A few months later MySpace signed a 1 Billion dollar Ad deal with Google.
Actually, no. Google signed a
$900 million deal with Fox Media, which has a number of major web sites. IGN, not Myspace, is generating most of the profitable ad displays.
What Google is getting from Myspace is incoming search traffic.
Ads on Myspace have very low value. Most of the advertisers are bottom-feeders (we collect statistics on this). Cost per click is very low.
Facebook is a really good business idea. Technically it's uninteresting, but socially it's brilliant.
It also has more revenue per employee than almost anything else. Facebook, the company, is tiny. For their growth period to a billion-dollar company, they were in a little 3-story building on Litton in Palo Alto, between a yoga studio and a beauty salon. (There
something about those few downtown blocks of Palo Alto. PayPal, Facebook, Alta Vista, and a host of other well-known names all started within a three block area. PayPal started above the bike shop. )
Facebook seems to hire based on Facebook. The women coming out of the building are good looking and the guys are hunks.
I have nothing against nuclear power, I just do not trust deregulation-happy business criminals to run them.
That's a very good comment. We need more nuclear power, and we need downright Draconian inspection and regulation to make it work. This is to the advantage of the industry. Remember, it wasn't Chernoybl that killed the US nuclear industry. It was the meltdown at Three Mile Island, Pennsylvania, in 1979. That was contained, but it was a close thing.
A good start would be for the Obama administration to appoint several nuclear whistleblowers to the Nuclear Regulatory Commission. The US also might have an exchange arrangement with French nuclear regulators (France gets 80% of its electricity from nuclear power) to get an outside look into regulation.
Longer term, we need a less-volatile energy industry. Going back to rate-of-return regulation would be worth it for the stabilty. Deregulation didn't bring end user prices down.
A DARPA "program manager" is often what Government procurement people call a "Contracting Officer's Technical Representative". This is someone who knows what the project is about, technically, and goes out to check on progress. Back at HQ, you write reports, go to meetings concerning what projects ought to go forward, and look at incoming proposals. You get to see a lot, and have some influence over research, but don't really do much yourself. The problem is finding people smart enough to do the job, willing to work for the Government not actually doing technical work, senior enough to tell companies and professors what to do, yet not has-beens.
Although many academics are unhappy with DARPA under Dr. Tony Tether, I think he's done good work. Academic robotics needed a serious butt-kick. DARPA had been putting money into robot vehicles since 1969 without getting anything usable. Tether dreamed up the DARPA Grand Challenge to light a fire under academic researchers. Early on, the big-name schools didn't want to field entries. It was quietly made clear to them that the gravy train was over - if they couldn't compete, they weren't getting further funding in robotics. Entire academic departments were devoted to that problem, and it got results.
More recently, Boston Dynamics' "Big Dog" robot has been demoed. Again, this was something far better than anything from decades of academic work. I can't speak for work outside robotics, but DARPA really has succeeded in forcing robotics groups to produce.
I'm still running Windows 2000 on the last Windows machine. It's so drama-free. No pushed updates, no annoying popups from Microsoft, no crashes in years.
You run Windows 2000. XP is tied to the mothership in Redmond. With Vista, Microsoft runs you.
Google
Google is really an ad agency. Search is just another traffic builder.
Most new hires outside of Mountain View are ad sales execs. Over time, this
will change the culture, as ad execs move up in the organization. Google has never had a second profitable product - AdWords generates all the profits. Despite all the new
"product" rollouts, none of the new stuff makes money.
When I've been over there, I get the feeling of "overfunded dot-com". There's all
this activity, but it's not contributing to their bottom line. The technical work
that does contribute to revenue revolves around ad optimization. The stock peaked
a while back, and it's way overpriced for the revenue, so don't rely on stock options.
Yes, they have free food, but the feeding schedule is designed to make people spend
their whole life at work.
Intel Dilbertland. By design. Tiny grey cubicles out to the horizon. Just visiting Intel HQ is depressing. Yet incredible CPU design work is done there, by huge teams.
Apple The Mac people hate the iPod people. The iPod people hate the iPhone people. If you're in a visible position, Steve Jobs yells at you. Some of that
attitude permeates the culture. Too much excess stress.
The next step is to encourage the makers of UMPCs to ship Wine with their units. Then users can run some of their legacy apps on the sub-$500 machines.
Interestingly, a Belgian company recently built two continuous flow breweries. 300 million litres per year (7000 bbl/day) each. 45 employees total, including the bottling operation. So continuous-flow fermentation from agricultural feedstock is working, doing a more difficult job than biofuel synthesis. This technology will be valuable for all the fuel-from-biomass schemes. Once somebody gets the genetically engineered microbes right, the plumbing and control systems are ready.
The second plant, incidentally, is producing most of the beer for the Olympics in China.
OK. It's another biomass to hydrocarbon conversion by fermentation with genetically engineered bacteria system. The company web site is all hype; it just mentions a "proprietary microbe", the only new part of the process. It's a lot like "cellulostic ethanol".
Vinod Khosla, a well-known venture capitalist, has been funding multiple startups in this space in hopes that someone will make a breakthrough.
There are many known ways to convert biomass to fuel, and most of them are expensive. You can't predict costs from lab-scale work. Until the process is working at pilot plant scale, cost predictions are hype.
In the lab, tests are typically run in batches, in glass containers, starting with fresh input materials. For commercialization of a low-cost product, the process has to work with a continuous flow. Continuous flow fermentation is hard to do; by-products may build up in the system, or contamination in the feedstock may mess up the process. They haven't dealt with those problems yet.
If the process has to be run in batches, like a brewery, with flushing and cleaning at the end of each cycle, the process is more tolerant of difficulties, but the operating cost goes up. It's possible to get the cost of a batch process down; beer production in bulk runs about $65/bbl. But beer is around 95% water, and for fuel applications, you don't get to count water as product.
Khosla has the right approach. He's placing little bets, in the tens of millions of dollars range, on many technologies. His experts check on how they're doing. The ones making progress get another round of funding, and the others don't. One or more of them will be a big win.
The Ford EEC IV engine controller from the 1980s was designed for a 30 year life span.
(My 1985 Bronco still has its original unit.) No software updates, either; the
program was masked onto the custom CPU chip. Never needed a recall. On the other
hand, my 2007 Jeep Wrangler had three separate major software updates in the first year.
Plenty of industrial hardware needs to be able to run for 30-40 years.
Jenkins Valves used to boast of century lifespans. They had pictures of a valve installed around 1900 which had been removed
during a water line replacement. It was still working fine, and after a few months of being photographed and shown off, it was re-installed in a new water line.
The original article is an ad for a service that looks at code for you.
But it's a real problem.
A basic problem with open source is that once you get beyond the top 50 or so projects, the quality is usually crap. Look at the source from a few random projects on SourceForge.
There aren't that many real "community" projects, where multiple programmers are working on the same code. The long tail isn't very good.
Machina Dynamica's Tru-Tone Duplex Cover is a special audiophile-grade cover for all duplex wall outlets; they are intended to replace all types of duplex covers - steel, plastic, wood, etc. - in the listening room -- including non-audio outlets and unused outlets. One or two Tru-Tone Covers can make a significant improvement, however we suggest a baseline of 3-4 Duplex Covers in the room for best results. Tru-Tone Duplex Outlet Covers produce a remarkable degree of focus, detail and presence.
The thing appears to be an absolutely standard plastic cover plate. This isn't about replacing the receptacle, which might be of some value if it was badly worn and causing voltage drops under load. It's just the plastic cover plate.
The only thing actually made by "Westinghouse" is nuclear reactors. The brand name is licensed out by CBS to Westinghouse Digital Electronics LLC, which is a front for Chi Mei Optoelectronics, a subsidiary of Chi Mei Corporation (Taiwan).
Chi Mei is probably the world's leading supplier of large LCD panels.
You've designed air conditioning units and don't get basic thermodymamics?
The short version is this. Heat engines are limited to a maximum efficiency E = (Tout - Tin)/ Tout. E is always < 1. The smaller the temperature difference, the lower the efficiency.
Temperatures are measured from absolute zero.
Refrigerators/heat pumps are heat engines in reverse. They have a maximum coefficient of performance (heat difference out / energy in) which is 1/E of the efficiency of a heat engine operating between the same temperature difference.
Coefficients of performance can be > 1; values like 9 can be achieved with real hardware, but only for small temperature differences.
If you use a refrigerator/heat pump to provide the temperature difference to drive a heat engine, the coefficient of performance of the refrigerator times the efficiency of the heat engine will always be less than 1. Which is why you can't build a perpetual motion machine by running a refrigerator's compressor from a heat engine.
There's nothing mysterious about this. There's a fairly obvious relationship between pressure, volume, and temperature of an ideal gas, and once you get that, everything else more or less follows.
Educational software is hard to write. Really hard. Except for very well defined skills, like typing or flying an aircraft, most educational software doesn't help much.
The OLPC should come with one or two really, really good applications for teaching reading or arithmetic, ones smart enough to self-adjust to the user's level and move them forward. That alone would justify the thing.
So Apple is going back to 64-bit x86. Good. The PowerPC machines were 64-bit, and moving to 32-bit Intel after 64-bit machines were out seemed a step backwards. It's disappointing that Apple didn't skip 32-bit Intel; now there will be a whole era of mixed-width code to get through.
It's not hard. One classic approach for use in schools is to take some political issue which was a big deal in its day but is forgotten now. Obtain material written about the subject from many points of view, some sensible, some totally bogus, and with various degrees of stridency.
Have students read through all the material and then write a brief evaluation of the various positions, listing the arguments, which ones they think are good, which ones seem bogus, and explain how they made that decision.
The Free Silver issue is a good example. Once upon a time, the "free and unlimited coinage of silver" was a big issue. This was an early attempt at an "economic stimulus package" in a hard-money system. There's a famous speech by William Jennings Bryan ("I will not allow
this nation to be crucified upon a cross of gold"), there were political cartoons, and there's plenty of material available. This is for high school level students.
In earlier grades, teach skepticism of advertising. Teach how to read an ad. What are they trying to sell you? What are they telling you? What aren't they telling you? Use old TV commercials from the Internet Archive as teaching tools. Teacher handbook: "Ogilvy on Advertising".
You don't need any extra kernel support for that. You just write the new version under a temporary name, and then atomically rename it over the old file. Fsync before renaming for extra credit.
It's not the default. It's hard to do portably. The ritual for doing it on Windows is quite complex and usually implemented wrong. Rename is only atomic on some UNIX/Linux file systems. One can end up with forgotten temp files lying around. File loss is possible on systems lacking an atomic rename function, although the data is still in a temporary file. Somewhere.
File systems should know more about file type.
Not "file type" in the extension sense, but file type in the sense of what the data written to the file needs for integrity.
There are only a few standard use cases:
The entire file is the commit unit. For most files, you either want the entire file written correctly or you don't want anything written. When nothing is written, the previous version, if any, should remain intact. Applications with "Save" functions need this model. For many binary file types, from images to executables, a partial file is totally useless. So the file should be committed when closed. IBM put this in some of their early UNIX versions, the ones based on UCLA Locus. Done right, if the program crashes before closing or committing the file, the file reverts to the old version. It's not necessary to update the metadata until file close, so this is the fastest mode, and should be the default.
Log files. Files are only extended; old data is not overwritten. Ordered journaling is desirable, so that after a crash, the file is intact out to the point of the crash. This is UNIX "open for append" mode.
Database files. The file is being used to support a database with read/write data structures. The database system needs to know when the file system has committed a write and may need to know about ordering of queued writes. This is the most complex case, but database implementors pay attention to file system details and are willing to make special calls if necessary to tell the file system when to commit and to wait for commits to complete.
If those three cases are properly supported, you should never see a garbled file from an unexpected shutdown. Some of the file systems out there have approximately the right
feature set for this, but there's no standardized interface and set of expectations that
corresponds to these use cases.
Except the current study doesn't address airliners. The current study is for vehicles less that 40ft. long.
No, the 40' limitation is for the section on boats, not aircraft. That's a different project in the same solicitation. It's a response to the USS Cole incident. The U.S. Navy would like to have the capability to do something about unexpected approaching small boats, something less lethal than machine-gunning them.
There's a way to do this right. Read this article about the F-16 GCAS. This is a ground collision avoidance system that works so well it can be used on combat missions, including flying through mountain passes at 500 knots, 200 feet from rock. Pilots call it "You can't fly any lower". When the Auto-GCAS decides a ground collision is imminent, it takes over the aircraft, rolls to wings-level and initiates a pull-up. (In an F-16, the roll is at 180 degrees per second and the pull-up is at 5G; for an airliner, much lower numbers would be configured and recovery would be initiated much sooner.) Read the article; fighter jocks liked the thing, and those guys usually hate letting the automation take over the aircraft.
This would prevent most "controlled flight into terrain" accidents (there are about three of those involving commercial jets per year, worldwide), so there's a big win in having this independent of military/terrorism worries. Once you have a system like this, it can be given "no-fly" areas into which it won't let the plane go. If you're going to enforce "no-fly" zones via hardware, it's better to do it through a system that knows about terrain and is looking at it with radar.
The way to do overrides would be to give the pilot a switch to turn off the system in an emergency, but doing so sends out an emergency transponder signal that this has been done. The ground then has the option of sending up a suitably encrypted signal to turn it back on. This gives a way to handle system failures. If the ground sees a plane heading somewhere it shouldn't be, the ground can force the system back on.
I wouldn't be at all surprised if Airbus starts offering something like this. (Airbus takes the position that the aircraft should protect itself against pilot errors. Boeing has the philosophy that the pilot should always be able to override the automation. The Boeing approach worked better back when the typical airline pilot had 10,000 hours, a previous military flying career, and was chosen competitively from a big pool of applicants.)
First, it's a Roland the Plogger story, so it's going to be wrong.
GPS accuracy is a serious problem for users who need high precision. More applications are assuming that GPS is precise to a few meters, which, often, it isn't. It's always good enough if you just need to find an airport. Below that level, error can be a problem.
Local high-precision systems, like GPS-based systems for landing, use a pseudolite, a receiver on the ground in a known location that receives GPS and broadcasts small corrections. The pseudolite is usually located near the end of the active runway, so as aircraft get closer to the runway, the error approaches zero. There's a similar setup for
"precision farming", where the tractor precision is precisely known but there's a psuedolite at the side of the field.
Without a pseudolite, it's harder.
Part of the problem is that there aren't enough satellites. To get a GPS lat/long fix, you need to see at least three sats. To get lat/log/elevation, you need to see four. For high-precision work (down to 15cm), you need five, plus correction signals from receiving stations (see Omnistar) that are monitoring propagation. You're lucky to see four in a built-up area, because you can only see part of the sky. If you can see five, you can measure error. Some systems use both GPS and GLONASS sats; now that Russia is building up the GLONASS constellation again, this works better. By 2009, the GLONASS constellation should be fully populated, and systems that use both GPS and GLONASS will have a better chance of seeing five sats.
Propagation problems always add delay; they never subtract from it. Propagation problems come from what the ionosphere is doing, and from reflections from big metal surfaces like buildings.
In urban canyons, you're seeing mostly bounces.
This is an issue for civilian uses that assume the system has more precision than it really does. Car navigation systems that try to tell whether a car is on a freeway or an adjacent side street from GPS data alone are likely to have problems. The same problem applies to GPS systems for railroad signalling (these make me nervous) which try to tell on which track a train is running.
Slashdot Mirror
A few months later MySpace signed a 1 Billion dollar Ad deal with Google.
Actually, no. Google signed a $900 million deal with Fox Media, which has a number of major web sites. IGN, not Myspace, is generating most of the profitable ad displays. What Google is getting from Myspace is incoming search traffic.
Ads on Myspace have very low value. Most of the advertisers are bottom-feeders (we collect statistics on this). Cost per click is very low.
You realize that the "billion-dollar" number is a valuation, it doesn't have anything to-do with their revenue.
Excellent point. They really don't have much of a revenue model, which is very Dot Com 1.0.
Facebook is a really good business idea. Technically it's uninteresting, but socially it's brilliant.
It also has more revenue per employee than almost anything else. Facebook, the company, is tiny. For their growth period to a billion-dollar company, they were in a little 3-story building on Litton in Palo Alto, between a yoga studio and a beauty salon. (There something about those few downtown blocks of Palo Alto. PayPal, Facebook, Alta Vista, and a host of other well-known names all started within a three block area. PayPal started above the bike shop. )
Facebook seems to hire based on Facebook. The women coming out of the building are good looking and the guys are hunks.
I have nothing against nuclear power, I just do not trust deregulation-happy business criminals to run them.
That's a very good comment. We need more nuclear power, and we need downright Draconian inspection and regulation to make it work. This is to the advantage of the industry. Remember, it wasn't Chernoybl that killed the US nuclear industry. It was the meltdown at Three Mile Island, Pennsylvania, in 1979. That was contained, but it was a close thing.
A good start would be for the Obama administration to appoint several nuclear whistleblowers to the Nuclear Regulatory Commission. The US also might have an exchange arrangement with French nuclear regulators (France gets 80% of its electricity from nuclear power) to get an outside look into regulation.
Longer term, we need a less-volatile energy industry. Going back to rate-of-return regulation would be worth it for the stabilty. Deregulation didn't bring end user prices down.
A DARPA "program manager" is often what Government procurement people call a "Contracting Officer's Technical Representative". This is someone who knows what the project is about, technically, and goes out to check on progress. Back at HQ, you write reports, go to meetings concerning what projects ought to go forward, and look at incoming proposals. You get to see a lot, and have some influence over research, but don't really do much yourself. The problem is finding people smart enough to do the job, willing to work for the Government not actually doing technical work, senior enough to tell companies and professors what to do, yet not has-beens.
Although many academics are unhappy with DARPA under Dr. Tony Tether, I think he's done good work. Academic robotics needed a serious butt-kick. DARPA had been putting money into robot vehicles since 1969 without getting anything usable. Tether dreamed up the DARPA Grand Challenge to light a fire under academic researchers. Early on, the big-name schools didn't want to field entries. It was quietly made clear to them that the gravy train was over - if they couldn't compete, they weren't getting further funding in robotics. Entire academic departments were devoted to that problem, and it got results. More recently, Boston Dynamics' "Big Dog" robot has been demoed. Again, this was something far better than anything from decades of academic work. I can't speak for work outside robotics, but DARPA really has succeeded in forcing robotics groups to produce.
I'm still running Windows 2000 on the last Windows machine. It's so drama-free. No pushed updates, no annoying popups from Microsoft, no crashes in years.
You run Windows 2000. XP is tied to the mothership in Redmond. With Vista, Microsoft runs you.
What is all this computing infrastructure doing that's useful? Other than advertising delivery?
The next step is to encourage the makers of UMPCs to ship Wine with their units. Then users can run some of their legacy apps on the sub-$500 machines.
Interestingly, a Belgian company recently built two continuous flow breweries. 300 million litres per year (7000 bbl/day) each. 45 employees total, including the bottling operation. So continuous-flow fermentation from agricultural feedstock is working, doing a more difficult job than biofuel synthesis. This technology will be valuable for all the fuel-from-biomass schemes. Once somebody gets the genetically engineered microbes right, the plumbing and control systems are ready.
The second plant, incidentally, is producing most of the beer for the Olympics in China.
OK. It's another biomass to hydrocarbon conversion by fermentation with genetically engineered bacteria system. The company web site is all hype; it just mentions a "proprietary microbe", the only new part of the process. It's a lot like "cellulostic ethanol".
Vinod Khosla, a well-known venture capitalist, has been funding multiple startups in this space in hopes that someone will make a breakthrough.
There are many known ways to convert biomass to fuel, and most of them are expensive. You can't predict costs from lab-scale work. Until the process is working at pilot plant scale, cost predictions are hype.
In the lab, tests are typically run in batches, in glass containers, starting with fresh input materials. For commercialization of a low-cost product, the process has to work with a continuous flow. Continuous flow fermentation is hard to do; by-products may build up in the system, or contamination in the feedstock may mess up the process. They haven't dealt with those problems yet.
If the process has to be run in batches, like a brewery, with flushing and cleaning at the end of each cycle, the process is more tolerant of difficulties, but the operating cost goes up. It's possible to get the cost of a batch process down; beer production in bulk runs about $65/bbl. But beer is around 95% water, and for fuel applications, you don't get to count water as product.
Khosla has the right approach. He's placing little bets, in the tens of millions of dollars range, on many technologies. His experts check on how they're doing. The ones making progress get another round of funding, and the others don't. One or more of them will be a big win.
The Ford EEC IV engine controller from the 1980s was designed for a 30 year life span. (My 1985 Bronco still has its original unit.) No software updates, either; the program was masked onto the custom CPU chip. Never needed a recall. On the other hand, my 2007 Jeep Wrangler had three separate major software updates in the first year.
Plenty of industrial hardware needs to be able to run for 30-40 years.
Jenkins Valves used to boast of century lifespans. They had pictures of a valve installed around 1900 which had been removed during a water line replacement. It was still working fine, and after a few months of being photographed and shown off, it was re-installed in a new water line.
The original article is an ad for a service that looks at code for you. But it's a real problem.
A basic problem with open source is that once you get beyond the top 50 or so projects, the quality is usually crap. Look at the source from a few random projects on SourceForge. There aren't that many real "community" projects, where multiple programmers are working on the same code. The long tail isn't very good.
Machina Dynamica's Tru-Tone Duplex Cover. Yes, an audiophile-grade wall outlet cover plate. $30.
Machina Dynamica's Tru-Tone Duplex Cover is a special audiophile-grade cover for all duplex wall outlets; they are intended to replace all types of duplex covers - steel, plastic, wood, etc. - in the listening room -- including non-audio outlets and unused outlets. One or two Tru-Tone Covers can make a significant improvement, however we suggest a baseline of 3-4 Duplex Covers in the room for best results. Tru-Tone Duplex Outlet Covers produce a remarkable degree of focus, detail and presence.
The thing appears to be an absolutely standard plastic cover plate. This isn't about replacing the receptacle, which might be of some value if it was badly worn and causing voltage drops under load. It's just the plastic cover plate.
The only thing actually made by "Westinghouse" is nuclear reactors. The brand name is licensed out by CBS to Westinghouse Digital Electronics LLC, which is a front for Chi Mei Optoelectronics, a subsidiary of Chi Mei Corporation (Taiwan).
Chi Mei is probably the world's leading supplier of large LCD panels.
You've designed air conditioning units and don't get basic thermodymamics?
The short version is this. Heat engines are limited to a maximum efficiency E = (Tout - Tin)/ Tout. E is always < 1. The smaller the temperature difference, the lower the efficiency. Temperatures are measured from absolute zero.
Refrigerators/heat pumps are heat engines in reverse. They have a maximum coefficient of performance (heat difference out / energy in) which is 1/E of the efficiency of a heat engine operating between the same temperature difference. Coefficients of performance can be > 1; values like 9 can be achieved with real hardware, but only for small temperature differences.
If you use a refrigerator/heat pump to provide the temperature difference to drive a heat engine, the coefficient of performance of the refrigerator times the efficiency of the heat engine will always be less than 1. Which is why you can't build a perpetual motion machine by running a refrigerator's compressor from a heat engine.
There's nothing mysterious about this. There's a fairly obvious relationship between pressure, volume, and temperature of an ideal gas, and once you get that, everything else more or less follows.
Educational software is hard to write. Really hard. Except for very well defined skills, like typing or flying an aircraft, most educational software doesn't help much.
The OLPC should come with one or two really, really good applications for teaching reading or arithmetic, ones smart enough to self-adjust to the user's level and move them forward. That alone would justify the thing.
So Apple is going back to 64-bit x86. Good. The PowerPC machines were 64-bit, and moving to 32-bit Intel after 64-bit machines were out seemed a step backwards. It's disappointing that Apple didn't skip 32-bit Intel; now there will be a whole era of mixed-width code to get through.
It's not hard. One classic approach for use in schools is to take some political issue which was a big deal in its day but is forgotten now. Obtain material written about the subject from many points of view, some sensible, some totally bogus, and with various degrees of stridency. Have students read through all the material and then write a brief evaluation of the various positions, listing the arguments, which ones they think are good, which ones seem bogus, and explain how they made that decision.
The Free Silver issue is a good example. Once upon a time, the "free and unlimited coinage of silver" was a big issue. This was an early attempt at an "economic stimulus package" in a hard-money system. There's a famous speech by William Jennings Bryan ("I will not allow this nation to be crucified upon a cross of gold"), there were political cartoons, and there's plenty of material available. This is for high school level students.
In earlier grades, teach skepticism of advertising. Teach how to read an ad. What are they trying to sell you? What are they telling you? What aren't they telling you? Use old TV commercials from the Internet Archive as teaching tools. Teacher handbook: "Ogilvy on Advertising".
You don't need any extra kernel support for that. You just write the new version under a temporary name, and then atomically rename it over the old file. Fsync before renaming for extra credit.
It's not the default. It's hard to do portably. The ritual for doing it on Windows is quite complex and usually implemented wrong. Rename is only atomic on some UNIX/Linux file systems. One can end up with forgotten temp files lying around. File loss is possible on systems lacking an atomic rename function, although the data is still in a temporary file. Somewhere.
File systems should know more about file type. Not "file type" in the extension sense, but file type in the sense of what the data written to the file needs for integrity.
There are only a few standard use cases:
If those three cases are properly supported, you should never see a garbled file from an unexpected shutdown. Some of the file systems out there have approximately the right feature set for this, but there's no standardized interface and set of expectations that corresponds to these use cases.
Except the current study doesn't address airliners. The current study is for vehicles less that 40ft. long.
No, the 40' limitation is for the section on boats, not aircraft. That's a different project in the same solicitation. It's a response to the USS Cole incident. The U.S. Navy would like to have the capability to do something about unexpected approaching small boats, something less lethal than machine-gunning them.
There's a way to do this right. Read this article about the F-16 GCAS. This is a ground collision avoidance system that works so well it can be used on combat missions, including flying through mountain passes at 500 knots, 200 feet from rock. Pilots call it "You can't fly any lower". When the Auto-GCAS decides a ground collision is imminent, it takes over the aircraft, rolls to wings-level and initiates a pull-up. (In an F-16, the roll is at 180 degrees per second and the pull-up is at 5G; for an airliner, much lower numbers would be configured and recovery would be initiated much sooner.) Read the article; fighter jocks liked the thing, and those guys usually hate letting the automation take over the aircraft.
This would prevent most "controlled flight into terrain" accidents (there are about three of those involving commercial jets per year, worldwide), so there's a big win in having this independent of military/terrorism worries. Once you have a system like this, it can be given "no-fly" areas into which it won't let the plane go. If you're going to enforce "no-fly" zones via hardware, it's better to do it through a system that knows about terrain and is looking at it with radar.
The way to do overrides would be to give the pilot a switch to turn off the system in an emergency, but doing so sends out an emergency transponder signal that this has been done. The ground then has the option of sending up a suitably encrypted signal to turn it back on. This gives a way to handle system failures. If the ground sees a plane heading somewhere it shouldn't be, the ground can force the system back on.
I wouldn't be at all surprised if Airbus starts offering something like this. (Airbus takes the position that the aircraft should protect itself against pilot errors. Boeing has the philosophy that the pilot should always be able to override the automation. The Boeing approach worked better back when the typical airline pilot had 10,000 hours, a previous military flying career, and was chosen competitively from a big pool of applicants.)
First, it's a Roland the Plogger story, so it's going to be wrong.
GPS accuracy is a serious problem for users who need high precision. More applications are assuming that GPS is precise to a few meters, which, often, it isn't. It's always good enough if you just need to find an airport. Below that level, error can be a problem.
Local high-precision systems, like GPS-based systems for landing, use a pseudolite, a receiver on the ground in a known location that receives GPS and broadcasts small corrections. The pseudolite is usually located near the end of the active runway, so as aircraft get closer to the runway, the error approaches zero. There's a similar setup for "precision farming", where the tractor precision is precisely known but there's a psuedolite at the side of the field.
Without a pseudolite, it's harder. Part of the problem is that there aren't enough satellites. To get a GPS lat/long fix, you need to see at least three sats. To get lat/log/elevation, you need to see four. For high-precision work (down to 15cm), you need five, plus correction signals from receiving stations (see Omnistar) that are monitoring propagation. You're lucky to see four in a built-up area, because you can only see part of the sky. If you can see five, you can measure error. Some systems use both GPS and GLONASS sats; now that Russia is building up the GLONASS constellation again, this works better. By 2009, the GLONASS constellation should be fully populated, and systems that use both GPS and GLONASS will have a better chance of seeing five sats.
Propagation problems always add delay; they never subtract from it. Propagation problems come from what the ionosphere is doing, and from reflections from big metal surfaces like buildings. In urban canyons, you're seeing mostly bounces.
This is an issue for civilian uses that assume the system has more precision than it really does. Car navigation systems that try to tell whether a car is on a freeway or an adjacent side street from GPS data alone are likely to have problems. The same problem applies to GPS systems for railroad signalling (these make me nervous) which try to tell on which track a train is running.
The link provided leads to an all-Flash page. Suspicious.