Indeed. It's an adequate enough conductor that plating something with gold won't ruin its conductivity(at least not as badly as a layer of tarnish/oxidation will), and it's good enough to use for wirebonding in chip packaging; but it's merely adequate. Corrosion resistance and extreme ductility are the really neat tricks.
I would have liked to; but(thankfully) I wasn't on the procurement side in the case of that particular job(given the nightmare that was our PO system, I am most grateful). Ultimately pragmatism won out over principle and(for the small subset of affected users) we got some $5 c-media USB sound cards, which proceeded to work perfectly regardless of CPU frequency.
It was one of those situations where the villains skated free; but the number of hours it would have taken on the phone, fighting it out between the vendor of the software that most notably exposed the problem, the PC OEM, the maker of the onboard sound device, and possibly Microsoft, just would have been hell. Not the desired outcome; but the slog to get something better probably would have cost more than entire new systems, never mind new soundcards.
For me, it was mostly a wake-up call about how bloody awful some peripherals and peripheral drivers still are, and how long assumptions that were recognized as dangerous and hacky back when 8MHz was a respectable CPU clock can stick around. Not a pleasant learning experience; but not all are. Somewhat amusing troubleshooting, though. You don't usually expect 100% CPU load to make audio playback better.
SLC is too rich for my blood, so I don't really have the luxury of comparison; but (just as a mixture of RAID and backups has proven cheaper than absolutely bulletproof engineering in HDDs), it wouldn't entirely surprise me if MLC's cost advantage, combined with ongoing improvements in masking its deficiencies, ultimately relegate SLC to relatively niche applications that don't have space for lots of redundancy; but do have the budget for classy hardware. (After all, just look at how well NOR flash's superiority over NAND flash has...mostly not saved it... from being replaced by stacks of cheap NAND behind a controller designed to make it looks like it doesn't suck as much as it does).
I'd be the last to deny that SLC is, in fact, objectively better; but you can buy so much MLC for the price of a given amount of SLC that, given decent controller design, there tends to be room for nontrivial redundancy and greater usable capacity. I can see why it offends purists; but practically all computer equipment today is the direct descendant of inferior crap that beat beautifully engineered and overtly superior systems on price.
My concern is less about that than about the possibility that(if any judge from any district will do) the tendency of the judiciary to rubber-stamp warrants will be markedly increased.
If the FBI has to deal with Judge X, Y, or Z; because they are in district 61, it is at least possible that they'll have to put together a convincing warrant request, lest it be denied. If they can pick any federal district judge, it won't exactly be a big secret which judge you want to talk to if you need some utter bullshit approved.
When it is the job of the judiciary to be at least skeptical, if not somewhat adversarial, when law enforcement comes knocking with a request to go break into something looking for evidence, increasing the pool of candidates for an "If you won't, someone else will." approach to getting a warrant approved really isn't much of a virtue.
Maybe they could turn it around, they at least have enough assets to burn to give them some time; but I'd be a little pessimistic about Sony just because they don't seem to have a clue when it comes to software. Some of the products from their glory days definitely have some firmware burned into them, so it's not as though they are utterly incapable of writing any kind of code; but UI/UX and user-facing software are more or less uniformly horrific on Sony products. Unfortunately for them, that's increasingly the part of the product that isn't entirely commodified, and where there is a real difference between companies.
It's somewhat like watching Nintendo try to comprehend what happened to the console market, now that "Having online services that actually works" has become something of a requirement.
They don't directly generate any sound(unlike systems, like CPU/GPU power supply circuitry, that have enough magnetics to really whine under the right load); but basically any digital bus puts out some EMI(not the litigious one, the noisy one) so if your audio player hardware is total shit the analog signal lines may pick some up and feed it into the amp, producing a variety of deeply unpleasant effects.
The root problem in such a case is that the analog lines for the sound output are grossly under-isolated from the rest of the system(and it's likely that one or more other high speed digital busses are scribbling on them) so trying to solve the problem with fancy SD cards is a bit of a waste; but electrical noise seeping into the audio output is certainly a real thing, especially on lousy gear with space and cost constraints.
Given how ghastly things can get if you try to use a really terrible SD card to do an SSD's job(eg. find the cheapest thing that a camera won't spit out in horror, and then write a liveCD image to it and see how much fun you have), I can understand Sony's desire to guarantee a minimum performance level for an expansion card that they knew would be doing almost nothing but storing executables and art assets.
What is inexcusable is the fact that they decided to spin an entirely new format for that purpose, rather than just telling people that "If you don't use one of these, blessed, microSD cards, on your own head be it if the games stutter".
Assuming that they aren't simply lying(sometimes a safe assumption...sometimes not so much) "Class 2", "Class 4", "Class 6", and "Class 10" are supposed to be guarantees of 2, 4, 6, and 10 MB/s minimum write speed, respectively, while UHS1 is supposed to be 10MB/s with support for UHS bus operation, and U3 is supposed to be 30MB/s with support for UHS-II bus operation.
How they store bits, internally, is not specified; but minimum write speed is obviously fairly important to people shooting video or enough large still images, quickly enough, that dumping them to flash can be a bottleneck. Unfortunately, despite the increasingly common case of SD/MMC-bus connected devices being used in situations where read performance matters as well(eg. storing programs in cellphones, being rPi root filesystems) standards for testing and labeling read speed, random I/O performance, and similar SSD-enthusiast stats are more or less nonexistent.
As bulk storage for music files, a Class 4 is likely to do just fine( even golden-ears FLAC is what, 5-6 megabytes/minute?); but unless the stickers are pure lies, a Class 10 is likely to be a rather nicer card.
It's actually pretty hard to fake gold(only Uranium and Tungsten have the right density; but are wrong on everything else, the brasses and bronzes that tepidly approach 'gold' in color don't have the density or chemical properties); but it is much, much, less difficult to deposit it in very, very, thin layers.
Some newer PCs actually made things worse: with the adoption of better power-saving features, it became quite common for the CPU frequency to move up and down depending on whether it was idling or you were throwing something at it at the time, rather than always running at the same frequency.
I had one 'delightful' system where the lousy onboard sound chip was apparently using the CPU clock as a timebase, despite the system being new enough that dynamic clock speed adjustment had been routine for several years. Until the vendor eventually issued a BIOS update that allowed us to turn off clock speed adjustment, we had to rig a low-priority Prime95 process to keep the CPU from idling into a lower speed state if we wanted the playback speed and pitch to be correct. And I though that using the CPU as a timebase died out about the same time as the 'turbo' button...
Gold is a bit overblown in advertising; but (aside from its real physical scarcity, and the unfortunate competition from finance assholes who want to carefully dig it up, refine it, and then have it sit in vaults), it is a genuinely nice ingredient for electrical applications. Adequate thermal and electrical conductivity, immune to most common causes of corrosion, not too difficult to electroplate, fairly easy to tailor from 'shockingly malleable and ductile' to 'adequately hard' just by adding or withholding a few % copper... Good stuff.
The proposed mechanism is at least in agreement with the laws of physics(which is a nice change by audiophile standards); but I have to wonder what kind of terrifyingly awful crap people are playing music on if noise from the SD/SDIO bus is a large enough portion of the problem that even a 100% ideal perfectly silent microSD card would make much of a difference.
Higher end cards have pushed the spec a bit; but SD is not a particularly fast or high-energy bus. It's ubiquitous, cheap, low power, and fast enough, and thus wildly popular; but if somebody's SD interface is causing serious audio issues, the mere thought of what that designer's RAM bus looks like would probably cause the FCC to send out their crack team of death commandos.
"Parallel Construction"... What good is a cool, powerful, sinister toy if you don't have a cover story that allows you to lie about the origins of evidence that would otherwise be inadmissible?
You don't have to like or trust Google(and you shouldn't) to agree that "Hey, let's quietly change rule 41 so that all you need to 'remote search'(by means tactfully unspecified) a computer anywhere is the approval of a judge, doesn't much matter which, from one of the 94 federal districts, rather than one at least vaguely related to the matter at hand!" is...perhaps...a bad move.
As the US demonstrated during the recent massive-clusterfuck-in-a-casino financial meltdown, advances in technology and worker productivity now allow the production of enough fraud to supply the entire industrialized world by a relatively small number of highly trained knowledge workers!
Why, then, should we have an inefficient, unproductive, labor force of blue collar criminals laboriously committing fraud, by hand, like some sort of pre-industrial master/apprentice nonsense, when we have massively more efficient fraud production technology available?
Some do, some don't. SSDs don't draw terribly heroic amounts of power, so shoving enough supercaps into the enclosure to let a drive put its affairs in order when power is pulled is conceptually unproblematic; but it does add cost and bulk, so if they don't promise that they do, they probably don't.
It's not something to freak out too much about(most HDDs have somewhere between 8 and 64MB of cache RAM, and make no particular guarantees about not just dropping its contents on the floor when the power goes out, and somehow we all survived); but a backup is never a bad idea.
The AC at the top specified SLC flash. I'm not sure that Apple even sells anything with an SSD based on that, and, among vendors that do, paying 3-5 times as much per unit capacity would count as a pretty good deal compared to MLC based devices, with higher prices being a definite possibility if the name on the sticker is right.
The odds that they could get paid better to do the same work if the business of crafting really neat attacks and managing defense ends up being handled by contractors?
I hardly theorize that geeks are just too edgy and anti-authoritarian to work for the man, man. That's nonsense. Some repulsive percentage of silicon valley is currently fighting like animals to see who can invade customer privacy faster, a markedly more sordid business; convincing them to supply services to the feds will not be difficult. I'm just not as clear on how enticing them to supply those services as enlisted personnel, rather than as contractors, is going to work; given both the substantial private sector demand for similar skills and the DoD's nontrivial problem with avoiding substantial overpayment for services it buys externally.
My skepticism is not that you'll be able to find people to do the job(outfits like Vupen already do, and I doubt such will become any less common); but that an armed forces recruiter would (barring substantial changes in what they are able to offer) have a particularly good time trying to poach the worthwhile talent from such groups.
If you are of the necessary talent and expertise to be of use for this sort of work, why would you go for the relatively lousy salary, potential to have your career advancement tied to your perceived ability in infantry combat, and comparatively strict rules when you could do the same job for either some private sector outfit, or for the DoD; but as an expensive contractor?
I can imagine why the people hatching these plans might want to have a cooler 'cyber command' than one that simply writes checks to contractors; but I'm having a hard time imagining how they plan to get the people that they need to go along with the idea.
Are the armed services types swarming over this just because if it has 'warrior' in the name they have to get a piece of the action, or do they actually have something resembling a coherent plan for being able to make a convincing pitch to the people they are hoping to attract?
Buying their services as consultants, or as civilian employees of DoD agencies, sure; cut them a check and they'll show right up; but some of these plans actually seem to involve enlisted geeks wearing hilariously incongruous camo in front of banks of monitors and 'cyber warrior'-ing. How is selling that going to work?
Chrome Remote Desktop - Access your desktop from another device. Punches through firewalls and routers automatically.
Unfortunately, it does that by making Google a 3rd party(I think that they even handle the authentication) in every connection you make between two of your own computers. They aren't privy to the actual content of the interaction, to the best of my knowledge; but that still creeps me the hell out.
It's unfortunate, really. An architecturally-modern successor to VNC(ie. same platform-agnostic low level approach; but taking advantage of the fact that most devices can, often with dedicated coprocessors, pump out a very nice H.264 stream or similar as easily or more easily than retro JPEG tiling stuff, along with a dose of some sort of remotely modern authentication) would be fantastic; but CRD doesn't even offer a 'the host is right on the same damn subnet, no, I don't need Google looking over my shoulder to connect to it!' mode.
Honestly, the security is the real slam dunk for me.
If ads were served through the same channels as the rest of the page, and from the same sources, with the same basic level of trustworthiness, I'd be inclined to be at least slightly conflicted about the poor starving site operator; but that's not how it works anymore. Even relatively 'respectable' ad networks are an architectural nightmare; practically designed to make malicious injection easy. The less respectable ones are no better and don't even bother to try to restrain bad actors.
Whatever arguments there are to be made about some 'implicit contract' to put up with ads as part of the ad-supported-model there may be, there is nothing that justifies the security clusterfuck that is ad distribution. You might as well just scavenge for used needles and shove them into your neck hoping for some left-over drugs as accept ads injected into a page.
Slashdot Mirror
Just ask the Holy Spirit, he's consubstantial with both the Son and the Father; despite begetting the son(yet not being the father).
Indeed. It's an adequate enough conductor that plating something with gold won't ruin its conductivity(at least not as badly as a layer of tarnish/oxidation will), and it's good enough to use for wirebonding in chip packaging; but it's merely adequate. Corrosion resistance and extreme ductility are the really neat tricks.
I would have liked to; but(thankfully) I wasn't on the procurement side in the case of that particular job(given the nightmare that was our PO system, I am most grateful). Ultimately pragmatism won out over principle and(for the small subset of affected users) we got some $5 c-media USB sound cards, which proceeded to work perfectly regardless of CPU frequency.
It was one of those situations where the villains skated free; but the number of hours it would have taken on the phone, fighting it out between the vendor of the software that most notably exposed the problem, the PC OEM, the maker of the onboard sound device, and possibly Microsoft, just would have been hell. Not the desired outcome; but the slog to get something better probably would have cost more than entire new systems, never mind new soundcards.
For me, it was mostly a wake-up call about how bloody awful some peripherals and peripheral drivers still are, and how long assumptions that were recognized as dangerous and hacky back when 8MHz was a respectable CPU clock can stick around. Not a pleasant learning experience; but not all are. Somewhat amusing troubleshooting, though. You don't usually expect 100% CPU load to make audio playback better.
SLC is too rich for my blood, so I don't really have the luxury of comparison; but (just as a mixture of RAID and backups has proven cheaper than absolutely bulletproof engineering in HDDs), it wouldn't entirely surprise me if MLC's cost advantage, combined with ongoing improvements in masking its deficiencies, ultimately relegate SLC to relatively niche applications that don't have space for lots of redundancy; but do have the budget for classy hardware. (After all, just look at how well NOR flash's superiority over NAND flash has...mostly not saved it... from being replaced by stacks of cheap NAND behind a controller designed to make it looks like it doesn't suck as much as it does).
I'd be the last to deny that SLC is, in fact, objectively better; but you can buy so much MLC for the price of a given amount of SLC that, given decent controller design, there tends to be room for nontrivial redundancy and greater usable capacity. I can see why it offends purists; but practically all computer equipment today is the direct descendant of inferior crap that beat beautifully engineered and overtly superior systems on price.
My concern is less about that than about the possibility that(if any judge from any district will do) the tendency of the judiciary to rubber-stamp warrants will be markedly increased.
If the FBI has to deal with Judge X, Y, or Z; because they are in district 61, it is at least possible that they'll have to put together a convincing warrant request, lest it be denied. If they can pick any federal district judge, it won't exactly be a big secret which judge you want to talk to if you need some utter bullshit approved.
When it is the job of the judiciary to be at least skeptical, if not somewhat adversarial, when law enforcement comes knocking with a request to go break into something looking for evidence, increasing the pool of candidates for an "If you won't, someone else will." approach to getting a warrant approved really isn't much of a virtue.
Maybe they could turn it around, they at least have enough assets to burn to give them some time; but I'd be a little pessimistic about Sony just because they don't seem to have a clue when it comes to software. Some of the products from their glory days definitely have some firmware burned into them, so it's not as though they are utterly incapable of writing any kind of code; but UI/UX and user-facing software are more or less uniformly horrific on Sony products. Unfortunately for them, that's increasingly the part of the product that isn't entirely commodified, and where there is a real difference between companies.
It's somewhat like watching Nintendo try to comprehend what happened to the console market, now that "Having online services that actually works" has become something of a requirement.
They don't directly generate any sound(unlike systems, like CPU/GPU power supply circuitry, that have enough magnetics to really whine under the right load); but basically any digital bus puts out some EMI(not the litigious one, the noisy one) so if your audio player hardware is total shit the analog signal lines may pick some up and feed it into the amp, producing a variety of deeply unpleasant effects.
The root problem in such a case is that the analog lines for the sound output are grossly under-isolated from the rest of the system(and it's likely that one or more other high speed digital busses are scribbling on them) so trying to solve the problem with fancy SD cards is a bit of a waste; but electrical noise seeping into the audio output is certainly a real thing, especially on lousy gear with space and cost constraints.
Given how ghastly things can get if you try to use a really terrible SD card to do an SSD's job(eg. find the cheapest thing that a camera won't spit out in horror, and then write a liveCD image to it and see how much fun you have), I can understand Sony's desire to guarantee a minimum performance level for an expansion card that they knew would be doing almost nothing but storing executables and art assets.
What is inexcusable is the fact that they decided to spin an entirely new format for that purpose, rather than just telling people that "If you don't use one of these, blessed, microSD cards, on your own head be it if the games stutter".
Assuming that they aren't simply lying(sometimes a safe assumption...sometimes not so much) "Class 2", "Class 4", "Class 6", and "Class 10" are supposed to be guarantees of 2, 4, 6, and 10 MB/s minimum write speed, respectively, while UHS1 is supposed to be 10MB/s with support for UHS bus operation, and U3 is supposed to be 30MB/s with support for UHS-II bus operation.
How they store bits, internally, is not specified; but minimum write speed is obviously fairly important to people shooting video or enough large still images, quickly enough, that dumping them to flash can be a bottleneck. Unfortunately, despite the increasingly common case of SD/MMC-bus connected devices being used in situations where read performance matters as well(eg. storing programs in cellphones, being rPi root filesystems) standards for testing and labeling read speed, random I/O performance, and similar SSD-enthusiast stats are more or less nonexistent.
As bulk storage for music files, a Class 4 is likely to do just fine( even golden-ears FLAC is what, 5-6 megabytes/minute?); but unless the stickers are pure lies, a Class 10 is likely to be a rather nicer card.
It's actually pretty hard to fake gold(only Uranium and Tungsten have the right density; but are wrong on everything else, the brasses and bronzes that tepidly approach 'gold' in color don't have the density or chemical properties); but it is much, much, less difficult to deposit it in very, very, thin layers.
Some newer PCs actually made things worse: with the adoption of better power-saving features, it became quite common for the CPU frequency to move up and down depending on whether it was idling or you were throwing something at it at the time, rather than always running at the same frequency.
I had one 'delightful' system where the lousy onboard sound chip was apparently using the CPU clock as a timebase, despite the system being new enough that dynamic clock speed adjustment had been routine for several years. Until the vendor eventually issued a BIOS update that allowed us to turn off clock speed adjustment, we had to rig a low-priority Prime95 process to keep the CPU from idling into a lower speed state if we wanted the playback speed and pitch to be correct. And I though that using the CPU as a timebase died out about the same time as the 'turbo' button...
Gold is a bit overblown in advertising; but (aside from its real physical scarcity, and the unfortunate competition from finance assholes who want to carefully dig it up, refine it, and then have it sit in vaults), it is a genuinely nice ingredient for electrical applications. Adequate thermal and electrical conductivity, immune to most common causes of corrosion, not too difficult to electroplate, fairly easy to tailor from 'shockingly malleable and ductile' to 'adequately hard' just by adding or withholding a few % copper... Good stuff.
The proposed mechanism is at least in agreement with the laws of physics(which is a nice change by audiophile standards); but I have to wonder what kind of terrifyingly awful crap people are playing music on if noise from the SD/SDIO bus is a large enough portion of the problem that even a 100% ideal perfectly silent microSD card would make much of a difference.
Higher end cards have pushed the spec a bit; but SD is not a particularly fast or high-energy bus. It's ubiquitous, cheap, low power, and fast enough, and thus wildly popular; but if somebody's SD interface is causing serious audio issues, the mere thought of what that designer's RAM bus looks like would probably cause the FCC to send out their crack team of death commandos.
Are you seriously telling me that if you slightly de-cripple a general-purpose computer that it can be used as a general purpose computer?
Truly, I am living in the goddamn future now...
"Parallel Construction"... What good is a cool, powerful, sinister toy if you don't have a cover story that allows you to lie about the origins of evidence that would otherwise be inadmissible?
You don't have to like or trust Google(and you shouldn't) to agree that "Hey, let's quietly change rule 41 so that all you need to 'remote search'(by means tactfully unspecified) a computer anywhere is the approval of a judge, doesn't much matter which, from one of the 94 federal districts, rather than one at least vaguely related to the matter at hand!" is...perhaps...a bad move.
I, for one, welcome this innovation!
As the US demonstrated during the recent massive-clusterfuck-in-a-casino financial meltdown, advances in technology and worker productivity now allow the production of enough fraud to supply the entire industrialized world by a relatively small number of highly trained knowledge workers!
Why, then, should we have an inefficient, unproductive, labor force of blue collar criminals laboriously committing fraud, by hand, like some sort of pre-industrial master/apprentice nonsense, when we have massively more efficient fraud production technology available?
Given the cost per gram of contemporary flash memory, I'm going to guess "Really, really, heroically expensive".
Some do, some don't. SSDs don't draw terribly heroic amounts of power, so shoving enough supercaps into the enclosure to let a drive put its affairs in order when power is pulled is conceptually unproblematic; but it does add cost and bulk, so if they don't promise that they do, they probably don't.
It's not something to freak out too much about(most HDDs have somewhere between 8 and 64MB of cache RAM, and make no particular guarantees about not just dropping its contents on the floor when the power goes out, and somehow we all survived); but a backup is never a bad idea.
The AC at the top specified SLC flash. I'm not sure that Apple even sells anything with an SSD based on that, and, among vendors that do, paying 3-5 times as much per unit capacity would count as a pretty good deal compared to MLC based devices, with higher prices being a definite possibility if the name on the sticker is right.
The odds that they could get paid better to do the same work if the business of crafting really neat attacks and managing defense ends up being handled by contractors?
I hardly theorize that geeks are just too edgy and anti-authoritarian to work for the man, man. That's nonsense. Some repulsive percentage of silicon valley is currently fighting like animals to see who can invade customer privacy faster, a markedly more sordid business; convincing them to supply services to the feds will not be difficult. I'm just not as clear on how enticing them to supply those services as enlisted personnel, rather than as contractors, is going to work; given both the substantial private sector demand for similar skills and the DoD's nontrivial problem with avoiding substantial overpayment for services it buys externally.
My skepticism is not that you'll be able to find people to do the job(outfits like Vupen already do, and I doubt such will become any less common); but that an armed forces recruiter would (barring substantial changes in what they are able to offer) have a particularly good time trying to poach the worthwhile talent from such groups.
That's exactly why I'm confused.
If you are of the necessary talent and expertise to be of use for this sort of work, why would you go for the relatively lousy salary, potential to have your career advancement tied to your perceived ability in infantry combat, and comparatively strict rules when you could do the same job for either some private sector outfit, or for the DoD; but as an expensive contractor?
I can imagine why the people hatching these plans might want to have a cooler 'cyber command' than one that simply writes checks to contractors; but I'm having a hard time imagining how they plan to get the people that they need to go along with the idea.
Are the armed services types swarming over this just because if it has 'warrior' in the name they have to get a piece of the action, or do they actually have something resembling a coherent plan for being able to make a convincing pitch to the people they are hoping to attract?
Buying their services as consultants, or as civilian employees of DoD agencies, sure; cut them a check and they'll show right up; but some of these plans actually seem to involve enlisted geeks wearing hilariously incongruous camo in front of banks of monitors and 'cyber warrior'-ing. How is selling that going to work?
Chrome Remote Desktop - Access your desktop from another device. Punches through firewalls and routers automatically.
Unfortunately, it does that by making Google a 3rd party(I think that they even handle the authentication) in every connection you make between two of your own computers. They aren't privy to the actual content of the interaction, to the best of my knowledge; but that still creeps me the hell out.
It's unfortunate, really. An architecturally-modern successor to VNC(ie. same platform-agnostic low level approach; but taking advantage of the fact that most devices can, often with dedicated coprocessors, pump out a very nice H.264 stream or similar as easily or more easily than retro JPEG tiling stuff, along with a dose of some sort of remotely modern authentication) would be fantastic; but CRD doesn't even offer a 'the host is right on the same damn subnet, no, I don't need Google looking over my shoulder to connect to it!' mode.
Honestly, the security is the real slam dunk for me.
If ads were served through the same channels as the rest of the page, and from the same sources, with the same basic level of trustworthiness, I'd be inclined to be at least slightly conflicted about the poor starving site operator; but that's not how it works anymore. Even relatively 'respectable' ad networks are an architectural nightmare; practically designed to make malicious injection easy. The less respectable ones are no better and don't even bother to try to restrain bad actors.
Whatever arguments there are to be made about some 'implicit contract' to put up with ads as part of the ad-supported-model there may be, there is nothing that justifies the security clusterfuck that is ad distribution. You might as well just scavenge for used needles and shove them into your neck hoping for some left-over drugs as accept ads injected into a page.