log☇︎
564300+ entries in 0.363s
asciilifeform: to china, or to me, and you, it's $0
asciilifeform: the point is not the cost.
decimation: yeah the bell labs guy said that they charge him $500k per year per seat
decimation: but doesn't give them a single step toward 'making their own'
decimation: asciilifeform: what it means is that they can enrich themselves privately by stamping out copies and selling shit on ebay
asciilifeform: what does it mean if asia has all of the factories, but is utterly dependant on winblows and a stack of monopoly turdware as tall as empire state building ?
decimation: yeah the bell labs guy has issues hiring because very few schools are willing to pay the bezzlars required to actually experiment with high-end process
asciilifeform: mircea is mistaken when he writes that china controls ic manufacture.
asciilifeform: aha and they won't even take your job if you aren't using cadence's cell libs.
decimation: asciilifeform: yeah and there are some esoteric telco applications too
decimation: as I recall, the guy claimed that most of the high-end fabs use cadence tools
asciilifeform: afaik the only customer is usg.
assbot: An Interview with Shahriar from The Signal Path - Quisquous Quivering Quadripole | The Amp Hour Electronics Podcast ... ( http://bit.ly/1A78JTP )
decimation: actually that reminds me, this is an interesting podcast: http://www.theamphour.com/228-an-interview-with-shahriar-from-the-signal-path-quisquous-quivering-quadripole/
asciilifeform: but, it is conceivable that - at this point - all new designs.
asciilifeform: and other 'on the cheap' jobs.
asciilifeform: just the bitcoin miners.
decimation: plus it's cash in their pocket
asciilifeform: also to prevent taking the design to another fpga.
asciilifeform: i.e. to prevent you from curing the ethernet card.
asciilifeform: at the risk of repeating the last 100+ xilinx threads - the closed architecture of -all- fpga vendors is specifically to enable this 'business model'
decimation: lol I didn't realize they had 'trialware' like that
asciilifeform: most of the more complicated ones being unusable in practice (e.g. ethernet controller that 'expires' after N frames)
thestringpuller: asciilifeform: i made an app, called the asciilifeform turd counter
asciilifeform: xilinx ships a set of identially-functioning turdlibraries for both languages.
decimation: all of which is abused by the vendor tools & magically wedged with vendor ip
asciilifeform: verilog is at least tolerable
decimation: yeah, that's pretty much the case. the vhdl codebase is littered with 'don't touch this' crazy code
asciilifeform: aka a turd.
asciilifeform: decimation: months of experimentation << the end result is a product which cannot be understood, function of which cannot be rationally explained, and which will break on the slightest deviation from the parent hardware
asciilifeform: conform to the timing diagram - or die.
asciilifeform: incidentally, fpga work shits straight into the faces of folks who think that 'anything can be slow-prototyped'
decimation: it can be done, it just takes many man-months of experimentation with a particular set of hardware
asciilifeform: virtually impossible to do well from scratch (vs vendor turdware) without the unobtainable internal docs
decimation: I've heard from folks who have that it is a pain in the ass
asciilifeform had that one printed & bound
asciilifeform: ask a 'bios tweak' aficionado.
asciilifeform: normally these are determined experimentally.
decimation: yeah w.r.t. exact timings, refresh rates, things
decimation: who knows if the authors had access to such knowledge
decimation: asciilifeform: there's probably an nda'ed talmud of errata that is supplied to the actual chipset dealers
decimation: ideally one would find a simultaneous set of 'hits' over an area so one could backtrack the cosmic rays
asciilifeform: not to my satisfaction.
decimation: yeah, that's a fair point
asciilifeform: just forget a refresh sometimes. or violate one of the many mandatory command sequences.
asciilifeform wrote a ddr2 controller for 'xilinx' chip once. it is amazingly easy to create a dysfunctional one with behaves like, for instance, the one pictured in that paper.
asciilifeform: decimation: overpowers << skeptical until demonstrated in field (i.e. x86 pc) rather than laboratory (fpga with custom dram controller)
decimation: asciilifeform: re: ecc dram << the paper claims that they can cause so many errors that it overpowers 'correct one detect two' style ECC dram
decimation: slur << that's the dumbest thing I've heard yet
thestringpuller: http://en.wikipedia.org/wiki/1992_Fiesta_Bowl << so this was really a thing?
thestringpuller: "The richest 500 addresses have continued to accumulate bitcoin through all the highs and lows."
assbot: The Toxoplasma Of Rage | Slate Star Codex ... ( http://bit.ly/13ZGWIM )
asciilifeform: dignork: forget code. the thing, as stated on own site, doesn't even threaten to make elementary sense.
cazalla: really just seems a way to scam donations
Adlai: oh crap, they slapped a GPL on the null codebase. now every new project must be GPLed
cazalla: i remember reading about these u99 guys, have another project named coinmesh
asciilifeform: it's dreary and insults the reader's intelligence merely by existing.
asciilifeform: i deliberately didn't even bring up the whole thing's (almost certain) reliance on tor and related idiocy
asciilifeform: arbitrators won't immediately run off and do whatever the fuck they like with the secret because... because what? magical drm goggles glued onto their skulls?
asciilifeform: no. why would they do that. we have l33t crypt0-d00dz now.
asciilifeform: what it actually takes in real life, to establish the kind of relationships needed
asciilifeform: 'slur' << i love how these nitwits show no symptoms whatsoever of having bothered to study how secrets were sold historically, from babylon to cold war, etc
Adlai: hm. haven't poked around this corner of the internet much lately
Adlai: https://news.ycombinator.com/item?id=8794707 << we're about to find out the width of the venn diagram intersection between #startups and here
Adlai: i'm almost curious to see what this download is
Adlai: https://github.com/u99/slur/stargazers <- who are these lemmings
asciilifeform: and disputes settled by showing the actual secret to five randomly selected lusers
Adlai likes how they talk about it in present tense, then later ask for money so they can build it
assbot: CDC reports potential Ebola exposure in Atlanta lab - The Washington Post ... ( http://bit.ly/13wJdtY )
asciilifeform: memory paper << snore. 1) serious folks using dram - use ecc dram. 2) serious folks do not allow opponent to execute arbitrary strange on their von neumann box 3) serious folks doing truly serious things use sram.
scoopbot: New post on Trilema by Mircea Popescu: http://trilema.com/2014/the-boy-blog-network/
assbot: Devuan - the GNU/Linux by Veteran Unix Admins. ... ( http://bit.ly/13wyXBY )
jurov: https://devuan.org/newsletter_22dec.html << they claim to be self-hosting already
cazalla: thestringpuller: I guess it technically is already XMas for cazalla <<< 830am xmas morning but still, cuppa tea and read logs before my first family xmas opening gifts
dignork: jurov: thanks for the link, haven't read the actual method before, it's awesome.
jurov: https://www.cs.princeton.edu/~appel/papers/memerr.pdf << there was already research on using much more random memory errors than this
decimation: at any rate, it's proof of the kind of error against which asciilifeform regularly pontificates, that is, a hardware error that is known only to the vendors of the chips
dignork: Oh, reading, even with noise, border cells might be usefull though, haven't seen it.
dignork: so you have some more than frequent pattern, it should match the actual data that you want to write.
dignork: chances of sucessfull execution drop to 0
dignork: we cause a random pattern error at this position...
jurov: why not? if js in certain version of firefox can reliably cause fast repeated reads at address X, thus affecting addresses Y,Z
dignork: so it normally cannot read your /dev/whatever to check the physical location, but sure, it's potentially DOS/crash, not an effective code execution.
decimation: certainly the js memory is not randomly moved every context switch
jurov: js could corrupt browser core memory few pages away in this scenario
decimation: dignork: are you implying that javascript doesn't use ram?
jurov: because there are clever interleaving schemes to allow for faster reading
dignork: well, sure, but these reads need to happen nearby executed code, which by random error turns into something executable, etc. Just can't construct any scenario it realisticly happens. But I might be wrong.
jurov: there's certainly a way how to remotely cause already present software to repeatedly read memory
dignork: well, same difference, you still run on his machine for this.
decimation: don't need to write, only read
dignork: well, the thing is, if you already can write into his memory, at very specific locations, you probably already running there :)
decimation: umm, you can corrupt the victim's memory by executing a loop?
decimation: note that ECC is only a partial mitigation
dignork: decimation: cool paper, scary, otherwise I didn't bother to calcualte probabilities for uncontrolled random scenario, or non-DOS attack vector.
decimation: asciilifeform: look at that, some academics actually studying hardware
decimation: https://www.ece.cmu.edu/~safari/pubs/kim-isca14.pdf << "By reading from the same address in DRAM, we show that it is possible to corrupt data in nearby addresses"
gribble: Bitstamp BTCUSD ticker | Best bid: 322.76, Best ask: 323.91, Bid-ask spread: 1.15000, Last trade: 323.93, 24 hour volume: 6248.93897115, 24 hour low: 322.0, 24 hour high: 338.99, 24 hour vwap: 332.111093959
scoopbot: New post on Qntra.net by thestringpuller: http://qntra.net/2014/12/coinbase-tracing-user-transactions/
BingoBoingo: I thought CoinDesk was still a VC thing