09:57 |
unpx |
http://logs.nosuchlabs.com/log/pest/2023-03-22#1025449 <<< I'm curious about the meaning of ``lubyisms'' and any resource that was ``painfully complicated to implement'' |
| |
↖ |
09:57 |
dulapbot |
Logged on 2023-03-22 12:30:29 asciilifeform: looked at a # of 'lubyisms', found that the moar theoretically-optimal sorts tend to be 1) painfully complicated to implement 2) very high constant cost, such that they aint likely to win in re pest, where most common use case is likely to be simply one station sending to anuthe |
| |
~ 34 minutes ~ |
10:32 |
unpx |
And are there any stats on loss rate? Because I guess the encoding shall be calibrated based on that |
10:46 |
asciilifeform |
unpx: atm seems that most packet loss happens on the segment of pipe nearest to people's desks (e.g. phf seems to have a high-loss connection at his station) and so aint remotely 'same for erryone' |
10:47 |
asciilifeform |
http://logs.bitdash.io/pest/2023-03-23#1025086 << 1st such code asciilifeform looked at (and invited other folx to do so) was the luby transform |
10:47 |
bitbot |
Logged on 2023-03-23 09:57:54 unpx[jonsykkel|deedbot|signpost]: http://logs.nosuchlabs.com/log/pest/2023-03-22#1025449 <<< I'm curious about the meaning of ``lubyisms'' and any resource that was ``painfully complicated to implement'' |
10:47 |
dulapbot |
Logged on 2023-03-22 12:30:29 asciilifeform: looked at a # of 'lubyisms', found that the moar theoretically-optimal sorts tend to be 1) painfully complicated to implement 2) very high constant cost, such that they aint likely to win in re pest, where most common use case is likely to be simply one station sending to anuthe |
10:53 |
PeterL |
Is "lubyism" a term used by other people across the industry or is that just something we use here for this type of thing? |
10:53 |
asciilifeform |
PeterL: afaik there's no traditional term that fully covers the subj |
10:54 |
asciilifeform |
'fountain codes' is prolly the nearest |
| |
~ 17 minutes ~ |
11:12 |
unpx |
I missed how simple it is to use MDS error-correcting codes as erasure codes. So afaik what we are looking for is something that does not require heavy decoding (e.g. Gaussian elimination ~O(n^3)). |
| |
~ 49 minutes ~ |
12:02 |
signpost |
unpx: decoding is of linear complexity in the one I wrote |
12:03 |
signpost |
iiuc encoding is also. |
12:05 |
signpost |
it would be a good use of time to at least say why that encoding scheme is unsuitable rather than just reaching for another. |
| |
↖ |
12:06 |
* |
signpost intends to find his way back to writing it in a language with saner parallelism soon. possibly via the WIP common lisp one I've got going. |
| |
~ 59 minutes ~ |
13:05 |
asciilifeform |
signpost: can plz post a link to your most recent prototype of that thing? ( asciilifeform was digging in logs, not found for sumreason ) |
13:07 |
signpost |
https://trinque.org/2023/01/18/ocpy_squirt_slurp/ |
13:07 |
asciilifeform |
signpost: ty! |
13:07 |
signpost |
yw |
13:15 |
asciilifeform |
signpost: (maybe someone already pointed out?) the genesis doesn't press as given, on acct of weirdness with file paths ( --- null/manifest false +++ ocpy/manifest c810...... ) |
13:21 |
signpost |
not sure, works over here http://paste.deedbot.org/?id=lTUo |
13:21 |
signpost |
with freshly wget'd patches and seals |
13:22 |
asciilifeform |
what asciilifeform sees . ( which vtron was signpost using ? ) |
13:24 |
signpost |
this lineage as built by pentacle, iirc phf's plus work by bvt and possibly others |
13:25 |
signpost |
yeah phf is right there in the filenames, derp |
| |
↖ |
13:25 |
asciilifeform |
aa |
13:26 |
signpost |
all I modified of v.sh was making WOTDIR, PATCHDIR, and SEALDIR env vars so I could aim the thing from pbuild |
13:27 |
* |
signpost really owes a pentacle website since whole point was eliminating per-desk toolchain ??? of this kind |
13:28 |
asciilifeform |
meanwhile, in sumthing completely different: this d00d apparently still in biz, sweating to bake own FG ( complains about 'unreadable schematics' and 'missing BOM', lol, could've asked? it ai |
13:28 |
asciilifeform |
nt a seekrit ) |
13:29 |
signpost |
but then where to get the "I made this" |
13:29 |
dulapbot |
(asciilifeform) 2022-05-20 asciilifeform: last asciilifeform knew, jfw et al's kit made FG look like 'bestseller' in comparison; but whoknows, maybe nao finally rich&famous... |
13:29 |
signpost |
guy stinks of inherited "how to make a million out of ten million" |
| |
↖ |
13:30 |
asciilifeform |
iirc d00d had an ee diploma. ought've been done in a week after starting, lol |
13:31 |
asciilifeform |
could've made own faster design, would've fit in with his ' asciilifeform's wares such shite! ' push. but instead this. |
13:36 |
asciilifeform |
2022, lol. |
13:39 |
* |
asciilifeform mind boggles, not sure how the recipe could've been made any easier. |
13:41 |
asciilifeform |
'I admit I don't yet fully understand why it's done the way it is, especially what the second comparator stage is doing and what considerations govern the particular choices of component values.' |
13:42 |
asciilifeform |
... wainot crack a b00k ? or at least breadboard the thing (it worx a++ on breadboard) and fiddle empirically ? but mno, gotta go straight to baking pile of smt boards ? nfi |
13:44 |
PeterL |
The thing reads like it is something he invented (no link to asciilifeform's page abou the thing?), but then he has that line you just quoted where he admits he does not know why it was done the way it was |
13:44 |
PeterL |
kinda sleazy not to give credit where due |
13:44 |
signpost |
yeah but he shows how to light an LED at the end; what do you want?! |
13:44 |
asciilifeform |
PeterL: does say 'reproduce design' but lulzily with no link to, y'know, the page with the schematics etc |
13:45 |
asciilifeform |
'let's pretend it never happened' |
13:49 |
asciilifeform |
d00d had... 4+y (!!) since 1st proclaimed 'fuckyou, i'ma make own rng' |
13:50 |
* |
asciilifeform naively actually expected that he would. esp. considering that asciilifeform littered the l0gz with over9000 tips re places to improve in the original |
13:52 |
asciilifeform |
'the bad dancer is hindered by his own arse'(tm)(r)(j00 proverb) |
13:58 |
asciilifeform |
http://logs.bitdash.io/pest/2023-03-23#1025118 << bbut, biznisman! not dirty sarariman! etc |
13:58 |
bitbot |
Logged on 2023-03-23 13:29:32 signpost: guy stinks of inherited "how to make a million out of ten million" |
14:00 |
asciilifeform |
could've, say, hired some chinese d00d who stayed awake in school? but mno. 'i'ma figure out with my 20$ scope!' etc |
14:01 |
asciilifeform |
( not even speaking of 'could've asked right here', can picture that aint an option in his head ) |
14:04 |
asciilifeform |
meanwhile, in other mega-titans of industry. |
| |
↖ |
14:13 |
* |
asciilifeform wonders re wtf concretely happens to the brains of the '10 $mil' folx. almost seems as if at certain magick threshhold in pile of dough, people become functionally retarded?! |
| |
↖ |
14:14 |
asciilifeform |
... possibly simply substance abuse ? nfi |
14:14 |
dulapbot |
(asciilifeform) 2020-02-04 asciilifeform: or, for that matter, his (admitted!) cocaine. imho still shows typical effects, even tho supposedly quit. |
14:15 |
asciilifeform |
e.g. at 1e9+$, evidently, one goes and buys twatter. |
14:17 |
asciilifeform |
or remember this d00d, at, dunno, 1e7+$ took up shitcoins etc |
| |
~ 22 minutes ~ |
14:40 |
* |
asciilifeform suspects: moolah unlocks latent narcissism, as lsd supposedly sometimes unlocks schizo. fella who conceives of self as 'aristocrat' will burn piles of dough for ~nuffin, but feel great bragging about how it is over9000x what plebe makes in lifetime |
14:40 |
dulapbot |
(trilema) 2017-08-29 kanzure: nmz787 took an electron microscope. the gigascope never materialized though. |
14:41 |
asciilifeform |
with conceit of 'aristocrat' comes the 'aristotelian' contempt for productive labour |
14:41 |
dulapbot |
Logged on 2022-10-31 17:20:10 asciilifeform[busybot]: sees the 'aristotelian' cosmography as product of 'thug aint happy to simply dominate others, needs to believe that he is chosen by gods'. ditto the dimwit, not happy to simply dominate via numbers, but 'hephaestus has to be a cripple and cuck' |
14:41 |
dulapbot |
(trilema) 2019-10-30 asciilifeform: http://logs.nosuchlabs.com/log/trilema/2019-10-30#1948740 << right, the only trooly intentional act is to sit in cafe and smoke cigar made of benjies. whereas to try an' sew some sorta usable purse outta 30 yrs of 'tech market' sow's ear , is what furniture does. ( at least mp_en_viaje i |
14:43 |
asciilifeform |
... not to mention, failure -- no longer stings, can piss away 1e6$ and still think of self as 'great biznisman' |
14:43 |
asciilifeform |
bankman, 'hero' of the current generation, famously: 'i would never consider reading a book' |
14:44 |
asciilifeform |
( opening b00k, for these folx -- hurts! might stumble across sumthing that takes actual effort, and 'no longer feel like genius' ) |
| |
~ 58 minutes ~ |
15:43 |
signpost |
value's inverted at the lowest of levels in this empire. only sound you can hear is the chaotic whoosh down the drain. |
15:43 |
signpost |
everyone's personal interest as he sees it is aligned with activity that accelerates it. |
15:43 |
signpost |
just a long way of saying suicide. |
15:45 |
signpost |
but then! |
15:45 |
signpost |
!!ticker btc usd |
15:45 |
deedbot |
$28,238 |
15:46 |
* |
signpost still thinks one of the cardinal sins of tmsr was impatience, rather than arrogance/delusion. |
15:50 |
signpost |
in related sandwich prices, gold popped above $2k today. feds probably "selling" paper like mad. |
15:53 |
signpost |
http://logs.bitdash.io/pest/2023-03-19#1024859 << apparently his cost basis is approx $30k. |
15:53 |
bitbot |
Logged on 2023-03-19 20:17:41 awt: Some big buyers I am aware of are Saylor (109,000), Tesla has 10k. I know a lot of non-tech non-middle-class buyers go though places like Swan. |
15:53 |
signpost |
the apparent lack of attacks on Saylor is suspicious as all hell imho. |
15:54 |
* |
signpost still thinks he might be part of the usg hodl. he's holding one hell of a sell-bomb. |
| |
~ 36 minutes ~ |
16:31 |
asciilifeform |
signpost: they never seem to lack a sell bomb when need one ( given as it's rechargeable ) |
16:31 |
dulapbot |
(asciilifeform) 2021-04-23 asciilifeform: a major moving part behind the price swings are the people who run this heat engine -- 1) spread fud re bitcoin 2) 'buy the dip' 3) drum up enthusiasm among morons 4) sell dear 5) goto 1, nao w/ moar coin, extracted from morons on both strokes of the engine piston |
16:35 |
asciilifeform |
on the rare occasions when plebes buy actual (i.e. on chain) btc, it finds, like clockwork, its way back into the 'bomber' (via 'silk roads', via shitcoin 'ico's, etc) , from which dropped again, ad infinitum. |
16:36 |
asciilifeform |
(and for that matter, miners are stuck selling some large % of their take ~immediately.) |
16:37 |
asciilifeform |
( soon to be 100% possib. ) |
16:37 |
bitbot |
Logged on 2023-03-19 18:27:40 asciilifeform[jonsykkel]: and latest rumour is iirc that miners (who, precisely as one'd expect from hill-climbing amoebas, moved en masse from cn to usa) are aboutta be taxed (i.e. forced to sell 30-40% of coin right off the bat), further keeping the waterfall wet |
| |
~ 25 minutes ~ |
17:02 |
awt |
Saylor was sued for back taxes in DC recently, if that counts. |
| |
~ 15 minutes ~ |
17:18 |
* |
asciilifeform not knows the story of saylor, except in context of l0gz, i.e. presumably a rockefeller type who bragged about btc hodl |