Kurzweil is a ground-floor card-carrying member of the Extrapolarian Society. I've been following his shtick forever.
He actually was, once upon a time, as smart as he thinks he is, but then he flunked Latin, and now he's become Exhibit A for hominem unius tunius timeo.
The actual challenge here isn't to figure out how much he's wrong. The challenge is to figure out how much he's right. And he's more right than most people think. But they can't get past how wrong he is, and still there shooting fish in a barrel, entirely missing the main event.
After reading up, I'm getting the feeling that the home run THROBBING BONER victory condition is to cannibalise what they perceive as Bitcoin's inherent scalability limit.
I suspect this premonition/pretense/pretext is the money-bag money-shot behind the scenes.
CT: Do you see a possibility where (intentionally or not) your appcoin becomes an altcoin, and competes with Bitcoin as a currency?
MV: Yes, any appcoin has the potential to overtake Bitcoin. In the commodity world, it's hard to find something that can compete with gold as money because of gold's unique physical properties and distinctiveness [money-bug blather redacted]. There is a strong argument that you do want to have a resource that is used primarily as money. LBRY Credits are not designed to outcompete Bitcoin in that role.
However, if Bitcoin adoption levels off and LBRY apps are used by billions of people, then cryptocurrency speculators and users may decide that they feel more comfortable holding and using an asset that has a more widely-demanded end use.
But surely that's only a stretch goal. The next level down is harder to pinpoint.
Bitcoin will still be a payment option on the LBRY app, but it won't power the network. Fortunately, services like ShapeShift.io will make it easy to convert LBRY Credits to Bitcoin and back.
Oops. Now the claim from ss.io is that they had fully effectively firewalled user assets and that this is not the hack you're looking for. Okay, sure.
The boundary to the real economy is no small matter. I could be earning LBRY Ponzi credits tomorrow. Oh, yes, they are a Ponzi credit (on the production side) until you have a valid plan to get them back out again (and a lot can happen between here and there). As things stand, appears that the main road out exits through the Fire Swamp known as Bitcoin. Nobody ever gets burned or sandbagged or ROUSed to death en route there. Sign me up.
I do kind of like this new era of kinder, gentler, reduced friction, liberal-values, neoliberal Ponzi schemes (we'll not discuss the environmental Death Star of sweaty appcoin minting minions.)
But... bottom line, end of the vine, to get out, there has to be an equal and opposite demand to get in. Well, that's 90% of the iceberg here, and the sticky end of the wicket, too.
That doesn't work all that well, especially on videos with a lot of specialist jargon in it. Like university lectures.
A Netflix-style competition with sizeable pot at stake (a dime per U.S. citizen?) would address this problem PDQ.
Academic lectures, above all things, would quickly succumb to preconditioning on the right bag of words. Speech technology is advancing by leaps and bounds. It mainly needs improvement in automatically zeroing in on the appropriate jargon domain. Wikipedia is already a topic modelling gold mine just waiting to be fully exploited (of course, you'd have to cleverly cut through the mess, but that's what the big prize is there to expedite).
This small sum of money in the grand scheme of things would about solve this problem permanently, with spin-off advances in speech recognition for all involved.
And it's not like hearing loss is just for the deaf. It's a universal progressive condition exacerbated by good diet, exercise, and otherwise exemplary health.
I'm still waiting to discover the on-chip interconnect. I'd imagine that some kind of on-chip TSV would provide a hefty upgrade in the width of the data path between the CCX modules, that would still have some penalty, though far less than the historic standard.
I really can't see how Naples is going to use the same CCX if AMD hasn't done something interesting here, but what do I know? Also, one active core per CCX helps with heat spreading and available boost.
Furthermore, AMD producing chips with four CCX modules as an economy product is suggestive of some kind of packaging constraint. Like have a TSV substrate that's designed to straddle the corner of each of four CCX modules and not wanting—for now, while defect rates remain adequate—to produce a special tandem TSV substrate, and tool up for its assembly. Perhaps the game plan is to drive down the price of the hexacore modules until these become standard at the low end, and not bother ramping on a twin CCX assembly process for volume production at all (maybe just niche mobile products, and maybe then with a somewhat thinned CCX module a year down the road; or—a slightly bulked low-voltage CCX module with four cores and no substrate interface at all).
But I'm just using logic here, rather than digging around for revealed information, and that has only a modest track record.
In the least surprise ever, turns out pajamas man-child develops tight-loop benchmark suites for the trade press. Normally. Except for this one time.
—
Setting: One unusual sunny morning.
Right at the crack of too-damn-early, there's a loud, surprising knock on the door. Curious, he shambles in sloppy slippers to the front door, where he's greeted by a slight man in a slick seersucker suit, who warmly extends a cold hand, and exclaims "my good man, you are just who we need".
"And who are you, again," asks pajamas man-child, with maximal crack of too-damn-early rhetorical sarcasm.
"I'm from Butler, Shine & White, department of Natural Born Unusual Suspects."
He lavishes upon his smooth introductory move a practiced pump on each of 'Butler', 'Shine', and 'White', Vaseline vise-grip apexing right on the 'na' in 'natural', relaxing on 'orn', then releasing precisely on second 'su'.
"Me?" pyjamas man-child replies meekly, meaty ham agog and drifting.
"True to form, true to form. Ewww, what's that sooty smell?"
"Shit, you caught me mid-spread. Must have left a large, hot lump."
"Well that's just the thing we'll be speaking about."
"What is?"
"Here's the thing. Here's the thing. We have it on good local authority that you're the king of shinola soliloquy."
"Local authority? Man, I'm so going to sue that pesky early-bird arborist."
"Don't be hasty. Let me tell you what we have in mind."
Pajamas man-child scratches behind his hairy pinna for a moment. "Sure, okay, fire away. Do tell me about this soliloquy shinola business."
"No, no, no! You've got that bass ackwards. Trust me, we've got all the soliloquy shinola money can buy. What we don't have... yet... is the natural born shinola soliloquy."
"Uh, if I catch your drift... what I mean is... uh... you know... the spread... it answers back."
"For sure, we'll dub that in. Now how about let's discuss terms."
"Really?"
"In all high-margin, commodity seriousness."
"Okay then, come on in. Want some toast?"
"Uh, thanks but no thanks. Just in case, I brought us some fresh croissants." BS&W holds up large brown bag with hand-lettered accent marks on every vowel.
"Looks like you brought the entire continental buffet."
"Truth is, I'm here to see you spread."
"That's going to take a lot of spread."
"We'll use the big tub."
"Uh... you just said 'tub' right? Not, uh, 'tooh' as in 'toothbr—'."
"—aw shucks, just between us, what's the big difference?"
"Uh, tubes come with a screw top... or so I've heard."
"Yes, we did consider novel packaging, but it just doesn't say 'butter'."
Apparently, the number of people out there who profess a passionate desire that the world accord to the scintillating schema that the ease of a thing is best measured two hours after first exposure greatly exceeds the number of people willing to invest two hours to behave this way in practice.
After 6 years in school I still couldn't read a Latin text without thoroughly examining it first to figure out which parts of the multi-line sentences went with which other parts, and which of many different uses of the same word was intended.
Feel free to practice this skill on my sentence above. (No, not at all, it was my pleasure.)
[*] If you study neurology and development, you'll realize that hardly any of our core competences in life were easily acquired, and that what's most easily acquired remains forever shallow.
The most likely scenario is that they felt their window opportunity was rapidly closing on safely ridding themselves of a new hire who wasn't entirely above board during the interview process.
Legal to HR: Don't let this guy work another hour! Every additional hour will potentially add thousands of dollars to our final litigation expense.
Feeling rushed, they didn't fully think through the optics.
BAE's defense will be that this wet-paperwork employee was not yet entitled to their vaunted work-life balance.
Furthermore, in this line of work (highly sensitive), the issue of being above board throughout the hiring process won't go unnoticed by the judiciary.
The other side of this is that, had they known, they probably would have quietly let the man disappear, so as not to get themselves into this pickle. Now that this whole mess is in the public eye, the implication that BAE would quietly disqualify a candidate who's wife has terminal cancer—with absolutely nothing written down internally—is hard to ignore for the discerning reader.
This isn't exactly news. What's news here is that we now have a convenient handle for lashing out at the prevailing state of affairs in the modern corporatocracy, one step removed, yet connected at the hip: it's much easier, is it not, to promise a great work-life balance if your hiring process is 100% effective as screening out anyone strongly motivated to demand such a thing (many might want it a little bit, biting their lip, biding their 24/7 probation officer, with a main eye on their promotion prospect).
"volkswagon" here, "volks wagon" there but still you managed to type THEN in all-caps. Function key assignment? Here's a suggestion: add another function key for VW.
Seriously, after "Volkwagon" in the article title, the powers that be don't need any further assistance.
I can't tell you how many times I've seen competent developers fired and replaced with new-fangled JavaScript developers who could code a dynamic website in one day.
Error: type "management problem" assigned to type "programming language".
You could argue that these were all people already well aware and into BDSM, and the movie just raised a desire in them. I seriously doubt that, however.
Doh! We're a hundred miles down the road into The One True Cause and most people can't even distinguish rate from order.
Time is short. Right?
List of shit I'd like to poke into. * item * another item * many more items
Some item from laundry list comes up in everyday circumstance. That particular item becomes momentarily top of mind. Return home. Ten minutes to kill. What's top of mind? Bingo. That thing that came up around the water cooler that was on your shits and giggles list anyway.
News headline the following day: in recent trends, the BDSM zeitgeist is hot, hot, hot.
No, actually. It just went through a minor flush of suggestibility-sort synchronicity. No kittens were renamed.
One basic axiom is that on critical systems you don't change anything unless absolutely necessary.
Which is why your master clock, if you have any competence whatsoever, is permanently set to GMT.
Every system utility that function in local time performs a conversion to local time.
It's the conversion from GMT to local time that jumps twice a year. No changed settings required. One side of this is worse than the other, because one side breaks monotonicity.
And it's so hard to program this correctly, my brain positively throbs for the humanity.
if (!done(once_only_task) &&
once_only_task.start_local <= now_local()) {
do (once_only_task);
mark_done (once_only_task); }
Wow! Wasn't that hard? Most other exercises in translating the least clue about the vagaries of time into correct code are equally daunting.
Any quality programming language should cough up a hairball when local times are subtracted naively: "use TAI or UTC directly, you complete idiot".
The unpredictable future relationship between TAI and UTC is far more problematic.
But you wanted time to be simple, and earth not to warm/cool, shuffle, spindle, or lose angular momentum, and money not to be a fiat currency, and periods to be unambiguous in all human locales over all time and space.
As we all know, small requests from small minds are always correct, so get the lead out, Batman.
This entire DST debate is just the next station on the train track of nuance removal.
No Java, approximate floating-point arithmetic does not give a reproducible answer unless the value of every operation is mandated—for all platforms and all supported operations—down to the value of the last ULP.
And even then, if the user reorders the rows of the input screen (a supposed invariant) oops there goes perfect consistency.
Why, why, why cruel world?
Because—you might want to sit down here—in truth, infinite precision arithmetic is difficult to implement efficiently and almost every efficient implementation makes tiny little trade-offs in silicon different than before.
William Kahan coined the term "The Table-Maker's Dilemma" for the unknown cost of rounding transcendental functions:
"Nobody knows how much it would cost to compute y^w correctly rounded for every two floating-point arguments at which it does not over/underflow. Instead, reputable math libraries compute elementary transcendental functions mostly within slightly more than half an ulp and almost always well within one ulp. Why can't y^w be rounded within half an ulp like SQRT? Because nobody knows how much computation it would cost... No general way exists to predict how many extra digits will have to be carried to compute a transcendental expression and round it correctly to some preassigned number of digits. Even the fact (if true) that a finite number of extra digits will ultimately suffice may be a deep theorem."
Okay, we knock that off before breakfast, and after lunch we find a fabulous solution to Arrow's impossibility theorem that no-one ever thought of before.
Like when GitHub goes down, and now all of these git users are shit out of luck (they'll always point out they could push to or pull from each others repos, yet this is a real pain in the ass in practice and they never actually do it).
Hey, let's pull the curtains off the rest of the matrix.
With git you really arebarely three square meals away from anarchy, and GitHub knows this.
Hence the outages tend to be brief, by all the earthly powers of mice and men and backup diesel generators.
Or maybe yields are better than expected. Or maybe the market analysis says that this price point results in more profit based on the marginal cost curve. Maybe they just have an aggressive new manager whose bonus is tied to units sold. To say "Nvidia is clearly worried" is putting an awful lot of certainty onto a purely speculative statement.
The pricing dance is extremely complex, but there's no question here about whether Nvidia's Magic Margin spreadsheet includes a giant column of anticipated AMD price-performance points, or that those estimated numbers have been trending up lately, on the back of AMD's strong showing with their new core design (which soon projects to drive the APU end of their product line, too).
Whether this be a Nervous Nonchalant Nelly magic-margin spreadsheet or a Cold Cowering Calculus magic-margin spreadsheet is another matter entirely.
Uh, wait a minute... I think I just got my Magic Eight Ball attributional wires crossed with those authoritative asspull alliterations.
A narrow, low-entropy strip around an excluded disk of 10^15 most typical password strings isn't necessary greater than 10^15 in size itself (and an iterative, self-referential problem suggests itself: given these 10^15 excluded passwords, what's the next set of 10^15 most typical patterns?)
But in practice, I don't think that's the biggest fish to fry, here.
You just need to assume the the disk filter will force the average password length upward and make sure that all passwords shy of the predicted equilibrium length are proportionately represented in the model, then you'll be close enough.
Woe to the person who analyzes the filter to find the shortest string not included, then deliberately uses that.
But OBVIOUSLY password rules force the user to avoid the common pitfalls in password selection and will more likely cause your users to have passwords that are not easily cracked.
I prefer my "obvious" shaken, not stirred.
The only way that password rules enhance password entropy is by forcing the user to supply more entropy. This is the same person who didn't supply enough entropy in the first place, probably someone who really likes A1 sauce.
And now you've got this: habaneroA1!
Sure, a list of the one hundred favourite steak sauces will add six or seven bits of entropy (note: the list won't be uniformly distributed). If you're adding that entropy to a bare word from the English language (rough entropy 13 to 14 bits) you've almost reliably made twenty full bits.
Congratulations! Nelson Mandela can no longer crack your password by hand given 19 years and a heavy rock-hammer slate.
What you really need here is to generate a (conceptual) list of 10^15 strings that best resemble common passwords, and then reject all strings from that list. There are standard methods from information theory to construct such a list given the statistical properties of a large list of previously exposed password strings (requires aptitude). Or you might even be able to train a neural network for this task, using 100% automated pattern inference, and arrive pretty close to the same place.
End result: Joe Sixpack gets rebuffed ten times in a row (he's not willing to do much more to his beloved low-entropy authentication burger than add the ketchup before the lettuce) and then he blows his top, pulls out the sharpest pen he owns, and simultaneously carves his fucking strong password onto a handy strip of paper and the wooden desk underneath it.
Shearing off the low-hanging 10^15 excludes almost every short pattern that half the population regards as even vaguely memorable.
And short patterns are the longest patterns that half the population can type reliably (without a pat on the back halfway through).
The pat on the back system could work. You'd have multiple password inputs providing a mandatory twenty bits each.
Remove six of those bits as a validity congruence (false positives: about 1.5%). You'd need to repeat this four rounds to get 50+ bits of true entropy (4*14=56).
And you'd need to ensure that none of the rounds were simple manipulations of other rounds, or derived from common sequences.
janfebmar / aprmayjun / julaugsep / octnovdec
Here's the problem. By the time the cracker receives two pats on the back for janfebmar / aprmayjun he's probably already onto a shrewd guess about the continuation.
So where you arrive:
There are many people out there where the shortest sequence they can reliably remember with sufficient entropy (which I take as 50 bits) is longer than the longest sequence they can type reliably.
twoshakesofalamb'stail is pretty easy to memorize, but a lot of people couldn't reliably type that b***d better than 10% of the time.
6uldv8!!! is pretty easy to type, but no chance it makes it under the 10^15 bar on any viable model of human psychology. For the sexless, x1k3c3d7 probably doesn't make it under that bar, either (given how well machine translation already works, I think the neural network is onto all of your cheap tricks—up to and including geometric patterns based on common keyboard layouts).
~Oj6ojEb}will make it under the bar and (with practice) is fairly quick to type.
Start memorizing NOW.
Do NOT repeat on any other system.
Prepare to memorize another dozen twisty little passages, all entirely alike in their extreme differentness.
Right. Because all financial processes attribute to the current administration so precisely (and immediately) that we need to write "full month" rather than just "month".
Among the numerate, the earliest possible report card on the new administration is the end of the first year, and we might by then be celebrating the accomplishments of President Pence, who might by then be one fat month into his presidency. Given these larger uncertainties, I'm not going to take any macro-economic wiggles as usefully reflective on Life Under New Management for a full two full years.
On the other hand, for those of you who buy into this ridiculous one-month report card, just imagine how fabulous life will be once the new administration manages to fully staff up. They accomplished this small, immediate miracle with half their chairs empty. That's quite something, and portends greatness, does it not?
Microsoft, Intel, Dell, Cisco, Sun/Oracle, HP/Compaq and even Apple/IBM were all pretty powerful in their own corners of the industry during the nineties.
That said, social+mobile really was a game changer on ultimate scale.
There's no evolutionary need for nine hours a week. What every sexologist has discovered is that people are different. While I wouldn't necessarily call it an addiction, I wouldn't rule it out, either.
Besides, most people informally use the word 'addiction' to imply that moderation has departed the building (presumably under a dark cloud) and I think that's what was intended here.
If sex caused weight gain the same way excess sugar intake causes weight gain, you wouldn't be chirping about "basic human needs". You'd be croaking something about basic human needs instead, and sizing up some Princess Leia to service your bariatric desires, only to find yourself thwarted by Hans Solo.
Launching an enterprise product that gains even a few points of market share from the very large blue incumbent can implement billions of dollars to the bottom line, as well as provided some innovation as there are now two big players '''on the field.
Sad, Ian, sad.
While not specifically mentioned in the announcement today, we do know that Naples is not a single monolithic die on the order of 500 mm2 or up. Naples uses four of AMD's Zeppelin dies (the Ryzen dies) in a single package.
But first I had to wade through that other sentence. (I want a Purple Heart!)
—
Once upon a time you would have a pair of quad-CPU servers, with two cores per CPU and two threads per core.
Now you have a pair of quad-die CPUs, with two cores per die, and two threads per core.
Box to box latency is replaced by CPU to CPU latency.
CPU to CPU latency is replaced by die to die latency within a single CPU package.
I wonder if AMD is bonding the four die modules using some form of TSV (through-silicon via).
—
The last time I had die modules was a dual Pentium Pro 166 MHz (with 512 KB of L2 cache per CPU as a second die module). Fast processors, but the DRAM was dog slow. Eventually replaced it with a dual-cartridge Pentium III 750 (L2 cache as separate chips on the cartridge). I liked the dualies back then. It made for a creamier GUI with the system under load.
I had a great Opteron system at work (a 24/7 affair) and barely used my home system at all for about four years. Intended to replace it with an AMD system, but then the Opteron product line went insane, and Intel came out with the CoreDuo, and that was all she wrote for my AMD loyalty card.
It's so nice to see AMD finally back in the game, but presently I can't flop to AMD until they properly validate FreeBSD and get themselves off the FreeBSD shit list.
It's conceptually well thought out and doesn't require a lot of corner cases. It's stable and common to every current ZFS implementation. I haven't looked at all the feature bits subsequently added, but I don't think many of them complicate recovery of basic files whatsoever.
This is notZFS Internals for Dummies, but do note how the available tools are first rate. Plus, the on-disk structure is integrity checked all the way down. Most likely, any misconception about bit-patterns will be brutally put to rest by the next disk block you fetch.
Finally, there's very little fundamental churn here, because ZFS was designed for the future on day one.
—
ZFS has one Achilles heel: the absence of block-pointer rewrite. Basically, the integrity layer is overly rigid about block placement, and thus certain kinds of desired flexibility are off the table, now and forever, until ZFS 2.0 comes with a different on-disk placement record (which might never happen, as the principals all refer to BPR in hushed voices as some large, daunting project—probably because maintaining the historical testing standard requires industrial-strength support).
—
Contrary to your ludicrously uninformed tar-pit scenario, ZFS is a paragon of long-term, binary-format stability.
You must somehow think the future can barely start a fire by rubbing two sticks together.
Have you ever turned the pages of Office "Open" XML?
This is potentially a hundred times more baroque, Byzantine, and baffling to some rub-stick deprived future generation, that wakes up severely hung-over and back-to-the-buggies as the dust settles on the AI apocalypse.
(How do we finally beat the AI uprising? Probably though nefarious virus bearing an OOXML payload, which even the ascendant AI-powered globally-distributed firewall fatally misclassifies.)
—
For true geeks, ZFS origin story from the horse's mouth.
BP-rewrite mentioned at 18:00. Something technical about DVA (data virtual address). Then a horrible "bolt on" is mentioned. Then a mike drop.
Circa 7m: it's not going to be a team of 80 people, it's just going to be me and Matt at the white board all day, every day. (And y'all knows how that goes. Turns out "tank" is a character from the Matrix. )
Circa 11m20: ztest and zloop
So we realized very quickly we could just create files in/tmp, pretend that they're disk drives, and then issue all the read and write commands and transactions from our test program and exercise all the code from userland.
The really big advantage of that was that it gave us the ability not just to test the datapath, which is pretty easy to do with normal stress tests, but also to test the administrative path.
So you could have things like: What happens if one thread is trying to attach a mirror, while another thread is trying to detach it, while another is trying to write, while somebody else is trying to resilver?"
In practice, those things are very hard to test with actual hardware, because it just takes a long time to do those things, but because we had very small devices that were really just/tmp files which were super fast, and because the cost of killing ztest and restarting it was measured in milliseconds, it meant we could crank through the
There's a word for this. It's called "conurbation". Practice saying this out loud three times every morning in front of the bathroom mirror and I promise you that this small ritual will cure your pedanticism in record time.
Slashdot Mirror
Kurzweil is a ground-floor card-carrying member of the Extrapolarian Society. I've been following his shtick forever.
He actually was, once upon a time, as smart as he thinks he is, but then he flunked Latin, and now he's become Exhibit A for hominem unius tunius timeo .
The actual challenge here isn't to figure out how much he's wrong. The challenge is to figure out how much he's right. And he's more right than most people think. But they can't get past how wrong he is, and still there shooting fish in a barrel, entirely missing the main event.
After reading up, I'm getting the feeling that the home run THROBBING BONER victory condition is to cannibalise what they perceive as Bitcoin's inherent scalability limit.
I suspect this premonition/pretense/pretext is the money-bag money-shot behind the scenes.
The Appcoin Revolution: Interview with Mike "Buttercup" Vine of LBRY:
But surely that's only a stretch goal. The next level down is harder to pinpoint.
Sounds easy.
Shapeshift.io has been hacked
Oops. Now the claim from ss.io is that they had fully effectively firewalled user assets and that this is not the hack you're looking for. Okay, sure.
The boundary to the real economy is no small matter. I could be earning LBRY Ponzi credits tomorrow. Oh, yes, they are a Ponzi credit (on the production side) until you have a valid plan to get them back out again (and a lot can happen between here and there). As things stand, appears that the main road out exits through the Fire Swamp known as Bitcoin. Nobody ever gets burned or sandbagged or ROUSed to death en route there. Sign me up.
I do kind of like this new era of kinder, gentler, reduced friction, liberal-values, neoliberal Ponzi schemes (we'll not discuss the environmental Death Star of sweaty appcoin minting minions.)
But ... bottom line, end of the vine, to get out, there has to be an equal and opposite demand to get in. Well, that's 90% of the iceberg here, and the sticky end of the wicket, too.
A Netflix-style competition with sizeable pot at stake (a dime per U.S. citizen?) would address this problem PDQ.
Academic lectures, above all things, would quickly succumb to preconditioning on the right bag of words. Speech technology is advancing by leaps and bounds. It mainly needs improvement in automatically zeroing in on the appropriate jargon domain. Wikipedia is already a topic modelling gold mine just waiting to be fully exploited (of course, you'd have to cleverly cut through the mess, but that's what the big prize is there to expedite).
This small sum of money in the grand scheme of things would about solve this problem permanently, with spin-off advances in speech recognition for all involved.
And it's not like hearing loss is just for the deaf. It's a universal progressive condition exacerbated by good diet, exercise, and otherwise exemplary health.
I'm still waiting to discover the on-chip interconnect. I'd imagine that some kind of on-chip TSV would provide a hefty upgrade in the width of the data path between the CCX modules, that would still have some penalty, though far less than the historic standard.
I really can't see how Naples is going to use the same CCX if AMD hasn't done something interesting here, but what do I know? Also, one active core per CCX helps with heat spreading and available boost.
Furthermore, AMD producing chips with four CCX modules as an economy product is suggestive of some kind of packaging constraint. Like have a TSV substrate that's designed to straddle the corner of each of four CCX modules and not wanting—for now, while defect rates remain adequate—to produce a special tandem TSV substrate, and tool up for its assembly. Perhaps the game plan is to drive down the price of the hexacore modules until these become standard at the low end, and not bother ramping on a twin CCX assembly process for volume production at all (maybe just niche mobile products, and maybe then with a somewhat thinned CCX module a year down the road; or—a slightly bulked low-voltage CCX module with four cores and no substrate interface at all).
But I'm just using logic here, rather than digging around for revealed information, and that has only a modest track record.
The time for digging is after Naples.
AMD-K6 3D (90 bogoids)
<=
Intel Core i7-4771 @ 3.50GHz (9940 bogoids) * 1%
<=
Via C3 Ezra (100 bogoids)
<=
Intel Pentium III Mobile 750MHz (103 bogoids)
<=
AMD Athlon 64 2000+ (116 bogoids)
<=
Intel Pentium 4 1300MHz (119 bogoids)
Wow, a couple of clown chips, and a searing indictment of Passmark, all rolled up together.
You can really see how Passmark should have been properly named Parkay Malarkey Spinmark.
Parkay Pentium 4, you are so busted.
[*] Cooking instructions: apply Parkay to soggy white bread, wait five minutes, LET THERE BE TOAST.
Source.
—
In the least surprise ever, turns out pajamas man-child develops tight-loop benchmark suites for the trade press. Normally. Except for this one time.
—
Setting: One unusual sunny morning.
Right at the crack of too-damn-early, there's a loud, surprising knock on the door. Curious, he shambles in sloppy slippers to the front door, where he's greeted by a slight man in a slick seersucker suit, who warmly extends a cold hand, and exclaims "my good man, you are just who we need".
"And who are you, again," asks pajamas man-child, with maximal crack of too-damn-early rhetorical sarcasm.
"I'm from Butler, Shine & White, department of Natural Born Unusual Suspects."
He lavishes upon his smooth introductory move a practiced pump on each of 'Butler', 'Shine', and 'White', Vaseline vise-grip apexing right on the 'na' in 'natural', relaxing on 'orn', then releasing precisely on second 'su'.
"Me?" pyjamas man-child replies meekly, meaty ham agog and drifting.
"True to form, true to form. Ewww, what's that sooty smell?"
"Shit, you caught me mid-spread. Must have left a large, hot lump."
"Well that's just the thing we'll be speaking about."
"What is?"
"Here's the thing. Here's the thing. We have it on good local authority that you're the king of shinola soliloquy."
"Local authority? Man, I'm so going to sue that pesky early-bird arborist."
"Don't be hasty. Let me tell you what we have in mind."
Pajamas man-child scratches behind his hairy pinna for a moment. "Sure, okay, fire away. Do tell me about this soliloquy shinola business."
"No, no, no! You've got that bass ackwards. Trust me, we've got all the soliloquy shinola money can buy. What we don't have ... yet ... is the natural born shinola soliloquy."
"Uh, if I catch your drift ... what I mean is ... uh ... you know ... the spread ... it answers back."
"For sure, we'll dub that in. Now how about let's discuss terms."
"Really?"
"In all high-margin, commodity seriousness."
"Okay then, come on in. Want some toast?"
"Uh, thanks but no thanks. Just in case, I brought us some fresh croissants." BS&W holds up large brown bag with hand-lettered accent marks on every vowel.
"Looks like you brought the entire continental buffet."
"Truth is, I'm here to see you spread."
"That's going to take a lot of spread."
"We'll use the big tub."
"Uh ... you just said 'tub' right? Not, uh, 'tooh' as in 'toothbr—'."
"—aw shucks, just between us, what's the big difference?"
"Uh, tubes come with a screw top ... or so I've heard."
"Yes, we did consider novel packaging, but it just doesn't say 'butter'."
And your logic is?
Here's how my equation falls out:
* States without coal: safety paramount.
* States with coal: extraction paramount.
Hence, disaggregation of oversight guarantees extraction.
Or—wait for it!—we can draw a BIG circle around the ENTIRE externality all at once (and one for all).
But that would actually lead to broad discussion, and horse trading, and the sound exercise of restraint, and the wrong kind of green.
So you're right. Let's close all the windows and let the states do it.
Apparently, the number of people out there who profess a passionate desire that the world accord to the scintillating schema that the ease of a thing is best measured two hours after first exposure greatly exceeds the number of people willing to invest two hours to behave this way in practice.
Feel free to practice this skill on my sentence above. (No, not at all, it was my pleasure.)
[*] If you study neurology and development, you'll realize that hardly any of our core competences in life were easily acquired, and that what's most easily acquired remains forever shallow.
The most likely scenario is that they felt their window opportunity was rapidly closing on safely ridding themselves of a new hire who wasn't entirely above board during the interview process.
Legal to HR: Don't let this guy work another hour! Every additional hour will potentially add thousands of dollars to our final litigation expense.
Feeling rushed, they didn't fully think through the optics.
BAE's defense will be that this wet-paperwork employee was not yet entitled to their vaunted work-life balance.
Furthermore, in this line of work (highly sensitive), the issue of being above board throughout the hiring process won't go unnoticed by the judiciary.
The other side of this is that, had they known, they probably would have quietly let the man disappear, so as not to get themselves into this pickle. Now that this whole mess is in the public eye, the implication that BAE would quietly disqualify a candidate who's wife has terminal cancer—with absolutely nothing written down internally—is hard to ignore for the discerning reader.
This isn't exactly news. What's news here is that we now have a convenient handle for lashing out at the prevailing state of affairs in the modern corporatocracy, one step removed, yet connected at the hip: it's much easier, is it not, to promise a great work-life balance if your hiring process is 100% effective as screening out anyone strongly motivated to demand such a thing (many might want it a little bit, biting their lip, biding their 24/7 probation officer, with a main eye on their promotion prospect).
"volkswagon" here, "volks wagon" there but still you managed to type THEN in all-caps. Function key assignment? Here's a suggestion: add another function key for VW.
Seriously, after "Volkwagon" in the article title, the powers that be don't need any further assistance.
Error: type "management problem" assigned to type "programming language".
This entire debate is rife with just-so soup of the day.
Doh! We're a hundred miles down the road into The One True Cause and most people can't even distinguish rate from order.
Time is short. Right?
List of shit I'd like to poke into.
* item
* another item
* many more items
Some item from laundry list comes up in everyday circumstance. That particular item becomes momentarily top of mind. Return home. Ten minutes to kill. What's top of mind? Bingo. That thing that came up around the water cooler that was on your shits and giggles list anyway.
News headline the following day: in recent trends, the BDSM zeitgeist is hot, hot, hot.
No, actually. It just went through a minor flush of suggestibility-sort synchronicity. No kittens were renamed.
Which is why your master clock, if you have any competence whatsoever, is permanently set to GMT.
Every system utility that function in local time performs a conversion to local time.
It's the conversion from GMT to local time that jumps twice a year. No changed settings required. One side of this is worse than the other, because one side breaks monotonicity.
And it's so hard to program this correctly, my brain positively throbs for the humanity.
if (!done(once_only_task) &&
once_only_task.start_local <= now_local()) {
do (once_only_task);
mark_done (once_only_task);
}
Wow! Wasn't that hard? Most other exercises in translating the least clue about the vagaries of time into correct code are equally daunting.
Any quality programming language should cough up a hairball when local times are subtracted naively: "use TAI or UTC directly, you complete idiot".
The unpredictable future relationship between TAI and UTC is far more problematic.
But you wanted time to be simple, and earth not to warm/cool, shuffle, spindle, or lose angular momentum, and money not to be a fiat currency, and periods to be unambiguous in all human locales over all time and space.
As we all know, small requests from small minds are always correct, so get the lead out, Batman.
This entire DST debate is just the next station on the train track of nuance removal.
No Java, approximate floating-point arithmetic does not give a reproducible answer unless the value of every operation is mandated—for all platforms and all supported operations—down to the value of the last ULP.
And even then, if the user reorders the rows of the input screen (a supposed invariant) oops there goes perfect consistency.
Why, why, why cruel world?
Because—you might want to sit down here—in truth, infinite precision arithmetic is difficult to implement efficiently and almost every efficient implementation makes tiny little trade-offs in silicon different than before.
Okay, we knock that off before breakfast, and after lunch we find a fabulous solution to Arrow's impossibility theorem that no-one ever thought of before.
Jesus fuck, are we programmers here, or pussies?
Hey, let's pull the curtains off the rest of the matrix.
With git you really are barely three square meals away from anarchy, and GitHub knows this.
Hence the outages tend to be brief, by all the earthly powers of mice and men and backup diesel generators.
Our Lady of Potted Fern Beside Left-Rear 5.1 Don't Fail Me Now:
[_] Robert Crowley (printer)
[_] Robert Crowley (CIA)
[_] Bob Crowley (Survivor contestant)
[_] it's a trap
Damn! I so much want to add that one to my bucket list, but then it would almost kill me to have to cross it off.
1. have sex
2. have sex again
So here's my big question: can you make bucket list items out of sheep's intestine? Cause I'd want to reuse this one a lot.
The pricing dance is extremely complex, but there's no question here about whether Nvidia's Magic Margin spreadsheet includes a giant column of anticipated AMD price-performance points, or that those estimated numbers have been trending up lately, on the back of AMD's strong showing with their new core design (which soon projects to drive the APU end of their product line, too).
Whether this be a Nervous Nonchalant Nelly magic-margin spreadsheet or a Cold Cowering Calculus magic-margin spreadsheet is another matter entirely.
Uh, wait a minute ... I think I just got my Magic Eight Ball attributional wires crossed with those authoritative asspull alliterations.
Small mathematical error from my previous post.
A narrow, low-entropy strip around an excluded disk of 10^15 most typical password strings isn't necessary greater than 10^15 in size itself (and an iterative, self-referential problem suggests itself: given these 10^15 excluded passwords, what's the next set of 10^15 most typical patterns?)
But in practice, I don't think that's the biggest fish to fry, here.
You just need to assume the the disk filter will force the average password length upward and make sure that all passwords shy of the predicted equilibrium length are proportionately represented in the model, then you'll be close enough.
Woe to the person who analyzes the filter to find the shortest string not included, then deliberately uses that.
I prefer my "obvious" shaken, not stirred.
The only way that password rules enhance password entropy is by forcing the user to supply more entropy. This is the same person who didn't supply enough entropy in the first place, probably someone who really likes A1 sauce.
And now you've got this: habaneroA1!
Sure, a list of the one hundred favourite steak sauces will add six or seven bits of entropy (note: the list won't be uniformly distributed). If you're adding that entropy to a bare word from the English language (rough entropy 13 to 14 bits) you've almost reliably made twenty full bits.
Congratulations! Nelson Mandela can no longer crack your password by hand given 19 years and a heavy rock-hammer slate.
What you really need here is to generate a (conceptual) list of 10^15 strings that best resemble common passwords, and then reject all strings from that list. There are standard methods from information theory to construct such a list given the statistical properties of a large list of previously exposed password strings (requires aptitude). Or you might even be able to train a neural network for this task, using 100% automated pattern inference, and arrive pretty close to the same place.
End result: Joe Sixpack gets rebuffed ten times in a row (he's not willing to do much more to his beloved low-entropy authentication burger than add the ketchup before the lettuce) and then he blows his top, pulls out the sharpest pen he owns, and simultaneously carves his fucking strong password onto a handy strip of paper and the wooden desk underneath it.
Shearing off the low-hanging 10^15 excludes almost every short pattern that half the population regards as even vaguely memorable.
And short patterns are the longest patterns that half the population can type reliably (without a pat on the back halfway through).
The pat on the back system could work. You'd have multiple password inputs providing a mandatory twenty bits each.
Remove six of those bits as a validity congruence (false positives: about 1.5%). You'd need to repeat this four rounds to get 50+ bits of true entropy (4*14=56).
And you'd need to ensure that none of the rounds were simple manipulations of other rounds, or derived from common sequences.
janfebmar / aprmayjun / julaugsep / octnovdec
Here's the problem. By the time the cracker receives two pats on the back for janfebmar / aprmayjun he's probably already onto a shrewd guess about the continuation.
So where you arrive:
There are many people out there where the shortest sequence they can reliably remember with sufficient entropy (which I take as 50 bits) is longer than the longest sequence they can type reliably.
twoshakesofalamb'stail is pretty easy to memorize, but a lot of people couldn't reliably type that b***d better than 10% of the time.
6uldv8!!! is pretty easy to type, but no chance it makes it under the 10^15 bar on any viable model of human psychology. For the sexless, x1k3c3d7 probably doesn't make it under that bar, either (given how well machine translation already works, I think the neural network is onto all of your cheap tricks—up to and including geometric patterns based on common keyboard layouts).
~Oj6ojEb} will make it under the bar and (with practice) is fairly quick to type.
Start memorizing NOW.
Do NOT repeat on any other system.
Prepare to memorize another dozen twisty little passages, all entirely alike in their extreme differentness.
Then multiply by Ultimate IDIOT Winter FAIL.
For bonus marks: take into accoun
Right. Because all financial processes attribute to the current administration so precisely (and immediately) that we need to write "full month" rather than just "month".
Among the numerate, the earliest possible report card on the new administration is the end of the first year, and we might by then be celebrating the accomplishments of President Pence, who might by then be one fat month into his presidency. Given these larger uncertainties, I'm not going to take any macro-economic wiggles as usefully reflective on Life Under New Management for a full two full years.
On the other hand, for those of you who buy into this ridiculous one-month report card, just imagine how fabulous life will be once the new administration manages to fully staff up. They accomplished this small, immediate miracle with half their chairs empty. That's quite something, and portends greatness, does it not?
Microsoft, Intel, Dell, Cisco, Sun/Oracle, HP/Compaq and even Apple/IBM were all pretty powerful in their own corners of the industry during the nineties.
That said, social+mobile really was a game changer on ultimate scale.
There's no evolutionary need for nine hours a week. What every sexologist has discovered is that people are different. While I wouldn't necessarily call it an addiction, I wouldn't rule it out, either.
Besides, most people informally use the word 'addiction' to imply that moderation has departed the building (presumably under a dark cloud) and I think that's what was intended here.
If sex caused weight gain the same way excess sugar intake causes weight gain, you wouldn't be chirping about "basic human needs". You'd be croaking something about basic human needs instead, and sizing up some Princess Leia to service your bariatric desires, only to find yourself thwarted by Hans Solo.
There are moments when I understand what other people are griping about.
AMD Prepares 32-Core Naples CPUs for 1P and 2P Servers: Coming in Q2
Sad, Ian, sad.
But first I had to wade through that other sentence. (I want a Purple Heart!)
—
Once upon a time you would have a pair of quad-CPU servers, with two cores per CPU and two threads per core.
Now you have a pair of quad-die CPUs, with two cores per die, and two threads per core.
I wonder if AMD is bonding the four die modules using some form of TSV (through-silicon via).
—
The last time I had die modules was a dual Pentium Pro 166 MHz (with 512 KB of L2 cache per CPU as a second die module). Fast processors, but the DRAM was dog slow. Eventually replaced it with a dual-cartridge Pentium III 750 (L2 cache as separate chips on the cartridge). I liked the dualies back then. It made for a creamier GUI with the system under load.
I had a great Opteron system at work (a 24/7 affair) and barely used my home system at all for about four years. Intended to replace it with an AMD system, but then the Opteron product line went insane, and Intel came out with the CoreDuo, and that was all she wrote for my AMD loyalty card.
It's so nice to see AMD finally back in the game, but presently I can't flop to AMD until they properly validate FreeBSD and get themselves off the FreeBSD shit list.
Please, AMD, do us a solid and make it rock.
There are already quite a few tools in computational journalism to automate the early assessment of a large data dump.
What do Journalists do with Documents?
C+J 2016: Documents, Data Mining and Discovery
As with all things, I'm sure the 20-80 rule applies.
The ZFS on-disk format is extremely well documented and not that terribly hard to understand.
ZFS On-Disk Specification
It's conceptually well thought out and doesn't require a lot of corner cases. It's stable and common to every current ZFS implementation. I haven't looked at all the feature bits subsequently added, but I don't think many of them complicate recovery of basic files whatsoever.
0:06 / 43:27 Examining ZFS On-Disk Format Using mdb and zdb: Max Bruning
This is not ZFS Internals for Dummies, but do note how the available tools are first rate. Plus, the on-disk structure is integrity checked all the way down. Most likely, any misconception about bit-patterns will be brutally put to rest by the next disk block you fetch.
Finally, there's very little fundamental churn here, because ZFS was designed for the future on day one.
—
ZFS has one Achilles heel: the absence of block-pointer rewrite. Basically, the integrity layer is overly rigid about block placement, and thus certain kinds of desired flexibility are off the table, now and forever, until ZFS 2.0 comes with a different on-disk placement record (which might never happen, as the principals all refer to BPR in hushed voices as some large, daunting project—probably because maintaining the historical testing standard requires industrial-strength support).
—
Contrary to your ludicrously uninformed tar-pit scenario, ZFS is a paragon of long-term, binary-format stability.
You must somehow think the future can barely start a fire by rubbing two sticks together.
Have you ever turned the pages of Office "Open" XML?
This is potentially a hundred times more baroque, Byzantine, and baffling to some rub-stick deprived future generation, that wakes up severely hung-over and back-to-the-buggies as the dust settles on the AI apocalypse.
(How do we finally beat the AI uprising? Probably though nefarious virus bearing an OOXML payload, which even the ascendant AI-powered globally-distributed firewall fatally misclassifies.)
—
For true geeks, ZFS origin story from the horse's mouth.
The Birth of ZFS by Jeff Bonwick
BP-rewrite mentioned at 18:00. Something technical about DVA (data virtual address). Then a horrible "bolt on" is mentioned. Then a mike drop.
Circa 7m: it's not going to be a team of 80 people, it's just going to be me and Matt at the white board all day, every day. (And y'all knows how that goes. Turns out "tank" is a character from the Matrix. )
Circa 11m20: ztest and zloop
There's a word for this. It's called "conurbation". Practice saying this out loud three times every morning in front of the bathroom mirror and I promise you that this small ritual will cure your pedanticism in record time.