The diamond industry has dug itself into a hole in the gemstone area, by valuing diamonds by lack of flaws. The "ideal diamond" is a perfect crystal.
This is not where you want to be positioned when going up against an industrial manufacturing process. Especially against a process borrowed from the semiconductor industry.
Expect PR campaigns emphasizing "the natural flaws of diamonds".
Yeah, and Gemesys, the new synthetic diamond gem manufacturer, caved in to deBeers and are now engraving "Gemesis created" and a serial number on each stone.
But in time, the bottom will fall out. In 1943, Linde Chemical created the first synthetic sapphire. Now, you can buy 200mm sapphire bar stock. Big 16mm gem-quality sapphires are available for about $10. The same thing happened to rubies and emeralds.
What the NT kernel does is well understood. The object code is widely available, and key parts, like file system formats, have been reverse engineered. There's plenty of documentation. A few major development shops have access to the source anyway.
If you're into kernel architecture, it might be interesting, but otherwise, so what?
Fifty years ago, in the heyday of the great department stores, yes. But high-end retail is too fragmented now. Today, it's the other way round. People hire personal shoppers to go out and do their shopping for them. They're called "personal shoppers" in NY and "stylists" in LA. Much more convenient than scheduling visits from suppliers.
Maya is one of those stories that should appear in business school textbooks. SGI acquired Alias and Wavefront, and much to the surprise of everybody, came out with a product that really was better than both. It promptly displaced Softimage|3D as the top package for film work.
At the time, Microsoft owned Softimage. Microsoft, having achieved their goal of moving 3D graphics onto Windows, sold off Softimage to Avid. Avid was the leader in 2D editing, but was starting to feel price pressure from below and was threatened by Softimage's move into that area. So they really bought Softimage to get the Softimage 2D editing package, and didn't really know what to do with the 3D product. Avid also had the problem that they were a high-end hardware vendor in a market where the high end was about to be eaten by the low end. As a result, the new Softimage 3D product, XSI, was years late.
So Maya took over. But it didn't help SGI sell expensive hardware. The low-end graphics boards were gaining on SGI. Maya was a software-only product, and didn't require SGI hardware. Maya is still available for Irix, but nobody buys SGI workstations to run it anymore. In fact, nobody buys SGI workstations for much of anything any more.
So it makes sense for SGI to sell off Maya.
Of course, SGI doesn't have much of a core business left ("We're a graphics company! No, we're a workstation company! No, we're a server company! No, we're a Linux company!"). Their core business is selling expensive hardware, and that's not a good business to be in.
This is good enough for rough surveys, but not good enough to allow collecting random images, locating them in space, and building 3D models. Another generation or two, and the expensive tripod-mounted Reigl scanners used for model-building will be replaced by handheld devices.
There are several other biographies of T. J. Watson Sr. The "official biography" is "The Lengthening Shadow" (1962). It's terrible. The "unofficial biography" is "Think, the Biography of the Watsons and IBM" (1969). That's quite good. Both were written while many people who knew Watson could still be interviewed.
Watson was a salesman, and was at one point NCR's top salesman, working for Patterson, the head of National Cash Register. The whole Patterson/NCR story is worth understanding. NCR's entire top management was convicted of criminal antitrust violations.
Their tactics make Microsoft look like small timers. NCR built defective duplicates of competing cash registers and sold them to make the competition look bad. Their sales reps were instructed on how to sabotage competing cash registers.
Big deal. This info is time-sensitive and decays rapidly.
Where the ship will be tomorrow is TOP SECRET.
Where the ship is today is SECRET. Where the ship was yesterday is CONFIDENTIAL. Where the ship was last week is UNCLASSIFIED.
- U.S. Navy guidance, circa 1970.
Eclipse for QNX is in a wierd state. This, roughly, is the situation. I think.
On QNX, it's the only IDE available. It's used primarily for C/C++ development, not Java development. QSSL (the company behind QNX) supports the Eclipse project, and donated the C/C++ support. But Eclipse has dropped support for QNX becauee QNX doesn't support the Java engine that newer versions of Eclipse need.
So QSSL sells a commercial version of Eclipse,
which is somehow made to work. There's a free version of QNX, but no Eclipse for it, which is a wierd situation for open source software.
Any comments on this from the Eclipse side of the world?
I'd like to see full Eclipse support of QNX, without having to run the commercial version of QNX. We have the commercial version of QNX, but some of what we develop is open source, and if we develop that with the IDE, nobody can build it with the free version.
Mashiro Mori contends that the progression
from a non-realistic to realistic portrayal of
a living thing is non-linear. In particular, there is an
"uncanny valley" as similarity becomes
almost, but not quite perfect.
That's not an original observation. The computer graphics industry struggled with that problem for most of the last decade, and made it out the other side of that valley a few years ago. The better film CG houses have that problem pretty much solved. Game CG is still working on it. It's harder for games; they need more automation and have to work in real time.
Motion capture data driving photorealistic characters looks good, but generated motion still isn't right. Look at something like "EA Football", and watch the movement go back and forth from realistic to terrible.
There are many, many SIGGRAPH papers in this area.
Synthetic parts should be reserved for synthetic actors. - Sextone for President spot, 1988.
The right solution is to put a GPS receiver in the camera and tag photos with time, date, and location.
No user action is required at picture-taking time.
Ricoh is already selling such a camera in Japan.
Kodak has a camera that plugs into an external GPS, but that's too clunky.
Pros would love this; often you want to search some big image archive for pictures of a specific location. Tourists would find their photos self-organizing.
Lookup can then be by address, or using a map or globe. Think MapQuest.
This offers the possibility of a new (and totally legitimate) peer-to-peer application - location based picture-sharing. See the pictures others took of tourist locations.
I actually saw one of those once, back in the 1980s. Very cute. A dead end, but cute. More crunch power with fewer gates than anybody else. In the first implementation, divide didn't work right for odd divisors.
Claim 1 does have some breadth. Arguably, though, this only covers systems with on-chip main memory (not cache) using more than half the real estate, some cache, and a variable-speed clock. Some microcontrollers fit those criteria, but they're not the most common ones. Bigger CPUs have off-chip memory, and low-end microcontrollers often have no cache system at all.
Exactly. This approach can actually work. It's worked before, in some DoD systems not well known outside the military and intelligence communities.
(I worked on one of those.)
This is now a communications problem.
Most of the literature on this kind of security is too theoretical. We need a "NSA Secure Linux For Dummies" book.
With mandatory security with levels and compartments going mainstream, we need apps designed to use it properly.
Mail handling is a good example. Each receive process should be running in a separate jail, with a net connection to the incoming port, a limited connection to the mail database, and no privilege to open files or network connections. Then it doesn't matter what happens in the receive process.
The software that passes data across security boundaries has to be carefully written and audited. But it doesn't have to do much.
Software has to be divided into two kinds - big, untrusted programs that do the work, and little, carefully audited security-critical programs that do very little.
The job of the OS is to keep each program in its own security box.
Mail, DNS, and web servers need to be broken up in this way. Now that Red Hat is going with SE Linux, it's time to do this. Get busy.
What I'm talking about is NASA's tax-supported foray into the industrial park business. They have all these tenant organizations at their oversized "centers". That's not a proper business for the Government to be in.
And they're a lousy landlord, one of their former tenants tells me.
This is not about downloading. This is about record stores being crushed by Wal-Mart. The same thing happened to local appliance stores years ago. It's happening to toy stores right now. (FAO Schwartz and Toys r Us both went under in January.) Up next, grocery stores and video stores. Blockbuster is scared.
There's way too much NASA for the amount of metal it puts into space.
NASA needs to close and downsize a few centers.
Ames should be cut back to a wind tunnel operation.
Slidell (now "the Stennis Space Center", a "multi-agency center for 30 resident agencies"), should be sold off to a private developer. The "Independent Verification and Validation Facility" in West Virginia should be consolidated with some NASA facility that needs its services.
Goddard needs some major cutbacks. (Goddard just awarded a $34 million contract for "conference support, duplicating, computer graphics, publication, and documentation" on a cost plus award fee basis. Then they issued a press release about it.)
NASA's non-flight research should be funded through the National Science Foundation. Environmental resarch should be moved to the EPA. In fact, even space science should go through NSF. NASA's job should be limited to flight hardware and support systems.
If NASA got rid of about half its organization, and insisted that the remaining half build stuff that flies, they might get somewhere.
The real status of this case is that SCO is under a court order to "describe with specificity" the infringing material, if any. IBM says SCO has not done so in their status report to the judge. SCO hasn't claimed otherwise. Both sides presented their motions in court today, and the judge will issue rulings on them in a few days.
If you read through the notes from the hearing, it's clear that SCO continues to refuse to, or is unable to, identify specific infringing code, and the judge doesn't like it. The judge
said "The problem is, unless you identify those codes, then IBM is
not in a position to have a response. We're at an impasse, and the case cannot continue with an impasse, that's why there was a court
order". That's a clear indication from the judge. The judge isn't buying SCO's nebulous theory of general infringement.
Cravath is slowly boxing in SCO. Notice that the trade secret claim has been dropped. The copyright claim isn't in the case yet, and IBM can probably insist that it doesn't go in without SCO showing the original and the purported copy side by side.
We have an Eaton VORAD radar on our DARPA Grand Challenge vehicle, and I have another one on my desk, looking out at an intersection. Physically, it's a rectangular black panel about the size of a license plate, flat in front and with cooling fins on the back.
This isn't a dumb speed gun radar.
It's a beam-steered phased array Doppler radar, reporting range, azimuth, and range rate on all the vehicles going by.
Multiple vehicles show as multiple targets. There's a Visual Basic app that comes with the unit, if purchased in the R&D configuration with the serial port adapter. We run it under QNX, and have our own interface programs.
The Eaton VORAD (stock version) is on tens of thousands of heavy trucks. It's intended to keep trucks from rear-ending other vehicles, which it is reported to do quite well. In some configurations, it will initiate braking on its own.
The vehicle version also stores data for accident reconstruction. Unlike simple airbag recorders, this unit shows what the other vehicles were doing. ("And here's where the other car ran the stop sign.")
The VORAD technology is priced around $2000, which seems to be limiting adoption. But it's a reliable box with millions of miles on truck bumpers.
Slashdot Mirror
Ah, yes. Download, decompress, untar, compile, discover missing package, download, decompress, untar, compile, discover missing package...
Expect PR campaigns emphasizing "the natural flaws of diamonds".
Gemesis URL (corrected)
Does this mean that telemarketing calls to VoIP phones are legal? Uh oh.
But in time, the bottom will fall out. In 1943, Linde Chemical created the first synthetic sapphire. Now, you can buy 200mm sapphire bar stock. Big 16mm gem-quality sapphires are available for about $10. The same thing happened to rubies and emeralds.
They're just rocks, people.
Is this a stellar mass squeezed down to planetary size, or what?
What the NT kernel does is well understood. The object code is widely available, and key parts, like file system formats, have been reverse engineered. There's plenty of documentation. A few major development shops have access to the source anyway. If you're into kernel architecture, it might be interesting, but otherwise, so what?
Fifty years ago, in the heyday of the great department stores, yes. But high-end retail is too fragmented now. Today, it's the other way round. People hire personal shoppers to go out and do their shopping for them. They're called "personal shoppers" in NY and "stylists" in LA. Much more convenient than scheduling visits from suppliers.
At the time, Microsoft owned Softimage. Microsoft, having achieved their goal of moving 3D graphics onto Windows, sold off Softimage to Avid. Avid was the leader in 2D editing, but was starting to feel price pressure from below and was threatened by Softimage's move into that area. So they really bought Softimage to get the Softimage 2D editing package, and didn't really know what to do with the 3D product. Avid also had the problem that they were a high-end hardware vendor in a market where the high end was about to be eaten by the low end. As a result, the new Softimage 3D product, XSI, was years late.
So Maya took over. But it didn't help SGI sell expensive hardware. The low-end graphics boards were gaining on SGI. Maya was a software-only product, and didn't require SGI hardware. Maya is still available for Irix, but nobody buys SGI workstations to run it anymore. In fact, nobody buys SGI workstations for much of anything any more.
So it makes sense for SGI to sell off Maya. Of course, SGI doesn't have much of a core business left ("We're a graphics company! No, we're a workstation company! No, we're a server company! No, we're a Linux company!"). Their core business is selling expensive hardware, and that's not a good business to be in.
This is good enough for rough surveys, but not good enough to allow collecting random images, locating them in space, and building 3D models. Another generation or two, and the expensive tripod-mounted Reigl scanners used for model-building will be replaced by handheld devices.
Watson was a salesman, and was at one point NCR's top salesman, working for Patterson, the head of National Cash Register. The whole Patterson/NCR story is worth understanding. NCR's entire top management was convicted of criminal antitrust violations. Their tactics make Microsoft look like small timers. NCR built defective duplicates of competing cash registers and sold them to make the competition look bad. Their sales reps were instructed on how to sabotage competing cash registers.
Where the ship will be tomorrow is TOP SECRET. Where the ship is today is SECRET. Where the ship was yesterday is CONFIDENTIAL. Where the ship was last week is UNCLASSIFIED. - U.S. Navy guidance, circa 1970.
People get jailed for stuff like that in the US now.
On QNX, it's the only IDE available. It's used primarily for C/C++ development, not Java development. QSSL (the company behind QNX) supports the Eclipse project, and donated the C/C++ support. But Eclipse has dropped support for QNX becauee QNX doesn't support the Java engine that newer versions of Eclipse need.
So QSSL sells a commercial version of Eclipse, which is somehow made to work. There's a free version of QNX, but no Eclipse for it, which is a wierd situation for open source software.
Any comments on this from the Eclipse side of the world?
I'd like to see full Eclipse support of QNX, without having to run the commercial version of QNX. We have the commercial version of QNX, but some of what we develop is open source, and if we develop that with the IDE, nobody can build it with the free version.
-
Mashiro Mori contends that the progression
from a non-realistic to realistic portrayal of
a living thing is non-linear. In particular, there is an
"uncanny valley" as similarity becomes
almost, but not quite perfect.
That's not an original observation. The computer graphics industry struggled with that problem for most of the last decade, and made it out the other side of that valley a few years ago. The better film CG houses have that problem pretty much solved. Game CG is still working on it. It's harder for games; they need more automation and have to work in real time.Motion capture data driving photorealistic characters looks good, but generated motion still isn't right. Look at something like "EA Football", and watch the movement go back and forth from realistic to terrible.
There are many, many SIGGRAPH papers in this area.
Synthetic parts should be reserved for synthetic actors. - Sextone for President spot, 1988.
It's meta-top-hyper-control-Q.
Pros would love this; often you want to search some big image archive for pictures of a specific location. Tourists would find their photos self-organizing.
Lookup can then be by address, or using a map or globe. Think MapQuest.
This offers the possibility of a new (and totally legitimate) peer-to-peer application - location based picture-sharing. See the pictures others took of tourist locations.
Claim 1 does have some breadth. Arguably, though, this only covers systems with on-chip main memory (not cache) using more than half the real estate, some cache, and a variable-speed clock. Some microcontrollers fit those criteria, but they're not the most common ones. Bigger CPUs have off-chip memory, and low-end microcontrollers often have no cache system at all.
Exactly. This approach can actually work. It's worked before, in some DoD systems not well known outside the military and intelligence communities. (I worked on one of those.)
This is now a communications problem. Most of the literature on this kind of security is too theoretical. We need a "NSA Secure Linux For Dummies" book.
Mail handling is a good example. Each receive process should be running in a separate jail, with a net connection to the incoming port, a limited connection to the mail database, and no privilege to open files or network connections. Then it doesn't matter what happens in the receive process.
The software that passes data across security boundaries has to be carefully written and audited. But it doesn't have to do much. Software has to be divided into two kinds - big, untrusted programs that do the work, and little, carefully audited security-critical programs that do very little.
The job of the OS is to keep each program in its own security box.
Mail, DNS, and web servers need to be broken up in this way. Now that Red Hat is going with SE Linux, it's time to do this. Get busy.
What I'm talking about is NASA's tax-supported foray into the industrial park business. They have all these tenant organizations at their oversized "centers". That's not a proper business for the Government to be in.
And they're a lousy landlord, one of their former tenants tells me.
Resistance is futile. You will be assimilated.
Ames should be cut back to a wind tunnel operation. Slidell (now "the Stennis Space Center", a "multi-agency center for 30 resident agencies"), should be sold off to a private developer. The "Independent Verification and Validation Facility" in West Virginia should be consolidated with some NASA facility that needs its services. Goddard needs some major cutbacks. (Goddard just awarded a $34 million contract for "conference support, duplicating, computer graphics, publication, and documentation" on a cost plus award fee basis. Then they issued a press release about it.)
NASA's non-flight research should be funded through the National Science Foundation. Environmental resarch should be moved to the EPA. In fact, even space science should go through NSF. NASA's job should be limited to flight hardware and support systems.
If NASA got rid of about half its organization, and insisted that the remaining half build stuff that flies, they might get somewhere.
If you read through the notes from the hearing, it's clear that SCO continues to refuse to, or is unable to, identify specific infringing code, and the judge doesn't like it. The judge said "The problem is, unless you identify those codes, then IBM is not in a position to have a response. We're at an impasse, and the case cannot continue with an impasse, that's why there was a court order". That's a clear indication from the judge. The judge isn't buying SCO's nebulous theory of general infringement.
Cravath is slowly boxing in SCO. Notice that the trade secret claim has been dropped. The copyright claim isn't in the case yet, and IBM can probably insist that it doesn't go in without SCO showing the original and the purported copy side by side.
Look for some rulings unfavorable to SCO shortly.
This isn't a dumb speed gun radar. It's a beam-steered phased array Doppler radar, reporting range, azimuth, and range rate on all the vehicles going by. Multiple vehicles show as multiple targets. There's a Visual Basic app that comes with the unit, if purchased in the R&D configuration with the serial port adapter. We run it under QNX, and have our own interface programs.
The Eaton VORAD (stock version) is on tens of thousands of heavy trucks. It's intended to keep trucks from rear-ending other vehicles, which it is reported to do quite well. In some configurations, it will initiate braking on its own.
The vehicle version also stores data for accident reconstruction. Unlike simple airbag recorders, this unit shows what the other vehicles were doing. ("And here's where the other car ran the stop sign.")
The VORAD technology is priced around $2000, which seems to be limiting adoption. But it's a reliable box with millions of miles on truck bumpers.