Show Idle (>14 d.) Chans

← 2018-08-23 | 2018-08-25 →
Mocky: maybe i'm dense, what do you mean reduce a pile at least in half?
mircea_popescu: so far, the correct approach seems to be : 1. select the outlier piles (highest q, lowest q). 2. mix the smaller of the two with such a fraction of the larger of the two as can be done. 3. repeat. if 2 fails, move inwardly on the pair list.
mircea_popescu: Mocky he means the x% is never < 50% and rarelyt <80%
mircea_popescu: the ~other~ good approach is 1. calculate average quality as a real ; 2. calculate the integer piles of consecutive qualities this reduces to (you'll never get an irrational value through 1, which in itself is a very itneresting lemma) ; 3. try all pile pairs in some (random ?) order for producing either of the two expected qualities.
Mocky: ok, I see it
asciilifeform: currently i suspect that any 2 miscible-q piles can be merged in either 1 or 2 steps. but too tired to attempt proof just nao
mircea_popescu: anyway, i expect the "hard" case would revolve around powers of two, so i'm guessing checkermark pairings probably deliver the hardest case (for current pairwise mixing). say in 8 bit : 85 items q170, 146 items q 73, 219 items q109 and so on.
a111: Logged on 2018-05-22 21:47 mircea_popescu: so far that's even the reason empire crossed the threshold into oblivion ( i'm too lazy to dig up ye olde trilema where i say "state may exist for as long as it so closely mimmicks my will i can't discern it's there, and not one second past that" ) and now must be destroyed. had it had the sense (rather, the capacity) to maintain both forms, i'd nwever have even noticed it, and it could have continued to exist therefore.
mod6: mornin'
mod6 tired.
mod6: need coffee
asciilifeform: mircea_popescu: i solved yer puzzler, will post sln later today
mod6: !!invoice BingoBoingo 0.01341464 One month rental of rockchip: Through September 2018
mod6: !!v DC1B15C060167472D1E14D037637116C56F6EC6D92BD265DD8C64281D7F45030
deedbot: Invoiced BingoBoingo 0.01341464 << One month rental of rockchip: Through September 2018
mircea_popescu: "Solution (.Sln) File - The new home for Visual Studio documentation is Visual Studio 2017 Documentation on The latest version of this topic can be found at Solution (.Sln) File. A solution is a structure for organizing projects in Visual Studio. The solution maintains the state information for projects ..."
mircea_popescu: way to go alf !!1 >D
mircea_popescu: (holy shit think about it, state information ~for projects~)
asciilifeform: mircea_popescu: it's standard microshit turd, 'what files luser has open and to what pos' etc
asciilifeform: massive binarygarbagepile
mircea_popescu: anyway. what did you mean, special linear group ?
mircea_popescu never met a lie-ician before!
asciilifeform: wait an' see, i suspect mircea_popescu will like the answ
mircea_popescu: alrighty!
asciilifeform: know what, i can post the answer, and somebody else can turn into proggy if he feels like
asciilifeform: cuz it is not complicated
asciilifeform: observe that the final 2 piles always satisfy the form n1*q1 + n2*q2 = S , where S is the sum of the masses (n_x * q_x for all x)
asciilifeform: in turn both of the piles that went into n1 (and likewise n2) satisfy same recurrence, n1_a * q1_a + n1_b + q1_b = n1*q1 ; n2_a * q2_a + n2_b + q2_b = n2*q2 .
asciilifeform: err, n2_a * q2_a + n2_b * q2_b = n2*q2
asciilifeform: mircea_popescu : do i need to go on, or has it clicked ?
asciilifeform: all you gotta do is to take { floor(sum(mass) / sum(qty)) - 1 , floor(sum(mass) / sum(qty)) } as the initial guess for the split of qualities of the final 2 piles
asciilifeform: in the case of Mocky's example, this will be initially 172,173
asciilifeform: then attempt to permute the list into two sublists which satisfy the recurrence. if the recursion comes up empty, you slide the guess tuple forward by 1. (there is no need to do this more than once, avg of the final 2 piles cannot exceed that of the entire thing)
asciilifeform: one of the two attempts will turn up a valid decomposition.
asciilifeform: if you want optimal decomposition, you keep all valid ones and pick the shortest.
deedbot: << Qntra - USG Lich McCain Surrenders In War On Cancer Will Spend Rest of Days As Cancer POW
BingoBoingo: ^ Phylactry has been found!
mircea_popescu: asciilifeform this is what proposes. but no, not "random", recursive instead.
a111: Logged on 2018-08-24 04:04 mircea_popescu: the ~other~ good approach is 1. calculate average quality as a real ; 2. calculate the integer piles of consecutive qualities this reduces to (you'll never get an irrational value through 1, which in itself is a very itneresting lemma) ; 3. try all pile pairs in some (random ?) order for producing either of the two expected qualities.
asciilifeform: mircea_popescu: aha , no randomness in my pill, it converges deterministically
asciilifeform: and never goes in circles
mircea_popescu: BingoBoingo good god, are they now gonna do a bunchga herp derp about how important the moron is and how irreplaceable the scrapping ?
BingoBoingo: mircea_popescu: The latest round appears to already have started
mircea_popescu: spurious fucktards.
BingoBoingo: But all comments are being attributed to "the family" suggesting he is already very close to Potato if not already there
asciilifeform: evaluating proposed split of set is very easy, you check that the sum(n_...) and weighed avg of the component Qs are in spec.
mircea_popescu: collective personage best personage, nobody has authority top say "this wasn't said".
mircea_popescu: asciilifeform let's put this in proper algo form.
asciilifeform: i'ma write it up in cl if nobody else wants to.
mircea_popescu: so 1. calculate q so that n q + m (q+1) = S (the sum of the count-quality products for the entire starting set)
mircea_popescu: 2. you split the starting set into two sub sets by the criteria ? and without splitting stacks ?
asciilifeform: stacks get split via the recurrence given in the start
asciilifeform: it describes all valid could-have-merged's
mircea_popescu: how do you choose between 7+5 and 6+6 ?
asciilifeform: you recurse
mircea_popescu: suppose i have stacks of 10k each with qualities from 50 to 150. average q is 100. now, how do i split these 100 stacks into two subsets ?
mircea_popescu: specifically, how do i pick among alternatives 1. "all stacks with odd index, left, all stacks with even index, right" and 2. "all stacks under 75 or over 125 left, all stacks within 75 to 125, right" ?
mircea_popescu shall proceed to coffee with diana_coman in a few, but thereafter will be back, so no worries.
asciilifeform: there's a strong chance this won't make sense until shown as lisp proggy.
mircea_popescu: considering how much lisp i read... possibly not even then!
mod6: :D
asciilifeform: but the way it goes, is to recurse along all valid paths, and mark the traversals that result in initial-state piles being successfully 'eaten'
mod6: same
mircea_popescu: seems to me the complexity of this will be massive.
asciilifeform: O(bitness of finalmass)
mircea_popescu: no fucking wai.
mod6: asciilifeform: Good to go for us to buy another round of SSDs? (apparently this time we can only get 3 at a time)
asciilifeform: the 'cheat' is that for ~any~ pile in the tree, you know that it had to result from one of 2 possible merges.
asciilifeform: mod6: why only 3 ?!
mod6: shitazon limits, i've seen it before myself.
mod6: we do what we can. i was pretty impressed with the speed at which they arrived.
mircea_popescu: mod6 conceivably (besides ip registration thing, which how's it going ?) might also get vendor status ? open a little shop of actual stuff in uruguay ?
mircea_popescu: a shop of used computer parts is a great complement for a dc op.
asciilifeform: that would rock
mod6: Was kinda thinking that myself mircea_popescu... at least, in concept.
mod6: I haven't even considered IP registration.. was there a inititive to get a block of our own from ICANN?
mircea_popescu: yeah. you can sell all your shit that's out of amortization at half price, become huge. well, "huge", south america huge.
mircea_popescu: mod6 yeah, back in like... february ?
mod6: oh, I'll have to talk to ben_vulpes and read the logs on that. thanks for the nudge.
mircea_popescu: anyhways bbl
mod6: c-ya
mod6: !Qcalc 900/6200
lobbesbot: mod6: 0.145161290323
mod6: !!pay BingoBoingo 0.14516129
mod6: !!v 945B7F7A6D4F4AEC6760362343EAB65C65D3A49F973D59E066287DEF2841727E
deedbot: mod6 paid BingoBoingo 0.14516129
mod6: BingoBoingo: whatever you don't spend, just hang on to it, and we'll tally it up at EOM for the books.
ben_vulpes: iirc lacnic wants to see the existing block more heavily used than it is right now before allocating more, BingoBoingo was there something else incorporation that LACNIC wanted to see?
ben_vulpes: incorporation-related*
asciilifeform: ben_vulpes: if it's just a matter of having all the ips answer when pinged -- we can have that cured without much sweat
asciilifeform: BingoBoingo: feel free to route all unrouted ips to dulap
BingoBoingo: asciilifeform: Aite
ben_vulpes: if it's that simple, that'd be spiffy indeed
mod6: It sounds like, "Be legally established within the region and use the resources in LACNIC's coverage area."
mod6: So we might need to incorporate first. *might*
trinque: "hurr they are for xen tenements"
trinque: mod6: seems like a valuable thing to have, local corp
trinque: I dunno the tax laws down there, but might be beneficial in regards to deducting expenses?
trinque: I hear only the states audit people too, so all the more reason :P
mod6: Well, it's ~$4500 to buy a pre-built corp. Then alf & myself would have to fly down and sign the articles of organization.
mod6: We think that we would probably have a higher base tax liability -- right BingoBoingo ?
mod6: $4500 is quite a bit for where we currently stand financially.
trinque: doesn't fit with my experience of running businesses, typically the thing doesn't post any substantial income in the startup phase
mod6: *shrug*
trinque: (it could be that already tax is not a significant slice of your costs, dunno how it works, but would read)
BingoBoingo: <mod6> So we might need to incorporate first. *might* << Or I can apply as a unipersonal
BingoBoingo: Or try
BingoBoingo: <trinque> doesn't fit with my experience of running businesses, typically the thing doesn't post any substantial income in the startup phase << THe big tax incurred would be VAT
mod6: Is the 'unipersonal' like a sole-propritorship ?
mod6: *proprietorship
trinque: a quick google suggests the rate for distribution of profit is lower than the rate for regular income, which is what I'd have expected, so conceivably BingoBoingo could be compensated in part through his stake in the local corp as profit
trinque: dunno what the "home office" rules are there either, but maybe some rent gets thrown in too
BingoBoingo: mod6: It is
BingoBoingo: trinque: THe substantial VAT here makes things complicated
trinque: yeah, seems like they're income-tax light and sales-tax heavy, eh?
trinque: (which seems pretty damned sane to me, actually)
trinque: er VAT, sales-tax wouldn't be properly descriptive
BingoBoingo: Right. For a comsumer VAT and sales tax look the same, but... ther is a substantial difference
asciilifeform: BingoBoingo: exactly how many vacant ip do we have ?
BingoBoingo: asciilifeform: I'll do a count
BingoBoingo: asciilifeform: I don't have the complete list of assigned ip's, but from .204 through .246 should be unallocated
mircea_popescu: trinque> mod6: seems like a valuable thing to have, local corp << a recurrent observation.
mod6: !!sent-invoices
mod6: !!ledger
mod6: nice
mod6: anyone else having trouble getting to : ?
mod6 is getting 504
asciilifeform: aha times out
mod6: jurov: halp ^
mod6: !~later tell jurov We're getting 504 & timeouts from the ML
mod6: !Q later tell jurov We're getting 504 & timeouts from the ML
lobbesbot: mod6: The operation succeeded.
lobbes: times out here as well
mircea_popescu: o hey lobbes ; how's life been treatin ya ?
mod6: thx lobbes seems something with amazon perhaps
lobbes: mircea_popescu: ohey. good!
lobbes: in fact, was going to announce: I'm getting closer to finishing what will be a genesis for a 'logbot-command-router' in python that does the talking and listening with postgres (an older but relevant thread for context:
a111: Logged on 2016-08-29 13:54 trinque: right. in this case I challenge anyone to find a need for a second patch on those modules themselves, rather than creating a genesis vpatch for their own module which talks to the db in postgres.
lobbes: I've tried to design it to be 'modular' so that the operator can easily add their own custom commands, and to be 'vpatch friendly' in that folx can release 'whateverbot' as a vpatch that just extends it (in fact, this is my plan for the auctionbot)
lobbes: still need to test the hell out of it, so I'm aiming for maybe a week (which, still gives me 2 months for auctionbot ETL. eh. will see)
mircea_popescu: so it relies on logbot and interfaces with it via postgres db ?
lobbes: exactly
lobbes: just listens to the notification channel in postgres
lobbes: and then does the command handling
lobbes will bake blawg post once complete
← 2018-08-23 | 2018-08-25 →