actually get 4x as many possible primes
asciilifeform: nao ideally one would have a http://btcbase.org/log/2017-11-07#1733382 i.e. 4096b of ~possible prime~ phase space
a111: Logged on 2017-11-07 16:36 asciilifeform: let's model the ideal prime-shitter. it would be an item that takes integer N , of whatever bitness, and produce the Nth prime ( or eggog if the Nth prime is bigger than the register bitness permitted. )
asciilifeform: but nobody's gonna.
asciilifeform: ( it's a 3000yr unsolved megaproblem )
asciilifeform: tho here's a somewhat barbaric method :
asciilifeform: 1) calculate what a certain b is, such that there are likely to be 2^4096 primes below 2^b-1
asciilifeform: 2) generate a random k, k < 2^b
asciilifeform: 3) if composite(k) goto 2
asciilifeform: 4) if log2(k) > 4096 goto 2
asciilifeform: 5) you have a winner: a prime selected from 2^4096 possibles.
mod6: i must be missing something
asciilifeform: actually yer not missing anything, above algo is an absurdity
asciilifeform: it will do exactly same thing as traditional one, but take 1000x as long.
mod6: when you say 'lost bottom bits' worth of entropy -- you save the discarded bits and use them later', are you talking about the highest order 2 digits, and the lowest 1, saving their original lower-order half and using that?
asciilifeform: they are the only ones you MUST set to 1 (i.e. lose the entropy of)
asciilifeform: because you gotta have odd p and q
asciilifeform: the shaving of the ~highest~ bits is an idiot kochism on the other hand,
asciilifeform: there is no legitimate reason to do it.
mod6: yeah, /me re-read and Mr. P. said they discard the higest 2 digits
asciilifeform: aha, koch does
mod6: so in your algo above, you're saying that you can work that magic with just the ~lowest~ discarded digit
asciilifeform: ignore the 5step thing
asciilifeform: but yes you can use each of the 2 discarded bottom bits to double the primespace available
mod6: sorry if this is obvious, wasn't to me.
asciilifeform: ( 1 caveat is that this is a leaking operation , theoretically )
asciilifeform bbl, teatime
BingoBoingo: !~ticker --market all
jhvh1: BingoBoingo: Bitstamp BTCUSD last: 6539.77, vol: 28531.98250185 | Bitfinex BTCUSD last: 6455.9, vol: 117240.23947972 | Kraken BTCUSD last: 6558.2, vol: 8497.52824841 | Volume-weighted last average: 6477.04655989
BingoBoingo: OMG Re-Crashing!
trinque: surely the tickerbot has been switched to bitcoin crash? what do I make of this?
BingoBoingo: Pie? Caek? Lemoncaek?
BingoBoingo: If making pies, you have to make the best, and best means round.
lobbes: Good news on archive front; archive.is d00d has agreed to add my ips to his cloudflare whitelist
trinque: nice lobbes. this means you'll not be ratelimited? or what was the problem?
lobbes: trinque, yeah. Hopefully will be able to bypass the cloudflare js/cookie challenge wall
lobbes: Very accommodating d00d indeed. I invited him here, as well, but you know how that goes
ben_vulpes: http://btcbase.org/log/2017-11-12#1736892 << some years ago, several people got together and worked through 'an incremental approach to compiler construction', one nick fitzgerald worked through it in ada: https://github.com/fitzgen/ada-scheme
a111: Logged on 2017-11-12 23:39 mircea_popescu: does an ada lisp ~even exist~ as far as anyone knows ?
ben_vulpes made some headway through lisp in small pieces, mind appropriately blown
ben_vulpes: and in ancients, dusted off mpfhf benchmarker, finished the bit-banging of inputs, fired off a run late last week that is *still hashing*
trinque: "bitch, this hash function is hard in *both* directions!"
asciilifeform: use Ada.Strings.Unbounded; << mno ben_vulpes this is ~specifically~ a Do Not Want
asciilifeform: ( it uses implicit heaptronics for everything )
ben_vulpes: suresure
a111: Logged on 2017-11-13 16:51 lobbes: Good news on archive front; archive.is d00d has agreed to add my ips to his cloudflare whitelist
asciilifeform: lobbes: does this mean that you can mirror the whole zip collection nao ?
BingoBoingo: lobbes: So modest is becoming less modest?
asciilifeform: http://btcbase.org/log/2017-11-13#1737245 << if you apply the bound we found in http://btcbase.org/log/2017-07-06#1679483 thread, and the http://btcbase.org/log/2017-08-15#1698509 trick, mphf a not-especially-slow hash
a111: Logged on 2017-11-13 17:47 ben_vulpes: and in ancients, dusted off mpfhf benchmarker, finished the bit-banging of inputs, fired off a run late last week that is *still hashing*
a111: Logged on 2017-07-06 00:26 asciilifeform: S grows by 1 or 0 bits per cycle.
a111: Logged on 2017-08-15 22:51 asciilifeform: but instead flipping a single bit that gets xored with the result every time you read from the would-have-been-flipped reg.
asciilifeform: though asciilifeform will admit to still being at a loss re what the appeal is , after these...
asciilifeform: but if you want to make a fast mphftron, for experimentation, the recipe is 1) compute upper bound of the scratch space length and preallocate. NEVER realloc 2) NEVER flip-all-the-bits, flip a 'did-we-flip' bit instead, and the latter always get xor'd with whatever bit you read from the flippablespace.
ben_vulpes: asciilifeform: absolutely, have a benchmarking in place, will be implementing those two changes and recording improvements
asciilifeform: also you don't want to cons. at. all.
phf: http://btcbase.org/log/2017-11-13#1737247 << it looks like a properly structured scheme evaluator, but it's ~explicitly~ lacking a native cons, which might be a very good exercise for whoever™ adding a static allocation space, adding mark-and-sweep, then all those To_Unbounded_String look like they can be simply search/replaced
a111: Logged on 2017-11-13 18:13 asciilifeform: use Ada.Strings.Unbounded; << mno ben_vulpes this is ~specifically~ a Do Not Want
asciilifeform: phf: ideally i'd get rid of Ada.Strings , full stop
phf: well, right. i'm not sure what ada.strings is (i.e. is it a protocol or concrete datatype), so i can't really comment further
asciilifeform: and rewrite the parser per se in scheme ( have it be present as commented bytecode constant )
phf: there be dragons. i mean, if you're rewriting a parser in lisp, then you might as well have proper readtables, rather then hardcoded sexp hack
asciilifeform: and get rid of the pointers.
asciilifeform: ( there's no particular reason why you can't have a schemetron use strictly arrays and integer indices into same )
phf: i'm not sure how you're planning on doing that, unless you mean ada level pointers. you'd have to have objects with values that are offsets into your virtual heap
phf: ah, k
asciilifeform: indices. as seen in ffa.
asciilifeform: ( the operative difference is that indices are bounded , and you can reason meaningfully about'em )
phf: right, i suspect that scheme.adb doesn't touch on it, because, again, no cons
asciilifeform: https://github.com/fitzgen/ada-scheme/blob/master/scheme.adb#L134 << the faux cons. observe, they use pointers for the car/cdr
asciilifeform: whole thing reads like straight translation from c
phf: somewhat relatedly one handy thing i saw on CADR is named cons regions, i.e. explicit memory regions where you can cons and every allocation function having a *-in-region equivalent, like (cons-in-region x y region). i'm not sure if that's there, but you presumably can do some kind (with-cons-region (region ...) body) thing. naturally those regions can be saved (preserved referential integrity) or cleared, etc.
phf: the array instead of pointers approach gives you free save (in fact you can run it against a mmaped region and have a ghetto core file)
a111: Logged on 2017-07-13 15:42 asciilifeform: phf: contrary to appearances, asciilifeform is not fixated on ada lang per se, but rather on the style of thinking it leads the operator into.
asciilifeform: phf: you can run your entire heap of a mmap'd region , neh
phf: that's what i said.
asciilifeform: no particular reason why it needs custom 'regions' support
phf: it's unrelated
asciilifeform: ( though they are useful for cache locality )
phf: asciilifeform: i'm using "memory management" meaning of cons, not like lisp 101 take on it. they don't have cons meaning that there's no managed heap, there's no gc on that heap, and you can't allocate things into the heap and let it be managed by heap machinery. so they have "cons", but their ~actual~ cons is ada's "new ..."
a111: Logged on 2017-11-12 23:12 asciilifeform: i'm not fully convinced that a scripting lang ~needs~ a gc
phf: like in php model? allocate as much as you want and then "free" on termination?
asciilifeform: aha. free with death.
asciilifeform: and not 'as much as you want' but up to B bytes, with B given on commandline and stackframed on warmup.
phf: right, maxheap
asciilifeform: to put it in libctronic terms, the resulting linux binary will call setbrk() ~exactly once~ in its life
phf: right, so that scheme.adb would benefit from a way to cons onto an arbitrary sized array, and then later someone can bolt a gc on top of that. can even implement it as an explicit function call rather than a threshold thing
asciilifeform: imho it dun particularly make sense to have gc in this application
asciilifeform: but i already described why.
phf: well, i'm thinking in terms of a TMSR MACHINE. scheme.adb linked against ffa linked against that com1 hack you posted some time ago :p
asciilifeform: if 'machine' i'd rather have handwritten 32kb asm thing, than whatever 'best effort' gcc shits out.
asciilifeform: but sure.
diana_coman: I can't seem to find in the logs any discussion re duplex construction/duplexing the sponge i.e. keccak's authors own proposal of using keccak for authenticated encryption; did anyone look into this?
asciilifeform: diana_coman: iirc it was in the original paper
asciilifeform: keccak is immune to length-extension attack so it is pretty straightforward to convert it into a cipher
diana_coman: well, I have several papers on keccak and one of them is this "Duplexing the sponge: single-pass authenticated encryption and other applications"
diana_coman: however so far I focused on the reference paper on keccak itself (The Keccak reference version 3.0)
asciilifeform: diana_coman: what are you contemplating making ?
diana_coman: asciilifeform, I'm still looking/exploring potential solutions for client-server communication needs in eulora
asciilifeform: well yes but moar specifically
asciilifeform: iirc you were gonna use mircea_popescu's algo ( use rsa to send otp pages, then later use'em )
diana_coman: asciilifeform, the second part is not so well defined/fixed yet
asciilifeform: what's hard re using otp ? it's a xor
diana_coman: yes; but it's unclear if a simple bitfield xor is best option
asciilifeform: for so long as you're actually using otp (i.e. 1 byte of key used for exactly 1 byte of payload) it's the only logical option
diana_coman: I meant the choice of specific, concrete way to expand the original bitfield i.e. "reuse the otp"
asciilifeform: how atrociously slow does the 'never reuse' variant look ?
asciilifeform: if you use actual one-time -- you then dun have to reinvent symmetric ciphering
diana_coman: heh, true that; I think first trouble there is that "never-reuse" choice means "no-knob" for client who pays however for the traffic; the whole point was precisely to let player choose their own level of compromise between cost and security (otps are generated on the server for good reason)
asciilifeform: the boojum is that neither i nor anybody else knows of any rational way to quantify the compromise.
asciilifeform: ( we dun have a scientific approach to symmetric ciphering. )
diana_coman: yees, but conceivably there might be one in the future; if no knob then no point as it were, entirely
asciilifeform: tbh i dun expect to live to see such a thing
diana_coman: asciilifeform, re mp's algo, it was this: http://btcbase.org/log/2017-10-06#1722036
a111: Logged on 2017-10-06 23:13 mircea_popescu: basically the scheme is, you rsa a random bitfield, then you expand that into as much otp as you want by doing recursively Fi = hash(bitfield + Fi-1). there's a limit on i, obviously, which can be set to 1.
asciilifeform: yea that's hash-as-blockcipher
asciilifeform: makes sense then
diana_coman: at a first pass this duplex thing based on keccak seems to be a similar attempt really, hence my question if anyone looked at it more closely (I'm still trying to fully grasp it, not there yet)
asciilifeform: as i understand, ordinary keccak suffices for this scheme
BingoBoingo: !~ticker --market all
jhvh1: BingoBoingo: Bitstamp BTCUSD last: 6545.26, vol: 30220.26606995 | Bitfinex BTCUSD last: 6413.3, vol: 121231.82335321 | Kraken BTCUSD last: 6564.5, vol: 8688.60323938 | Volume-weighted last average: 6446.40578362
asciilifeform: ^ pheeature idea : why not have ticker autofire when the number moves >10% from last tick
asciilifeform: meanwhile, in very vintage lulz, https://archive.is/I5JC0 >> 'Secretary of State Baker said Washington would not object to military intervention in Romania by Soviets or the Warsaw Pact.'
asciilifeform: 'The Chinese miners were instructed to continue mining the coin, even at great financial loss, to support a pretension of value and use, minimally sustaining its life. When the price troughed, those who were in the know about the plan accumulated it in large quantities' etc.
jhvh1: BingoBoingo: The operation succeeded.
mircea_popescu: http://btcbase.org/log/2017-11-13#1737193 << in no sense more terminal than say apple.
a111: Logged on 2017-11-13 14:41 asciilifeform: in other lulz, https://seekingalpha.com/article/4122890-tesla-approaches-terminal-decline
asciilifeform: exactly same item
mircea_popescu: right. "has cash" usg printed for it.
asciilifeform: moar spectacular tho, candle burns at both ends
mircea_popescu: but for the nitpick : it's not the job of the ceo to know why robots have limits.
asciilifeform: it is if he opens mouth re robots
mircea_popescu: this totally orc bullshit where everyone is an expert in everyone's field already...
mircea_popescu: asciilifeform ceo is an operational management position.
mircea_popescu: cto's job is to understand why robots have limits, if the company is doing robots.
asciilifeform: gensec is also managerial position. hruschev and his corn speeches remain lulzy.
asciilifeform: fella oughta chat with his cto, vizier, etc before becoming a public clown.
mircea_popescu: indeed. but here's the hidden truth : the idiots TALKING TO HIM about it are no different from the people asking beautician re politics.
mircea_popescu: ie, he's an idiot for answering, but someone somewhere was an idiot for asking.
asciilifeform: these are always on hand
mircea_popescu: THAT someone isn't admitting to idiocy, preferring isntead to derp about how "Ceo is supposed to know all".
mircea_popescu: if it were workable for ceo to know all why would corporations even be needed in first place ?
mircea_popescu: http://btcbase.org/log/2017-11-13#1737203 << i'm happier with properly reporting keys as 4090 bits, and pointedly explaining WHY the difference to the user.
a111: Logged on 2017-11-13 15:09 asciilifeform: but you can trivially show that using the bottom bits in this way lets you actually get 4x as many possible primes
mircea_popescu: politically preferable.
mircea_popescu: im not going to have my tech people do backflips to seamlessly bridge imperial idiocy into reality when i could just have the marketing people point out to how the empire lied by making the difference a point of difference.
asciilifeform: mircea_popescu: i made a stab of computing a lower bound of bitness for hypothetical '4096b of possible prime' but ran out of juice.
mircea_popescu: waste of juice.
asciilifeform: it's an answerable question, euler could have made short work of it
mircea_popescu: euler is dead.
mircea_popescu: http://btcbase.org/log/2017-11-13#1737223 << so what, have 0-led p, q ? what are you on aboot.
a111: Logged on 2017-11-13 15:20 asciilifeform: there is no legitimate reason to do it.
asciilifeform: mircea_popescu: the only case where this is a problem is 0-led p + 0-led q
a111: Logged on 2017-11-13 16:51 lobbes: Good news on archive front; archive.is d00d has agreed to add my ips to his cloudflare whitelist
asciilifeform: if only 1 -- then high bits of p,q remain seekrit
mircea_popescu: asciilifeform 0 led either one, or rather, non 11 led either one gives you a sub 4096 N
mircea_popescu: http://btcbase.org/log/2017-11-13#1737241 << tell him that if he regs a name ima donate to his project.
a111: Logged on 2017-11-13 17:04 lobbes: Very accommodating d00d indeed. I invited him here, as well, but you know how that goes
asciilifeform: i can actually see the 1. but why 11
mircea_popescu: asciilifeform there's nothing in principle wrong with the 2045 bit primes, except, of couyrse, the lying about it.
mircea_popescu: plenty of entropy left as it is.
asciilifeform: ( and why not then 111, 1111.... )
mircea_popescu: asciilifeform 11 x 11 = 1001 ; 10 x 10 = 0100
mircea_popescu: you need two bits to make 4, not 3. that's it.
mircea_popescu: 111 doesn't add anything, 111 x 111 = 110001 ie 6
asciilifeform: you don't actually get a 10x10=0100 because carry bits ripple up
asciilifeform: (last digit of a product is not a straight product of the last-digit-of-p and last-digit-of-q )
asciilifeform: ( by 'last' here is meant, of course, leftmost )
mircea_popescu: 101111 x 101111 = 100010100001, if you're lucky ; but 100000 x 101111 = 010111100000 which is no good.
mircea_popescu: asciilifeform we're discussing first digits not last digits here.
asciilifeform: the way i'd implement the whole shebang, is simply to reject both primes if the highest bit of pq is not 1 .
asciilifeform: rinse, repeat.
asciilifeform: no reason to lose that 1bit of entropy.
mircea_popescu: you lose it by this impl as well.
mircea_popescu: as you'll reject the primes and end up with the same 2045 bits of entropy
asciilifeform: you lose 1. but in koch's variant you lose 2 .
asciilifeform: ( plus the 2 bottoms )
asciilifeform: 4 all in all
mircea_popescu: ok, you'd conceivably squeeze a little more, but again, see http://btcbase.org/log/2017-11-14#1737360
a111: Logged on 2017-11-14 01:02 mircea_popescu: im not going to have my tech people do backflips to seamlessly bridge imperial idiocy into reality when i could just have the marketing people point out to how the empire lied by making the difference a point of difference.
asciilifeform: at any rate i haven't any argument against http://btcbase.org/log/2017-11-14#1737376
a111: Logged on 2017-11-14 01:06 mircea_popescu: asciilifeform there's nothing in principle wrong with the 2045 bit primes, except, of couyrse, the lying about it.
mircea_popescu: clarity is more valuable than a nearer asimptote, in many contexts.
asciilifeform: honestly labeled box imho suffices.
mircea_popescu: it's still an unreturned function "what other bits are lost what other places". so this 4090 still an upper bound.
mircea_popescu: but... working on.
asciilifeform: afaik diana_coman exhaustively showed the places
a111: Logged on 2017-11-13 17:47 ben_vulpes: and in ancients, dusted off mpfhf benchmarker, finished the bit-banging of inputs, fired off a run late last week that is *still hashing*
mircea_popescu: asciilifeform quintuplechecking never hurt anything.
asciilifeform: ( she is using my sanitized gpg bignum. but i did not preserve koch's faux-rng atrocity ; so anything pertaining to entropy, is new )
asciilifeform: iirc she is pumping it straight from fg.
asciilifeform: afaik the only remaining, and most obvious 'loss' is the one implicit in prime number theorem ( where , wat, ~10k possible rng outputs correspond to same prime output )
mircea_popescu: http://btcbase.org/log/2017-11-13#1737262 << this is actually a pretty good approach.
a111: Logged on 2017-11-13 19:29 phf: http://btcbase.org/log/2017-11-13#1737247 << it looks like a properly structured scheme evaluator, but it's ~explicitly~ lacking a native cons, which might be a very good exercise for whoever™ adding a static allocation space, adding mark-and-sweep, then all those To_Unbounded_String look like they can be simply search/replaced
mircea_popescu: asciilifeform afaik "nextprime" or anything like it is not used
asciilifeform: somewhat counterintuitively, you still get same result ( minus the time sidechannel leak, naturally )
asciilifeform: ( there are still fewer primes than 2^4096bit phase space )
a111: Logged on 2017-11-13 19:40 asciilifeform: whole thing reads like straight translation from c
asciilifeform: ( 2048 rather. but you get the idea )
mircea_popescu: asciilifeform no, because see, if you don't use nextprime you lack the "nop bridge" so to speak. rolling number 6 does not take you to 7. to get 7 you need a natural 7, and this is equiprobable to rolling a natural 2^74207281-1 on the space of (0,2^74207281-1).
mircea_popescu: (above quoted, largest prime known, found last year.)
asciilifeform: now i recall having argued this myself, lol
mircea_popescu: very important NOT to use any kind of nextprime, as it turns out .
asciilifeform somewhat addled with viral fevers, and strange dreams at night ( featuring such colourful characters as tov. ceaușescu )
mircea_popescu: i get odd viral influenzas in the first months of mingling with the whores of a new land also.
asciilifeform: that at least has up-side
mircea_popescu: but supposedly this makes you stronger.
mircea_popescu: !!up hubud
deedbot: hubud voiced for 30 minutes.
mircea_popescu: allegedly teh soviets were so impressed with it packed the whole assembly line took it to russia in 1945
mircea_popescu: (that site pretty decent ref of romanian vehicle production, incl de teardrop lulzitem)
ben_vulpes: http://btcbase.org/log/2017-11-14#1737405 << currently working the 4th permutation of a 65536 byte message for a 32 bit hash
a111: Logged on 2017-11-14 01:12 mircea_popescu: http://btcbase.org/log/2017-11-13#1737245 < ha!
mircea_popescu: http://btcbase.org/log/2017-11-13#1737304 << was never discussed nor did i ever really grok it.
a111: Logged on 2017-11-13 20:35 diana_coman: I can't seem to find in the logs any discussion re duplex construction/duplexing the sponge i.e. keccak's authors own proposal of using keccak for authenticated encryption; did anyone look into this?
mircea_popescu: http://btcbase.org/log/2017-11-13#1737321 << this is a problem ; but perhaps opening it up to the market may be helpful.\
a111: Logged on 2017-11-13 20:57 asciilifeform: the boojum is that neither i nor anybody else knows of any rational way to quantify the compromise.
mircea_popescu: if we had a way to quantify we could just decide.
mircea_popescu: http://btcbase.org/log/2017-11-13#1737334 << nah, autospeaking bots to be kept at a minimum which is 0.
a111: Logged on 2017-11-13 21:10 asciilifeform: ^ pheeature idea : why not have ticker autofire when the number moves >10% from last tick
mircea_popescu: BingoBoingo shaping up.
hanbot: <mircea_popescu> (asciilifeform's ticker idea) << nah, autospeaking bots to be kept at a minimum which is 0. << for the curious, why is say deedbot's rss announcer a non-auto event whereas a market movement isn't?
mircea_popescu: history, pretty much.
mircea_popescu: but if you wish to argue it in substance, the fiat valuation of bitcoin is broadly irrelevant -- to bitcoin, to the actrual things the fiat turds misclaim to represent, etcetera. whereas the penmanship of the l1 is relevant by definition.
hanbot: that seems sound.
asciilifeform: http://btcbase.org/log/2017-11-14#1737435 << keep in mind, this was not a high bar -- ru soldiers took home even toilets from germany
a111: Logged on 2017-11-14 01:44 mircea_popescu: allegedly teh soviets were so impressed with it packed the whole assembly line took it to russia in 1945
asciilifeform: http://btcbase.org/log/2017-11-14#1737451 << i see the q of whether a coin buys a rowboat, a battleship, or entire flotilla, as broadly interesting one -- but mebbe that's just me
a111: Logged on 2017-11-14 02:07 mircea_popescu: but if you wish to argue it in substance, the fiat valuation of bitcoin is broadly irrelevant -- to bitcoin, to the actrual things the fiat turds misclaim to represent, etcetera. whereas the penmanship of the l1 is relevant by definition.
asciilifeform: ( i could even readily believe that an , e.g., 25x rise in the heathenbux:btc exch rate would make no practical diff to mircea_popescu . but i suspect that i am not the only one here for whom it would make a palpable diff. )
mircea_popescu: 10% dun buy any flotillas.
mircea_popescu: afaic, i still have coins i paid like $2.10 for.
asciilifeform: pretty sure i still have a coin i paid 0 for, lel
asciilifeform: ( it could buy a veeeery small, in a bottle , flotilla... )
lobbes: http://btcbase.org/log/2017-11-13#1737252 << I do plan to walk-back the logs and pull what still exists. Hard-drive space willing for being able to serve up the things on-demand (only ~40GB to work with on the VPS)
a111: Logged on 2017-11-13 18:21 asciilifeform: lobbes: does this mean that you can mirror the whole zip collection nao ?
lobbes: http://btcbase.org/log/2017-11-13#1737253 << time will tell. Depends on how much the logs-to-date worth of archives end up being in drive space. Bandwidth also a factor. Many things left to be sussed out.
a111: Logged on 2017-11-13 18:32 BingoBoingo: lobbes: So modest is becoming less modest?
jhvh1: BingoBoingo: The operation succeeded.
BingoBoingo: It looks like I'm going to be spending 2U on an Ubiquity Edgerouter Pro and an Ebuquity Edgeswitch lite, unless there are other ideas for networking hardware
BingoBoingo: Under the guise of when visiting Shaman, buy cheapest that can move the traffic
← 2017-11-12 | 2017-11-14 →