Forum logs for 12 Jun 2019

Monday, 16 March, Year 12 d.Tr. | Author:
feedbot: http://trilema.com/2019/the-exceptionalist/ << Trilema -- The Exceptionalist [02:49]
spyked: http://trilema.com/forum-logs-for-11-jun-2019#2539631 <-- lol, I had precisely the same experience! [03:30]
a111: Logged on 2019-06-11 21:38 diana_coman: well, the 18yo me at programming exam-to-get-qualification-diploma got told directly (out of prof's sheer exasperation I suppose): "next time, don't think that much!" [03:30]
spyked: I'd made a small toy proggy to draw a mandelbrot set, zoom in etc. when the time came to show it, no one in the room even knew what a fractal was, although ~all~ the teachers there were math-informatics graduates [03:31]
mircea_popescu: jaysus [03:42]
mircea_popescu: in other news, pillars of eternity is a fucking great "good old Infinity engine days". up there with planescape : torment i'd say (not as deliberately-artsy-quirky-???). [09:11]
mircea_popescu: absolutely must play. [09:11]
diana_coman: me is basically playing planeshift : torment [09:17]
diana_coman: can't beat it for that part really :P [09:18]
mircea_popescu: lmao [09:25]
mircea_popescu: moar liek 0x706c 616e 6573 6869 6674 203a 2074 6f72 amirite. [09:26]
mircea_popescu: anyone wanna do a small paypal for me ? [10:52]
mircea_popescu: danielpbarron, ? [10:52]
mircea_popescu: trinque, http://p.bvulpes.com/pastes/DmWFl/?raw=true [12:01]
feedbot: http://qntra.net/2019/06/rowhammer-being-used-to-read-from-vulnerable-ram/ << Qntra -- ROWHAMMER Being Used To Read From Vulnerable RAM [12:22]
mircea_popescu: noo ?! [12:31]
mircea_popescu: and here i thought that's just what it's for. severe case of who could've predicted! [12:32]
asciilifeform: orig for write. [12:32]
asciilifeform: also claims to work on ecc boxen, but asciilifeform not read $item yet and has nfi how substantial is claim [12:33]
mircea_popescu: nfw ?! [12:33]
asciilifeform: on the surface, seems to use the ecc corrector delay as sidechan [12:34]
* asciilifeform still doesn't grasp why is it that no one sells mobos where can use redundant (majoritated, rather than failover, the latter is dime a dozen) banks of ram. [12:35]
mircea_popescu: cuz made in china, and chinese man only does what granpa did. [12:43]
BingoBoingo: Similarly there's a lack of SRAM and boards supporting it. [12:46]
diana_coman: Unfortunately, as Cojocar et al. [13] show, the synchronous [12:52]
diana_coman: nature of the ECC correction algorithm typically exposes such [12:52]
diana_coman: information through a timing channel, where memory accesses [12:52]
diana_coman: that require error correction are measurably slower than normal [12:52]
diana_coman: ha [12:52]
diana_coman: from the paper so apparently yes, on ecc too [12:52]
asciilifeform: BingoBoingo: i dug into this last yr, it is worse than uninitiated can picture, nobody's even making reasonably-sized sram (deliberately not counting here various 'pseudo-s' etc abominations) [12:57]
asciilifeform: diana_coman: it makes sense, if it can't correct in O(1) cuz it has to interrupt a burst (all dram since mid-90s runs in long bursts much of the time, typically fill 1 or moar cache lines) then leaks via time. [12:58]
asciilifeform: even if doesn't interrupt a burst, still leaks 'fact of' [12:59]
diana_coman: aha [13:01]
asciilifeform: BingoBoingo: i actually went through coupla hundred datashits in late '17, for various srams. after rejecting various 'pseudo' (i.e. internally dram, with built-in refresher) , left with buncha items, typically <=64MB, where in small print, if you actually use it as random access (i.e. not in long burst) it runs at 1995 speeds (7-15ns typically access) [13:01]
asciilifeform: there's nothing physically impossible or even especially difficult about making gigantic sram -- you simply need 6 transistors for erry 1 that'd be in the dram yer replacing. simply, no one does. [13:04]
asciilifeform: cuz britneykonsoomer dun need it for anyffin. [13:05]
BingoBoingo: Once again mass market makes actual market products impossibru because BrittneyChain [13:05]
asciilifeform: BingoBoingo: you can still buy all the 1980s-style srams (incl. rad-hard) you like, 128kB 10ns chip is about 4bux. these suffice for the icbms and wherever else folx actually care about bits staying put. what no one makes is sram that'd fit the bloat of 'modern' pc os etc. [13:07]
asciilifeform: could even say that erry piece of soft gets the irons that it deserves. [13:08]
asciilifeform: ( why should 8GB of 'drepper-os' or winblowz get 8GB of sram ? it'd be rather a penicillin for a corpse ) [13:09]
BingoBoingo: And yet, plenty of fab for shitcoin ASICs [13:10]
asciilifeform: cuz they ~pay, lol [13:10]
asciilifeform: see mircea_popescu's earlier pt re why. [13:11]
a111: Logged on 2019-06-06 07:23 mircea_popescu: http://btcbase.org/log/2019-06-05#1917398 << nothing to laugh, it's exactly correct. the long standing slight margin on mining over trading is exactly this, "market participant voluntarily set cost of privacy". [13:11]
BingoBoingo: Meanwhile in local lulz https://www.montevideo.com.uy/Noticias/Desde-su-implementacion-regulacion-de-Cannabis-arrebato-unos-US-25-millones-al-narco-uc721097 << Nothing of the sort. The folks who used to sling one dope now sling another dope. [13:13]
asciilifeform: BingoBoingo: from asciilifeform's archaeologies, got hypothesis that market for standalone srams is 99% for repair/reproduction of '80s-'90s designs -- current-day ic which require sram tend to 1) need a few kB, maybe a MB or 2 , and 2) include it on-die [13:14]
asciilifeform: e.g., last pc that i personally opened that had standalone sram for l2 cache, was a 486 dx2. [13:15]
asciilifeform: errything after that, rolled it onto die [13:16]
asciilifeform: ( recall that light in vacuum goes 30m / ns , and charge in solid conductor -- yet slower ) [13:17]
asciilifeform: *30cm [13:18]
asciilifeform: lol [13:18]
asciilifeform: btw this figure is in errybody's kindergarten textbook, but is ~extremely~ overoptimistic for electrical machine, where if you have long conductor, it also has capacitance & inductance, if you want to vary voltage on opposite end of it, you have to let it discharge... ) [13:20]
asciilifeform: in asciilifeform's various benchmarkisms (ffalogy etc) found, unsurprisingly, that just about all observed variation in measured performance among current decade of x86 irons, is almost entirely proportional to cache size. [13:22]
asciilifeform: serious 'non-vonneuman' box would have no standalone rams at all, but simply buncha cpu , each with small mipslike item surrounded by die 99% fulla sram. [13:24]
feedbot: http://thetarpit.org/posts/y05/093-hunchentoot-i.html << The Tar Pit -- A review of Hunchentoot's code history [14:17]
feedbot: http://qntra.net/2019/06/ongoing-ebola-outbreak-crosses-border-into-uganda/ << Qntra -- Ongoing Ebola Outbreak Crosses Border Into Uganda [15:17]
trinque: mircea_popescu: http://p.bvulpes.com/pastes/ScL9b/?raw=true [15:25]
mircea_popescu: trinque, pretty cool. no big rush on this side, jsut tryna keep tabs pn things. [15:58]
trinque: np at all [16:11]
BingoBoingo: In US election meddling abroad https://archive.is/1ViEg [16:13]
BingoBoingo: "En ese punto, la mujer encuestada le preguntó quién realiza la encuesta, a lo que el interlocutor asegura que la Fundación Latinoamericana para la Promoción de la Democracia. A continuación ella le preguntó si la fundación tenía sede en Uruguay, a lo que el encuestador respondió que sí. Pero esa fundación no tiene sitio web, ni forma alguna de contacto disponible. " [16:14]
mircea_popescu: heh [16:23]
BingoBoingo: The confirmation though is the "español neutro" which means pointedly not Uruguayan [16:31]
feedbot: http://bimbo.club/2019/06/the-adventure-of-goosey-boy/ << Bimbo.Club -- The adventure of Goosey boy. [21:15]
Category: Logs
Comments feed : RSS 2.0. Leave your own comment below, or send a trackback.
Add your cents! »
    If this is your first comment, it will wait to be approved. This usually takes a few hours. Subsequent comments are not delayed.