The airbag control unit has two "slots" in EEPROM for stored events, the "Deployment Event" and "Near Deployment Event" slot. The "Deployment Event" slot stores the last five seconds of data when the control unit fires the airbag. This is a one-time event - once this has happened, the airbag control unit cannot be used again. (It's replaced with the airbag, if the car is repairable.) "Near Deployment Events" represent situations where the airbag unit started the "fire the air bag decision" process, but decided not to fire the bag. Two successive accelerometer samples of 2G or greater wake up the air bag control algorithm. The biggest delta-V near-deployment event is stored; a bigger one replaces the old one. After 250 engine starts (at least in GM vehicles) the "near deployment event" is erased.
There's local power storage in the airbag unit, so that even if battery power is lost, the airbag can still fire. So the data usually gets stored, too.
The real purpose of this unit is to fine-tune the "fire the air bag decision" algorithm. Early airbags were going off in accident situations that didn't really require airbag deployment. The current generation is doing better. The NTSB collects this data. This found at least one defect. A few false deployments had occured on gravel roads when a big rock happened to hit the
sensing unit. That's been fixed in current models.
This is one of the dumber ideas to come along. First, it's "security by obscurity". If it becomes at all popular, it will be added to lists of standard attacks. Second, it's subject to playback attacks. Third, it generally involves a "key" that belongs to the software, not an individual. Fourth, listening in promiscuous mode eats substantial resources on a busy network. Fifth, the "port knocking" server runs at a high enough privilege level that it itself may introduce security holes.
Atriks is a mailing list company. "Atriks offers targeted public record data that comes entirely from publicly available Internet sources. We collect, compile, aggregate and provide the most high-quality, complete, and up-to-date data possible for every individual and business with a presence on the Internet." They're a member of the Direct Marketing Association. They claim a server farm with 330 servers and seven terabytes of data. Here are some of the lists they offer:
Atriks Broadband Consumers "1,000,000+ consumers who have demonstrated a thirst for better technology and a willingness to spend money for enhanced products and services are included."
Atriks Personal Domain Owners with Credit Cards "7,000,000+ consumers have registered a domain for their own personal use and have created Web sites that share everything from jokes to family pictures. A key part of their registration is supplying credit card information, resulting in a file with all major credit card selects available."
Atriks Subscribers by ISP "6,700,000+ subscribers identified by ISP are included in this database. Mailers can target these subscribers by more than 100 selectable ISP providers."
Those are just the "consumer" lists. They also have business-to-business lists.
Atriks is co-located with a local ISP, MV Communications.. MV has been in business for many years. They have modest backbone connections for an ISP; 6Mb/s to Global Crossing, 12Mb/s to Level 3, and 12Mb/s to Paetec.
Unclear at this time if MV and Atriks have common ownership.
They're what the DMA would call a "legitimate spammer".
Supercomputer CPUs are dead. It's been a long time since Cray made CPUs. Current-generation supercomputers are composed of large numbers of commercial microprocessors.
The only major exception is the current fastest supercomputer, the Earth Simulator in Japan. It uses custom vector processors. They seem vaguely similar in architecture to Playstation 2 vector units, but they probably are unrelated.
Every other machine in the top 10 is built from standard processors. The old DEC Alpha, PowerPCs, and IA-32 predominate, with a few Itanium machines.
Because supercomputers today have several thousand processors, they can't even be big shared-memory multiprocessors. Speed of light lag in the interconnects would slow everything down. It just takes too long for the signals to make it across the room.
So all supercomputers today are clusters of one kind or another, fast machines with slower interconnects between them. The hardware architecture revolves around interconnect schemes.
The software architecture revolves around working around the limitations of the interconnect schemes.
Tightly coupled problems don't map well to such machines.
Bear in mind that we're talking about clusters of uniform machines located near each other with gigabit or better interconnects. We're not talking about "clusters" consisting of spare-time programs out at the end of Internet connections. Those are useful only for problems with almost no coupling between parts. Such problems are usually low hit rate search problems, like cryptanalysis, SETI@HOME, and such.
Yes, there's the Cray X1, the last of the liquid-cooled monsters, but it looks like the only customers who bought one were Government agencies with old Cray machines.
Why would one want to do this, except as a demo? Either OS can now crash the machine, so the MTBF gets worse. You get to pay both Microsoft and Red Hat. And few people run Linux because they like the desktop applications.
This sounds like one of those "I'm l33t" toys.
The ability to run Windows apps on Linux is far more useful.
Pacific Shores Center This huge, strikingly beautiful bayfront office park, built at the end of the dot-com boom, stands complete but empty. Great place to film an interview.
The trailer park next to Moffett Field Facing the intake of the huge wind tunnel at NASA Ames is a trailer park. Take 101 to Shoreline in Mountain View, turn east, go about three blocks, turn right opposite the movie theaters, and drive to the end of the street. The trailer park is right in front of the 100-foot high air intake.
The abandoned FMC manufacturing complex in San Jose. The Bradley Fighting Vehicle was built there. There's a test track for the things.
Downtown San Jose at rush hour Little traffic, plenty of free parking, half the stores are closed. It's like Sunday, every day.
It's not hard to do. It costs $335. And you can do it online in about fifteen minutes.
If you register a trademark and it's rejected for the "principal register" as not being unique enough, you may be able to get registration on the "supplemental register". If your trademark is on the supplemental register, you can't enforce it against others, but you're protected against claims by others. So it's worth doing in cases like the one here.
There's a general misconception about NSA Secure Linux. It has a tough security model, but it's not developed to high security standards. The whole point of NSA Secure Linux is to find out if useful applications can be built on an OS with a mandatory security model. NSA has had tough OSs built for them before (I worked on one), but they were so restrictive that very few applications were developed for them.
What developers can do to help is to modify a web server, a mail server, and a DNS server (the most attacked server side software) to run under NSA Secure Linux, partitioned into levels of integrity.
The idea is that just because somebody attacks, say, the mail server receive program, that doesn't get them power over the whole system. All it should do is let them run their attack code in a jail where it can't do anything except burn CPU cycles, and maybe generate phony incoming mail. Mail associated with the known IP address from which that copy of the receiver was launched, so the problem can be tracked.. When they disconnect, or some monitor program kills the corrupted component off, all the damage should be flushed.
You need to rework key server apps so that about 95% of the code is untrusted and jailed, while the 5% that has to do security-related functions is isolated, identified, and carefully examined.
OK, you're doing a desktop. Mostly GUI elements, no hard real time requirements, lots of pointers, many developers.
Java seems appropriate here, if you can get the performance. It's a memory-safe language, and you don't have to obsess on memory management correctness.
Garbage collection is acceptable. There's a big pool of Java developers. There's a hard-code open source compiler. Microsoft doesn't control the language or the environment.
Whether the rather clunky Java libraries add negative value is something you have to think about, hard. The language itself is OK.
Back then, people didn't have race condition detectors for VLSI designs.. The 6502 was a 1970s design. CAD barely existed. ICs were laid out using Kodalith and X-acto knives. There's been progress since then.
The color is fine. Brightness is the problem
on
The Blues for LEDs
·
· Score: 4, Informative
On the annoyance front, there was a fad about a year ago for really bright power-on indicators. I have a Shuttle PC and a DVD player that will light up a dark room with their blue power-on LEDs. That's excessive.
The color, though, is correct. The standard NEMA rules for indicator colors, used on industrial gear for decades, are
GREEN Normal status.
AMBER Abnormal status. Operator should be aware that an amber lamp is lit.
RED Trouble status. Operator should take action to make the red light go out.
BLUE or WHITE On, or other non-specific meaning.
So blue and white are actually the default colors.
Red should be used only for trouble indicators. We're still getting over the cheap red LED glut of the 1980s, when everything had red LEDs.
Anything that rackmounts should follow these rules. It's not only annoying, but a headache, to have red lights for non-trouble conditions in a rack of equipment. IBM always has.
Actually, this is somewhat misleading. The top players listed are Microsoft (Windows CE), Wind River, Symbian, Palm. QNX, OSE, and Green Hills.
But Microsoft, Symbian, and Palm are really selling into handheld devices, not hard real-time control. (The phone and PDA markets are much bigger than real time control, though.) Wind River's VRTX is the dominant player by a big margin, especially in low-end embedded control. QNX is next, and is usable on a broad range of platforms. Wind River is more of a specialist maker catering to the military Ada market.
Following these seven come LinuxWorks and MonteVista, who are moving up. These are the main Linux-based offerings.
Also confusing the issue is Windows XP Embedded, which is basically a Windows XP from which you can delete stuff you don't need. This sells more into point-of-sale applications than hard real time control.
Reverse osmosis plants work just fine, but the cost is still around $1 per cubic meter. This is about 264 gallons. Current prices for water for Boston suburbs are around $1 per 100 cubic feet, or $0.35 per cubic meter. So desalinization is about 3x the current Boston-area price of water.
In Los Angeles, which is in a desert, water prices are over twice that of Boston. So desalinization starts to look good.
Power consumption is a problem, but not a huge one. Water must be pressurized to about 1000PSI to get it through the membrane, so you can work out the energy cost from that. (Hint: it's the same as the energy cost to lift the water to the top of a column with 1000PSI at the bottom, which is about 2000 feet.)
Reverse osmosis membranes now last for five years, so maintenance is less of a headache than it used to be.
There's no particular advantage in producing "low quality" water via reverse osmosis. You have to get most of the solids out to prevent crudding up the membranes, and the membrane stops almost everything bigger than a water molecule.
Reverse osmosis plants also generate brine, which is not too much of a problem if you're alongside an ocean, but can be a huge headache if you're drawing water from a brackish well and want to purify it.
it's theoretically impossible to write a general test to find all race conditions in code.
Baloney. It is possible to write programs for which race conditions are undecideable. Such programs are broken. It is possible to write programs for which race condition detection is NP-hard. Such programs are broken if N is large. It is also possible to write programs for which race conditions can be proven to be absent. That's what you want to do.
Actually, it's straightforward to design software to be free of race conditions on a single machine. You then have a deadlock avoidance problem, but deadlocks are easily detected when they occur.
Hardware is routinely designed to be free of race conditions, after all.
Actually, Java 3D (which never really worked) was too low-level in some respects. Not enough for a game engine, too much for a graphics API.
The collision detection primitives, to give an example I know something about, were badly chosen. There are ways to do collision detection very efficiently, but not using Java 3D's collision primitives.
Aristotle thought that what differentiated humans from the animals was that humans could do arithmetic. Now, we know that, fundamentally, arithmetic isn't hard. It doesn't take that many gates to make an ALU, and that's totally understood.
Vision, on the other hand, is very tough.
Chess is beginning to look like that. It yields to brute force. And by modern computational standards, not very much brute force. "Deep Fritz" tied 2:2 with Kasparov running on a desktop 4-processor IA-32 machine. Kasparov says it plays better than Deep Blue, which was a roomful of custom IBM hardware. If you haven't been on the cover of "Chess Life", "Deep Fritz", running on a standard PC, can trounce you. And it's only US$110.23.
I can't imagine this is a new idea. I would have expected this to have been tried around 1955 or so, around the time reliable traffic detectors were developed.
It only generates 150KW. That's not much. Typical wind turbines generate 200KW to 700KW each, on windy days. (Average values are much lower.)
Typical nuclear power plants generate 1,000,000KW.
Powering a home takes about 1-2KW on average, so 10,000 homes require perhaps 15,000KW.
The SNAIL people want to move up to the 750KW range or so. That's more reasonable. As wind power people have discovered, having huge numbers of little turbines isn't cost effective. But
somewhere around a few hundred KW per turbine, the economics start to work. If you can find a good site with steady wind. As with dams, there aren't that many good sites.
It will probably take several decades of operating experience to turn this into a reliable technology, just as it did with windpower. It's been half a century since the Grandpa's Knob loss of blade accident. The first big power-generating wind turbine oversped and threw a blade several hundred feet. For many years, nobody built one that big again. Gradually, the aerodynamics and control problems were figured out. It's taken that long to make large wind turbines work reliably and profitably.
Anything with moving parts in the ocean is likely to be high-maintenance. Making one of these things work reliably for decades will be tough. Maintenance will be costly. There's no guarantee of success.
In short, there's no breakthrough here until it's been running for a few years without breaking.
Has anyone figured out the ClearPlay "filter lists" yet? The Internet player for PCs downloads them via the Internet. What does the standalone player do?
Can you create your own filter lists? For example, could you express the "Star Wars Phantom Edit" (the one that deletes that Jar Jar characte) as a ClearPlay filter list? This has potential.
Why aren't we hearing words like "knowingly, willfully negligent", or "reckless endangerment", or "conspiracy to violate the Computer Fraud and Abuse Act"?
Slashdot Mirror
The airbag control unit has two "slots" in EEPROM for stored events, the "Deployment Event" and "Near Deployment Event" slot. The "Deployment Event" slot stores the last five seconds of data when the control unit fires the airbag. This is a one-time event - once this has happened, the airbag control unit cannot be used again. (It's replaced with the airbag, if the car is repairable.) "Near Deployment Events" represent situations where the airbag unit started the "fire the air bag decision" process, but decided not to fire the bag. Two successive accelerometer samples of 2G or greater wake up the air bag control algorithm. The biggest delta-V near-deployment event is stored; a bigger one replaces the old one. After 250 engine starts (at least in GM vehicles) the "near deployment event" is erased.
There's local power storage in the airbag unit, so that even if battery power is lost, the airbag can still fire. So the data usually gets stored, too.
The real purpose of this unit is to fine-tune the "fire the air bag decision" algorithm. Early airbags were going off in accident situations that didn't really require airbag deployment. The current generation is doing better. The NTSB collects this data. This found at least one defect. A few false deployments had occured on gravel roads when a big rock happened to hit the sensing unit. That's been fixed in current models.
Next idea.
If you want to cool by immersion, get Fluorinert, with a boiling point well above 100C.
Whois:
55 Bridge Street
Manchester, NH 03101-1188
Phone- 603-624-7008
Fax- 603-624-9089
hostmaster@atriks.com
Atriks is a mailing list company. "Atriks offers targeted public record data that comes entirely from publicly available Internet sources. We collect, compile, aggregate and provide the most high-quality, complete, and up-to-date data possible for every individual and business with a presence on the Internet." They're a member of the Direct Marketing Association. They claim a server farm with 330 servers and seven terabytes of data. Here are some of the lists they offer:
-
Atriks Broadband Consumers "1,000,000+ consumers who have demonstrated a thirst for better technology and a willingness to spend money for enhanced products and services are included."
-
Atriks Personal Domain Owners with Credit Cards "7,000,000+ consumers have registered a domain for their own personal use and have created Web sites that share everything from jokes to family pictures. A key part of their registration is supplying credit card information, resulting in a file with all major credit card selects available."
-
Atriks Subscribers by ISP "6,700,000+ subscribers identified by ISP are included in this database. Mailers can target these subscribers by more than 100 selectable ISP providers."
Those are just the "consumer" lists. They also have business-to-business lists.Atriks is co-located with a local ISP, MV Communications.. MV has been in business for many years. They have modest backbone connections for an ISP; 6Mb/s to Global Crossing, 12Mb/s to Level 3, and 12Mb/s to Paetec. Unclear at this time if MV and Atriks have common ownership.
They're what the DMA would call a "legitimate spammer".
I see a day coming when, in one day, half the computers in the US have their disks erased.
Every other machine in the top 10 is built from standard processors. The old DEC Alpha, PowerPCs, and IA-32 predominate, with a few Itanium machines.
Because supercomputers today have several thousand processors, they can't even be big shared-memory multiprocessors. Speed of light lag in the interconnects would slow everything down. It just takes too long for the signals to make it across the room.
So all supercomputers today are clusters of one kind or another, fast machines with slower interconnects between them. The hardware architecture revolves around interconnect schemes. The software architecture revolves around working around the limitations of the interconnect schemes. Tightly coupled problems don't map well to such machines.
Bear in mind that we're talking about clusters of uniform machines located near each other with gigabit or better interconnects. We're not talking about "clusters" consisting of spare-time programs out at the end of Internet connections. Those are useful only for problems with almost no coupling between parts. Such problems are usually low hit rate search problems, like cryptanalysis, SETI@HOME, and such.
Yes, there's the Cray X1, the last of the liquid-cooled monsters, but it looks like the only customers who bought one were Government agencies with old Cray machines.
You need to send them a DMCA takedown notice. That will get their attention.
Either OS can now crash the machine, so the MTBF gets worse. You get to pay both Microsoft and Red Hat. And few people run Linux because they like the desktop applications.
This sounds like one of those "I'm l33t" toys.
The ability to run Windows apps on Linux is far more useful.
If you register a trademark and it's rejected for the "principal register" as not being unique enough, you may be able to get registration on the "supplemental register". If your trademark is on the supplemental register, you can't enforce it against others, but you're protected against claims by others. So it's worth doing in cases like the one here.
What developers can do to help is to modify a web server, a mail server, and a DNS server (the most attacked server side software) to run under NSA Secure Linux, partitioned into levels of integrity.
The idea is that just because somebody attacks, say, the mail server receive program, that doesn't get them power over the whole system. All it should do is let them run their attack code in a jail where it can't do anything except burn CPU cycles, and maybe generate phony incoming mail. Mail associated with the known IP address from which that copy of the receiver was launched, so the problem can be tracked.. When they disconnect, or some monitor program kills the corrupted component off, all the damage should be flushed.
You need to rework key server apps so that about 95% of the code is untrusted and jailed, while the 5% that has to do security-related functions is isolated, identified, and carefully examined.
That's what NSA Secure Linux is for.
Java seems appropriate here, if you can get the performance. It's a memory-safe language, and you don't have to obsess on memory management correctness. Garbage collection is acceptable. There's a big pool of Java developers. There's a hard-code open source compiler. Microsoft doesn't control the language or the environment.
Whether the rather clunky Java libraries add negative value is something you have to think about, hard. The language itself is OK.
Back then, people didn't have race condition detectors for VLSI designs.. The 6502 was a 1970s design. CAD barely existed. ICs were laid out using Kodalith and X-acto knives. There's been progress since then.
The color, though, is correct. The standard NEMA rules for indicator colors, used on industrial gear for decades, are
-
GREEN Normal status.
-
AMBER Abnormal status. Operator should be aware that an amber lamp is lit.
-
RED Trouble status. Operator should take action to make the red light go out.
-
BLUE or WHITE On, or other non-specific meaning.
So blue and white are actually the default colors. Red should be used only for trouble indicators. We're still getting over the cheap red LED glut of the 1980s, when everything had red LEDs.Anything that rackmounts should follow these rules. It's not only annoying, but a headache, to have red lights for non-trouble conditions in a rack of equipment. IBM always has.
Actually, this is somewhat misleading. The top players listed are Microsoft (Windows CE), Wind River, Symbian, Palm. QNX, OSE, and Green Hills. But Microsoft, Symbian, and Palm are really selling into handheld devices, not hard real-time control. (The phone and PDA markets are much bigger than real time control, though.) Wind River's VRTX is the dominant player by a big margin, especially in low-end embedded control. QNX is next, and is usable on a broad range of platforms. Wind River is more of a specialist maker catering to the military Ada market.
Following these seven come LinuxWorks and MonteVista, who are moving up. These are the main Linux-based offerings.
Also confusing the issue is Windows XP Embedded, which is basically a Windows XP from which you can delete stuff you don't need. This sells more into point-of-sale applications than hard real time control.
In Los Angeles, which is in a desert, water prices are over twice that of Boston. So desalinization starts to look good.
Power consumption is a problem, but not a huge one. Water must be pressurized to about 1000PSI to get it through the membrane, so you can work out the energy cost from that. (Hint: it's the same as the energy cost to lift the water to the top of a column with 1000PSI at the bottom, which is about 2000 feet.) Reverse osmosis membranes now last for five years, so maintenance is less of a headache than it used to be.
There's no particular advantage in producing "low quality" water via reverse osmosis. You have to get most of the solids out to prevent crudding up the membranes, and the membrane stops almost everything bigger than a water molecule.
Reverse osmosis plants also generate brine, which is not too much of a problem if you're alongside an ocean, but can be a huge headache if you're drawing water from a brackish well and want to purify it.
Baloney. It is possible to write programs for which race conditions are undecideable. Such programs are broken. It is possible to write programs for which race condition detection is NP-hard. Such programs are broken if N is large. It is also possible to write programs for which race conditions can be proven to be absent. That's what you want to do.
Actually, it's straightforward to design software to be free of race conditions on a single machine. You then have a deadlock avoidance problem, but deadlocks are easily detected when they occur.
Hardware is routinely designed to be free of race conditions, after all.
The collision detection primitives, to give an example I know something about, were badly chosen. There are ways to do collision detection very efficiently, but not using Java 3D's collision primitives.
Chess is beginning to look like that. It yields to brute force. And by modern computational standards, not very much brute force. "Deep Fritz" tied 2:2 with Kasparov running on a desktop 4-processor IA-32 machine. Kasparov says it plays better than Deep Blue, which was a roomful of custom IBM hardware. If you haven't been on the cover of "Chess Life", "Deep Fritz", running on a standard PC, can trounce you. And it's only US$110.23.
If this stuff spreads off the golf course, does the maker come after you for a patent violation?
About 20 years ago, there was a fad for this, using dozens of effects pedal modules and noise generators. It was called "power electronics".
It sucked.
I can't imagine this is a new idea. I would have expected this to have been tried around 1955 or so, around the time reliable traffic detectors were developed.
An article in the Scottish press has more useful info.
It only generates 150KW. That's not much. Typical wind turbines generate 200KW to 700KW each, on windy days. (Average values are much lower.) Typical nuclear power plants generate 1,000,000KW. Powering a home takes about 1-2KW on average, so 10,000 homes require perhaps 15,000KW.
The SNAIL people want to move up to the 750KW range or so. That's more reasonable. As wind power people have discovered, having huge numbers of little turbines isn't cost effective. But somewhere around a few hundred KW per turbine, the economics start to work. If you can find a good site with steady wind. As with dams, there aren't that many good sites.
It will probably take several decades of operating experience to turn this into a reliable technology, just as it did with windpower. It's been half a century since the Grandpa's Knob loss of blade accident. The first big power-generating wind turbine oversped and threw a blade several hundred feet. For many years, nobody built one that big again. Gradually, the aerodynamics and control problems were figured out. It's taken that long to make large wind turbines work reliably and profitably.
Anything with moving parts in the ocean is likely to be high-maintenance. Making one of these things work reliably for decades will be tough. Maintenance will be costly. There's no guarantee of success.
In short, there's no breakthrough here until it's been running for a few years without breaking.
Has anyone figured out the ClearPlay "filter lists" yet? The Internet player for PCs downloads them via the Internet. What does the standalone player do? Can you create your own filter lists? For example, could you express the "Star Wars Phantom Edit" (the one that deletes that Jar Jar characte) as a ClearPlay filter list? This has potential.
A Cisco exec should do hard time for this.