#pest | 2024-01-29

Logged on 2024-01-28 23:15:53 crtdaydreams[jonsykkel]: http://logs.nosuchlabs.com/log/asciilifeform/2024-01-21#1115698 << nfi what copypaste had in mind dragging up reiser from the bottom of the barrel

11:06

in ~15y+ -- 0 problems. asciilifeform doesn't give a nanoshit who reiser whacked with an axe.

11:07

$ticker btc usd

11:07

busybot

Current BTC price in USD: $42642.97

~ 1 hours 35 minutes ~

12:42

meanwhile, a delish lul from these fine folx.

12:42

Logged on 2023-07-26 13:32:22 asciilifeform[jonsykkel|deedbot|signpost]: signpost: matters, evidently, to the fella who too proud to write in re fg schems when 'i have no idea why was done this way' etc

12:44

*

asciilifeform notsure whether this needs detailed commentary, beyond 'sleep of reason begets monsters'(tm)(r)

~ 1 hours 40 minutes ~

14:24

... re: upstack, key oldthread on subj.

14:24

(trilema) 2016-02-21 asciilifeform: the fact that reiser was jailed cemented it as permanently high-quality, no one wants to touch it, no 'progressors' progressate at it, no gnomes, no popularizers!111, etc.

~ 2 hours 40 minutes ~

17:05

http://logs.nosuchlabs.com/log/pest/2024-01-29#1032465 << i had concluded this response was a troll, beyond efficient storage of a buncha small files it was just another fs. is there something I'm missing?

~ 31 minutes ~

17:36

http://logs.bitdash.io/trilema/2018-01-22#1774011 << this is something that's bothered me a little, asciilifeform. maybe you misunderstand? it doesn't have an alu because you're supposed to be able to implement it.

17:36

(trilema) 2018-01-22 asciilifeform: scheme79 was a prototype, did not even have an alu!

17:40

In chapter 3.3.4 of SICP they describe a "Simulator for Digital Circuits"

~ 17 minutes ~

17:58

in the actual MIT lecture 5B Sussman states that it is how it's done.

17:59

The whole purpose of the scheme-79 chip was to demonstrate generational logic using an embedded language in lisp

~ 3 hours 1 minutes ~

21:01

http://logs.bitdash.io/pest/2024-01-29#1031944 << consider: what do you, personally, expect from a file system ?

21:01

Logged on 2024-01-29 17:05:46 crtdaydreams[jonsykkel|signpost]: http://logs.nosuchlabs.com/log/pest/2024-01-29#1032465 << i had concluded this response was a troll, beyond efficient storage of a buncha small files it was just another fs. is there something I'm missing?

21:01

it's an 'unsexy' item with 'unsexy' design goals -- reasonable speed under commonplace use cases, and reliability (incl. under damage). reiser -- delivered.

21:03

... and the manner of his disappearance kept the apparently bug-free fs from suffering 'improvements' by idiots.

21:04

http://logs.bitdash.io/pest/2024-01-29#1031945 << picture the performance characteristics of e.g. multiplier implemented 'from inside' that thing.

21:04

Logged on 2024-01-29 17:36:57 crtdaydreams[jonsykkel|signpost]: http://logs.bitdash.io/trilema/2018-01-22#1774011 << this is something that's bothered me a little, asciilifeform. maybe you misunderstand? it doesn't have an alu because you're supposed to be able to implement it.

21:06

(if actually added in iron, using s79 simply as layout language -- different conversation, naturally. but if implemented via the equiv. of church numerals -- ugh)

21:07

it's possible to pipeline it, more importantly, verify the integrity of the implementation, perhaps not necissarily alu, but other items like hashing algos, symmetric encryption algos

21:08

they also made a scheme-81

21:11

I agree, ofc there would be far too much overhead without pipelining, implementation could fork output to a configurable bus for pipelining and to register for debugging and non mission-critical objectives

21:13

the stuff they did was pretty advanced

21:13

i can link relevant docs if interested

21:15

s86

21:17

https://apps.dtic.mil/sti/pdfs/ADA201258.pdf[s86 arch]

21:17

s86 arch

21:20

Parent doc listing everything in relation to daedalus and the scheme vlsi project as a whole, I can't find anything on PI (likely absorbed into NS), CMTPG or SCHEMA

21:21

apparently even rms had his fair share of contributions to the project too

21:23

Many other familiar names, weste, knight, moon etc

~ 42 minutes ~

22:05

*

asciilifeform had actually read iirc all of the material linked above. tho it's been many yrs.

22:06

crtdaydreams: asciilifeform not says 'no point to s8x'. simply that it does not economically use whatever avail. silicon , and if baking it today one would want to depart from the original in certain predictable ways

22:07

s8x is arguably analogous to the 'egyptian' algo collection in ffa. readable, 'elegant', but slow.

22:08

there are even cases where that's ~all~ you want.

22:09

e.g. there are times where a 100sec (vs msec) rsa op is fine.

22:11

the rub is that if you don't already know that your use case falls into this category -- the answer is rather likely to be 'nope'

22:12

We can agree on that. Imo it's the best thing with out a $24e6 fab (that's how much it cost for s86!)

22:12

observe that the lispm people did not implement s8x. it is worth thinking about wainot

22:13

(and they weren't insisting on 'displaying lolcats' or the 200MB www sites of today, etc, either)

22:13

sussman stated he ditched the scheme project because he wanted to pursue AI

22:13

others, nfi, would have to ask them

22:14

they wanted e.g. fast routing ('ns' & co), which needs, among other things, reasonably efficient numerics

22:14

but knight, moon, holloway, et al evidently chose to build interlisp, zetalisp and derivatives therof because ^

22:14

aha

22:15

once yer actually trying to ~do~ sumthing with the lispm, suddenly it no longer suffices to 'here's existence proof, and if you wanna multiply, do it with church numerals'

22:15

nothing from weste btw

22:16

well bolix went away from integrated vlsi, I think there's still promise in generational hw

22:16

the raw cost of fab is imho a red herring tho

22:16

the problem aint the sticker price, but that there's no known way to recover it

22:16

heh, market for lispm?

22:16

(if you knew of one, could talk sumbody into fronting it, conceivably)

22:17

best thing I can think of for basic scheme chip is teaching programming

22:17

all the way down

22:17

pray that a new generation of competent programmers will work something out given a push in the right direction

22:18

the orig lm folx conned pentagon into coughing up the dough. which worked all the way up to the point where better talkers came along and promised 'moar for the $'

22:18

(asciilifeform) 2020-08-12 asciilifeform: and if a ~better liar~ comes along, he will win, and you -- lose. the technical merits of the proj being ~irrelevant.

22:18

(asciilifeform) 2022-06-11 asciilifeform: the most grim lesson there imho is that the '80s lispm corps were funded in exactly same way. and once the 'better liars' came along with a 'sexier' spiel, the lispms kaput.

22:18

other applications conecievably could be generational ai, something like a morphological hardware neural net

22:19

makes you wonder about nsas motivations for shunting attempts at ressurection, exhibit a being mallery

22:19

crtdaydreams: the 'ai' aficionados seem to be interested exclusively in 'flops per $', which is wai they're doing it on gaming gpu

22:20

any niche product inescapably loses on that param

22:20

reducing ai to a numbers game proves results tho

22:22

*

asciilifeform recs linked thrd as a whole re 'where's lispm', imho covers key points

22:22

(asciilifeform) 2020-08-12 asciilifeform: at same time, various vendors (e.g. franz, sold lisp for sun ws) pushed 'our margarine is just as good as real butter' scamola. and fooled enuff users to burn away what little of commercial market there was for lispm

22:24

http://logs.bitdash.io/pest/2024-01-29#1031996 << 'long game'

22:24

Logged on 2024-01-29 22:19:49 crtdaydreams[jonsykkel|signpost]: makes you wonder about nsas motivations for shunting attempts at ressurection, exhibit a being mallery

22:24

Logged on 2022-10-31 17:53:58 asciilifeform[6]: ( starting with mallery, who 'i bought for $mil, but Worth Billions!' -- or alternatively, per asciilifeform's hypothesis, simply happy to sit on the whole thing for nsa & co. who not eager to see even toy example of cpu-and-os-sans-overflows rise from bottom of ocean )

22:25

oughta collate all of this, not only for comprehension, but for future readers, I don't think we'll be getting that promised essay out of phf

22:25

(trilema) 2018-01-22 phf: scheme-79 is not "whole thing published", which is something that i said in the logs multiple times also. there's AI memoes of variying detali, but the actual toolset, something called daedalus, and the corresponding daedalus files that actually describe the chip, are nowhere to be found. also dies will not be particularly interesting because the whole point of daedalus was semi assisted layout

22:27

nsa et al are in the 'maintain perma-insecurity' biz, and not even need to do very much at all, the useful idiots do most of their work for'em

22:27

( see e.g. & elsewhere )

22:30

their methods worked even on soviet bean counters, who didn't need much in the way of convincing to copy pdp11, so as to run (conventiently available) us warez, complete with exploitable defects

22:31

fast-fwd to current-day ru, where 'elbrus' is a sparc (!) with x86 emulation mode (to conveniently run winblows warez, shit thee not) and whole thing in fact made in tw

22:32

simply cuz that's the only approach to life that makes sense to a bean counter

22:34

not exactly about to make an appeal to putin kek

22:34

try selling any of these on the notion that software can be correct + comprehensible -- and will get same reaction as if said 'one can park tank up yer arse'

22:36

I'll posit tho that regardless of end goal being ns 2.0 there needs to be a somewhat more sane basis to build off

22:36

anything more sane than above .ru 'elbrus'

22:36

crtdaydreams: point being, even folx who currently the loudest re 'sovereignty' and sticking it to the reich -- still stepping on same rake they've been jumping up and down on since ~1970, without interruption, and then seem surprised when their irons fail in various convenient (for reich) ways

22:37

nobody's aboutta 'sane foundations' because the payoff for it (and that's if you somehow get hold of people who will actually do the job, rather than convert the dough into yachts) is on order of 30-40yrs

22:38

and current-day mandarins don't think on that time scale

22:39

so what, wait for the tower to topple from complexity?

22:40

arguably happened long ago

22:40

or is french renaissance flavour needed 'behead the royalty'

22:41

nfi what concretely 'needed' in this sense. historically 'tower falls' results in smaller towers built from the pieces of the old, along same idiot pattern that sank the original

22:42

neh, don't underestimate the capacity for people to pile shit on shit, instead of digging another whole simply build a taller tower

22:42

*

asciilifeform must bbl.

22:42

ta

22:45