Yeah. When I buy a hard drive, I do not want it prefilled with crapware. In fact, there should be a prominent security warning on the package for any media that comes with executable content. Microsoft OSs still tend to execute any executable content that comes within range of the machine.
Did they put an autorun file on the hard drive? I'd regard that as "exceeds authorized access".
The most popular location in the top 10 is Chicago, IL, with three of the largest data centers. 350 E. Cermak is across from Chicago's convention center. Elk Grove Village is west of O'Hare Airport.
Why Chicago? It's a central US location. It doesn't get too hot. Power reliability is good. Transportation access is good. Earthquakes, hurricanes, and floods are rare. There are many big, solid industrial buildings available. It's not as depressed as Detroit or Cleveland.
There's real 3D directed by James Cameron, where the movie was shot in 3D. And there's Disney Fake3D(tm), where "3D" is added in post-production. The latter is like watching a multi-layered Flash animation.
Cameron uses depth to add reality, not generate in-your-face "effects". At no time in Avatar does an object come "through the screen". That looks fake, and breaks the mood.
Alice in Wonderland, on the other hand, was shot in 2D, and converted to 3D by In-Three. The 2D images go through a program called "In3gue", which segments the images and adds depth info for each pixel.
Once this has been done, the image layers can be manipulated in depth.
Even if the original content was shot in 3D, extensive post-production is required. See this guide to 3D post.
Understand how fake stereoscopic 3D is. In the real world, beyond a few meters, there are no noticeable stereo effects.
An interesting point made by In-Three is that content for kids requires a narrower eye separation. It's well known that objects placed "beyond infinity" by stereoscopy produce much unhappiness in viewers, up to and including eyestrain and stomach distress. Content created for adult human eye separation inflicts this effect on children.
An important problem that hasn't been faced yet is that 3D viewing on smaller home screens viewed at close range requires quite different parallax adjustments than on big distant screens. Merely transferring theatrical 3D to Blu-Ray disks is really going to suck.
It's a very broad shop. It's not the only public shop in Silicon Valley. If you want to do electronics, Hacker Dojo has better workstations, where surface-mount work is possible. If you're building furniture, The Sawdust Shop has a better wood shop. But TechShop has both sewing machines and CNC milling machines, a stereolithography machine and a plasma cutter, which you usually don't find under one roof.
There are Silicon Valley companies which buy memberships and send their employees over to use the machine tools. The four big manual Bridgeport mills, the big lathes, and the stereolithography machine are usually being used by pros.
TechShop gives classes constantly, but most of the people who use TechShop already have considerable familiarity with tools. They just need access to the bigger machines. It's a good place to learn how to use CNC machine tools. CNC software is quite good today, and TechShop has reasonable midrange CNC design software (Vectrix Cut2D/Cut3D, SolidWorks, etc.) installed on their rather sluggish Windows Vista desktop machines.
There's not much electronics and robotics work. Although TechShop gives Arduno programming classes, and people take them and build the projects, not much electronics gets built there. They have power supplies, meters, scopes, and soldering stations, but they're 1980s technology.
The most popular activity is cutting decorative patterns with the laser cutter. It's easy to do, and two laser cutters are busy doing it almost continuously. Those machines just need line art in CorelDraw; you don't have to learn SolidWorks or Vectrix and do real CNC programming.
No one activity dominates, though; there are people building birdhouses and people building rocket engines.
Does Google's measurement include delays from off-site ad servers? That's a big issue. For many sites (including Slashdot), the off-site ad servers are the big bottleneck.
Web site programmers will now have to avoid ad code that delays page loading until the ads come in.
I expect to see ad code that measures the response time of the ad server, and if the ad server doesn't respond fast enough, drops the ad and reports the fail to a monitoring site.
Then we'll see sites gaming the system. If Google is using information from their "Google Toolbar" to affect search results, we'll probably see attempts to pump fake data into the Google Toolbar server. Google is going to have to learn the lesson well known to developers of networked games - "never trust the client".
It's hard getting the attention of some vendors. I see vulnerabilities in a slightly different context - hacked web sites hosting phishing pages.
We distribute a list of major domains being exploited by active phishing scams. This is obtained by processing PhishTank data, and we do this because we want to reduce the collateral damage from a tough blacklist system. At any given time, there are about 30 to 80 domains on the list.
Some sites get themselves off the list quickly. By now, most of the better free hosting services and short-URL services are automatically checking PhishTank and the APWG blacklist to see when they've been hit. Today, if you run a service where anybody can put up a page that could be used for phishing (i.e. it's not full of your own headers and banners), you need automation to deal with attacks. I've been in contact with the abuse guy at "t35.com", which is a free hosting service. They've recently been hit by a flood of phishing attacks, with several hundred new reports in PhishTank per day. The attacks were coming in faster than the abuse guy could clean them out. They're now gaining on the problem, but haven't squashed it yet. Take-away lesson: automate this.
The ones near the top of the list have been there for a while. Note the dates, which are the date that the oldest phishing report still online and active appeared in PhishTank. Some just need help. Typically, these are small organizations like churches and nonprofits that have had a break-in and were partially taken over by a phishing site. I send them the Anti-Phishing Working Group's "What To Do if your Site Has Been Hacked". Sometimes I give them a phone call. They deserve sympathy.
Then there are the hard cases. These are sites with no visible contact address, or a clueless abuse department.
At the moment, Google Sites and Google Spreadsheets are being used for phishing. Google is new to the free hosting business, and the phishers have discovered some tricks that Google can't yet handle. While Google puts a "report abuse" link on their site pages, it's possible to set up a file for downloading on Google Sites, and an HTML page can be served that way, without Google's abuse checking. There's also an exploit of Google Spreadsheets. That one is an example of Habbo Hotel phishing. We've reported these to Google several times, but they haven't been fixed yet.
We've been seeing a new type of attack recently - a phishing operation breaks into a shared hosting server and plants phishing pages on multiple domains on a single server. One of these hit one of the mysterious "*.websitewelcome.com" servers, which has "cloaked domain registration" and no useful default web page. These seem to be associated with "ThePlanet.com", but whether ThePlanet operates them, is providing wholesale hosting, is providing colocation, or is just the upstream connectivity provider is not clear.
Hiding the contact information of a hosting provider is legally unwise. The hosting provider may lose the "safe harbor" protection of the the DMCA. The "safe harbor" provision for "Information Residing on Systems or Networks At Direction of Users" only applies if "the service provider has designated an agent to receive notifications of claimed infringement... by making available through its service, including on its website in a location accessible to the public, and by providing to the Copyright Office, substantially the following information: the name, address, phone number, and electronic mail address of the agent." So when the RIAA or the MPAA come calling, a likely event for a hosting service, they get
The experiments so far indicate that paying students for results improves only the results paid for. Pay for attendance, you get attendance. Pay for grades on quizzes, you get grades on quizzes. End of year scores don't improve much, if at all. And when the money stops, so does the improvement.
Can this robot hold a position, or return to a position upon surfacing and learning its position? Or is at the mercies of the ocean currents as to where it ends up?
No, it can't. It can adjust its depth; that's all.
Compare the Wave Glider, from Liquid Robotics. This is a privately funded product. It has two parts, a surface "floater" that looks like a surfboard, and a tethered "glider", which hangs below it, about 10m underwater. Wave action on the floater pulls the glider up, and gravity brings it down. Spring-loaded ailerons move the glider forward, powered by the wave motion, and it tows the floater. A rudder on the glider allows steering. The floater has solar panels, a GPS, and an Iridium satellite data link.
The Wave Glider is not only autonomous and self-powered, but can make long trips under control. First they sent one all the way around the Big Island in Hawaii. Worked fine. Then they sent it from Hawaii to California. This took a while; it averages around 1 knot; more in storms, less in calm weather.
In storms, the floater is pulled through waves, like a surfboard, and comes out unharmed. They picked it up in Monterey Bay, saw that it was in good condition, and sent it back out again. They parked it in Monterey Bay for a while, circling in a 50 meter circle. Then they sent it back out again on an trip to Alaska and back.
The Wave Glider generally stays within about 50m of its programmed course. The Coast Guard treats it as "floating debris", and it doesn't show lights. If something hits it, it's like running over a surfboard. The control center on shore (a laptop with an Iridium phone) gets ship tracking data, and they guide the Wave Gliders out of the way of large ships.
Harmon already did. QNX used to run well on most desktop machines and some laptops, so you could self-host and develop for QNX on QNX.
That's gradually been broken over the last few years. Harmon is an audio electronics company. Owning an OS company was out of their league. At least RIM is in the right industry.
I used QNX extensively from 2002-2005, and my desktop machine, plus all the machines in a robot vehicle, ran QNX. QNX runs many of the high-end robots; BigDog, for example, uses QNX. Several DARPA Grand Challenge vehicles ran QNX. When timing really matters, and the problem is far too complex for PLCs and low-end microcontrollers, you need something like QNX.
It's really sad. QNX as an OS was very good. QNX as a company, especially their marketing operation, managed to anger developers, customers, and employees, resulting in a major brain drain out of the QNX community.
Despite Torvalds, mainstream computing is moving to message passing between separate processes. Gnome, Chrome, MacOS, and WebKit are all multi-process message passing systems. QNX does message passing far better than UNIX or Linux does. Message passing in QNX works like a subroutine call. Under UNIX and Linux, a message pass involves the excess baggage of turning a subroutine call into a write to a socket, with marshaling for the conversion from message to stream, and with several unnecessary trips through the scheduler.
There used to be much concern about copying overhead in microkernels, but that's less of an issue in modern CPUs. If you're copying material that was just created, as with most message passing, it will be in the cache, and copying will be fast. "Zero copy" systems that play games with the MMU are generally a lose today, because CPU and cache flushing is required to bring main memory into sync before the MMU setup is changed. Unless you're copying gigabytes, that costs more than the data copy.
This isn't a bug. This is a backdoor inserted by someone at Sun.
The article says there is an "undocumented parameter" which allows specifying, on the command line, which run-time system to load. That allows loading arbitrary executable code. It's a built-in backdoor.
The current generation of machining software finally has constructive solid geometry that really works. The software can predict where the surface of the work is, as material is removed from it, and can reliably calculate clearances to the tools. I'm very impressed. This really works for arbitrary convex objects now. I've worked on collision detection enough to understand how hard that is.
Coordinating the multiple axes isn't the hard part. That's just relative transformation matrices, which has been done in computer graphics for many years. (Although the newer robot and machining systems understand some of the machine dynamics, and consider inertia. That's new.) It's the modeling of the surface as it changes that's hard.
This is very expensive software, but it's worth it. You need both HyperMill and either SolidWorks or Inventor. You design the part in SolidWorks or Inventor, then use HyperMill to generate the commands for the CNC machine. Total cost is upwards of $10,000. The CNC machine tool itself is relatively dumb; it's just running previously computed moves. The newer machine tools have software to display the 3D model and the tool, so you can check the planned moves against the actual ones when setting up.
Nobody machines consumer products out of solid blocks of metal except as a demo, of course. It takes hours to machine something that can be made in seconds by stamping or molding. Machine tools are used mostly to make stamping and molding dies, and one-off parts. Also, even in modest volumes, you don't start with plain blocks of metal. You cast or forge a blank and machine off the excess.
What may happen is that most of the people who used credit (not debit) cards demand a chargeback from their bank, EA gets hit with thousands of chargeback fees, and EA's merchant bank kicks them into a higher cost credit card category for excessive chargebacks.
There are Visa procedures for this. This is a chargeback code 82 - "Duplicate Processing". Likely cause: "Electronically submitted the same batch of transactions to the merchant bank more than once". See "The Chargeback Life Cycle", page 71, for an overview.
Generally, if chargebacks exceed 100 chargebacks and 1% of transactions, the chargeback penalty provisions kick in. Thereafter, the merchant is charged $100 per chargeback by the merchant's bank. The merchant is forced into Visa's "High Risk Chargeback Monitoring Program", a $5000 "review fee" is charged to the merchant for the first month, and even higher fees are charged if the problem continues.
Even big merchants have to pay. The banks have to deal individually with each customer to straighten out the mess. They charge the merchant for that.
Incidentally, "No Chargeback" sales receipts are prohibited by Visa rules and will not be enforced by banks.
EA is telling their customers to contact their financial institution before calling EA. It would probably be cheaper for EA if EA dealt with the problems themselves, but their call center may be too small.
Some users are complaining that EA charged them partway through the billing cycle, when they didn't owe EA a payment.
Anyway, EA will be getting a big bill from their bank.
Neither WHOIS information nor IP address block allocation (ARIN's remit) should be private.
Neither businesses nor anonymous web sites are entitled to anonymity in most of the developed world. Europe, in fact, is tougher on this than the US. Europe has the European Privacy Directive, but that's for individuals acting in their private capacity. Businesses come under the European Directive on Electronic Commerce.
1. In addition to other information requirements established by Community law, Member States shall ensure that the service provider shall render easily, directly and permanently accessible to the recipients of the service and competent authorities, at least the following information: (a) the name of the service provider; (b) the geographic address at which the service provider is established; (c) the details of the service provider, including his electronic mail address, which allow him to be contacted rapidly and communicated with in a direct and effective manner;
"Service provider" here means web site owner/operator. So even in an area with strong privacy laws, businesses don't have the right to run anonymous web sites.
California has a similar law for sites that accept credit cards. It's a criminal offense in California to accept credit cards from an anonymous web site.
At SiteTruth, our demo search site, we use this requirement to filter out "bottom-feeder" sites from search results. If it looks commercial, and we can't figure out who owns the site after trying about five different approaches, it's down-rated, and we move this down in search results. This puts teeth into fighting "search engine spam".
Sites can put up phony address info, of course, but that's a felony in many jurisdictions. It's generally treated as fraud, and if it's someone else's address, identity theft. That's a line most "bottom feeders" don't want to cross. Also, much such fraud is reported to sites like PhishTank, so there are red flags to check.
If you want to put up a personal site to express your political opinions, fine. But if it's selling something, it can't be anonymous. Deal with it.
Little Web based games, chat, e-mail, social networking, word processing, image editing, and hundreds of other incredibly popular Web technologies are currently limited by the rendering speed as often as by bandwidth.
Rendering speed is rarely the bottleneck. Slow response from ad networks on ad-heavy pages is a more common bottleneck.
Pixlr, the browser-based image editor, uses Flash to provide a Photoshop-like tool. It will go compute-bound if you use the Clone tool, but most other operations won't hit 100% CPU utilization. Filters take about 100ms or so on modest size images. It's about the same speed as Photoshop itself.
If you need hardware graphics acceleration to implement a word processor or "social networking", something has been very badly designed. Yes, there are a few social networking sites that hit 100% CPU utilization, but that's from terrible implementation. Myspace did a site redesign to clean up their act in that area.
What worries me is overdependence on GPS. There are a small number of GPS satellites, there aren't as many on-orbit spares as there are supposed to be, and there's one central GPS control center. Migration to GPS as the primary air traffic navigational system is risky.
The satellites can survive 14 days of control center downtime, and the newer satellites with "autonav" capability can operate on their own for 180 days. If the USAF launches the ten additional satellites now being built on schedule, the system robustness will increase. But they're not up yet.
I've driven a real vehicle through a remote link. We could run our DARPA Grand Challenge vehicle through a WiFi link. Originally, we tried using a joystick, which worked very badly. Everybody overcontrolled. We had to get a Logitech USB steering wheel and pedals. With that, the vehicle could be driven remotely.
Driving through a game pad is hopeless. Most video game cars on consoles have their CG below ground level (which is physically impossible in the real world) to make them stable when overcontrolled.
The fundamental problem with user review sites is that they don't work unless the number of reviewers is large compared to the number of things reviewed. Yelp needs a statistically significant sample size per thing reviewed to work. Movies, yes. Resorts, probably. Major restaurants, maybe. Local plumbers, no. Joe's Plumbing will be reviewed by Joe, Joe's brother in law, Joe's plumbing supply house rep, and maybe a customer.
That's bad enough, but Yelp sales reps "making you an offer you can't refuse" is worse.
There are small businesses that live in terror of Yelp.
Back in the 1990s, the SF Bay Area had a "rating service" which got into trouble for extortion. This was before the Web got big. They used window stickers in participating businesses, and heavily promoted their ratings. Their sales pitch to businesses was "your competitor has one of our stickers, shouldn't you buy one too?". They were shut down.
This is looking like a "wait for Service Pack 1" or "wait for version 2" situation.
This could be a nice little device in a year, when the software is debugged, the cellular interface is in and works, the cellular networks have provisioned enough capacity to serve video to the thing, there's decent support for business documents, and the price has dropped by 50% or so.
Document work sounds just awful
on
iPad Progress Report
·
· Score: 2, Insightful
Doing anything with documents on an iPad sounds awful. You apparently have to "sync" with a Mac using iTunes. In any business environment, you'd want to talk to some server.
Apparently the iPad is incompatible with Google Docs, although this may just be a bug.
I vaguely remember a trailer from a year or two ago for some Star Wars like movie that looked like a bad game trailer. Did that ever come out, or was it just killed as embarrassing?
This will be popular for game bots. No more patching the code.
I'm waiting for a game 'bot which uses a camera to look at the screen. That would work on consoles, even ones that used HDCP displays. The future of gold farming is
a dimly lit room with hundreds of shelves of game consoles, and cameras pointed at small screens, tied to a farm of 1U servers playing the games.
Slashdot Mirror
Yeah. When I buy a hard drive, I do not want it prefilled with crapware. In fact, there should be a prominent security warning on the package for any media that comes with executable content. Microsoft OSs still tend to execute any executable content that comes within range of the machine.
Did they put an autorun file on the hard drive? I'd regard that as "exceeds authorized access".
The most popular location in the top 10 is Chicago, IL, with three of the largest data centers. 350 E. Cermak is across from Chicago's convention center. Elk Grove Village is west of O'Hare Airport.
Why Chicago? It's a central US location. It doesn't get too hot. Power reliability is good. Transportation access is good. Earthquakes, hurricanes, and floods are rare. There are many big, solid industrial buildings available. It's not as depressed as Detroit or Cleveland.
The gjy is using an engineering sample CPU, he has a motherboard set up for overclocking, and he doesn't have a set of CPU diagnostics. Clueless.
There's real 3D directed by James Cameron, where the movie was shot in 3D. And there's Disney Fake3D(tm), where "3D" is added in post-production. The latter is like watching a multi-layered Flash animation.
Cameron uses depth to add reality, not generate in-your-face "effects". At no time in Avatar does an object come "through the screen". That looks fake, and breaks the mood.
Alice in Wonderland, on the other hand, was shot in 2D, and converted to 3D by In-Three. The 2D images go through a program called "In3gue", which segments the images and adds depth info for each pixel. Once this has been done, the image layers can be manipulated in depth.
Even if the original content was shot in 3D, extensive post-production is required. See this guide to 3D post.
Understand how fake stereoscopic 3D is. In the real world, beyond a few meters, there are no noticeable stereo effects.
An interesting point made by In-Three is that content for kids requires a narrower eye separation. It's well known that objects placed "beyond infinity" by stereoscopy produce much unhappiness in viewers, up to and including eyestrain and stomach distress. Content created for adult human eye separation inflicts this effect on children.
An important problem that hasn't been faced yet is that 3D viewing on smaller home screens viewed at close range requires quite different parallax adjustments than on big distant screens. Merely transferring theatrical 3D to Blu-Ray disks is really going to suck.
solidworks works with vista?!!!
I've seem it take ten minutes to launch on TechShop's machines, but yes.
As someone who uses the Menlo Park TechShop, a few comments.
It's a very broad shop. It's not the only public shop in Silicon Valley. If you want to do electronics, Hacker Dojo has better workstations, where surface-mount work is possible. If you're building furniture, The Sawdust Shop has a better wood shop. But TechShop has both sewing machines and CNC milling machines, a stereolithography machine and a plasma cutter, which you usually don't find under one roof.
There are Silicon Valley companies which buy memberships and send their employees over to use the machine tools. The four big manual Bridgeport mills, the big lathes, and the stereolithography machine are usually being used by pros. TechShop gives classes constantly, but most of the people who use TechShop already have considerable familiarity with tools. They just need access to the bigger machines. It's a good place to learn how to use CNC machine tools. CNC software is quite good today, and TechShop has reasonable midrange CNC design software (Vectrix Cut2D/Cut3D, SolidWorks, etc.) installed on their rather sluggish Windows Vista desktop machines.
There's not much electronics and robotics work. Although TechShop gives Arduno programming classes, and people take them and build the projects, not much electronics gets built there. They have power supplies, meters, scopes, and soldering stations, but they're 1980s technology.
The most popular activity is cutting decorative patterns with the laser cutter. It's easy to do, and two laser cutters are busy doing it almost continuously. Those machines just need line art in CorelDraw; you don't have to learn SolidWorks or Vectrix and do real CNC programming. No one activity dominates, though; there are people building birdhouses and people building rocket engines.
Does Google's measurement include delays from off-site ad servers? That's a big issue. For many sites (including Slashdot), the off-site ad servers are the big bottleneck.
Web site programmers will now have to avoid ad code that delays page loading until the ads come in. I expect to see ad code that measures the response time of the ad server, and if the ad server doesn't respond fast enough, drops the ad and reports the fail to a monitoring site.
Then we'll see sites gaming the system. If Google is using information from their "Google Toolbar" to affect search results, we'll probably see attempts to pump fake data into the Google Toolbar server. Google is going to have to learn the lesson well known to developers of networked games - "never trust the client".
It's hard getting the attention of some vendors. I see vulnerabilities in a slightly different context - hacked web sites hosting phishing pages. We distribute a list of major domains being exploited by active phishing scams. This is obtained by processing PhishTank data, and we do this because we want to reduce the collateral damage from a tough blacklist system. At any given time, there are about 30 to 80 domains on the list.
Some sites get themselves off the list quickly. By now, most of the better free hosting services and short-URL services are automatically checking PhishTank and the APWG blacklist to see when they've been hit. Today, if you run a service where anybody can put up a page that could be used for phishing (i.e. it's not full of your own headers and banners), you need automation to deal with attacks. I've been in contact with the abuse guy at "t35.com", which is a free hosting service. They've recently been hit by a flood of phishing attacks, with several hundred new reports in PhishTank per day. The attacks were coming in faster than the abuse guy could clean them out. They're now gaining on the problem, but haven't squashed it yet. Take-away lesson: automate this.
The ones near the top of the list have been there for a while. Note the dates, which are the date that the oldest phishing report still online and active appeared in PhishTank. Some just need help. Typically, these are small organizations like churches and nonprofits that have had a break-in and were partially taken over by a phishing site. I send them the Anti-Phishing Working Group's "What To Do if your Site Has Been Hacked". Sometimes I give them a phone call. They deserve sympathy.
Then there are the hard cases. These are sites with no visible contact address, or a clueless abuse department. At the moment, Google Sites and Google Spreadsheets are being used for phishing. Google is new to the free hosting business, and the phishers have discovered some tricks that Google can't yet handle. While Google puts a "report abuse" link on their site pages, it's possible to set up a file for downloading on Google Sites, and an HTML page can be served that way, without Google's abuse checking. There's also an exploit of Google Spreadsheets. That one is an example of Habbo Hotel phishing. We've reported these to Google several times, but they haven't been fixed yet.
We've been seeing a new type of attack recently - a phishing operation breaks into a shared hosting server and plants phishing pages on multiple domains on a single server. One of these hit one of the mysterious "*.websitewelcome.com" servers, which has "cloaked domain registration" and no useful default web page. These seem to be associated with "ThePlanet.com", but whether ThePlanet operates them, is providing wholesale hosting, is providing colocation, or is just the upstream connectivity provider is not clear.
Hiding the contact information of a hosting provider is legally unwise. The hosting provider may lose the "safe harbor" protection of the the DMCA. The "safe harbor" provision for "Information Residing on Systems or Networks At Direction of Users" only applies if "the service provider has designated an agent to receive notifications of claimed infringement... by making available through its service, including on its website in a location accessible to the public, and by providing to the Copyright Office, substantially the following information: the name, address, phone number, and electronic mail address of the agent." So when the RIAA or the MPAA come calling, a likely event for a hosting service, they get
The experiments so far indicate that paying students for results improves only the results paid for. Pay for attendance, you get attendance. Pay for grades on quizzes, you get grades on quizzes. End of year scores don't improve much, if at all. And when the money stops, so does the improvement.
That's useful info.
Can this robot hold a position, or return to a position upon surfacing and learning its position? Or is at the mercies of the ocean currents as to where it ends up?
No, it can't. It can adjust its depth; that's all.
Compare the Wave Glider, from Liquid Robotics. This is a privately funded product. It has two parts, a surface "floater" that looks like a surfboard, and a tethered "glider", which hangs below it, about 10m underwater. Wave action on the floater pulls the glider up, and gravity brings it down. Spring-loaded ailerons move the glider forward, powered by the wave motion, and it tows the floater. A rudder on the glider allows steering. The floater has solar panels, a GPS, and an Iridium satellite data link.
The Wave Glider is not only autonomous and self-powered, but can make long trips under control. First they sent one all the way around the Big Island in Hawaii. Worked fine. Then they sent it from Hawaii to California. This took a while; it averages around 1 knot; more in storms, less in calm weather. In storms, the floater is pulled through waves, like a surfboard, and comes out unharmed. They picked it up in Monterey Bay, saw that it was in good condition, and sent it back out again. They parked it in Monterey Bay for a while, circling in a 50 meter circle. Then they sent it back out again on an trip to Alaska and back.
The Wave Glider generally stays within about 50m of its programmed course. The Coast Guard treats it as "floating debris", and it doesn't show lights. If something hits it, it's like running over a surfboard. The control center on shore (a laptop with an Iridium phone) gets ship tracking data, and they guide the Wave Gliders out of the way of large ships.
Harmon already did. QNX used to run well on most desktop machines and some laptops, so you could self-host and develop for QNX on QNX. That's gradually been broken over the last few years. Harmon is an audio electronics company. Owning an OS company was out of their league. At least RIM is in the right industry.
I used QNX extensively from 2002-2005, and my desktop machine, plus all the machines in a robot vehicle, ran QNX. QNX runs many of the high-end robots; BigDog, for example, uses QNX. Several DARPA Grand Challenge vehicles ran QNX. When timing really matters, and the problem is far too complex for PLCs and low-end microcontrollers, you need something like QNX.
It's really sad. QNX as an OS was very good. QNX as a company, especially their marketing operation, managed to anger developers, customers, and employees, resulting in a major brain drain out of the QNX community.
Despite Torvalds, mainstream computing is moving to message passing between separate processes. Gnome, Chrome, MacOS, and WebKit are all multi-process message passing systems. QNX does message passing far better than UNIX or Linux does. Message passing in QNX works like a subroutine call. Under UNIX and Linux, a message pass involves the excess baggage of turning a subroutine call into a write to a socket, with marshaling for the conversion from message to stream, and with several unnecessary trips through the scheduler.
There used to be much concern about copying overhead in microkernels, but that's less of an issue in modern CPUs. If you're copying material that was just created, as with most message passing, it will be in the cache, and copying will be fast. "Zero copy" systems that play games with the MMU are generally a lose today, because CPU and cache flushing is required to bring main memory into sync before the MMU setup is changed. Unless you're copying gigabytes, that costs more than the data copy.
This isn't a bug. This is a backdoor inserted by someone at Sun.
The article says there is an "undocumented parameter" which allows specifying, on the command line, which run-time system to load. That allows loading arbitrary executable code. It's a built-in backdoor.
Very nice. But not that unusual for a modern machine tool. Here's a Matsura mill doing much the same thing. It's the software that's interesting.
The current generation of machining software finally has constructive solid geometry that really works. The software can predict where the surface of the work is, as material is removed from it, and can reliably calculate clearances to the tools. I'm very impressed. This really works for arbitrary convex objects now. I've worked on collision detection enough to understand how hard that is.
Coordinating the multiple axes isn't the hard part. That's just relative transformation matrices, which has been done in computer graphics for many years. (Although the newer robot and machining systems understand some of the machine dynamics, and consider inertia. That's new.) It's the modeling of the surface as it changes that's hard.
This is very expensive software, but it's worth it. You need both HyperMill and either SolidWorks or Inventor. You design the part in SolidWorks or Inventor, then use HyperMill to generate the commands for the CNC machine. Total cost is upwards of $10,000. The CNC machine tool itself is relatively dumb; it's just running previously computed moves. The newer machine tools have software to display the 3D model and the tool, so you can check the planned moves against the actual ones when setting up.
Nobody machines consumer products out of solid blocks of metal except as a demo, of course. It takes hours to machine something that can be made in seconds by stamping or molding. Machine tools are used mostly to make stamping and molding dies, and one-off parts. Also, even in modest volumes, you don't start with plain blocks of metal. You cast or forge a blank and machine off the excess.
What may happen is that most of the people who used credit (not debit) cards demand a chargeback from their bank, EA gets hit with thousands of chargeback fees, and EA's merchant bank kicks them into a higher cost credit card category for excessive chargebacks.
There are Visa procedures for this. This is a chargeback code 82 - "Duplicate Processing". Likely cause: "Electronically submitted the same batch of transactions to the merchant bank more than once". See "The Chargeback Life Cycle", page 71, for an overview.
Generally, if chargebacks exceed 100 chargebacks and 1% of transactions, the chargeback penalty provisions kick in. Thereafter, the merchant is charged $100 per chargeback by the merchant's bank. The merchant is forced into Visa's "High Risk Chargeback Monitoring Program", a $5000 "review fee" is charged to the merchant for the first month, and even higher fees are charged if the problem continues.
Even big merchants have to pay. The banks have to deal individually with each customer to straighten out the mess. They charge the merchant for that.
Incidentally, "No Chargeback" sales receipts are prohibited by Visa rules and will not be enforced by banks.
EA is telling their customers to contact their financial institution before calling EA. It would probably be cheaper for EA if EA dealt with the problems themselves, but their call center may be too small.
Some users are complaining that EA charged them partway through the billing cycle, when they didn't owe EA a payment.
Anyway, EA will be getting a big bill from their bank.
Neither WHOIS information nor IP address block allocation (ARIN's remit) should be private. Neither businesses nor anonymous web sites are entitled to anonymity in most of the developed world. Europe, in fact, is tougher on this than the US. Europe has the European Privacy Directive, but that's for individuals acting in their private capacity. Businesses come under the European Directive on Electronic Commerce.
"Service provider" here means web site owner/operator. So even in an area with strong privacy laws, businesses don't have the right to run anonymous web sites.
California has a similar law for sites that accept credit cards. It's a criminal offense in California to accept credit cards from an anonymous web site.
At SiteTruth, our demo search site, we use this requirement to filter out "bottom-feeder" sites from search results. If it looks commercial, and we can't figure out who owns the site after trying about five different approaches, it's down-rated, and we move this down in search results. This puts teeth into fighting "search engine spam".
Sites can put up phony address info, of course, but that's a felony in many jurisdictions. It's generally treated as fraud, and if it's someone else's address, identity theft. That's a line most "bottom feeders" don't want to cross. Also, much such fraud is reported to sites like PhishTank, so there are red flags to check.
If you want to put up a personal site to express your political opinions, fine. But if it's selling something, it can't be anonymous. Deal with it.
Little Web based games, chat, e-mail, social networking, word processing, image editing, and hundreds of other incredibly popular Web technologies are currently limited by the rendering speed as often as by bandwidth.
Rendering speed is rarely the bottleneck. Slow response from ad networks on ad-heavy pages is a more common bottleneck.
Pixlr, the browser-based image editor, uses Flash to provide a Photoshop-like tool. It will go compute-bound if you use the Clone tool, but most other operations won't hit 100% CPU utilization. Filters take about 100ms or so on modest size images. It's about the same speed as Photoshop itself.
If you want real 3D hardware acceleration in a browser, Shockwave has had it for a decade. Here's a 3D flythrough in Shockwave. There's Quake in Shockwave.
If you need hardware graphics acceleration to implement a word processor or "social networking", something has been very badly designed. Yes, there are a few social networking sites that hit 100% CPU utilization, but that's from terrible implementation. Myspace did a site redesign to clean up their act in that area.
I read that you cannot activate the iPad from Linux.
You can't activate the iPad from the iPad, either. Which is just silly.
What worries me is overdependence on GPS. There are a small number of GPS satellites, there aren't as many on-orbit spares as there are supposed to be, and there's one central GPS control center. Migration to GPS as the primary air traffic navigational system is risky.
The satellites can survive 14 days of control center downtime, and the newer satellites with "autonav" capability can operate on their own for 180 days. If the USAF launches the ten additional satellites now being built on schedule, the system robustness will increase. But they're not up yet.
I've driven a real vehicle through a remote link. We could run our DARPA Grand Challenge vehicle through a WiFi link. Originally, we tried using a joystick, which worked very badly. Everybody overcontrolled. We had to get a Logitech USB steering wheel and pedals. With that, the vehicle could be driven remotely.
Driving through a game pad is hopeless. Most video game cars on consoles have their CG below ground level (which is physically impossible in the real world) to make them stable when overcontrolled.
The fundamental problem with user review sites is that they don't work unless the number of reviewers is large compared to the number of things reviewed. Yelp needs a statistically significant sample size per thing reviewed to work. Movies, yes. Resorts, probably. Major restaurants, maybe. Local plumbers, no. Joe's Plumbing will be reviewed by Joe, Joe's brother in law, Joe's plumbing supply house rep, and maybe a customer.
That's bad enough, but Yelp sales reps "making you an offer you can't refuse" is worse. There are small businesses that live in terror of Yelp.
Back in the 1990s, the SF Bay Area had a "rating service" which got into trouble for extortion. This was before the Web got big. They used window stickers in participating businesses, and heavily promoted their ratings. Their sales pitch to businesses was "your competitor has one of our stickers, shouldn't you buy one too?". They were shut down.
K&R's book on C is wordy. The true classic is the Revised Report on the Algorithmic Language ALGOL-60. In its original typeset form, it is 19 pages.
Languages which need 1000-page books are badly designed.
This is looking like a "wait for Service Pack 1" or "wait for version 2" situation.
This could be a nice little device in a year, when the software is debugged, the cellular interface is in and works, the cellular networks have provisioned enough capacity to serve video to the thing, there's decent support for business documents, and the price has dropped by 50% or so.
Doing anything with documents on an iPad sounds awful. You apparently have to "sync" with a Mac using iTunes. In any business environment, you'd want to talk to some server.
Apparently the iPad is incompatible with Google Docs, although this may just be a bug.
I vaguely remember a trailer from a year or two ago for some Star Wars like movie that looked like a bad game trailer. Did that ever come out, or was it just killed as embarrassing?
This will be popular for game bots. No more patching the code.
I'm waiting for a game 'bot which uses a camera to look at the screen. That would work on consoles, even ones that used HDCP displays. The future of gold farming is a dimly lit room with hundreds of shelves of game consoles, and cameras pointed at small screens, tied to a farm of 1U servers playing the games.