02:29 |
verisimilitude |
I want to pay someone with a cryptocurrency. However, he doesn't accept Bitcoin, but Ethereum. What are your suggestions for a geographical service of which I may make use to send money to an Ethereum wallet? |
| |
~ 25 minutes ~ |
02:55 |
adlai|text |
verisimilitude: search around for "instant exchange", these spring up like weeds. obviously they have counterparty risk, they always work in the way most favorable to their operators, rather than actually using scripts to make the two payments mutually continguent. |
02:56 |
adlai|text |
the original ones did not require registration of any sort: on first page, HTML form for source and target coins, target address, and amount; click, receive ephemeral sending address for source coin, and amount to pay. |
02:57 |
* |
adlai|text has not used any of these recently, although a few years ago they did all seem to begin requiring some amount of additional info, e.g. email address, or worse, an entire account registration process. |
02:58 |
* |
adlai|text may have misunderstood what verisimilitude meant by 'geographical' |
02:58 |
verisimilitude |
I meant like meeting with someone to do this. |
02:59 |
adlai|text |
aha. localbitcoins.com is/was a centralized site for this kind of matchmaking; and these days, there is a decentralized java hairball whose name I'm inexplicably blanking on. |
03:00 |
adlai|text |
bitsquare.io |
03:01 |
verisimilitude |
I'll look into it; I appreciate it, adlai|text. |
03:01 |
adlai|text |
I think the original idea of BitSquare was to matchmake traders who wanted to trade coin in return for bank transfer of fiat, although it is definitely possible to do one coin for a different one. |
03:03 |
* |
adlai|text was disappointed to find it not useful for physically locating traders for a cash payment, a few years ago, although in retrospect this is not too surprising. it is designed with the assumption that the payments in both directions are electronic and reasonably immediate, even if they take some time to confirm. |
03:04 |
adlai|text |
if able to trade cash... why do you need the java hairball? simply each point pistol at counterparty, smile, and trade; holster upon reaching sufficient confirmations. |
03:09 |
adlai|text |
if not own pistol, 'not ready for bitcoin' (tm) (/s) (r) |
| |
~ 13 hours 1 minutes ~ |
16:11 |
phf |
http://logs.nosuchlabs.com/log/asciilifeform/2022-06-19#1107606 << was just idle curiosity, i wasn't following the conversation too closely, so didn't realize it was a bug. i thought it's something i need to be aware in implementation |
16:11 |
dulapbot |
Logged on 2022-06-19 17:03:09 thimbronion: phf if you can figure out how to reliable reproduce I would be much obliged |
16:11 |
asciilifeform |
billymg: in what format is your db dump ? zcat log_db.gz | psql ... aint eatin' it: gzip: log_db.gz: not in gzip format |
16:13 |
billymg |
asciilifeform: it's produced by this command: http://paste.deedbot.org/?id=7O9J |
16:13 |
phf |
kek |
16:13 |
phf |
billymg: you might want to add a |gzip there in the pipe so it's actually a gz file |
16:14 |
asciilifeform |
phf: bug; [x|y|z..] normally marks receipt of multiple copies of hearsay msg when >1 peer sends. which obv. aint aboutta happen on a station with 1 peer (not to mention oughta mark distinct senders) |
16:14 |
billymg |
phf: i thought that's what the -Fc flag did |
16:14 |
billymg |
format: compressed |
16:14 |
phf |
billymg: disregard. i've forever been corrupted by plan9, don't believe in flags |
16:15 |
billymg |
`pg_restore -U nsabot -d nsalog < log_db.gz` works for me to import |
16:15 |
asciilifeform |
pg_dump --username yeruser yerdb | gzip > log_db.gz |
16:15 |
asciilifeform |
^ what is used on asciilifeform's box |
16:15 |
billymg |
asciilifeform: should've included in the readme if it was supposed to match 1:1 |
16:16 |
asciilifeform |
hm indeed omitted in readme lol |
16:16 |
phf |
billymg: apparently -Fc is not really compress, it's "custom" which also compresses. so it's probalby not a gz |
16:16 |
asciilifeform |
it defo aint a gz |
16:17 |
asciilifeform |
nor is the format an editable text ( was gonna grab the req'd lines ) |
16:18 |
billymg |
asciilifeform: if you just want the lines for pest to import with eat_dump.py you can grab them from my ilog |
16:18 |
billymg |
i have a script for that, one sec... |
16:19 |
* |
asciilifeform had a similar, lol, but can't seem to locate a working copy; been too long since last did this... |
16:19 |
billymg |
script for ripping ilogs: http://paste.deedbot.org/?id=hYsI |
16:20 |
billymg |
for bitdash.io run with e.g. : ./download_raw_logs.sh 1006546 1006561 http://logs.bitdash.io/log-raw/pest ml.txt |
16:20 |
billymg |
change s / e indexes to what you want |
16:20 |
asciilifeform |
ty billymg |
16:20 |
billymg |
np |
16:20 |
asciilifeform |
billymg: does your default index start at 1000000 like asciilifeform's ? |
16:20 |
billymg |
yes |
16:20 |
billymg |
it should |
16:21 |
billymg |
i'll update the format on the log dump too, apparently gzip compression has some gains over whatever pg_dump uses |
16:22 |
billymg |
yup: http://logs.bitdash.io/pest/2021-11-08#1000000 |
16:22 |
bitbot |
(pest) 2021-11-08 shinohai: That worked, ran /unpeer then /peer followed by /at with ip and now no double entry |
16:27 |
* |
billymg for some reason thought that pg_dump flag actually produced gzip format output |
16:29 |
asciilifeform |
billymg: ok; ate billymg's log, plugged in dulapbot once more. lessee how worx |
16:30 |
* |
asciilifeform still not saw reply on own station from the helloworld, oddly |
16:30 |
dulapbot |
(pest) 2022-06-20 asciilifeform: helloworld |
| |
~ 1 hours 54 minutes ~ |
18:24 |
asciilifeform |
meanwhile in #p. |
18:24 |
dulapbot |
(pest) 2022-06-20 asciilifeform: had a thought recently, that just about all of the gnarly embargo buffer logic in pest protocol is only req'd because irc frontend is a 'teletype', i.e. immutable |
18:25 |
asciilifeform |
^ 'heretical' but ultimately rather obv. observation imho |
18:26 |
asciilifeform |
... and observe that longbuffer is already not treated 'like teletype', i.e. it can be retrofilled by getdata |
18:27 |
* |
asciilifeform long suspected that the protocol can be simplified substantially, and this is a primo place for where |
| |
~ 57 minutes ~ |
19:25 |
billymg |
i'm going to rent a server as a stopgap until another dulap becomes available in asciilifeform's rack. i'm going to move the crawler (currently on ec2) and the logger (currently on an rk) to this box. additionally i'll spin up a trb node on it, wainot |
19:26 |
billymg |
centos 7 or debian 8? (the other options are newer versions of the former or ubuntu flavors) |
| |
~ 32 minutes ~ |
19:59 |
verisimilitude |
Yes, it's obviously so, asciilifeform. |
20:00 |
verisimilitude |
I'd consider renting a server from asciilifeform, but it hardly makes good sense to have our DDoS-resistant and distributed chat protocol use the same few machines. |
20:06 |
* |
asciilifeform encourages, naturally, folx to subscribe to asciilifeform's rack, but ~not~ to ~only~ asciilifeform's rack |
20:06 |
* |
billymg nods |
20:07 |
billymg |
my cr box will be used for pest |
20:07 |
billymg |
asciilifeform rack box for various www services |
20:07 |
billymg |
and until that one is ready, a shitbox running centos 7 or debian 8 |
20:08 |
billymg |
if both of those suck, i'll choose debian because i like the name more |
20:09 |
billymg |
i imagine both will have python 2.7, which is my only hard requirement (even the ec2 "amazon linux" box shipped with that as the default iirc) |
20:09 |
asciilifeform |
billymg: may be worth, if not feeling like hairshirting w/ gentoos, to try 'artix' ( a non-poetteringized 'arch' asciilifeform had used successfully in commercial works ) |
20:09 |
asciilifeform |
( an otherwise unremarkable binaristic linux ) |
20:10 |
billymg |
asciilifeform: this is really just to get something up asap, so i'm limited to OS options that are not their "client sends ISO" option |
20:10 |
asciilifeform |
( suffers from some ughs but afaik none pertinent to a headless use ) |
20:10 |
dulapbot |
Logged on 2021-05-17 09:30:33 asciilifeform: 'artix' indeed is systemd-free. but many of the packages (e.g. 'cups') do not work; and device hotplugging is mysteriously broken (mass storage -- worx; usb-to-serial -- not; and this despite a known-good kernel. it's whatever the thing has instead of udev that's a dud) |
20:10 |
asciilifeform |
a |
20:11 |
billymg |
i searched the logs re debian/centos, seems people prefer centos 6 (pre systemd apparently) |
20:11 |
billymg |
but not one of the choices |
20:11 |
asciilifeform |
billymg: iirc suffered from python 2.6 tho |
20:11 |
billymg |
centos 7? |
20:11 |
asciilifeform |
6 |
20:11 |
billymg |
ah |
20:13 |
billymg |
looks like debian 8 ("jessie") includes python 2.7 and 3.4, so should work |
20:15 |
* |
asciilifeform solidly in favour of depythonization, but realizes that it's a long way off |
20:15 |
dulapbot |
Logged on 2022-04-11 22:09:05 asciilifeform: sh, perlism, pythonism, 'must die'(tm) imho. |
20:20 |
verisimilitude |
The only reasonable way to kill sh is to stop using UNIX. |
20:21 |
asciilifeform |
verisimilitude: reminiscent of legendary reply of brit war ministry to churchill's 'how do we get rid of german uboats?' -- 'most immediately -- if could boil the ocean' |
20:25 |
verisimilitude |
It's easier to stop using UNIX. |
20:25 |
asciilifeform |
marginally |
20:26 |
asciilifeform |
( can, e.g., throw out the comp entirely, but presumably wasn't what verisimilitude meant ) |
20:26 |
dulapbot |
Logged on 2022-06-17 17:10:23 asciilifeform: verisimilitude: if you see yerself as involved in pure mathematics, rather than aiming to actually solve sumthing in physical reality w/ a comp on sumthing resembling human timescale -- the machine is arguably a distraction, and you oughta be working w/ pen & paper |
20:27 |
asciilifeform |
observe that there aint in fact any alternatives, if keeping the comp (e.g. microshit victims who attempt any serious works are still quite certainly 'using unix', arguably without most of the 'good parts' at that) |
20:30 |
asciilifeform |
atm errybody stuck with a unix (or crippled imitation), even the lolcats-on-pnoje folx. |
20:31 |
* |
asciilifeform recalls as recently as ~decade ago seeing e.g. macos9 ref hdrs in www log; but these have long ago vanished |
20:32 |
verisimilitude |
No, that's one way. |
20:32 |
asciilifeform |
for verisimilitude's purposes, asciilifeform suspects, phf's plan9 is 'a unix' |
20:32 |
verisimilitude |
Yes. |
20:32 |
verisimilitude |
It's the same vile shit, make no mistake. |
20:33 |
verisimilitude |
``Act from reason, and failure makes you rethink and study harder. Act from faith, and failure makes you blame someone and push harder.'' |
20:34 |
verisimilitude |
Erik Naggum wrote that, and it also applies to Plan 9. |
| |
↖ |
20:35 |
asciilifeform |
verisimilitude: asciilifeform strongly convinced that anyffin one could write for x86 (and similar 'c-machine' iron) will end up effectively 'a unix', like in the old su joak where 'i take home parts from sewing machine factory erryday, but evrything i assemble always ends up being a maxim mg' |
20:37 |
asciilifeform |
( the general idea here, but concrete particulars variously elsewhere in oldlogz ) |
20:37 |
dulapbot |
(trilema) 2017-04-03 asciilifeform: trinque: the flaw is that you gotta support a megatonne of liquishit for even nic to work -- dma, page tables, etc |
20:39 |
amberglint |
Hello everyone |
20:39 |
asciilifeform |
wb amberglint ! |
20:39 |
amberglint |
asciilifeform: Have you seen this? http://www.bitsavers.org/pdf/symbolics/Sunstone_Architecture_198711.pdf |
| |
↖ |
20:40 |
amberglint |
Someone dumped the spec for the unreleased Ivory successor |
20:40 |
asciilifeform |
amberglint: nope, not prev. seen |
20:40 |
asciilifeform |
amberglint: don't suppose the dumper also dumped a runnable os for it ? |
20:41 |
amberglint |
Doesn't look like there's anything else, I suspect it's what Scott McKay found earlier |
20:41 |
asciilifeform |
a |
20:41 |
asciilifeform |
still potentially interesting from archaeological pov |
20:41 |
* |
asciilifeform wonders whether phf seen it |
20:42 |
amberglint |
Gary Palter also mentioned something called the P-Machine, apparently Symbolics' unreleased multiprocessor hardware |
20:42 |
amberglint |
No docs yet |
20:42 |
verisimilitude |
I'm not convinced, asciilifeform. |
20:42 |
asciilifeform |
amberglint: iirc that one was obliquely hinted at in certain press releases, and there was a bit of hype re a planned use on a satellite |
20:42 |
verisimilitude |
Consider ITS. |
20:43 |
asciilifeform |
verisimilitude: i rec to see whether still unconvinced once you've tried revving up nic/disk/vga on x86/64 from bare iron and seen what's involved and how in particular brittle |
20:44 |
asciilifeform |
verisimilitude: its was 'a unix' in so far as asciilifeform concerned. |
20:44 |
verisimilitude |
No, I'd rather not. |
20:44 |
* |
asciilifeform unsurprised |
20:45 |
verisimilitude |
I should read more on the ice 40. |
20:46 |
amberglint |
Looks like the Sunstone is a register rather than a stack machine |
20:46 |
amberglint |
32 general purpose window registers |
20:47 |
amberglint |
Kinda like the SPARC I suppose |
20:48 |
amberglint |
58 defined instructions |
20:49 |
amberglint |
No microcode |
20:49 |
* |
asciilifeform always found interesting that the commercial lispm's were extremely 'conventional' internally, rather than e.g. 'scheme83'-style apparatus with iron gc , cons-only memory, etc |
| |
↖ |
20:49 |
dulapbot |
(trilema) 2017-01-07 phf: http://btcbase.org/log/2017-01-07#1598294 << scheme83 is like a "canticle for leibowitz" artifact. "published design" is overstatement of the century. scraps of published memos and reports spread over out of print conference proceedings, the bulk of actual technology needed to recreate probably somewhere on a TAPE. i don't know where you got that mask generator runs on scheme83. the entire production stack was for mit cadr |
20:50 |
asciilifeform |
not all that surprising , tho, iron makers have always been the cautious type, on acct of nre cost |
20:51 |
asciilifeform |
... nor were various paradigm shifts that theoretically 'easy' today available on '80s fab |
20:52 |
asciilifeform |
( e.g. 'orthogonal persistence' would've had to mean magnetic core mem , given speed of period disk, etc ) |
20:54 |
asciilifeform |
e.g. symbolics famously 'pushed the envelope' on iron in all kindsa ways (take simply the fact of their buying own glass factory to make the high-res crt, which was avail. nowhere off-shelf) but can only push 'so' far |
20:56 |
amberglint |
It took 10 to 12 man-years to design the Ivory while the MicroVAX took 70 to 80 man-years around the same time |
20:57 |
asciilifeform |
amberglint: well, considering that they were designing it on the 3600, and with fully lispized 'ns' etc cad orchestra -- unsurprise |
20:58 |
* |
asciilifeform wouldn't be astonished if learned that microvax , or parts thereof, were still drawn with ink, on whatman paper |
20:58 |
amberglint |
Have you tried to do anything with the cracked NS? |
20:58 |
amberglint |
It's a shame the docs for it are missing |
20:58 |
asciilifeform |
amberglint: nope. part of why was that not found any examples whatsoever to work with, nor docs |
20:58 |
phf |
http://logs.nosuchlabs.com/log/asciilifeform/2022-06-20#1107706 << you're such an insuffarable cunt, i'm prepared to defend php just to not be in the same camp with you. shit basic on altair is a superior system to the idea system that only exists in verisimilitude's head. did you know that in my system there are no flaws? did you also know that i've solved all practical problems by never doing anything practical? fuck. you. |
| |
↖ |
20:58 |
dulapbot |
Logged on 2022-06-20 16:31:51 verisimilitude: Erik Naggum wrote that, and it also applies to Plan 9. |
20:59 |
asciilifeform |
lol |
20:59 |
* |
asciilifeform expected almost exactly ^ eventually |
21:00 |
verisimilitude |
Calm down. |
21:00 |
asciilifeform |
phf: imho verisimilitude runs a very serious risk of ending up as an 'why won't they read my elementary proof of fermat's' fella |
21:00 |
dulapbot |
(trilema) 2017-07-06 erlehmann: http://web.mst.edu/~lmhall/WhatToDoWhenTrisectorComes.pdf |
| |
↖ |
21:01 |
verisimilitude |
I'll try to avoid a fate of useless mediocrity. |
21:04 |
* |
asciilifeform ~decade ago and formerly used to often get 'fan mail' of similar angle to phf's to verisimilitude -- 'stfu about notional systems that only in yer head!' ; naodays not gets these anymoar, but sometimes 'why aintcha writing about lispms! write about lispms!' now, lol |
21:04 |
verisimilitude |
I'll try to a more sufferable cunt eventually, phf. |
21:05 |
asciilifeform |
'what's with this boring bitcoin shite, and arithmetics in ada, where's the lispm' etc |
21:06 |
asciilifeform |
hard to please erryone. |
21:06 |
* |
asciilifeform when answers ' the lispm could be in this 'ice40' if it were 300x larger' -- not usually satisfies interlocutor |
21:07 |
* |
asciilifeform wouldn't be satisfied either, if it were sumbody else answering, lol |
21:08 |
verisimilitude |
All of the large enough FPGAs are locked down, right? |
21:08 |
asciilifeform |
verisimilitude: not merely 'locked down' but ungranular |
21:08 |
amberglint |
Isn't the ECP5 is a magnitude larger and mostly reverse-engineered except for the SerDes part? |
21:09 |
asciilifeform |
amberglint: ~10x larger |
21:09 |
asciilifeform |
(i.e. still unlikely to fit a i386, much less a lispm of any description ) |
21:10 |
asciilifeform |
subj for ref. |
21:10 |
asciilifeform |
tops out at 85k LUTs. |
| |
↖ |
21:10 |
verisimilitude |
How hard would it be to chain several together with message-passing? |
21:11 |
asciilifeform |
verisimilitude: massive bottleneck (in both time & space) |
21:12 |
amberglint |
Google fabs opensource 130nm chips for free apparently: https://efabless.com/open_shuttle_program |
21:12 |
verisimilitude |
Yes, with mystery meat added. |
21:12 |
asciilifeform |
'Projects must use a common test harness and padframe based on the Caravel repo. New projects should start by duplicating or forking the Caravel User Project repo and implementing their project using the user_project_wrapper. The Caravel repo is configured as a submodule in the project under the ‘caravel’ directory.' |
21:13 |
amberglint |
The Caravel is a problem, yeah |
21:13 |
asciilifeform |
seems like it's a 'build on our riscv' thing, rather than 'charity fab for ab initio' |
21:14 |
verisimilitude |
By the by, hello amberglint, I don't believe we've met before now. |
21:15 |
amberglint |
I was dropping by occasionally for many years |
21:15 |
asciilifeform |
( and see also re fabs ) |
21:15 |
dulapbot |
Logged on 2020-08-20 19:28:15 asciilifeform: actually went an' investigated a number of firms that offer 'small-run' ic fab in various processes. they all have these in common. |
21:15 |
dulapbot |
Logged on 2020-01-20 20:54:06 asciilifeform: there is a 2017 thread where i contacted various small-run fab houses, unfortunately dun have the link handy (and the #s are likely to be out of date nao) |
21:15 |
asciilifeform |
( in particular ) |
21:15 |
dulapbot |
Logged on 2020-01-20 21:32:07 asciilifeform: later threads w/ some detail re subj : 1 2 3 |
21:17 |
* |
asciilifeform one time coupla yr ago had a brief exchange with a cn group which initially resembled 'nigerian prince' in flavour but turned out to exist as-described; wanted asciilifeform to design a lispm to be fabbed in pekin; but after some back'n'forth turned out that they wanted asciilifeform to also ~finance~ part of it, lol |
21:17 |
asciilifeform |
... was the end of the conv. |
21:18 |
asciilifeform |
'talk to rockefeller, not me' |
21:19 |
amberglint |
Lol |
21:19 |
* |
asciilifeform never thought it'd be difficult to make clear that one aint rockefeller... |
21:19 |
verisimilitude |
Well, was the conversation written professionally, or as in this channel? |
21:20 |
asciilifeform |
verisimilitude: chinese-professionally lol |
21:20 |
asciilifeform |
not went far. |
21:21 |
asciilifeform |
for asciilifeform to take any such interlocutor seriously, the dough would have to flow in very opposite direction to what described above. |
21:22 |
* |
asciilifeform aint aboutta do some cn fella's phd thesis for him for phree |
21:25 |
verisimilitude |
Now that I mull over it, I can't even get this kind of e-mail. |
21:25 |
asciilifeform |
verisimilitude: didja need'em for sumthing, lol |
21:25 |
asciilifeform |
verisimilitude: when you bake sumthing in physical world, you'll prolly start getting'em , for all the good it does |
21:27 |
* |
asciilifeform walking around 'sunstone' doc, wonders 'why the stack pointers etc. exposed to the os/programmer?' -- but equally legit q re 'ivory' & friends |
21:27 |
dulapbot |
Logged on 2022-06-20 16:37:30 amberglint: asciilifeform: Have you seen this? http://www.bitsavers.org/pdf/symbolics/Sunstone_Architecture_198711.pdf |
21:28 |
asciilifeform |
( p. 112 & elsewhere ) |
21:29 |
asciilifeform |
this was historically asciilifeform's crit of the 'big' lispms -- 'screen door on submarine' |
21:30 |
amberglint |
That's a good question |
21:30 |
asciilifeform |
i.e. appears that symbolics (can't speak re xerox et al, and e.g. lmi/texas instruments irrelevant, as afaik only sold cadr clones) didn't actually have the confidence to bake a fully 'sealed' lisp cpu, where pointerism not exposed to programmer |
21:31 |
asciilifeform |
instead left room for all kindsa 'late night with pizzas' kludgery |
21:31 |
dulapbot |
Logged on 2022-06-11 11:41:04 phf: it's also ancient development practices, which i'm usually all for, but this is the case where some moved on made sense. global package level state in dynamic variables kind of things |
21:32 |
asciilifeform |
this is possibly 1 piece of the puzzle 'wai bolix died?' . ultimately was making slightly-fancier 'sparc' and the folx making unashamed 'regular' sparc etc. ate'em and laughed. |
21:34 |
* |
asciilifeform not aboutta 'hurrr, designers at bolix -- idjits!!' -- folx were working under obv. economic constraints, even at the peak of phree reagan moolah |
21:35 |
asciilifeform |
ic fab -- was (as today) expensive, and folx who dared to make 'martian' arch, which would need multiple revisions to become of any use whatsoever, and with access to the req. moolah -- few to nonexistent. |
21:36 |
asciilifeform |
afaik mit's 'scheme83'/84 were the 1st (and last) such attempts to be made public. |
21:39 |
phf |
asciilifeform: i think schemexx cpu was a novel idea, traditionally wild pointer problem was solved at the level of ucode. at least for bolix i suspect by the time they were making a cpu (and they've poached everyone they could from schemexx project), they already had to deal with legacy architecture. |
21:39 |
asciilifeform |
phf: aha, there wasn't any way they could 100% run away from 3600 legacy |
21:40 |
phf |
in fact i wouldn't even call it legacy architecture. their approach was very much a ball of mud kind of hacking on architecture level also. "take these common circuits and replace them with pals" later "take this one big circuit and put it vlsi", rather than architecture 2.0!11 |
21:40 |
asciilifeform |
with very minor cleanup -- yes |
21:41 |
asciilifeform |
( see e.g. ) |
21:41 |
dulapbot |
Logged on 2020-08-12 11:31:20 asciilifeform: both faced very similar problem in 1980 -- how to go from 'two tonnes of 74xxx' to vlsi. and at ~same time~ fund refinement of os and application programs (at the time, useful software was expected to ~ship with a comp~, like you expect tyres to come w/ a new auto ) |
21:41 |
verisimilitude |
The problem with load-gc-copy isn't immediately obvious to me. |
21:42 |
verisimilitude |
I likely lack context with the rest of the document is all. |
21:42 |
asciilifeform |
verisimilitude: you know how programmer doesn't expect to be able to change how ADD, SUB, DIV, MUL work ? |
21:43 |
asciilifeform |
per asciilifeform's lights, gc oughta be precisely like that. |
21:43 |
asciilifeform |
it either worx -- in which case belongs in the irons and wholly outside of purview of os -- or it's a kludge, in which case, well, what we have nao. |
21:44 |
asciilifeform |
'gc in operating system' is similar to the ancient lul of certain '80s micros where os had to manually trigger dram refresh |
21:44 |
asciilifeform |
(not even speaking of 'gc in user coad' idjicy there) |
21:44 |
verisimilitude |
Well, at least one model did allow changing how addition, subtraction, and others worked. |
21:45 |
verisimilitude |
But sure, I see. |
21:46 |
asciilifeform |
verisimilitude: until 'recently' , iron div/mul was rare, and these were normally implemented by users |
21:46 |
dulapbot |
Logged on 2022-05-16 17:50:04 asciilifeform: ( from ^ pov inclusion of #9' and #11 extravagant -- '80s iron made do with 'egyptians' virtually without exception ) |
21:46 |
asciilifeform |
but where included in iron -- typically used as given. |
21:48 |
asciilifeform |
... continuing with above -- per asciilifeform's lights, moving to/from ram and nonvolatile store -- also belongs in function of correctly-built iron rather than to be left to programmer. but already many times described this in depth, so won't repeat. |
21:48 |
asciilifeform |
current comp (and even '80s lispm) very much 8-pedal bus. |
21:48 |
dulapbot |
Logged on 2022-06-18 16:10:51 asciilifeform: verisimilitude: picture a bus w/ 4 brake and 4 gas pedals. 1 per wheel. |
21:49 |
asciilifeform |
( today in multicpu, sometimes over9000-pedal bus, lol ) |
21:51 |
verisimilitude |
I still like the idea of trying to use systolic arrays for everything. |
21:51 |
asciilifeform |
very easy to 'like idea'. nao draw a working machine... |
21:52 |
verisimilitude |
I'd like to, one day. |
21:53 |
asciilifeform |
'liking' is very, very cheap. then you start w/ physical objects, and find that even in absurdly simple circuit can spend month+ fighting metastability etc |
21:53 |
verisimilitude |
A systolic array really does seem to be something best used as a component of a larger system, however. |
21:55 |
asciilifeform |
iron makers (unlike programmers) historically not 'tards'. but extremely cautious, 'burned on hot, will blow on cold', because failure is very, very expensive. ( and if yer doing anyffin out of the ordinary -- virtually guaranteed for 1st n iterations ) |
21:56 |
asciilifeform |
and n can be arbitrarily large. |
21:58 |
asciilifeform |
today, even to 'try' is so expensive that only 'tbtf' can meaningfully play. and they make exclusively x86 and pnojes for 1e8 morons to lolcat on, outta sheer economic lock if nuffin else. |
21:58 |
verisimilitude |
I expect TBTF to expand to something Taiwanese. |
21:58 |
asciilifeform |
'too big to fail'(tm) |
21:59 |
asciilifeform |
see also e.g. |
21:59 |
dulapbot |
(trilema) 2015-07-26 asciilifeform: the coveted 'garden of eden' situation where ~there isn't a microshit~, when there is not a usg department-of-computing that has its tentacles in every orifice - won't be coming back until some quasi-mythical breakthrough decentralizes chip fabbing |
21:59 |
dulapbot |
(trilema) 2018-10-25 asciilifeform: imho the classical fab is an overwhelmingly incatronic tech, it centralizes unhealthily. |
21:59 |
dulapbot |
(trilema) 2019-04-17 asciilifeform: it's the most massively centralized industry that exists. there are actually moar independent satellite launchers than e.g. <= 70nm fabs. |
22:00 |
verisimilitude |
Yes. |
22:00 |
asciilifeform |
whole field is in a cultural trap from which there atm aint even a hypothetical exit. |
| |
↖ |
22:01 |
verisimilitude |
Isn't seventy nanometer small enough, with the fat trimmed? |
22:01 |
asciilifeform |
verisimilitude: likely even 1um is smallenuff (2um was good enuff for bolix) ; but it aint as if the old processes somehow cheap |
22:01 |
verisimilitude |
I was trying to recall just that, yes. |
22:03 |
asciilifeform |
and if you aint making '386' and rather want to drive modern drams, throw packets around at modern gb/s, etc. yer stuck w/ 'modern' densities |
22:03 |
asciilifeform |
the spittoon is in one strand!(tm)(r) |
22:04 |
verisimilitude |
Yes, we've already established how fucked everything is. |
22:04 |
asciilifeform |
dun hurt to elaborate re why. |
22:04 |
asciilifeform |
at any rate not negotiable part of any, even hypothetical & distant unfucking. |
22:08 |
* |
asciilifeform must bbl |
| |
~ 27 minutes ~ |
22:35 |
amberglint |
Speaking of slightly fancier sparcs, I've found a new MIT spin-off company that is selling a tagged RISC-V variant called Dover: https://www.dovermicrosystems.com/ |
| |
↖ |
22:36 |
amberglint |
All public info I managed to find: |
22:36 |
amberglint |
https://www.dropbox.com/s/4osfto8yljsxvgj/dover-sullivan2017.pdf?dl=0 |
22:36 |
amberglint |
https://riscv.org/wp-content/uploads/2016/01/Wed1430-dover_riscv_jan2016_v3.pdf |
22:36 |
amberglint |
https://web.cs.wpi.edu/~rjwalls/nesd/slides/SullivanDraperNESDFall2016Dover.pdf |
22:37 |
amberglint |
Someone's attempt to write an OS for it: https://web.mit.edu/ha22286/www/papers/MEng20_4.pdf |
22:37 |
amberglint |
It's not for workstations, but rather for "embedded" and "IoT" uses |
22:38 |
phf |
dat "thesis spacing" |
22:47 |
signpost |
http://logs.nosuchlabs.com/log/asciilifeform/2022-06-20#1107844 << worth mentioning the extent of the cultural trap, that garage fab is likely to get the same terrorist/pedo/tax-evader reception as garage-printed gun. |
22:47 |
dulapbot |
Logged on 2022-06-20 17:57:59 asciilifeform: whole field is in a cultural trap from which there atm aint even a hypothetical exit. |
23:02 |
verisimilitude |
It's even worse, because people expect the right to own guns. |
| |
~ 15 minutes ~ |
23:17 |
verisimilitude |
There's no constitutional right to unmolested computation. |
23:18 |
verisimilitude |
There's not even a constitutional right to being unmolested in practice. |
23:19 |
signpost |
a more intelligent civilization might've taken an abstract interpretation of the first and fourth amendments, but this is not that civilization. |
23:27 |
verisimilitude |
Perhaps the next civilization will. |
| |
~ 29 minutes ~ |
23:57 |
vex |
signpost, garage fab, is that where you park the truck on the driveway, do a quick hoover, and start baking ics? |