They evolved the most aggressive society in the world, started a war, got nuked, and discovered they were pacifists (with apologies to Neal Stephenson). The aggressiveness is still there, but it's being suppressed. The internal pressure results in... strangeness.
For the amount of email most people do, a few dozens of dollars into a Raspberry Pi would work just fine as an email server. $500 seems like huge overkill.
From the sound of things, they're concealing operating expenses in the price of the hardware, the exact opposite model for inkjet printers. No box "the size of a router" has $400 worth of hardware in it, let alone $(500*0.97) worth (typical 3% margin Asian manufacturers operate on). Not when a cell phone with a capacitive touchscreen and a battery in addition to all the required resources to run a low volume mailserver is $30.
Now if it's 100% redundant hardware, including two independent wall warts for power, externally accessible redundant hot-swappable flash storage, externally accessible redundant hot-swappable CPU/RAM cartridges, and dual embedded externally accessible hot-swappable lithium batteries, then it's worth $500. Basically two cell phones worth of hardware smooshed together in an electrically sensible way, with custom software that can deal with a CPU disappearing and reappearing.
Having described it, that is what I want. Why isn't someone making that? For less than $OUTRAGEOUS_SUM, in less than a 4U form factor.
802.11ay is supposed to enjoy sub-millisecond latency, definitely a prerequisite for VR.
The two highest resolution VR headsets are currently 2160x1200 (both eyes combined) and should run at 90 Hz.
A little arithmetic: 2160 * 1200 * 90 * 12 = 2799360000 bits per second. 2.8 gigabits (not gibibits) per second, near as damnit. That fits nicely in even 802.11ad. Put the transmitter in the ceiling and 90% of the VR use case is solved.
Rumor has it HTC is creating a WiGig accessory for the Vive and Vive Pro due out this year, but if they were going to make it in time for the holiday season, I would have expected them to start up a hype train, and I see no sign of one. Last articles on the subject are from January, claiming a Q3 release, and are fairly speculative. I'm betting they're trying to avoid requiring installation of the transceiver on the ceiling and it's not going well.
Sell it to a sleazy retailer who wants to track & identify people entering their stores?
They won't. They are that sleazy retailer. They certainly aren't going to give up their competitive advantage of identifying shoplifters the moment they walk in the door. Nor are they going to give up their competitive advantage of identifying the sucker walking in the door that responds well to a cleverly placed electronic coupon on an aisle display screen.
It's much easier to make a robot that picks and packs regularly shaped clothing packets (Uniqlo sells a lot of stuff in cardboard stiffened plastic packets for example) than one which can handle arbitrary shapes like a human can.
Walmart has had the clout to dictate packaging to its suppliers for many many years already. Most recently they demanded, and got, RFID chips in all pallets.
Amazon is plenty big enough to start dictating packaging, if they aren't already. The availability of these material handling robots gives them incentive to do just that.
I expect millions of consumer items to be subjected to the same sort of selection pressures that dry goods in grocery stores have been subjected to for decades. There aren't 47 different shapes of soup can on the shelf. There's 3. And Progresso was pushing it when they went against the grain and put out their own shape. There aren't 19 different cereal box shapes. There's 2. And General Mills had a fight on their hands when they pushed through that second size. For years there was one.
The presence of material handling robots is going to start putting enough pressure on consumer goods packaging to override the currently countervailing pressure against spending more on packaging than the absolute minimum required for legal reasons. There might still be a few hundred formfactors, but there won't be millions anymore. Automated retail will see to it.
If we could find out how to build houses in an automated fashion (and thus make housing cheaper while cutting out an awful lot of labour costs), we could drastically change the world... maybe then rents/mortgages wouldn't be in the "half your wages or more" category.
I can conceive of methods for automating 90% of building a house, but I don't have the capital to develop those ideas and have no interest in sucking up to the people who do because they're assholes.
I suspect a great many people with engineering degrees feel the same way.
They're asking for common knowledge, not common sense.
Instilling common knowledge into an intelligence takes approximately 18 years. You're allowed 8 hours per day of offline processing. Approximately 17% of intelligences can be expected to fail to complete the training.
I got involved as a sports fan for a while. It was a great Petri dish to explore human cognition. But then my favourite resource disappeared behind a paywall. Sure, I could pay. But now the discussion is limited to include only those people who choose to pay. The group structure is now inherently different. It's no longer such a great Petri dish for me to explore human cognition (having become far more captive and insular).
Dude, you are waaaaay overthinking this. It's baseball. There's effectively no cognition involved. That's rather the point, as far as I can tell.
P.S. every quotation mark should be two instances of a 32-character random nonce, never to be ever used again.
Dude, you are way way way waaaaay overthinking this. Exponential escape growth is telling you to refactor your script into functions, or failing that, abandon shell script for a proper programming language. That rabbit hole you are digging at does not have a rabbit at the bottom of it. The burrow was abandoned long ago.
So stop whimpering that your clever multiply-nested commands have more backslash escape characters than a Jupiter-scale Pine Barrens on Ringworld after a small asteroid hull breach that doesn't clear the upper atmosphere. GET OVER IT you irritating shit.
This sounds just awful. Alexa is going to pimp cough drops to me all day long, despite the fact that I already have two bags of them sitting in the cabinet?
Until you buy them from Amazon, yes.
Or the fact that I've already got a case of chicken soup in the pantry?
Better buy your soup from Whole Foods, comrade.
If I'm vegan, is Alexa still going to pimp the chicken soup, or is it going to push some BS homeopathic remedy instead?
That depends. Who paid them more this week, Campbell's or the bullshit merchant? Or why not both? I give it less than 2 years before the damn thing starts playing literal ad jingles.
It won't work. It can't work. I say this because it uses the same materials for the salt and piping as molten salt nuclear reactors and I've been told the salt will simply eat through the pipes and all you will have is an expensive mess.
Only for extremely high temperature use. Liquid fluoride salts may be useful for storage systems that operate between 700 and 800 degrees C. However that's not necessary to make an operational system. It's entirely possible to build one that operates below 600 degrees C, and in fact this is the most common because that's the operating temperature regime of steam turbines. Between 600 and 700 degrees, gas turbines are used.
Commercial molten salt storage is in operation today, using sodium nitrate, potassium nitrate, or calcium nitrate, running at an operating temperature on the hot side of 566 degrees C. No fluoride salt storage is in operation precisely because of the difficulties and dangers involved. Whereas fluoride salt is optional for solar-driven molten salt storage technology, it is not optional for fission reactors, which depend on the properties of liquid fluoride to chemically bind with uranium and its fission byproducts.
I only say that the same applies to nuclear power, it can be put underneath anything. It can be under an airport, a military base, or a bunch of solar collectors. Land use is effectively zero for both but the power output per area is very low for solar but nearly unlimited for nuclear.
Neither is true.
First, nuclear power density has very hard upper limits. Exceeding them is called a nuclear bomb.
Second, nuclear power requires some sort of cold side, the same as coal or gas, or indeed solar thermal molten salt power systems. For the vast majority of nuclear power plants, this is a large body of water. For those few built without a ready source of water, a very large amount of atmospheric cooling is required. This cooling tower is the most visible, most iconic part of nuclear power plant design, and it isn't optional when building high efficiency power generation facilities. People who propose containerized nuclear reactors always hand-wave away the cooling tower, though they all mention it in the fine print as being a necessary part of an installation. Even naval reactors still use ocean water for coolant, and sacrifice efficiency because they don't have room for a cooling tower, making up for some of it by being mobile.
Therefore I conclude that they don't believe CO2 emissions are any real threat to humanity or national security.
Correct. No government in the world believes CO2 emissions are any real threat. Some governments believe they can get free money from unfounded panic about CO2 emissions. Since those same governments have already gotten and continue to get free money because there might be Communists (with a capital C) under the bed, they are probably correct.
How ironic if we were to eliminate ageing just as the planet decides to rid itself of us irritating little ticks.
People who can actually live long enough to experience a real change of climate might be more inclined to pay attention to the possibility of a poor change.
Of all the articles on Slashdot, this one really WILL attract FUD from Russian trolls. It's Russia that is losing their NASA contract for ferrying US astronauts to the space station when SpaceX (and eventually Boeing) can do it.
The Trump administration had better invent some bullshit reason for continuing to funnel money to Roscosmos or a bunch of maintenance guys who know how to cobble together an ICBM out of spare parts are going to be out of a job.
That's why claiming Boeing is working on "flying cars" is a lot of "bullshit". They aren't. They're working on air taxis. They will have all the same restrictions as helicopters, but they won't have a pilot. None of them are going to be singlecopters, either, nor run on fossil fuels. They will all be battery multicopters. That makes them quieter than helis at all times, so while they still won't be fit to take off from driveways, they will be able to take off from more locations.
Now if only the US military would let them use the technology that makes stealth helicopter blades so quiet. Then they'd really have something.
That is also true. In fact, people who talk to strangers on the subway/bus/whatever are usually weirdos.
In the United States that varies radically by geography. In Michigan, no one ever talks to strangers. In Illinois, everyone does. Most of the Midwest is willing to talk to strangers. The northeast, not so much.
Yes we've landed on the moon but claims that there is no more we can learn by going again are manifestly absurd. There is a ton of engineering and science we could learn by going again.
More to the point, going and staying for a while. Flags and bootprints a second time is definitely useless. Flags, bootprints, and a place to live are considerably more useful. That requires solving new engineering problems.
The real issue is technical. How do we create a secure compute environment?
Starting with simplicity. The simplest CPU is really a very small thing. College students reimplement the MIPS instruction set on a regular basis. It's possible to know what literally every transistor is for in the chip design, build that system on a chip, and use it as the fundamental building blocks of your system. Even if everything else you purchase is suspicious, you can be absolutely certain of one device on your network. You probably can't fabricate the chip yourself, but since you know it inside and out, you can test it, independently, and verify it only ever does what you expect it to do. You also probably can't fabricate the carrier board for it, but you can probably design a single layer board to host it which is 100% visually and electrically auditable. Then you can add all the components to it yourself. The Maker movement has made solder paste a cheap off-the-shelf product, and you can buy a toaster oven anywhere.
The result won't be fast, and won't be running any Microsoft operating system, but you didn't want to do that anyway.
They told us we were too dumb to make our own stuff.
Then they told us that people are too expensive to make our own stuff.
Then they told us after automating the factory floor, making labor costs insignificant we have to have a monopoly or we can't compete.
I wonder what their excuse will be now why we can't make our own stuff?
The excuse, which isn't an excuse, is that we don't know how. It's quite literally true. Building a high frequency mainboard correctly is nontrivial, and while we know how to design them, and know how to set up automated tests for them, we don't know all the little tricks that actual manufacturers have learned by doing the job for decades.
Sparkfun has been finding that out, and documenting some of it publicly. They bought a pick and place machine so they could fabricate their own boards for some of the stuff they design. Getting it to work reliably was a journey, and not an easy one. And that's for crappy little $20 low frequency parts that work even on a breadboard, not gigahertz boards worth $1000 before you even drop a CPU onto them.
Somebody will be learning how again. You can bet that now that it's public, the US government acquisitions process will start mandating US assembly for boards it buys for use in classified environments. Somebody will jump on that, because they'll be able to charge a huge premium for a while, since there will be no other option. Monopoly pricing always attracts the US business man.
How is a single chip on a motherboard going to do the following and do it without someone noticing:
1: Intercept data on the server without knowledge of what OS is running and/or without a driver to facilitate OS access?
2: Send that data to some 3rd party, through a firewall, without the bandwidth usage being noticed?
The entire point of this article is that traffic was noticed. Amazon wanted to buy Elemental. Amazon was auditing Elemental. Amazon's auditors found unexplained network traffic. Not very much of it, but they found some. And in the process of trying to explain it, somebody got very intrigued, and shipped off the servers that were sending unexplained packets to a security firm in Canada, and that's how Bloomberg has a story to write about. The data was noticed.
There's a reason why US spy agencies prefer wireless. First because historically that's all there was, but also because of this. Any good sysadmin knows what traffic is flowing on their network, and notices traffic that's not supposed to be there.
Complete lack of any hard evidence to support Bloomberg's claims aside, if you were to take it at face value then you've got to hand the Amazon team some *serious* respect for noticing that there was an additional chip the size of a pencil tip on some of a their server boards that was not present on others or in the design spec. And that's before you consider that they didn't just blow it off and supposedly figured out at least some of the things that it was up to.
According to the article, Amazon saw unexplained network traffic during their due diligence inspections of Elemental's operations, couldn't explain it themselves, but did isolate it to specific machines, and shipped them off to a security firm in Canada to figure them out.
That Canadian company is the one who deserves serious respect for determining that nothing in the CPU or the OS was anything other than as expected and so the problem must be somewhere on the board. They're the ones who went over those boards literally component by component until they figured out what was going on. Presumably they guessed it was near the IPMI chips, which made the job easier (no mucking around with the CPU's power supply capacitors required), but it's still an impressive feat.
Everyone involved on both sides has come out publicly to say Bloomberg is wrong. Why are we still talking about it?
All parties involved have it in their vested interest to deny this.
All parties are required by law to deny this. It's a classified investigation which Bloomberg says is still open. According to Bloomberg's reporting, they don't just want to deny it—they have to deny it. With the Supermicro boards in question in use by the DOD and the CIA, it's quite literally a matter of national security.
Slashdot Mirror
Wolves will also kill and eat small humans.
Not if they're wearing a red hood.
You're as curious about Japanese Hip-hop's #1 as I am : https://www.youtube.com/watch?...
What the fuck is WRONG with these people?
They evolved the most aggressive society in the world, started a war, got nuked, and discovered they were pacifists (with apologies to Neal Stephenson). The aggressiveness is still there, but it's being suppressed. The internal pressure results in... strangeness.
You mean this that you can buy now?
Uh, yes, that. What is HTC scared of that they haven't been advertising it?
For the amount of email most people do, a few dozens of dollars into a Raspberry Pi would work just fine as an email server. $500 seems like huge overkill.
From the sound of things, they're concealing operating expenses in the price of the hardware, the exact opposite model for inkjet printers. No box "the size of a router" has $400 worth of hardware in it, let alone $(500*0.97) worth (typical 3% margin Asian manufacturers operate on). Not when a cell phone with a capacitive touchscreen and a battery in addition to all the required resources to run a low volume mailserver is $30.
Now if it's 100% redundant hardware, including two independent wall warts for power, externally accessible redundant hot-swappable flash storage, externally accessible redundant hot-swappable CPU/RAM cartridges, and dual embedded externally accessible hot-swappable lithium batteries, then it's worth $500. Basically two cell phones worth of hardware smooshed together in an electrically sensible way, with custom software that can deal with a CPU disappearing and reappearing.
Having described it, that is what I want. Why isn't someone making that? For less than $OUTRAGEOUS_SUM, in less than a 4U form factor.
802.11ay is supposed to enjoy sub-millisecond latency, definitely a prerequisite for VR.
The two highest resolution VR headsets are currently 2160x1200 (both eyes combined) and should run at 90 Hz.
A little arithmetic: 2160 * 1200 * 90 * 12 = 2799360000 bits per second. 2.8 gigabits (not gibibits) per second, near as damnit. That fits nicely in even 802.11ad. Put the transmitter in the ceiling and 90% of the VR use case is solved.
Rumor has it HTC is creating a WiGig accessory for the Vive and Vive Pro due out this year, but if they were going to make it in time for the holiday season, I would have expected them to start up a hype train, and I see no sign of one. Last articles on the subject are from January, claiming a Q3 release, and are fairly speculative. I'm betting they're trying to avoid requiring installation of the transceiver on the ceiling and it's not going well.
Sell it to a sleazy retailer who wants to track & identify people entering their stores?
They won't. They are that sleazy retailer. They certainly aren't going to give up their competitive advantage of identifying shoplifters the moment they walk in the door. Nor are they going to give up their competitive advantage of identifying the sucker walking in the door that responds well to a cleverly placed electronic coupon on an aisle display screen.
It's much easier to make a robot that picks and packs regularly shaped clothing packets (Uniqlo sells a lot of stuff in cardboard stiffened plastic packets for example) than one which can handle arbitrary shapes like a human can.
Walmart has had the clout to dictate packaging to its suppliers for many many years already. Most recently they demanded, and got, RFID chips in all pallets.
Amazon is plenty big enough to start dictating packaging, if they aren't already. The availability of these material handling robots gives them incentive to do just that.
I expect millions of consumer items to be subjected to the same sort of selection pressures that dry goods in grocery stores have been subjected to for decades. There aren't 47 different shapes of soup can on the shelf. There's 3. And Progresso was pushing it when they went against the grain and put out their own shape. There aren't 19 different cereal box shapes. There's 2. And General Mills had a fight on their hands when they pushed through that second size. For years there was one.
The presence of material handling robots is going to start putting enough pressure on consumer goods packaging to override the currently countervailing pressure against spending more on packaging than the absolute minimum required for legal reasons. There might still be a few hundred formfactors, but there won't be millions anymore. Automated retail will see to it.
If we could find out how to build houses in an automated fashion (and thus make housing cheaper while cutting out an awful lot of labour costs), we could drastically change the world... maybe then rents/mortgages wouldn't be in the "half your wages or more" category.
I can conceive of methods for automating 90% of building a house, but I don't have the capital to develop those ideas and have no interest in sucking up to the people who do because they're assholes.
I suspect a great many people with engineering degrees feel the same way.
First minutes of a real war and we will all be dead
Speak for yourself. I never leave my bunker.
They're asking for common knowledge, not common sense.
Instilling common knowledge into an intelligence takes approximately 18 years. You're allowed 8 hours per day of offline processing. Approximately 17% of intelligences can be expected to fail to complete the training.
Good luck.
I got involved as a sports fan for a while. It was a great Petri dish to explore human cognition. But then my favourite resource disappeared behind a paywall. Sure, I could pay. But now the discussion is limited to include only those people who choose to pay. The group structure is now inherently different. It's no longer such a great Petri dish for me to explore human cognition (having become far more captive and insular).
Dude, you are waaaaay overthinking this. It's baseball. There's effectively no cognition involved. That's rather the point, as far as I can tell.
P.S. every quotation mark should be two instances of a 32-character random nonce, never to be ever used again.
Dude, you are way way way waaaaay overthinking this. Exponential escape growth is telling you to refactor your script into functions, or failing that, abandon shell script for a proper programming language. That rabbit hole you are digging at does not have a rabbit at the bottom of it. The burrow was abandoned long ago.
So stop whimpering that your clever multiply-nested commands have more backslash escape characters than a Jupiter-scale Pine Barrens on Ringworld after a small asteroid hull breach that doesn't clear the upper atmosphere. GET OVER IT you irritating shit.
And take your meds.
Am I the only one who thinks these auto-reply systems are exactly as (in)capable as Dr. Sbaitso?
This sounds just awful. Alexa is going to pimp cough drops to me all day long, despite the fact that I already have two bags of them sitting in the cabinet?
Until you buy them from Amazon, yes.
Or the fact that I've already got a case of chicken soup in the pantry?
Better buy your soup from Whole Foods, comrade.
If I'm vegan, is Alexa still going to pimp the chicken soup, or is it going to push some BS homeopathic remedy instead?
That depends. Who paid them more this week, Campbell's or the bullshit merchant? Or why not both? I give it less than 2 years before the damn thing starts playing literal ad jingles.
It won't work. It can't work. I say this because it uses the same materials for the salt and piping as molten salt nuclear reactors and I've been told the salt will simply eat through the pipes and all you will have is an expensive mess.
Only for extremely high temperature use. Liquid fluoride salts may be useful for storage systems that operate between 700 and 800 degrees C. However that's not necessary to make an operational system. It's entirely possible to build one that operates below 600 degrees C, and in fact this is the most common because that's the operating temperature regime of steam turbines. Between 600 and 700 degrees, gas turbines are used.
Commercial molten salt storage is in operation today, using sodium nitrate, potassium nitrate, or calcium nitrate, running at an operating temperature on the hot side of 566 degrees C. No fluoride salt storage is in operation precisely because of the difficulties and dangers involved. Whereas fluoride salt is optional for solar-driven molten salt storage technology, it is not optional for fission reactors, which depend on the properties of liquid fluoride to chemically bind with uranium and its fission byproducts.
I only say that the same applies to nuclear power, it can be put underneath anything. It can be under an airport, a military base, or a bunch of solar collectors. Land use is effectively zero for both but the power output per area is very low for solar but nearly unlimited for nuclear.
Neither is true.
First, nuclear power density has very hard upper limits. Exceeding them is called a nuclear bomb.
Second, nuclear power requires some sort of cold side, the same as coal or gas, or indeed solar thermal molten salt power systems. For the vast majority of nuclear power plants, this is a large body of water. For those few built without a ready source of water, a very large amount of atmospheric cooling is required. This cooling tower is the most visible, most iconic part of nuclear power plant design, and it isn't optional when building high efficiency power generation facilities. People who propose containerized nuclear reactors always hand-wave away the cooling tower, though they all mention it in the fine print as being a necessary part of an installation. Even naval reactors still use ocean water for coolant, and sacrifice efficiency because they don't have room for a cooling tower, making up for some of it by being mobile.
Therefore I conclude that they don't believe CO2 emissions are any real threat to humanity or national security.
Correct. No government in the world believes CO2 emissions are any real threat. Some governments believe they can get free money from unfounded panic about CO2 emissions. Since those same governments have already gotten and continue to get free money because there might be Communists (with a capital C) under the bed, they are probably correct.
Switching from a totally swipe based screen interface on phone to track pad on the keyboard is the major stumbling block It think.
Pretty sure you're right, especially since Microsoft is experiencing exactly that problem from the opposite end.
How ironic if we were to eliminate ageing just as the planet decides to rid itself of us irritating little ticks.
People who can actually live long enough to experience a real change of climate might be more inclined to pay attention to the possibility of a poor change.
Of all the articles on Slashdot, this one really WILL attract FUD from Russian trolls. It's Russia that is losing their NASA contract for ferrying US astronauts to the space station when SpaceX (and eventually Boeing) can do it.
The Trump administration had better invent some bullshit reason for continuing to funnel money to Roscosmos or a bunch of maintenance guys who know how to cobble together an ICBM out of spare parts are going to be out of a job.
That's why claiming Boeing is working on "flying cars" is a lot of "bullshit". They aren't. They're working on air taxis. They will have all the same restrictions as helicopters, but they won't have a pilot. None of them are going to be singlecopters, either, nor run on fossil fuels. They will all be battery multicopters. That makes them quieter than helis at all times, so while they still won't be fit to take off from driveways, they will be able to take off from more locations.
Now if only the US military would let them use the technology that makes stealth helicopter blades so quiet. Then they'd really have something.
That is also true. In fact, people who talk to strangers on the subway/bus/whatever are usually weirdos.
In the United States that varies radically by geography. In Michigan, no one ever talks to strangers. In Illinois, everyone does. Most of the Midwest is willing to talk to strangers. The northeast, not so much.
Yes we've landed on the moon but claims that there is no more we can learn by going again are manifestly absurd. There is a ton of engineering and science we could learn by going again.
More to the point, going and staying for a while. Flags and bootprints a second time is definitely useless. Flags, bootprints, and a place to live are considerably more useful. That requires solving new engineering problems.
The real issue is technical. How do we create a secure compute environment?
Starting with simplicity. The simplest CPU is really a very small thing. College students reimplement the MIPS instruction set on a regular basis. It's possible to know what literally every transistor is for in the chip design, build that system on a chip, and use it as the fundamental building blocks of your system. Even if everything else you purchase is suspicious, you can be absolutely certain of one device on your network. You probably can't fabricate the chip yourself, but since you know it inside and out, you can test it, independently, and verify it only ever does what you expect it to do. You also probably can't fabricate the carrier board for it, but you can probably design a single layer board to host it which is 100% visually and electrically auditable. Then you can add all the components to it yourself. The Maker movement has made solder paste a cheap off-the-shelf product, and you can buy a toaster oven anywhere.
The result won't be fast, and won't be running any Microsoft operating system, but you didn't want to do that anyway.
They told us we were too dumb to make our own stuff.
Then they told us that people are too expensive to make our own stuff.
Then they told us after automating the factory floor, making labor costs insignificant we have to have a monopoly or we can't compete.
I wonder what their excuse will be now why we can't make our own stuff?
The excuse, which isn't an excuse, is that we don't know how. It's quite literally true. Building a high frequency mainboard correctly is nontrivial, and while we know how to design them, and know how to set up automated tests for them, we don't know all the little tricks that actual manufacturers have learned by doing the job for decades.
Sparkfun has been finding that out, and documenting some of it publicly. They bought a pick and place machine so they could fabricate their own boards for some of the stuff they design. Getting it to work reliably was a journey, and not an easy one. And that's for crappy little $20 low frequency parts that work even on a breadboard, not gigahertz boards worth $1000 before you even drop a CPU onto them.
Somebody will be learning how again. You can bet that now that it's public, the US government acquisitions process will start mandating US assembly for boards it buys for use in classified environments. Somebody will jump on that, because they'll be able to charge a huge premium for a while, since there will be no other option. Monopoly pricing always attracts the US business man.
How is a single chip on a motherboard going to do the following and do it without someone noticing:
1: Intercept data on the server without knowledge of what OS is running and/or without a driver to facilitate OS access?
2: Send that data to some 3rd party, through a firewall, without the bandwidth usage being noticed?
The entire point of this article is that traffic was noticed. Amazon wanted to buy Elemental. Amazon was auditing Elemental. Amazon's auditors found unexplained network traffic. Not very much of it, but they found some. And in the process of trying to explain it, somebody got very intrigued, and shipped off the servers that were sending unexplained packets to a security firm in Canada, and that's how Bloomberg has a story to write about. The data was noticed.
There's a reason why US spy agencies prefer wireless. First because historically that's all there was, but also because of this. Any good sysadmin knows what traffic is flowing on their network, and notices traffic that's not supposed to be there.
Complete lack of any hard evidence to support Bloomberg's claims aside, if you were to take it at face value then you've got to hand the Amazon team some *serious* respect for noticing that there was an additional chip the size of a pencil tip on some of a their server boards that was not present on others or in the design spec. And that's before you consider that they didn't just blow it off and supposedly figured out at least some of the things that it was up to.
According to the article, Amazon saw unexplained network traffic during their due diligence inspections of Elemental's operations, couldn't explain it themselves, but did isolate it to specific machines, and shipped them off to a security firm in Canada to figure them out.
That Canadian company is the one who deserves serious respect for determining that nothing in the CPU or the OS was anything other than as expected and so the problem must be somewhere on the board. They're the ones who went over those boards literally component by component until they figured out what was going on. Presumably they guessed it was near the IPMI chips, which made the job easier (no mucking around with the CPU's power supply capacitors required), but it's still an impressive feat.
Everyone involved on both sides has come out publicly to say Bloomberg is wrong. Why are we still talking about it?
All parties involved have it in their vested interest to deny this.
All parties are required by law to deny this. It's a classified investigation which Bloomberg says is still open. According to Bloomberg's reporting, they don't just want to deny it—they have to deny it. With the Supermicro boards in question in use by the DOD and the CIA, it's quite literally a matter of national security.