This sounds like some old waterfall-model guy complaining about a modern "agile" programmer. If you're just doing web crap, quality doesn't matter. Time to market does.
Predator drones use a common small aircraft engine used on about 30 other small aircraft. Similar Rotax engines are used in snowmobiles, jet skis, go-carts, and other small engine applications.
You're going to get an alarm every time something with a small engine goes by, and you probably won't pick up an aircraft flying high enough to not be blatantly annoying.
If you want to detect nearby aircraft, build a radar. There are automotive anti-collision radars that could be adapted.
The gist of it is that you can't write a big, sophisticated interrupt handler; you have to break it up into tiny evented pieces that can return before the hard deadline expires. That has lots of design implications.
You're not supposed to do much in a QNX interrupt handler except activate a thread to handle the event. I've written hot-plugging support for FireWire and USB devices, which is much easier to do when it's done in a user process.
You take about a 10%-15% performance hit for extra context switches and message passing. But it's all constant-time overhead. You don't have all those things that cause monolithic kernels to stall.
This looks like criminal activity under the Computer Fraud and Abuse Act. The "obtains anything of value" clause there seems to apply. When can we expect arrests?
Mozilla needs to focus on their core business. The number of Firefox bugs fixed remains lower than the number being reported, and the internals, which date back to the Netscape era, need replacement.
It is. It's a real-time microkernel based OS. The kernel is about 60K bytes. All it does is manage memory, timers, and message passing. Everything else is in user space. There is a hard upper bound on interrupt lockout time, and it's around a microsecond.
This is what you want to control complex real-time systems that need tight coordination. All the Boston Dynamics robots, BigDog, ATLAS, etc. run QNX. The servo loops are running at 1KHz on those robots. Tight real-time coordination of all those complex motions requires a true real-time OS. (The robots that run ROS/Linux are more sluggish.)
But, after totally botched marketing, the death of the main designer, two sales of the company (to Harmon and then RIM), and transitions from closed source to open source to closed source to open source to closed source again, QNX, the company, has blown it.
None of this has any bearing on smartphone sales. They're not a hard real time problem. You could write the entire UI in HTML/CSS/Javascript and it would work fine on current processors.
Analysis of the block chain indicates that somewhere around 5 million early Bitcoins have never been traded. Those were generated when the difficulty was so low that ordinary CPUs could generate thousands of Bitcoins. Some of them were probably lost by people running the software in the early days and not keeping the results, but there's a reasonable chance that, somewhere, the anonymous people behind this have a few million Bitcoins stored up. Someday they may cash out.
Bitcoins earned: 1.00
Value today: $133.58
Total time: 14.5 days
Rigs operating: 2-3
Towards the end of the run I was getting impatient and ExtremeTechâ(TM)s Joel Hruska helped me sprint to the finish by giving me access to some of his machines...
The era of Bitcoin mining with off-the-shelf hardware is over. The serious people are already using FPGAs, and if the people advertising ASIC hardware actually ship working product in quantity, even that will be obsolete. Like most Bitcoin-related businesses, the people selling ASIC mining hardware are flakes.
Bitcoin mining becomes exponentially harder as the 21 million Bitcoin limit is approached. Over 11 million Bitcoins have already been found, so this is already more than half over.
All miners are in competition. The rate of Bitcoin discovery is fixed, so the more people mining, the less each miner makes.
This has been known for some time. The biggest energy cost associated with many food products is moving the 2-ton family SUV to and from the grocery store to move 25 pounds of merchandise. Moving a fully loaded semi isn't that expensive per unit weight.
Webvan is coming back. Amazon owns it now. Webvan was popular, but the operating costs were too high. One of Webvan's executives realized that what they needed was robots. He went on to found Kiva Systems, which makes robots for warehouse operations. Kiva robots handle fulfillment at Walgreens.com, Gap.com, Staples.com, and many other big retailers.
Amazon recently bought Kiva. Amazon's CEO is an investor in Rethink Robotics, which makes robot arms and hands. (The Kiva robots move shelf units to human pickers, where a laser pointer shows them what to pick. It looks like eliminating the human pickers is next.) Amazon is opening local warehouses in major cities. Amazon is starting to offer same-day delivery. This time it will be profitable.
Sounds like an outsized ego in search of an audience. This guy should try a few comedy clubs in LA. If he gets invited back, maybe he has potential. If not, don't quit your day job.
OpenFlow is basically a way to turn a packet network into a rather dumb virtual circuit network. It works something like Tymnet, circa 1971. In Tymnet, all the virtual circuits were set up by a "supervisor" computer, which told each node where each flow was supposed to be forwarded. The supervisor also handled authentication, but data packets didn't have to pass through the supervisor once the connection was set up.
That's what OpenFlow does, mostly. The first packet of each new "flow" (IP/port/IP/port set, usually) is sent to Master Control, which decides whether that flow will be allowed. Master Control can also choose to monitor the flow. The implications are obvious.
DOCSIS 3, the cable modem traffic control architecture, can potentially do most of the same things, and offers better control over bandwidth. DOCSIS 3 tends to be run more to control users than to maximize throughput, but that's a marketing issue. (If your cable connection is throttling something, the commands to do it were probably sent to a DOCSIS node.) There's good QoS and fair queuing stuff in DOCSIS 3, but it's not always used intelligently. DOCSIS is less intrusive than OpenFlow; the nodes are sent rules to enforce, but there's no need to get permission of Master Control for every new flow.
The rest of "software defined networking" seems to involve adding another layer of indirection to Ethernet addresses so they can be moved around within the data center. ("There is no problem in computer science that cannot be solved by adding another layer of indirection.") That's a reasonable network management tool, but it's not exactly a profound concept.
Was the last century a fluke where a large middle class had power, which will soon revert to the more common system in human history where a tiny few live in splendor and the rest live under their heel?
Probably. When capitalism functions as designed, the price of labor drops to just above survival level. This is the "iron law of wages", and held for most of history. For much of the 20th century, in the developed world, it was different. When productivity went up, so did wages. That was driven by two factors - unions, and fear of communism.
Nobody has taken communism seriously in decades, even the remaining communists. But from the 1930s to the 1970s, it was seen as a serious threat to capitalism. In the 1930s, during the Great Depression, capitalism failed, while communism in the USSR was on the way up. There was real fear that communism might win economically. Fear of nationalization forced companies to increase wages and treat their workers better.
When the USSR started building atomic bombs, space satellites, and ICBMs, there was fear in the US that the USSR might pull ahead in technology. This fear drove the "space race", and is why the US set up NASA and funded the space program so heavily.
This all ended in the 1970s. The best year ever for blue collar workers in the US was 1973. The USSR no longer seemed to be an economic threat. So things gradually went back to normal, and real wages in the US went down for several decades thereafter.
"If you want a vision of the future, imagine a boot stamping on a human face - forever." - Orwell.
The design software isn't the problem. The problem is that the low-end 3D printers suck.
The ultraviolet stereolithography machines work fine, but so far, they cost too much. The Form 1 machine ($2300) is supposed to ship Real Soon Now. That's probably the first low-end machine that will really work.
The low-end plastic extruder approach (MakerBot, RepRap, Up, etc.) is fundamentally flawed. You're trying to weld a hot thing to a cold thing. That never works reliably. Cold solder joints and bad welds are the usual results of trying to do that in other materials. It sort of works for small objects where the previous layer doesn't have time to cool completely. But the time between one layer and the next being laid down has way too much effect on the weld quality.
You need some way to heat the layer below the weld just before the weld, like a laser or a hot air jet. It probably would only take a few watts of laser power aimed at the join. You'd have to enclose and interlock the build area, as with a laser cutter, but that's not hard.
The plastic extruder machines will probably go away once stereolithography gets cheaper. It's a sort-of-works technology. Printers went through this. There was wet electrostatic printing (Versatec), magnetic printing, ink jet printing by electrostatic deflection of a stream of ink drops, electrolytic printing (dates from the 19th century), and spark printing. Commercial products using all those technologies were manufactured and sold, but xerographic and ink jet technologies were just better.
We need full camera coverage at all hazardous industrial sites. The fertizer plant that blew up in Texas had 270 tons of ammonium nitrate on hand, but were only authorized one ton.
They acquired an unauthorized weapon of mass destruction which demolished most of the town. The owners should be punished accordingly.
Hazardous area cameras should be monitored by OSHA, Homeland Security, the Oil, Chemical and Atomic Workers Union, and Underwriters Laboratories. That would keep everyone honest.
Schmidt has been flogging this book all over the place. I had to drop a Facebook group to get rid of all the promotion for this.
From the excerpts I've read, this vision of the future is rather banal. It's a 1950s middle-management view of the future. Better collaborative PowerPoint-type presentations in the "cloud", better meeting and travel scheduling, stuff like that. In their future, people still wear suits, have meetings, and go to work, but in self-driving cars. It sounds like something AT&T and GM would have presented at the 1964 World's Fair.
Nope, first time in my life I ever heard that name.
Me too, and I make significant money as an inventor.
Looking at their website, it seems similar to Kickstarter.
No, it's much, much worse. The terms are awful. "While one aspect of the Site is to provide Users, including Idea Submitter's with the right to share in Product Sales Revenue, Users should not participate on the Site or in the Services primarily for financial gain, as any actual monetary compensation actually received by a User in connection with the Site, may or may not bear any relation to the actual time invested on the Site, or the quality or quantity of User Content submitted, or Influence allocated." They own your submitted idea even if they don't use it. If they don't use it, you can buy it back from them for a fee, plus a cut of future revenue.
There are resources for inventors. The USPTO has a page. There are startup incubators. There's Kickstarter. And then there's a bunch of slimeball "invention brokers", with success rates somewhere below 0.1%. These clowns are in that category.
I used all that gear in high school and high school summer jobs. I've wired panels for an IBM 402, an IBM 407 (the last of the electromechanical accounting machines and the best one), a 514/519 reproducer (a 519 has a mark sense reader option), and the 77 and 84 collators. And, of course, card sorters and punches. I was able to draw graphs with a 402 and generate poetry with an 84 collator. This is pushing the limits of those machines.
The normal processing cycle for a sales/billing operation looks like this:
Transactions are punched on cards by a keypunch operator in a fixed format, with customer number, item number, quantity, and price each. Date is automatically punched, copied from the previous card.
Payments are also punched on cards in a similar way.
There's another deck of cards with three lines of address info on 3 cards, kept by customer number and address line number.
Finally, there's a deck of cards with the previous month's balance for each customer.
End of month processing begins by running the transaction cards through a 602A Calculating Punch, which can multiply. It multiplies quantity by price each and punches that info into the same card.
All the transaction cards are then sorted by customer number and date.
The decks are merged together with a collator, which has two input hoppers and four output hoppers. This matches numbers and checks sequence between cards. The assembled deck has, for each customer, three address cards, a previous-balance card, and all the transaction cards, payments first. Each block of cards for one customer thus contains the data for one invoice. The collator kicks out anything that doesn't match and stops if a deck is out of sequence.
The merged deck then goes to the tabulator, which is filled with preprinted invoice forms, usually multi-part carbons. The tabulator can add, subtract, and print, but not multiply. The invoices are printed. For this operation, a reproducer (a big card punch used to copy card decks) is cabled up to the tabulator with a cable about 2 inches in diameter. At the end of each invoice, the reproducer is triggered to punch a new previous-balance card for that customer, which will be used in the next billing cycle.
After a successful tabulator run, the merged deck is sorted in one pass to separate the name and address cards, which will be used again next month. The transaction cards go into storage as backup. The previous-balance deck is stored for use next month.
The fan-fold invoices go through a decollator to pull the carbons apart, and a burster to separate the pages for the copy that gets mailed. Then there's folding, inserting into envelopes, a pass through a postage meter, and mailing.
Other reports can be produced from the same cards. The previous-balance deck and the address deck are used to produce reports such as who owes how much, and which customers are buying the most.
The card operations aren't that bad. All this stuff is slow, but automatic. The data entry is the labor-intensive part of the operation.
I thought the Computer History Museum got that IBM 402. There's one in the Computer History Museum now. They may have the machine the company was using for parts.
Major security breach. This is a device with unknown software, a WiFi radio, a cellular radio, a battery, and a processor, shipped covertly to offices. That provides a potential backdoor into internal WiFi networks. What could possibly go wrong?
Watch the IBM video. The robot is the supervisor. The robot tells the human what to do. Sometimes there's someone remotely controlling the robot; sometimes it's following canned instructions. The human is there to do the manual labor.
Wasn't that exactly what XHTML was supposed to become?
Sort of. In XHTML you can add new tags. But the web development community can't handle XHTML. Open and close tags must match, and they can't do that consistently.
(Real-world HTML is incredibly bad. Most major sites still won't go through an HTML validator. Even superficial parsing is hard. Incorrect comment delimiters are all over the place. HTML5 has well-defined parsing of bogus HTML, so now all major browsers interpret bad HTML the same way. Just trying to figure out what character set a page is really in requires a huge amount of machinery.
The W3C should have specified that browsers, upon detecting a syntax error in HTML, should put an error message bar in the document, and from that point forward, render in a "basic default mode", where all the content is displayed in the default font, color, and size. That would have cleaned up the web.)
If only they lasted. We've had LED traffic lights in areas of Silicon Valley for most of a decade, and many of them are failing, one section of LEDs at a time. There's a huge bin of partially failed traffic light round LED panels at Weird Stuff Warehouse in Sunnyvale. Mostly red and green; the yellows aren't on as much.
The problem is outright failure, not dimming. That's an indication of a manufacturing problem, like a joint that fails under thermal cycling. Many pre-2004 LEDs are prone to this problem
If the company repurchases its stock, the shares go away, thereby increasing the value of the remaining shares.
More significantly, it increases the value of stock options. Dividends don't do anything for option holders. (They're also taxed at a higher rate than the capital gain after a stock buyback. Buybacks and taking on debt are actions successful companies take primarily for tax reasons.)
Slashdot Mirror
This sounds like some old waterfall-model guy complaining about a modern "agile" programmer. If you're just doing web crap, quality doesn't matter. Time to market does.
Predator drones use a common small aircraft engine used on about 30 other small aircraft. Similar Rotax engines are used in snowmobiles, jet skis, go-carts, and other small engine applications.
You're going to get an alarm every time something with a small engine goes by, and you probably won't pick up an aircraft flying high enough to not be blatantly annoying.
If you want to detect nearby aircraft, build a radar. There are automotive anti-collision radars that could be adapted.
The gist of it is that you can't write a big, sophisticated interrupt handler; you have to break it up into tiny evented pieces that can return before the hard deadline expires. That has lots of design implications.
You're not supposed to do much in a QNX interrupt handler except activate a thread to handle the event. I've written hot-plugging support for FireWire and USB devices, which is much easier to do when it's done in a user process.
You take about a 10%-15% performance hit for extra context switches and message passing. But it's all constant-time overhead. You don't have all those things that cause monolithic kernels to stall.
This looks like criminal activity under the Computer Fraud and Abuse Act. The "obtains anything of value" clause there seems to apply. When can we expect arrests?
You could write the entire UI in HTML/CSS/Javascript and it would work fine on current processors.
Just after I wrote that, it's demoed.
Another layer on top of Android on top of Linux.
Mozilla needs to focus on their core business. The number of Firefox bugs fixed remains lower than the number being reported, and the internals, which date back to the Netscape era, need replacement.
First of all, QNX is awesome.
It is. It's a real-time microkernel based OS. The kernel is about 60K bytes. All it does is manage memory, timers, and message passing. Everything else is in user space. There is a hard upper bound on interrupt lockout time, and it's around a microsecond.
This is what you want to control complex real-time systems that need tight coordination. All the Boston Dynamics robots, BigDog, ATLAS, etc. run QNX. The servo loops are running at 1KHz on those robots. Tight real-time coordination of all those complex motions requires a true real-time OS. (The robots that run ROS/Linux are more sluggish.)
But, after totally botched marketing, the death of the main designer, two sales of the company (to Harmon and then RIM), and transitions from closed source to open source to closed source to open source to closed source again, QNX, the company, has blown it.
None of this has any bearing on smartphone sales. They're not a hard real time problem. You could write the entire UI in HTML/CSS/Javascript and it would work fine on current processors.
A founder at the top making a mint...
Analysis of the block chain indicates that somewhere around 5 million early Bitcoins have never been traded. Those were generated when the difficulty was so low that ordinary CPUs could generate thousands of Bitcoins. Some of them were probably lost by people running the software in the early days and not keeping the results, but there's a reasonable chance that, somewhere, the anonymous people behind this have a few million Bitcoins stored up. Someday they may cash out.
Bitcoins earned: 1.00
Value today: $133.58
Total time: 14.5 days
Rigs operating: 2-3
Towards the end of the run I was getting impatient and ExtremeTechâ(TM)s Joel Hruska helped me sprint to the finish by giving me access to some of his machines...
The era of Bitcoin mining with off-the-shelf hardware is over. The serious people are already using FPGAs, and if the people advertising ASIC hardware actually ship working product in quantity, even that will be obsolete. Like most Bitcoin-related businesses, the people selling ASIC mining hardware are flakes.
Bitcoin mining becomes exponentially harder as the 21 million Bitcoin limit is approached. Over 11 million Bitcoins have already been found, so this is already more than half over. All miners are in competition. The rate of Bitcoin discovery is fixed, so the more people mining, the less each miner makes.
This has been known for some time. The biggest energy cost associated with many food products is moving the 2-ton family SUV to and from the grocery store to move 25 pounds of merchandise. Moving a fully loaded semi isn't that expensive per unit weight.
Webvan is coming back. Amazon owns it now. Webvan was popular, but the operating costs were too high. One of Webvan's executives realized that what they needed was robots. He went on to found Kiva Systems, which makes robots for warehouse operations. Kiva robots handle fulfillment at Walgreens.com, Gap.com, Staples.com, and many other big retailers.
Amazon recently bought Kiva. Amazon's CEO is an investor in Rethink Robotics, which makes robot arms and hands. (The Kiva robots move shelf units to human pickers, where a laser pointer shows them what to pick. It looks like eliminating the human pickers is next.) Amazon is opening local warehouses in major cities. Amazon is starting to offer same-day delivery. This time it will be profitable.
Small retailers who are aware of this are very afraid.
Sounds like an outsized ego in search of an audience. This guy should try a few comedy clubs in LA. If he gets invited back, maybe he has potential. If not, don't quit your day job.
OpenFlow is basically a way to turn a packet network into a rather dumb virtual circuit network. It works something like Tymnet, circa 1971. In Tymnet, all the virtual circuits were set up by a "supervisor" computer, which told each node where each flow was supposed to be forwarded. The supervisor also handled authentication, but data packets didn't have to pass through the supervisor once the connection was set up. That's what OpenFlow does, mostly. The first packet of each new "flow" (IP/port/IP/port set, usually) is sent to Master Control, which decides whether that flow will be allowed. Master Control can also choose to monitor the flow. The implications are obvious.
DOCSIS 3, the cable modem traffic control architecture, can potentially do most of the same things, and offers better control over bandwidth. DOCSIS 3 tends to be run more to control users than to maximize throughput, but that's a marketing issue. (If your cable connection is throttling something, the commands to do it were probably sent to a DOCSIS node.) There's good QoS and fair queuing stuff in DOCSIS 3, but it's not always used intelligently. DOCSIS is less intrusive than OpenFlow; the nodes are sent rules to enforce, but there's no need to get permission of Master Control for every new flow.
The rest of "software defined networking" seems to involve adding another layer of indirection to Ethernet addresses so they can be moved around within the data center. ("There is no problem in computer science that cannot be solved by adding another layer of indirection.") That's a reasonable network management tool, but it's not exactly a profound concept.
LinkedIn used to be a job board for consultants. I'm not sure what it is now. I haven't logged in for months.
Is changing the features of a web site three times a day actually useful?
Was the last century a fluke where a large middle class had power, which will soon revert to the more common system in human history where a tiny few live in splendor and the rest live under their heel?
Probably. When capitalism functions as designed, the price of labor drops to just above survival level. This is the "iron law of wages", and held for most of history. For much of the 20th century, in the developed world, it was different. When productivity went up, so did wages. That was driven by two factors - unions, and fear of communism.
Nobody has taken communism seriously in decades, even the remaining communists. But from the 1930s to the 1970s, it was seen as a serious threat to capitalism. In the 1930s, during the Great Depression, capitalism failed, while communism in the USSR was on the way up. There was real fear that communism might win economically. Fear of nationalization forced companies to increase wages and treat their workers better.
When the USSR started building atomic bombs, space satellites, and ICBMs, there was fear in the US that the USSR might pull ahead in technology. This fear drove the "space race", and is why the US set up NASA and funded the space program so heavily.
This all ended in the 1970s. The best year ever for blue collar workers in the US was 1973. The USSR no longer seemed to be an economic threat. So things gradually went back to normal, and real wages in the US went down for several decades thereafter.
"If you want a vision of the future, imagine a boot stamping on a human face - forever." - Orwell.
The design software isn't the problem. The problem is that the low-end 3D printers suck.
The ultraviolet stereolithography machines work fine, but so far, they cost too much. The Form 1 machine ($2300) is supposed to ship Real Soon Now. That's probably the first low-end machine that will really work.
The low-end plastic extruder approach (MakerBot, RepRap, Up, etc.) is fundamentally flawed. You're trying to weld a hot thing to a cold thing. That never works reliably. Cold solder joints and bad welds are the usual results of trying to do that in other materials. It sort of works for small objects where the previous layer doesn't have time to cool completely. But the time between one layer and the next being laid down has way too much effect on the weld quality. You need some way to heat the layer below the weld just before the weld, like a laser or a hot air jet. It probably would only take a few watts of laser power aimed at the join. You'd have to enclose and interlock the build area, as with a laser cutter, but that's not hard.
The plastic extruder machines will probably go away once stereolithography gets cheaper. It's a sort-of-works technology. Printers went through this. There was wet electrostatic printing (Versatec), magnetic printing, ink jet printing by electrostatic deflection of a stream of ink drops, electrolytic printing (dates from the 19th century), and spark printing. Commercial products using all those technologies were manufactured and sold, but xerographic and ink jet technologies were just better.
We need full camera coverage at all hazardous industrial sites. The fertizer plant that blew up in Texas had 270 tons of ammonium nitrate on hand, but were only authorized one ton. They acquired an unauthorized weapon of mass destruction which demolished most of the town. The owners should be punished accordingly.
Hazardous area cameras should be monitored by OSHA, Homeland Security, the Oil, Chemical and Atomic Workers Union, and Underwriters Laboratories. That would keep everyone honest.
Schmidt has been flogging this book all over the place. I had to drop a Facebook group to get rid of all the promotion for this.
From the excerpts I've read, this vision of the future is rather banal. It's a 1950s middle-management view of the future. Better collaborative PowerPoint-type presentations in the "cloud", better meeting and travel scheduling, stuff like that. In their future, people still wear suits, have meetings, and go to work, but in self-driving cars. It sounds like something AT&T and GM would have presented at the 1964 World's Fair.
Nope, first time in my life I ever heard that name.
Me too, and I make significant money as an inventor.
Looking at their website, it seems similar to Kickstarter.
No, it's much, much worse. The terms are awful. "While one aspect of the Site is to provide Users, including Idea Submitter's with the right to share in Product Sales Revenue, Users should not participate on the Site or in the Services primarily for financial gain, as any actual monetary compensation actually received by a User in connection with the Site, may or may not bear any relation to the actual time invested on the Site, or the quality or quantity of User Content submitted, or Influence allocated." They own your submitted idea even if they don't use it. If they don't use it, you can buy it back from them for a fee, plus a cut of future revenue.
There are resources for inventors. The USPTO has a page. There are startup incubators. There's Kickstarter. And then there's a bunch of slimeball "invention brokers", with success rates somewhere below 0.1%. These clowns are in that category.
I used all that gear in high school and high school summer jobs. I've wired panels for an IBM 402, an IBM 407 (the last of the electromechanical accounting machines and the best one), a 514/519 reproducer (a 519 has a mark sense reader option), and the 77 and 84 collators. And, of course, card sorters and punches. I was able to draw graphs with a 402 and generate poetry with an 84 collator. This is pushing the limits of those machines.
The normal processing cycle for a sales/billing operation looks like this:
The card operations aren't that bad. All this stuff is slow, but automatic. The data entry is the labor-intensive part of the operation.
I thought the Computer History Museum got that IBM 402. There's one in the Computer History Museum now. They may have the machine the company was using for parts.
Here it is running in Conroe, TX in 2011. (Terrible video, though)
Major security breach. This is a device with unknown software, a WiFi radio, a cellular radio, a battery, and a processor, shipped covertly to offices. That provides a potential backdoor into internal WiFi networks. What could possibly go wrong?
Watch the IBM video. The robot is the supervisor. The robot tells the human what to do. Sometimes there's someone remotely controlling the robot; sometimes it's following canned instructions. The human is there to do the manual labor.
This is the future.
Machines should think. People should work.
Wasn't that exactly what XHTML was supposed to become?
Sort of. In XHTML you can add new tags. But the web development community can't handle XHTML. Open and close tags must match, and they can't do that consistently.
(Real-world HTML is incredibly bad. Most major sites still won't go through an HTML validator. Even superficial parsing is hard. Incorrect comment delimiters are all over the place. HTML5 has well-defined parsing of bogus HTML, so now all major browsers interpret bad HTML the same way. Just trying to figure out what character set a page is really in requires a huge amount of machinery.
The W3C should have specified that browsers, upon detecting a syntax error in HTML, should put an error message bar in the document, and from that point forward, render in a "basic default mode", where all the content is displayed in the default font, color, and size. That would have cleaned up the web.)
If only they lasted. We've had LED traffic lights in areas of Silicon Valley for most of a decade, and many of them are failing, one section of LEDs at a time. There's a huge bin of partially failed traffic light round LED panels at Weird Stuff Warehouse in Sunnyvale. Mostly red and green; the yellows aren't on as much.
The problem is outright failure, not dimming. That's an indication of a manufacturing problem, like a joint that fails under thermal cycling. Many pre-2004 LEDs are prone to this problem
If the company repurchases its stock, the shares go away, thereby increasing the value of the remaining shares.
More significantly, it increases the value of stock options. Dividends don't do anything for option holders. (They're also taxed at a higher rate than the capital gain after a stock buyback. Buybacks and taking on debt are actions successful companies take primarily for tax reasons.)