Then why are not all human officers required to have their eyes replaced with cybernetic implants?
The U.S. Army is already there with the Warfighter Refractive Eye Surgery Program.
"To increase combat readiness, the Department of Defense has established the Warfighter Refractive Eye Surgery Program (WRESP) which allows eligible active-duty soldiers to receive laser refractive eye surgery. The goal is to minimize or eliminate the need to wear corrective eyewear."
Re:that old canard about x86 complexity
on
Apple, ARM, and Intel
·
· Score: 4, Interesting
Far closer to the truth of the matter is that x86 has a much higher design cost than an orthogonal clean-sheet alternative.
True. Years ago I went to a talk where the head of the Pentium Pro design team showed a graph of the number of engineers working on the project. It peaked around 3,000. Nobody had ever had a CPU design team that big before.
The variable length instruction alignment problem of x86, although ugly, isn't a huge consumer of transistors. AMD dealt with it by expanding instructions to fixed length when loaded into cache. Intel dealt with it by sometimes starting ambiguous cases in parallel and discarding the bogus results later. The downside of fixed-length instructions, as in RISC machines, is code bloat - PowerPC code is about twice as big as x86 code, which impacts cache miss rate.
While one instruction per clock RISC CPUs (low-end MIPS and DEC Alpha parts, and the Atmel AVR series are examples) are simple, superscalar machines executing more than one instruction per clock are almost as complex as x86 CPUs. That's why RISC stopped being a win.
Harry Pyle was developing the instruction set for the Datapoint 2200 in his dorm room at Case Tech in Cleveland in the late 1960s. Same building I was in; different floor. That led to the 8008 and the 8080 and the 80286 and the 80386 and...
Laser cutters for sheet metal are widely used industrially, but need more power than this little one. 150W to 6000W CO2 lasers are used for metal.
Metal cutting is more difficult than cutting wood or plastic. Getting a clean cut is harder. On plastics or wood you can burn your way through slowly, but on metals, you need enough power for fast cutting or you get slag on the edges.
I was just cutting medium density fiberboard on a laser cutter last night. The problem is not the laser beam igniting the cabinet. That's hard to do. The problem is igniting the workpiece, which is easy for many materials, and the cabinet not being able to contain the resulting fire. The cutting process should take place in a nonflammable box with an exhaust to the outside.
Sheet steel is cheap. Spot welding is cheap. This is not rocket science.
This has nothing to do with processing the data from the rover (which comes in at rates a dial-up modem could handle). It's about the web hosting system that lets casual visitors look at the pretty pictures.
NASA could just upload the stills to Flickr and the videos to Youtube and save some money.
It's software for running an upload site. It has shiny graphics, shinier than most other upload systems. The page layouts are squares in a grid, with a plain background. so it looks phone-like and trendy.
As of right now, there is one (1) open MediaGoblin site. There are also two with registration disabled, and two of (non-porno) baby pictures.
The current install setup is too complicated and requires root privileges. If this thing was made as easy to install as Wordpress, it might be more useful.
ASUS discontinued their entire netbook line on September 4, 2012.
Low cost netbooks with large hard drives interfered with the "lock users into the cloud then raise the price and make ads more intrusive" strategy of Google, Facebook, and Microsoft.
Due to a court decision about ten years ago, patent holders now have to sue before they can discuss patent licensing. If you accuse someone of infringement, they can sue to have the patent invalidated, and they have some tactical advantages if they sue first. So patent holders don't send infringement letters any more. They have to start with litigation. That's the reason for the boom in litigation.
This makes sense for sales reps. Their job is to talk to people and convince them to buy. Maybe take the actual order. That doesn't require much input or local computing power, but it requires convenient access to catalog and customer data. This is a business application that maps well to small screen mobile devices.
The press release is deceptive. They did not build a
working turbofan engine with a 3D printer. They built a plastic scale model of a Rolls Royce turbofan engine with a Stratasys 3D printer. It will rotate if powered with compressed air. Rolls Royce gave U of VA a $2 million dollar grant which supported that effort.
The plane itself wasn't printed as one piece. It was more like printing the parts of a plane kit. Very slowly. 80 hour weeks are mentioned. Not sure where the $2000 cost figure comes from, but it doesn't include labor or 3D printer time. Maybe that's just the plastic cost.
The Science Fiction Museum in Seattle seemed a likely place for this to end up, but that, like most Paul Allen projects, went bust. It closed in 2011. There's no really good place to put this.
A device that 1) has a data link to the outside world, 2) has a GPS receiver, and 3) has a microphone ought to be far riskier to steal. Something is not right here.
You should be able to log into your phone account from another device and retrieve the location of the phone, Maybe listen in and record calls from your phone, too.
The "Silent Circle" uses their own "Silent Network", allowing centralized user tracking. Also, the code isn't open source, so you have no idea if the crypto key generation is any good or if there are backdoors.
The problem is microkernels suck in many ways and no one wants them.
The problem is that academic microkernels, Mach in particular, suck. Mach is a BSD derivative.
QNX is quite good. You pay some extra copying overhead, maybe 10-20% of CPU time, or a few months of Moore's Law, for message passing. The microkernel itself is about 60K of code. There's hope of getting something that small perfect. QNX, though, is not intended to be a high-security system; its purpose is to be a high reliability real-time system with repeatable response time.
Many real-time and automotive applications use QNX.
Linux + SELinux, (SELinux, which was originally built by the NSA for those who don't know enough history to realise) is an operating system with an immutable watchdog. What more do you want?
SELinux wasn't intended to be highly secure. It's an add-on to Linux, after all, not a new OS. The purpose of SELinux was to get a mandatory-security system out and widely used so that applications would be written to run under tight restrictions. Read what NSA originally wrote about it.
A big problem with secure operating systems is getting applications to run in a secure environment. That means saying "no" a lot. No, your game can't find out what else is running. No, Photoshop can't snoop the LAN for other instances of Photoshop with the same serial number. No, you can't run code in a spreadsheet attached to an email. No, you can't have a browser which has pages from multiple sites in the same memory space. That's what it means to have a secure OS.
The hope of SELinux was that applications would gradually be rewritten to run under tight restrictions like that. It didn't happen.
Look how much whining there is whenever Microsoft tightens up Windows. Users will choose ad-supported games that phone home over security.
Right; actually half were generated by the end of 2010 or so. The difficulty level, the amount of computation required to get a Bitcoin, was very low until the end of 2009. It has climbed by 7 orders of magnitude since then. That's why the people who set up Bitcoin have most of the Bitcoins.
Coursera makes their site look like they're a real academic institution. Yes, they say "free", but lots of sites say "free" and lie. It's not clear whether "free" is just a bait to get people to sign up, and then attempts are made to "upgrade" them to a paid account. Or there may be "fees". Their terms say "We reserve the right to change or modify the Terms of Use at our sole discretion at any time. Any change or modification to the Terms of Use will be effective immediately upon posting by us. " So they could add fees at any time.
There's a scam where online schools sign up students, and collect enough information that the school can apply for financial aid from the U.S. Department of Education. "The inspector general's office says participants in 42 different fraud rings have been convicted and more than $7.5 million in restitution and fines have been ordered in the past six years. This may be only a small portion of the problem."
See Belford University, which the BBB says is a scam that's generated hundreds of complaints.
It's quite proper for a consumer-protection agency to be concerned.
With reengineering this could be a beautiful thing to put together entirely hands-off. As it is, the design is only amenable to manual assembly. As you imply, it'd require different approaches to design of various parts to get good yields with automated assembly.
Right. And it might be slightly thicker, which Apple sees as a problem. Look at the teardown, and notice the cable harness with the tiny connectors going into the sides of the subassemblies. That implies a tough assembly problem. Assembly requires getting each part close to its final location, then, while holding the part, attaching the connector. Only then can the part be moved to its final location and attached. This is much harder than bolting down the parts and plugging in the wiring harness. (It's probably one of the places where Foxconn employes are screwing up. Two-handed tweezer work under a magnifier is not easy. Most people can't do it, and almost nobody can do it consistently when tired.)
If the harness were on top and connectors went in vertically, assembly would be easier but the device thickness would increase slightly. Redesigning to fix that problem would delay production. Using the back as a substrate and putting traces on it, then soldering everything to the back in ball-grid array style might work, but that increases fragility because stress on the back cover is applied directly to a soldered joint.
Compare the Droid Razr teardown. That's a design more suited for automated assembly. Put part in place, then attach connector. Uses more adhesives and is less maintainable, though. Those are the tradeoffs.
"GOOG" posted a third-quarter profit of $2.18 billion, or $6.53 a share, down from $2.73 billion, or $8.33 a share, a year earlier. Excluding stock-based compensation and other items, profit fell to $9.03 from $9.72 a share. Revenue, excluding traffic acquisition costs, improved to $11.33 billion.
Analysts surveyed by Thomson Reuters expected earnings of $10.65 a share and net revenue of $11.86 billion.
Total costs jumped 71%.
Not good. Profits are down, and costs are way up. Need to look at the 10-Q filing to find out what the "traffic acquisition costs" are. Earnings excluding stuff are a form of spin control. The real number is the GAAP earnings, which will appear on the 10-Q.
The numbers indicate that Google is buying traffic, and it's not helping profits. That's a sign of trouble. It's not yet clear how bad the trouble is.
Looking ahead, at some point, we're going to hit "peak online advertising", where total spend on online advertising stops growing. At that point, everybody whose business model is "ad-supported free" is in competition with everybody else in that model, fighting over a pie of fixed or declining size.
I wonder how many of these were generated early on and are being hoarded by the early adopters.
Probably most of them. In the early days of Bitcoin, the amount of computation needed to generate a Bitcoin was orders of magnitude less than it is now, and the number of Bitcoins that could be generated per unit time was higher. More than half of the Bitcoins in existence were generated prior to the end of 2009. The system has a huge early-adopter bias built into it.
Bitcoin generation is competitive; the compute load required to generate a Bitcoin is automatically adjusted about once a month based on the number of Bitcoins generated in the last time period. The originator of Bitcoin is still anonymous, and being the first adopter, generating coins with no competition, probably holds many of those cheap-to-generate Bitcoins.
But they can't cash out without crashing the market. The total daily transaction volume in Bitcoins is roughly that of a big supermarket or two. Most of that volume is between traders; actual goods and services sales are tiny.
That's the real problem with Bitcoin. As a currency, it went nowhere. It was supposed to compete with PayPal. Instead, it's mostly a speculative vehicle. By now, one would think that there would be games, music stores, and app stores using Bitcoins, just as a convenience for small transactions. Didn't happen.
Slashdot Mirror
I'd assumed that Apple had been streaming their product launches for years. They weren't?
Then why are not all human officers required to have their eyes replaced with cybernetic implants?
The U.S. Army is already there with the Warfighter Refractive Eye Surgery Program. "To increase combat readiness, the Department of Defense has established the Warfighter Refractive Eye Surgery Program (WRESP) which allows eligible active-duty soldiers to receive laser refractive eye surgery. The goal is to minimize or eliminate the need to wear corrective eyewear."
Far closer to the truth of the matter is that x86 has a much higher design cost than an orthogonal clean-sheet alternative.
True. Years ago I went to a talk where the head of the Pentium Pro design team showed a graph of the number of engineers working on the project. It peaked around 3,000. Nobody had ever had a CPU design team that big before.
The variable length instruction alignment problem of x86, although ugly, isn't a huge consumer of transistors. AMD dealt with it by expanding instructions to fixed length when loaded into cache. Intel dealt with it by sometimes starting ambiguous cases in parallel and discarding the bogus results later. The downside of fixed-length instructions, as in RISC machines, is code bloat - PowerPC code is about twice as big as x86 code, which impacts cache miss rate.
While one instruction per clock RISC CPUs (low-end MIPS and DEC Alpha parts, and the Atmel AVR series are examples) are simple, superscalar machines executing more than one instruction per clock are almost as complex as x86 CPUs. That's why RISC stopped being a win.
Harry Pyle was developing the instruction set for the Datapoint 2200 in his dorm room at Case Tech in Cleveland in the late 1960s. Same building I was in; different floor. That led to the 8008 and the 8080 and the 80286 and the 80386 and ...
Laser cutters for sheet metal are widely used industrially, but need more power than this little one. 150W to 6000W CO2 lasers are used for metal.
Metal cutting is more difficult than cutting wood or plastic. Getting a clean cut is harder. On plastics or wood you can burn your way through slowly, but on metals, you need enough power for fast cutting or you get slag on the edges.
I was just cutting medium density fiberboard on a laser cutter last night. The problem is not the laser beam igniting the cabinet. That's hard to do. The problem is igniting the workpiece, which is easy for many materials, and the cabinet not being able to contain the resulting fire. The cutting process should take place in a nonflammable box with an exhaust to the outside.
Sheet steel is cheap. Spot welding is cheap. This is not rocket science.
This has nothing to do with processing the data from the rover (which comes in at rates a dial-up modem could handle). It's about the web hosting system that lets casual visitors look at the pretty pictures.
NASA could just upload the stills to Flickr and the videos to Youtube and save some money.
It's software for running an upload site. It has shiny graphics, shinier than most other upload systems. The page layouts are squares in a grid, with a plain background. so it looks phone-like and trendy.
As of right now, there is one (1) open MediaGoblin site. There are also two with registration disabled, and two of (non-porno) baby pictures.
The current install setup is too complicated and requires root privileges. If this thing was made as easy to install as Wordpress, it might be more useful.
ASUS discontinued their entire netbook line on September 4, 2012.
Low cost netbooks with large hard drives interfered with the "lock users into the cloud then raise the price and make ads more intrusive" strategy of Google, Facebook, and Microsoft.
Due to a court decision about ten years ago, patent holders now have to sue before they can discuss patent licensing. If you accuse someone of infringement, they can sue to have the patent invalidated, and they have some tactical advantages if they sue first. So patent holders don't send infringement letters any more. They have to start with litigation. That's the reason for the boom in litigation.
It's a huge pain for everyone.
It's an ARM CPU, but with only 96KB of RAM. That's very small for a CPU of that power.
This Science Fiction Museum in Seattle is still open:
It's just a temporary exhibit (pictures) at what's really a rock music museum now. The permanent science fiction museum closed in 2011.
This makes sense for sales reps. Their job is to talk to people and convince them to buy. Maybe take the actual order. That doesn't require much input or local computing power, but it requires convenient access to catalog and customer data. This is a business application that maps well to small screen mobile devices.
The press release is deceptive. They did not build a working turbofan engine with a 3D printer. They built a plastic scale model of a Rolls Royce turbofan engine with a Stratasys 3D printer. It will rotate if powered with compressed air. Rolls Royce gave U of VA a $2 million dollar grant which supported that effort.
The plane itself wasn't printed as one piece. It was more like printing the parts of a plane kit. Very slowly. 80 hour weeks are mentioned. Not sure where the $2000 cost figure comes from, but it doesn't include labor or 3D printer time. Maybe that's just the plastic cost.
The Science Fiction Museum in Seattle seemed a likely place for this to end up, but that, like most Paul Allen projects, went bust. It closed in 2011. There's no really good place to put this.
A device that 1) has a data link to the outside world, 2) has a GPS receiver, and 3) has a microphone ought to be far riskier to steal. Something is not right here.
You should be able to log into your phone account from another device and retrieve the location of the phone, Maybe listen in and record calls from your phone, too.
The "Silent Circle" uses their own "Silent Network", allowing centralized user tracking. Also, the code isn't open source, so you have no idea if the crypto key generation is any good or if there are backdoors.
The problem is microkernels suck in many ways and no one wants them.
The problem is that academic microkernels, Mach in particular, suck. Mach is a BSD derivative. QNX is quite good. You pay some extra copying overhead, maybe 10-20% of CPU time, or a few months of Moore's Law, for message passing. The microkernel itself is about 60K of code. There's hope of getting something that small perfect. QNX, though, is not intended to be a high-security system; its purpose is to be a high reliability real-time system with repeatable response time.
Many real-time and automotive applications use QNX.
One of the great inventors. "Ovonics", amorphous-siliicon solar cells, batteries...
The article is the personal blog of an editor at Scientific American. Not, as we say on Wikipedia, a reliable source.
Linux + SELinux, (SELinux, which was originally built by the NSA for those who don't know enough history to realise) is an operating system with an immutable watchdog. What more do you want?
SELinux wasn't intended to be highly secure. It's an add-on to Linux, after all, not a new OS. The purpose of SELinux was to get a mandatory-security system out and widely used so that applications would be written to run under tight restrictions. Read what NSA originally wrote about it.
A big problem with secure operating systems is getting applications to run in a secure environment. That means saying "no" a lot. No, your game can't find out what else is running. No, Photoshop can't snoop the LAN for other instances of Photoshop with the same serial number. No, you can't run code in a spreadsheet attached to an email. No, you can't have a browser which has pages from multiple sites in the same memory space. That's what it means to have a secure OS.
The hope of SELinux was that applications would gradually be rewritten to run under tight restrictions like that. It didn't happen.
Look how much whining there is whenever Microsoft tightens up Windows. Users will choose ad-supported games that phone home over security.
Right; actually half were generated by the end of 2010 or so. The difficulty level, the amount of computation required to get a Bitcoin, was very low until the end of 2009. It has climbed by 7 orders of magnitude since then. That's why the people who set up Bitcoin have most of the Bitcoins.
Coursera makes their site look like they're a real academic institution. Yes, they say "free", but lots of sites say "free" and lie. It's not clear whether "free" is just a bait to get people to sign up, and then attempts are made to "upgrade" them to a paid account. Or there may be "fees". Their terms say "We reserve the right to change or modify the Terms of Use at our sole discretion at any time. Any change or modification to the Terms of Use will be effective immediately upon posting by us. " So they could add fees at any time.
There's a scam where online schools sign up students, and collect enough information that the school can apply for financial aid from the U.S. Department of Education. "The inspector general's office says participants in 42 different fraud rings have been convicted and more than $7.5 million in restitution and fines have been ordered in the past six years. This may be only a small portion of the problem."
See Belford University, which the BBB says is a scam that's generated hundreds of complaints.
It's quite proper for a consumer-protection agency to be concerned.
With reengineering this could be a beautiful thing to put together entirely hands-off. As it is, the design is only amenable to manual assembly. As you imply, it'd require different approaches to design of various parts to get good yields with automated assembly.
Right. And it might be slightly thicker, which Apple sees as a problem. Look at the teardown, and notice the cable harness with the tiny connectors going into the sides of the subassemblies. That implies a tough assembly problem. Assembly requires getting each part close to its final location, then, while holding the part, attaching the connector. Only then can the part be moved to its final location and attached. This is much harder than bolting down the parts and plugging in the wiring harness. (It's probably one of the places where Foxconn employes are screwing up. Two-handed tweezer work under a magnifier is not easy. Most people can't do it, and almost nobody can do it consistently when tired.)
If the harness were on top and connectors went in vertically, assembly would be easier but the device thickness would increase slightly. Redesigning to fix that problem would delay production. Using the back as a substrate and putting traces on it, then soldering everything to the back in ball-grid array style might work, but that increases fragility because stress on the back cover is applied directly to a soldered joint.
Compare the Droid Razr teardown. That's a design more suited for automated assembly. Put part in place, then attach connector. Uses more adhesives and is less maintainable, though. Those are the tradeoffs.
"GOOG" posted a third-quarter profit of $2.18 billion, or $6.53 a share, down from $2.73 billion, or $8.33 a share, a year earlier. Excluding stock-based compensation and other items, profit fell to $9.03 from $9.72 a share. Revenue, excluding traffic acquisition costs, improved to $11.33 billion.
Analysts surveyed by Thomson Reuters expected earnings of $10.65 a share and net revenue of $11.86 billion.
Total costs jumped 71%.
Not good. Profits are down, and costs are way up. Need to look at the 10-Q filing to find out what the "traffic acquisition costs" are. Earnings excluding stuff are a form of spin control. The real number is the GAAP earnings, which will appear on the 10-Q.
The numbers indicate that Google is buying traffic, and it's not helping profits. That's a sign of trouble. It's not yet clear how bad the trouble is.
Looking ahead, at some point, we're going to hit "peak online advertising", where total spend on online advertising stops growing. At that point, everybody whose business model is "ad-supported free" is in competition with everybody else in that model, fighting over a pie of fixed or declining size.
I wonder how many of these were generated early on and are being hoarded by the early adopters.
Probably most of them. In the early days of Bitcoin, the amount of computation needed to generate a Bitcoin was orders of magnitude less than it is now, and the number of Bitcoins that could be generated per unit time was higher. More than half of the Bitcoins in existence were generated prior to the end of 2009. The system has a huge early-adopter bias built into it.
Bitcoin generation is competitive; the compute load required to generate a Bitcoin is automatically adjusted about once a month based on the number of Bitcoins generated in the last time period. The originator of Bitcoin is still anonymous, and being the first adopter, generating coins with no competition, probably holds many of those cheap-to-generate Bitcoins.
But they can't cash out without crashing the market. The total daily transaction volume in Bitcoins is roughly that of a big supermarket or two. Most of that volume is between traders; actual goods and services sales are tiny.
That's the real problem with Bitcoin. As a currency, it went nowhere. It was supposed to compete with PayPal. Instead, it's mostly a speculative vehicle. By now, one would think that there would be games, music stores, and app stores using Bitcoins, just as a convenience for small transactions. Didn't happen.
(Anyone remember CyberCoin? DigiCash? Beenz? Didn't think so.)