This may be Microsoft's answer to Google. And, incidentally, to Linux.
Microsoft already has a tiered operating system - the "business model" version of Vista is significantly more expensive than the "consumer model". The logical extension of this is a free consumer model with ads, and an expensive business model without ads.
This has many advantages for Microsoft. First, it's an answer to Google, which is ad-supported but doesn't have much of a lock-in mechanism, like Microsoft does. Second, it's an answer to Linux - preloaded consumer grade Windows becomes ad-supported, which is probably worth more than the current preload charge. It even helps Microsoft deal with piracy. The consumer version will be free. The business version will come with mandatory online services (they'll call it something like "Managed Workgroups") that will tie it to the mothership in Redmond.
This isn't that novel. It's very much like Randall Beer's insect work from a decade ago. It's hierarchical control using controllers built from control blocks the authors call "neurons". It's a pure reflex system, with no explicit prediction.
Also notice that it's a planar biped, constrained so that it can't fall sideways.
There's better locomotion and balance work going on in Japanese hobbyist robotics.
It's good that people are working on this stuff again. There was some impressive work in the 1980s and early 1990s, then a big lull.
It's not the definition of "ad agency" that matters. It's the statutory immunities for Internet service providers and newspapers. These are narrow.
See Doctor's Associates vs. QIP Holders, which turns on the distinction between an "Internet service provider" and an "Internet content provider" in the Communications Decency Act. Content providers are liable; service providers are not. ""Information content provider" is defined as "any person or
entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service."
Ad agencies like Google are going to have to address this.
The law on this varies by country, but given that Google regulates the style, content, and format of ads, then charges for them, they're clearly not just a passive conduit. More significantly, Google acts as an ad agency when it places ads on the web sites of others. It determines where, when, and how often the ad runs. That's acting as an ad agency. Ad agencies are routinely held liable in false advertising lawsuits. Sites on which Google ads run probably qualify for a safe harbor, but Google, acting as an ad agency placing ads on the sites of others, does not.
It's not clear how much liability an Internet ad agency has for content, but failure to take basic steps to identify the advertiser running the ad looks like negligence.
Here's a summary of US false advertising federal law. "The FTC can pursue the advertiser, its agency, and their employees. It can fine,
and enjoin, them. If the advertiser or agency is a subsidiary of another company, the
FTC can go after the parent. The FTC can even impose liability for false advertising on
a merged successor."
Similar principles prevail in Australian law. "The Commission does not necessarily expect (advertising) agencies to independently check the
technical claims made about a product, but if they are complicit in an obviously
misleading presentation, and fine print is used to obscure an offer's restrictions, then
difficulties start to arise."
Scene graphs do sound a little off the wall to me, based on the description in the review.
Scene graphs are a very useful concept. The problem is that they're more useful for handling object behavior than drawing. Game engines have elaborate scene graphs, used by the game system to decide what's near what and what to do about it, the collision detection system to keep things from going through walls, and the physics engine. So if you're doing something that has automated behavior, the scene graph is the representation of the world.
It seriously makes my head hurt to think that in order to draw an object X at location Y and orientation N, I have to create an object X, create an object "Offset", create an object "Rotation", then link them all together.
Even in OpenGL you do something like that. Think about what happens when you move a character with subparts, like arms. Rendering is a tree operation with 4x4 rotation matrices at the nodes.
If you have a scene graph for behavior purposes, you have enough info around to drive a rendering system like OpenGL. You don't need a second scene graph in the graphics system. It just gets in the way.
There is information that stays within the graphics system, like textures and drawing geometry, and OpenGL has display lists for retaining that info. It's important to use those, because you can send that info to the graphics card and leave it there, which is a big performance win. But that's not a full scene graph; it's graphic detail associated with scene graph nodes.
Historically, scene graph APIs haven't been too useful. There are successful drawing APIs, like OpenGL, and game engines, but the in-between middleware hasn't been that useful. SGI Inventor, later Open Inventor, was the classic in that space, and it's not used much any more.
Java 3D is abandonware. Sun wrote it, badly, and it's now a "community source project", meaning Sun doesn't support it any more. I used it in its early days and wasn't impressed. The Java3D collision detection system was both badly designed and badly implemented. The general consensus was "give us an OpenGL binding and get this turkey out of the way".
I know, but that's what directors want. I used to do physics simulation for high-end animation. Directors want an end state - they want the character to end up in some specified position. Sometimes one that's unreachable in the physical universe, let alone achievable with human muscle power. That's tough to do with a physics engine.
The way this is usually done in production today is to motion capture lots of motion, splice the bits of motion together, and edit the result manually. The result is some good motion and some bogus motion tied together. It looks bogus, but it's become a cinematic convention.
This really shows up in sports games. When EA runs an EA Football ad during an NFL game, you can tell from way across the room that the motion looks wrong.
Game-like motion has become enough of a cliche to be parodied. The opening scene of Tomb Raider has Angelina Jolie moving like a video game character, tucking and rolling while staying in a single vertical plane, just like the game.
There are many cinematic motion conventions that don't work in the real world. The classic is a car jumping across a gap. In reality, once the front wheels go over the edge, the car starts to rotate forward in pitch at a high rate. When you see a car jump in a movie, there are guides, ramps, extra wheels, and even pneumatic rams involved.
As for "the way his body looks when he lands just doesnt look natural. looks as if he just fell 3 feet. his body should have crouched/sunk more.", that kind of thing is sometimes done with flying rigs and high-speed computer-controlled winches. "Underworld - Evolution" did that. They record and debug the motion in a heavily padded gym, then play it back on the set.
Today, when someone does a tough stunt for real, nobody notices. There's a minor SF film which shows a woman running down the face of a 40-story building with a cable paying out behind her for support. A stuntwoman is really doing that on a real building. And for the bottom 30 feet, the star of the picture is really doing that, twisting to land on her feet and come out shooting. On the screen, it looks no different than similar things done in CG in other movies.
My experience is that Bonded Spammer is essentially dead. If you have Spam Assassin set to tag Bonded Spammer mail, you'll get items in X-Spam-Status like "RCVD_IN_BSP_TRUSTED". I have Firefox set to dump all those into the Bonded Spammer folder. The last e-mail to come in with that tag was in January 2007. I used to get more Bonded Spammer e-mails back in 2004 and 2005, but in 2006 it tapered off, and now it seems to be gone.
Is anyone else still seeing that junk?
Highlighting phishing sites is nice, but weak
on
Firefox Quickies
·
· Score: 4, Interesting
Just highlighting domains of phishing sites isn't going to be enough.
Here's today's list of domains that "sort of look like Paypal". These are after subdomain truncation.
"paypal-checker.com"
"paypal-contact.net"
"paypal-customize.com"
"paypal-erreur2.com"
"paypal-security.com"
"paypal-web-dll-scrnupdateaccount.ici.st"
"paypal-web-scrn-dll-pl-dai-pl-webscrndllfs-wertyu i.ork.pl"
"paypal.powered.at"
"paypal.q.fm"
"paypalaccverify.com"
"paypalcomcgibinwebscrcmd.by.ru"
"paypalcomcgibinwebscrcmm.by.ru"
"paypalcomcgibinwebscre.by.ru"
"paypalconstomers.com"
"paypalct.com"
"paypall.ro"
"paypalmd.com"
"paypalobjects.us"
"paypalsecuritycenter.org"
"paypalverification.org"
"paypel-acc-5.com"
"paypilpal.com"
"paypll-wscr.com"
"paypluspl.com"
These are from PhishTank, which blacklists at the URL level based on manual reports. For SiteTruth", we're in the process of converting to blacklisting phishing sites by the entire base domain. That's because we now see hundreds of entries like "session-624333.nationalcity.com.userpro.tw", which has to be treated as a bad indicator for all of "userpro.tw".
There's collateral damage. There are days when "tinyurl.com" and "notlong.com" get blacklisted, because phishing sites use them. MSN gets complaints about this. Today, anybody running something like "tinyurl" needs to continually check the phishing databases for attempts to abuse their service, or their own reputation is toast.
The amazing thing about Facebook is that it's a tiny company. Facebook headquarters is in a little building at 170 Hamilton Avenue in Palo Alto, next to the yoga shop and nail salon, and across from the retro soda fountain. It doesn't take much in the way of staff to run the thing.
The servers are in Northern Virginia, but most of the staff is in that little building in Palo Alto.
Bandwidth isn't where the cost goes. The biggest costs for an ISP are in in marketing and customer support, not bandwidth. If you look at wholesale ISP rates, what an reseller pays per customer for the raw service, they're less than half the retail rates.
The key to cost control is keeping down the number of people involved. Fibre is cheap.
One of Keen's points is that blogs and Craigslist are killing newspaper reporting. There are fewer people whose day job it is to go out and find out what's going on. Most blogs rehash information collected by others; true reportage is rare. Pick up a newspaper and see how few stories were initiated by reporters, as opposed to starting from some form of publicity.
This is a long-term trend; it's taken decades to reach this point. Compare newspapers from 1920, 1940, 1960, 1980, and 2000.
I have heard, from a moderately credible source in the security industry, that the people behind BlueFrog stopped because they were sent threats, along with pictures of their home and children, by the spammers they were attacking.
Someone may try this again, but it may have to be one of the security companies that handle lethal threats, like Blackwater or the Kroll division of Garda.
That argument has been used for teaching Latin in public schools, plane geometry, abstract algebra, and various other disciplines of specialist interest. It's worthwhile to learn something that requires formalism for that reason. But it's more cost-effective to learn something which is both useful and uses formalism in a productive way.
It's useful to have the skill to read a paper in a computer science journal and be able to 1) tell if it's a good idea, and 2) turn it into working code.
How many people have the time for that in a work environment today, though? Most people just download code, even it it's flawed.
Shuttle did well with their innovative heat pipe system, which is a rigid, sealed unit connecting a heat exchanger atop the CPU with one near an air outlet and fan. The case and motherboard were designed around the cooling system. That's what makes their small form factor PCs workable without overheating problems. We used those things outdoors in summer, while field testing robots, and they held up well. I've never had a Shuttle PC overheat, even at 105F ambient.
But the new graphics card cooling technology looks like a tacky afterthought. Big hoses all over the place. Too much plumbing. It comes with a paint job that might look good on a pickup with a lift job. So you get a sense of the target market.
If you like this sort of thing, go read "Soon, I Will be Invincible!", the fictional memoir of an evil mad scientist who tries to take over the world. It's the classic dweeb fantasy, with appropriate interior decoration.
All this stems from a distinction in wiretap law that goes back to the dial telephone era. Listening to voice requires a warrant, because that info belongs to the parties of the call only. But information used by the telephone company itself to route the call, like dial digits, can be requested from the telephone company. A "pen register" was classically a little electromechanical gadget that recorded dial pulses as dashes on a paper tape. There was no way to extract voice info with a pen register.
Then came Touch-Tone. Now the switching data was in the voice channel. After some court decisions, it was established that listening to the voice channel and extracting tones was OK, if done with "minimal" access to the voice channel.
Over time, this led to the "pen register" exception being extended to content the telco didn't process, including tones sent during a call to third-party services like voice mail, packet headers, E-mail headers, cellular location data, etc. Then came a "lower standard for stored messages", which included SMS messages and E-mail. Then came bulk interception via CALEA. Then the Patriot Act.
What math do you need in computer science today? It's a tough call.
But today, I'd study number-crunching rather than discrite math.
I have a classical computer science education - automata theory, number theory, combinatorics, mathematical logic - all discrite math. That's what CS was supposed to be about in
the 1980s. It's hasn't been enormously useful, and I'm writing this as someone who ran a successful project to develop a proof of correctness system. Mathematical logic gets used a little, but tree-like algorithms are more important. I'm not sure automata theory is useful for much of anything. It's one of those things, like proofs in plane geometry, taught in schools because it was the first theory in the field.
Number-crunching, probability, and statistics seem to be more useful today.
If you do anything in graphics, you have to have the 3D transformation math down cold.
I've had to do tensor calculus and integration of non-linear differential equations (I used to do physics engines for animation) and that required plowing through difficult math books and getting consulting from university math departments. Bayesian statistics have become very important - it's used in spam filters, search engines, computer vision, and most things that have to intelligently reduce messy real-world data. I didn't get enough of that at Stanford.
On the other hand, where are you going to use this stuff? Outside of Google, Microsoft, and universities, who does real CS research any more? All the good research centers (DEC WRL, HP Labs, IBM Almaden, PARC, etc.) are gone, or a shadow of what they once were.
This guy has a web site, Theseus Research, where you can read some of his papers. There's a long, somewhat turgid paper on "null logic", which seems to be an approach to designing unclocked digital logic without race conditions. It's one of those off-in-its-own-world papers with no references and no indication of the ideas actually being tried.
He does have something called the "Theseus Warp Algorithm", which is a different approach to resizing images. The example looks OK, but he only tries it on one image, and that image is an unusual case.
The book is apparently an extension of his 1993 paper,
A Critical Review of the Notion of the Algorithm in Computer Science". The main point he's making is that much of computing isn't about "algorithms", which, traditionally, are computational functions which take in some set of inputs and produce a set of outputs.
Much of computing is about doing something about a stream of incoming events. Most modern programs, certainly anything with a GUI, are a set of interconnected working parts, not "algorithms" in the classic sense. That's reasonable enough, and that's as far as his 1993 paper got. He poses a question, but does not define a solution.
Does his book go further, or is he still just philosophizing? Don't know.
Painfully true. In particular, the religious wackos got to run the educational system. Over 90% of the doctorates issued in Saudi Arabia are in "religious studies". At the lower levels, most teaching time is devoted to religious subjects. The textbooks and teaching are standardized across the country.
One result is that few young Saudis learn how to do real-world jobs. Foreign workers do almost all the real jobs. The Government has a huge number of make-work jobs for their own citizens.
Because oil revenue per capita is dropping steadily, this isn't going to work much longer. There have been some attempts by the Saudi government to turn this around. But it's not really working.
The site's Whois history information is available from a site that archives that info. It costs $15 per month.
They show Whois changes on 2007-03-11,
2007-07-03,
2007-07-04,
2007-07-05, and
2007-07-06. So if anybody needs to prove anything, the truth is out there.
Clever little piece of early technology. Main article is link to a paysite. Video has too much talking head and not enough of the thing working.
Actual YouTube link. About two minutes of talking head (skip that part), five seconds of the thing working. The second sequence of the robot, where you see how the cord is wound around the axles, makes it obvious how the thing works.
The basic idea is that you have a 3-wheeled platform with two powered wheels and an idler. The powered wheels are on a split shaft, and each half of the shaft has a drum around which a cord is wound. The cords go up to a pole, through a pulley, and down to a weight.
If that was the whole thing, it would just go in a straight line. The clever part is that there are some pins in the rollers, and those are used to reverse the winding direction of the cords. So by putting in appropriate pins and winding the cord in alternate directions, you can set it to move forward, turn in place in either direction, or reverse. Simple, cute, and reasonably accurate, because the programming setup determines the distance in each mode, not the time.
You can't cycle it, though; it's a one-shot thing. Once it's run, you have to manually redo all the proper winding and setup. This is significant, because, unlike something cam-driven, it has no potential as an industrial technology. It doesn't lead to, say, an automatic weaving loom.
One of the most common military batteries is the BA 5590, a lithium/sulfur dioxide primary battery with a good combat record over fifteen years. 24VDC at 200mA for 28 hours, or 135 watt-hours in a 1 Kg package. There's an upgrade to lithium manganese dioxide in the same form factor, for 333 watt-hours. The original article says that the competition requires a battery with 1290 watt-hours in a 4Kg package, or 323 watt-hours per kilogram. That's where the primary batteries are now. So this isn't a big advance over existing technology. It's more of a packaging issue.
The competition makes no mention of cost. Those military non-rechargeable batteries cost about $100 each. They store about 3 cents worth of power at utility rates.
Interestingly, the competition doesn't require a rechargeable battery.
Rechargeable batteries aren't that popular for infantry equipment. Finding a power source, place, and time to charge batteries is tough in combat. While most military vehicles can provide external power, it ties down a vehicle and crew to use it as a power source.
6. The paedophile and predator perspective
Catholic churches are an El Dorado for paedophiles and predators who want to harm young people. The people behind the churches are not in control of safety and do not put enough effort into keeping predators out of the churches.
Myspace, Facebook, etc. aren't losing multiple multi million dollar lawsuits for deliberately assisting in covering up child molestation by their employees. The Catholic Church is. Let the kids go online, but never let them enter a Catholic church alone.
Slashdot Mirror
This may be Microsoft's answer to Google. And, incidentally, to Linux.
Microsoft already has a tiered operating system - the "business model" version of Vista is significantly more expensive than the "consumer model". The logical extension of this is a free consumer model with ads, and an expensive business model without ads.
This has many advantages for Microsoft. First, it's an answer to Google, which is ad-supported but doesn't have much of a lock-in mechanism, like Microsoft does. Second, it's an answer to Linux - preloaded consumer grade Windows becomes ad-supported, which is probably worth more than the current preload charge. It even helps Microsoft deal with piracy. The consumer version will be free. The business version will come with mandatory online services (they'll call it something like "Managed Workgroups") that will tie it to the mothership in Redmond.
The action is elsewhere. It's at the Game Developer's Conference for technology, and the Hollywood Games Summit for content.
Anyone with $799 can go to the Hollywood Games Summit. They even throw in subscriptions to both Game Developer and the Hollywood Reporter.
Also, here's the cited paper.
This isn't that novel. It's very much like Randall Beer's insect work from a decade ago. It's hierarchical control using controllers built from control blocks the authors call "neurons". It's a pure reflex system, with no explicit prediction.
Also notice that it's a planar biped, constrained so that it can't fall sideways.
There's better locomotion and balance work going on in Japanese hobbyist robotics.
It's good that people are working on this stuff again. There was some impressive work in the 1980s and early 1990s, then a big lull.
It's not the definition of "ad agency" that matters. It's the statutory immunities for Internet service providers and newspapers. These are narrow.
See Doctor's Associates vs. QIP Holders, which turns on the distinction between an "Internet service provider" and an "Internet content provider" in the Communications Decency Act. Content providers are liable; service providers are not. ""Information content provider" is defined as "any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service."
Ad agencies like Google are going to have to address this. The law on this varies by country, but given that Google regulates the style, content, and format of ads, then charges for them, they're clearly not just a passive conduit. More significantly, Google acts as an ad agency when it places ads on the web sites of others. It determines where, when, and how often the ad runs. That's acting as an ad agency. Ad agencies are routinely held liable in false advertising lawsuits. Sites on which Google ads run probably qualify for a safe harbor, but Google, acting as an ad agency placing ads on the sites of others, does not.
It's not clear how much liability an Internet ad agency has for content, but failure to take basic steps to identify the advertiser running the ad looks like negligence.
Here's a summary of US false advertising federal law. "The FTC can pursue the advertiser, its agency, and their employees. It can fine, and enjoin, them. If the advertiser or agency is a subsidiary of another company, the FTC can go after the parent. The FTC can even impose liability for false advertising on a merged successor."
Similar principles prevail in Australian law. "The Commission does not necessarily expect (advertising) agencies to independently check the technical claims made about a product, but if they are complicit in an obviously misleading presentation, and fine print is used to obscure an offer's restrictions, then difficulties start to arise."
Scene graphs do sound a little off the wall to me, based on the description in the review.
Scene graphs are a very useful concept. The problem is that they're more useful for handling object behavior than drawing. Game engines have elaborate scene graphs, used by the game system to decide what's near what and what to do about it, the collision detection system to keep things from going through walls, and the physics engine. So if you're doing something that has automated behavior, the scene graph is the representation of the world.
It seriously makes my head hurt to think that in order to draw an object X at location Y and orientation N, I have to create an object X, create an object "Offset", create an object "Rotation", then link them all together.
Even in OpenGL you do something like that. Think about what happens when you move a character with subparts, like arms. Rendering is a tree operation with 4x4 rotation matrices at the nodes. If you have a scene graph for behavior purposes, you have enough info around to drive a rendering system like OpenGL. You don't need a second scene graph in the graphics system. It just gets in the way.
There is information that stays within the graphics system, like textures and drawing geometry, and OpenGL has display lists for retaining that info. It's important to use those, because you can send that info to the graphics card and leave it there, which is a big performance win. But that's not a full scene graph; it's graphic detail associated with scene graph nodes.
Historically, scene graph APIs haven't been too useful. There are successful drawing APIs, like OpenGL, and game engines, but the in-between middleware hasn't been that useful. SGI Inventor, later Open Inventor, was the classic in that space, and it's not used much any more.
Java 3D is abandonware. Sun wrote it, badly, and it's now a "community source project", meaning Sun doesn't support it any more. I used it in its early days and wasn't impressed. The Java3D collision detection system was both badly designed and badly implemented. The general consensus was "give us an OpenGL binding and get this turkey out of the way".
I know, but that's what directors want. I used to do physics simulation for high-end animation. Directors want an end state - they want the character to end up in some specified position. Sometimes one that's unreachable in the physical universe, let alone achievable with human muscle power. That's tough to do with a physics engine.
The way this is usually done in production today is to motion capture lots of motion, splice the bits of motion together, and edit the result manually. The result is some good motion and some bogus motion tied together. It looks bogus, but it's become a cinematic convention.
This really shows up in sports games. When EA runs an EA Football ad during an NFL game, you can tell from way across the room that the motion looks wrong.
Game-like motion has become enough of a cliche to be parodied. The opening scene of Tomb Raider has Angelina Jolie moving like a video game character, tucking and rolling while staying in a single vertical plane, just like the game.
There are many cinematic motion conventions that don't work in the real world. The classic is a car jumping across a gap. In reality, once the front wheels go over the edge, the car starts to rotate forward in pitch at a high rate. When you see a car jump in a movie, there are guides, ramps, extra wheels, and even pneumatic rams involved.
As for "the way his body looks when he lands just doesnt look natural. looks as if he just fell 3 feet. his body should have crouched/sunk more.", that kind of thing is sometimes done with flying rigs and high-speed computer-controlled winches. "Underworld - Evolution" did that. They record and debug the motion in a heavily padded gym, then play it back on the set.
Today, when someone does a tough stunt for real, nobody notices. There's a minor SF film which shows a woman running down the face of a 40-story building with a cable paying out behind her for support. A stuntwoman is really doing that on a real building. And for the bottom 30 feet, the star of the picture is really doing that, twisting to land on her feet and come out shooting. On the screen, it looks no different than similar things done in CG in other movies.
My experience is that Bonded Spammer is essentially dead. If you have Spam Assassin set to tag Bonded Spammer mail, you'll get items in X-Spam-Status like "RCVD_IN_BSP_TRUSTED". I have Firefox set to dump all those into the Bonded Spammer folder. The last e-mail to come in with that tag was in January 2007. I used to get more Bonded Spammer e-mails back in 2004 and 2005, but in 2006 it tapered off, and now it seems to be gone.
Is anyone else still seeing that junk?
Just highlighting domains of phishing sites isn't going to be enough. Here's today's list of domains that "sort of look like Paypal". These are after subdomain truncation.u i.ork.pl"
"paypal-checker.com"
"paypal-contact.net"
"paypal-customize.com"
"paypal-erreur2.com"
"paypal-security.com"
"paypal-web-dll-scrnupdateaccount.ici.st"
"paypal-web-scrn-dll-pl-dai-pl-webscrndllfs-werty
"paypal.powered.at"
"paypal.q.fm"
"paypalaccverify.com"
"paypalcomcgibinwebscrcmd.by.ru"
"paypalcomcgibinwebscrcmm.by.ru"
"paypalcomcgibinwebscre.by.ru"
"paypalconstomers.com"
"paypalct.com"
"paypall.ro"
"paypalmd.com"
"paypalobjects.us"
"paypalsecuritycenter.org"
"paypalverification.org"
"paypel-acc-5.com"
"paypilpal.com"
"paypll-wscr.com"
"paypluspl.com"
These are from PhishTank, which blacklists at the URL level based on manual reports. For SiteTruth", we're in the process of converting to blacklisting phishing sites by the entire base domain. That's because we now see hundreds of entries like "session-624333.nationalcity.com.userpro.tw", which has to be treated as a bad indicator for all of "userpro.tw".
There's collateral damage. There are days when "tinyurl.com" and "notlong.com" get blacklisted, because phishing sites use them. MSN gets complaints about this. Today, anybody running something like "tinyurl" needs to continually check the phishing databases for attempts to abuse their service, or their own reputation is toast.
The amazing thing about Facebook is that it's a tiny company. Facebook headquarters is in a little building at 170 Hamilton Avenue in Palo Alto, next to the yoga shop and nail salon, and across from the retro soda fountain. It doesn't take much in the way of staff to run the thing. The servers are in Northern Virginia, but most of the staff is in that little building in Palo Alto.
Now that's successful "Web 2.0".
Bandwidth isn't where the cost goes. The biggest costs for an ISP are in in marketing and customer support, not bandwidth. If you look at wholesale ISP rates, what an reseller pays per customer for the raw service, they're less than half the retail rates.
The key to cost control is keeping down the number of people involved. Fibre is cheap.
For in-depth coverage of this issue, read "The Cult of the Amateur: How today's Internet is killing our culture", by Andrew Keen. That covers the subject much better than the usual blogodreck.
One of Keen's points is that blogs and Craigslist are killing newspaper reporting. There are fewer people whose day job it is to go out and find out what's going on. Most blogs rehash information collected by others; true reportage is rare. Pick up a newspaper and see how few stories were initiated by reporters, as opposed to starting from some form of publicity. This is a long-term trend; it's taken decades to reach this point. Compare newspapers from 1920, 1940, 1960, 1980, and 2000.
Unfortunately, they didn't have the resources
I have heard, from a moderately credible source in the security industry, that the people behind BlueFrog stopped because they were sent threats, along with pictures of their home and children, by the spammers they were attacking.
Someone may try this again, but it may have to be one of the security companies that handle lethal threats, like Blackwater or the Kroll division of Garda.
This is just a troll to drive traffic to an ad-heavy site.
That argument has been used for teaching Latin in public schools, plane geometry, abstract algebra, and various other disciplines of specialist interest. It's worthwhile to learn something that requires formalism for that reason. But it's more cost-effective to learn something which is both useful and uses formalism in a productive way.
It's useful to have the skill to read a paper in a computer science journal and be able to 1) tell if it's a good idea, and 2) turn it into working code. How many people have the time for that in a work environment today, though? Most people just download code, even it it's flawed.
Shuttle did well with their innovative heat pipe system, which is a rigid, sealed unit connecting a heat exchanger atop the CPU with one near an air outlet and fan. The case and motherboard were designed around the cooling system. That's what makes their small form factor PCs workable without overheating problems. We used those things outdoors in summer, while field testing robots, and they held up well. I've never had a Shuttle PC overheat, even at 105F ambient.
But the new graphics card cooling technology looks like a tacky afterthought. Big hoses all over the place. Too much plumbing. It comes with a paint job that might look good on a pickup with a lift job. So you get a sense of the target market.
If you like this sort of thing, go read "Soon, I Will be Invincible!", the fictional memoir of an evil mad scientist who tries to take over the world. It's the classic dweeb fantasy, with appropriate interior decoration.
This mess developed over time.
All this stems from a distinction in wiretap law that goes back to the dial telephone era. Listening to voice requires a warrant, because that info belongs to the parties of the call only. But information used by the telephone company itself to route the call, like dial digits, can be requested from the telephone company. A "pen register" was classically a little electromechanical gadget that recorded dial pulses as dashes on a paper tape. There was no way to extract voice info with a pen register.
Then came Touch-Tone. Now the switching data was in the voice channel. After some court decisions, it was established that listening to the voice channel and extracting tones was OK, if done with "minimal" access to the voice channel.
Over time, this led to the "pen register" exception being extended to content the telco didn't process, including tones sent during a call to third-party services like voice mail, packet headers, E-mail headers, cellular location data, etc. Then came a "lower standard for stored messages", which included SMS messages and E-mail. Then came bulk interception via CALEA. Then the Patriot Act.
What math do you need in computer science today? It's a tough call. But today, I'd study number-crunching rather than discrite math.
I have a classical computer science education - automata theory, number theory, combinatorics, mathematical logic - all discrite math. That's what CS was supposed to be about in the 1980s. It's hasn't been enormously useful, and I'm writing this as someone who ran a successful project to develop a proof of correctness system. Mathematical logic gets used a little, but tree-like algorithms are more important. I'm not sure automata theory is useful for much of anything. It's one of those things, like proofs in plane geometry, taught in schools because it was the first theory in the field.
Number-crunching, probability, and statistics seem to be more useful today. If you do anything in graphics, you have to have the 3D transformation math down cold. I've had to do tensor calculus and integration of non-linear differential equations (I used to do physics engines for animation) and that required plowing through difficult math books and getting consulting from university math departments. Bayesian statistics have become very important - it's used in spam filters, search engines, computer vision, and most things that have to intelligently reduce messy real-world data. I didn't get enough of that at Stanford.
On the other hand, where are you going to use this stuff? Outside of Google, Microsoft, and universities, who does real CS research any more? All the good research centers (DEC WRL, HP Labs, IBM Almaden, PARC, etc.) are gone, or a shadow of what they once were.
This guy has a web site, Theseus Research, where you can read some of his papers. There's a long, somewhat turgid paper on "null logic", which seems to be an approach to designing unclocked digital logic without race conditions. It's one of those off-in-its-own-world papers with no references and no indication of the ideas actually being tried.
He does have something called the "Theseus Warp Algorithm", which is a different approach to resizing images. The example looks OK, but he only tries it on one image, and that image is an unusual case.
The book is apparently an extension of his 1993 paper, A Critical Review of the Notion of the Algorithm in Computer Science". The main point he's making is that much of computing isn't about "algorithms", which, traditionally, are computational functions which take in some set of inputs and produce a set of outputs. Much of computing is about doing something about a stream of incoming events. Most modern programs, certainly anything with a GUI, are a set of interconnected working parts, not "algorithms" in the classic sense. That's reasonable enough, and that's as far as his 1993 paper got. He poses a question, but does not define a solution.
Does his book go further, or is he still just philosophizing? Don't know.
Painfully true. In particular, the religious wackos got to run the educational system. Over 90% of the doctorates issued in Saudi Arabia are in "religious studies". At the lower levels, most teaching time is devoted to religious subjects. The textbooks and teaching are standardized across the country.
One result is that few young Saudis learn how to do real-world jobs. Foreign workers do almost all the real jobs. The Government has a huge number of make-work jobs for their own citizens.
Because oil revenue per capita is dropping steadily, this isn't going to work much longer. There have been some attempts by the Saudi government to turn this around. But it's not really working.
For a cynical view on all this by a Saudi, see The Religious Policeman.
The site's Whois history information is available from a site that archives that info. It costs $15 per month.
They show Whois changes on 2007-03-11, 2007-07-03, 2007-07-04, 2007-07-05, and 2007-07-06. So if anybody needs to prove anything, the truth is out there.
Clever little piece of early technology. Main article is link to a paysite. Video has too much talking head and not enough of the thing working.
Actual YouTube link. About two minutes of talking head (skip that part), five seconds of the thing working. The second sequence of the robot, where you see how the cord is wound around the axles, makes it obvious how the thing works.
The basic idea is that you have a 3-wheeled platform with two powered wheels and an idler. The powered wheels are on a split shaft, and each half of the shaft has a drum around which a cord is wound. The cords go up to a pole, through a pulley, and down to a weight.
If that was the whole thing, it would just go in a straight line. The clever part is that there are some pins in the rollers, and those are used to reverse the winding direction of the cords. So by putting in appropriate pins and winding the cord in alternate directions, you can set it to move forward, turn in place in either direction, or reverse. Simple, cute, and reasonably accurate, because the programming setup determines the distance in each mode, not the time.
You can't cycle it, though; it's a one-shot thing. Once it's run, you have to manually redo all the proper winding and setup. This is significant, because, unlike something cam-driven, it has no potential as an industrial technology. It doesn't lead to, say, an automatic weaving loom.
One of the most common military batteries is the BA 5590, a lithium/sulfur dioxide primary battery with a good combat record over fifteen years. 24VDC at 200mA for 28 hours, or 135 watt-hours in a 1 Kg package. There's an upgrade to lithium manganese dioxide in the same form factor, for 333 watt-hours. The original article says that the competition requires a battery with 1290 watt-hours in a 4Kg package, or 323 watt-hours per kilogram. That's where the primary batteries are now. So this isn't a big advance over existing technology. It's more of a packaging issue.
The competition makes no mention of cost. Those military non-rechargeable batteries cost about $100 each. They store about 3 cents worth of power at utility rates.
Interestingly, the competition doesn't require a rechargeable battery. Rechargeable batteries aren't that popular for infantry equipment. Finding a power source, place, and time to charge batteries is tough in combat. While most military vehicles can provide external power, it ties down a vehicle and crew to use it as a power source.
6. The paedophile and predator perspective
Catholic churches are an El Dorado for paedophiles and predators who want to harm young people. The people behind the churches are not in control of safety and do not put enough effort into keeping predators out of the churches.
That's documented reality, established in hundreds of lawsuits.
Myspace, Facebook, etc. aren't losing multiple multi million dollar lawsuits for deliberately assisting in covering up child molestation by their employees. The Catholic Church is. Let the kids go online, but never let them enter a Catholic church alone.