Auto dealer franchise laws reflect a long history of auto manufacturers screwing dealers. Auto dealers were traditionally small businesses with one supplier, which put them firmly under the thumb of the manufacturer. Many dealers still are, although some are big mufti-manufacturer chains.
After looking at the New York and Massachusetts laws, it's not clear that they prohibit a manufacturer from selling entirely through their own stores. What the laws clearly prohibit is a manufacturer competing with its own dealers. If a manufacturer doesn't have any independent dealers, the law probably doesn't apply. The dealers are trying to stretch the law by arguing that the manufacturer is unfairly competing with their dealership, but that may not work.
California prohibits a manufacturer from opening a company store within 10 miles of a dealer, so Tesla has no problem there.
USB mice are standardized self-configuring devices. The handshake when an HID device plugs in tells the host what the device can do. There's no reason that an HID device needs a custom driver.
They're right - who in the US today studies how to set up a production line? Who gets a degree in production engineering? Production engineering is a complex problem. Failure can occur in non-obvious ways. Mistakes can lead to a plant which looks great, produces good product, and costs too much per unit produced. Or a good plant which needs huge capital expenditures and a long shutdown to change to a different product. Apple, in their manufacturing days, made both those mistakes.
I agree. The presentation is all hand-waving with PowerPoint. The claim is that their box uses about half the power and space of the least efficient company data center they used for comparison. However, an existing data center won't have all the latest, highest density drives - the average drive will probably be a few years old. So their average figures will be worse.
What it looks like they did is build a box where you can't replace drives without taking the box out of the rack. That yields higher density, and that's not totally unreasonable. You could operate your disks on the assumption that you replace at the box level, not the drive level. Provide enough installed spares to cover the anticipated life of the box, administer it remotely, and never open it. You'll probably have fewer maintenance-induced failures and you need fewer server monkeys. You need a storage management system like Google's, where drives are assumed to be unreliable and everything is duplicated, usually in a drive physically distant.
So this guy proposes to improve security by replacing web sites with executable applications on the user's machine? What's wrong with this picture?
The author argues that disallowing clicks on transparent objects would break too much. It would break some minor functions on a few pages at Google and Facebook, and Yahoo if anybody still cares. They can fix that; it's bad coding, not something that they needed to do. It would break thousands of annoying popups. Win. If the pixel clicked isn't at least 25% opaque, the image doesn't get the click. This enforces visual fidelity.
(I put that in my image buttons - if you click on the image rectangle, but not on the round button, nothing happens.)
What's impressive about Musk is that he's good at running manufacturing. Space-X designs and builds rockets and spacecraft in their own plants with their own employees. Same for Tesla. That's what impressed Automobile Magazine. The Tesla roadster was, in their opinion, just a Lotus Elise with an electric power plant. But the Tesla sedan is an all-new design and a well-executed one.
Apple is a design house and a marketing operation. The manufacturing is done by low-wage workers at Hon Hai Precision Industries in Shenzhen. Apple used to be a manufacturer, but they had trouble running plants efficiently and gave up.
This is the logic behind a flinch reflex. It's just enough approaching obstacle detection to avoid hitting stuff.
It's good to have in a UAV that has to operate near obstacles. It's not full SLAM, but it doesn't need to be.
casual discussions about a potential acquisition/hire agreement
That's a long way from an actual buyout. About half of big deals that are announced as done fall through. If you're not in a conference room at Wilson Sonsini on Page Mill talking to someone who reports to Zuckerberg, its not serious yet.
A pending patent is mentioned. That's not that big a deal. Anybody who's any good in Silicon Valley has a patent or two. I have six, two of which have produced significant revenue. It's unlikely that broad coverage can be achieved on a scheme for packing disk drives into racks. There's much prior art. You probably have a six month advantage over the competition. If that. Evtron doesn't seem to be actually shipping product. Compare what AmpliStore is actually shipping. The number of disk drives per rack is lower, but there are 40 computers in there, too. A storage farm with small ARM-based CPUs might cut the space needed for the compute power, but that's not exactly an original idea.
Go is primarily intended for writing server-side web applications that need to go fast. This is a huge market and a crucial part of Google's business. But it's a garbage-collected language. It's not suitable for writing the garbage collector itself, or for embedded applications. It's not a language you can use down at the bottom. Go itself is implemented in C, not Go.
Erlang can be used down at the bottom on bare metal, but is not popular outside the telecom area.
The only people who dismiss C++ templates as being in "la-la land" are those who don't understand templates.
That's not the problem. The problem is that C++ templates are a reasonably decent template system being abused as a compile-time programming language. Templates are a set of rewrite rules, and just because a term-rewriting system can be made Turing-complete doesn't mean that's a good way to code. A decade of misplaced effort has gone into making the C++ template system work as a compile-time programming language.
Compile-time programming has a long history, mostly bad. PL/1 had a compile-time language. There was interest in "extensible languages" in the 1970s and 1980s. The usual result was unreadable code. LISP macros weren't too bad; at least the same language was used at compile time and at run time. The UNIX world had a few go-rounds in that direction; UNIX still has m4, used for almost nothing except sendmail config files. C macros are extremely limited, which is probably a good thing.
In the more dynamic languages, this is less of an issue. You can usually do at run time what one struggles to do at compile time for hard-compiled languages.
C++ templates are touted as a language safety improvement. They don't fix the underlying problems in the language; they just wallpaper over them. Then the mold seeps through as raw pointers escape the collection classes.
Javascript is at last a decent object-oriented programming language, but much of the Javascript out there is miserably written by people who have no clue. Much of it is cut and pasted from older bad Javascript, with special cases for different browsers. Even worse are front ends to convert Java or something else into obfuscated Javascript.
C should have died decades ago. The problem is that all of the replacements were worse. Modula tanked because Wirth and DEC botched the marketing. Ada tanked because it was too verbose. All the languages with garbage collection are unsuitable for low-level work. The C++ committee went off into template la-la land and became irrelevant. So we still have buffer overflows, security breaches, and crashes all over the place because the key language of the infrastructure sucks. Treating arrays as pointers was a horrible mistake.
HTML browsers should have required, from the beginning, that the opening and closing brackets balance. Instead, we now have HTML5, with clearly defined semantics for broken HTML. Have you ever seen what has to go into an HTML 5 parser to make that work?
Machine learning is great, but the notation of the field sucks. Most of what's going on is better visualized geometrically.
Microsoft says the future of programming is adding trivial little "apps" to a Microsoft-provided core and being paid peanuts for them. Apple insists they get to monopolize anything worth doing, and others can only develop "apps" in areas Apple can't profit from. Not a good future.
System administration is a blue-collar job, like electricians. But without unions.
Check out The Animation Guild, IATSE Local 839, AFL-CIO. The Animation Guild represents artists and computer graphics workers in Southern California. Computer graphics people at Cartoon Network, Dreamworks, Fox, Hasbro, Marvel, Nicolodeon, Sony, Disney, and Warner are all in that union local.
What do they get for it? Here's a summary of current contracts. First, there's a union wage scale, but it's a minimum. Most workers are paid more than "scale". Second, hours worked and overtime pay are strictly enforced. More than 8 hours per day, overtime pay. More than 40 hours per week, overtime pay. More than 5 days per week, overtime pay. These multiply, so that if you work 14 hours on a Sunday, the hourly rate is huge. Movie projects have "crunches" too, and when they do, the employees get paid a lot of money. This is why production scheduling and budgeting are taken very seriously in Hollywood. So seriously that there are completion bond companies which, if a project gets too far behind, have the authority to fire the director and producer and put in their own people.
The Animation Guild also runs a pension fund. They point out that the Guild has been around longer than all but two animation studios. Hanna-Barbera (Flintstones, Jetsons, etc. and Walter Lanz (Woody Woodpecker), once big names in animation, are long gone; the Animation Guild is still there.
I've run into an IATSE organizer at SIGGRAPH meetings. They've tried to organize the video game industry, but so far, without success. In Redwood City, Electronic Arts and Dreamworks have adjacent buildings. Dreamworks is union; EA is not. The working conditions at EA are much worse.
Google's. Their "anal probe" approach to acquiring customer information doesn't work for in-person purchases. They're trying to roll up the "affinity card" business. Payment is just something they handle to get their hooks into your transaction data.
Unless you have really crappy credit, why would you need more than one credit card? I have one personal credit card, one business credit card, and an ATM card. If you want to borrow money, credit cards are a terrible loan deal.
If they actually contacted you, report that to the FBI. They're probably contacting other people, too. A pattern will emerge.
A useful technical solution that seems not to be used much is to make web site services "fair", rather than first-in, first out. If something has a queue, and you're handling an request from source X, take the next work item from a source other than X. The result is the volume of attacks coming from an individual IP address doesn't matter. Only the number of attacking IP addresses matters. Your real users will still get through, although there will be degradation in proportion to the number of hostile IP addresses.That really should be a feature in Apache.
We use this for a free API service we offer. If you make a request, it may either be satisfied immediately if we have the data available, or the request is queued for processing (this involves examining and rating a web site) and the caller gets a "try again later" status. The processing queue is "fair", so no single source can overwhelm it. (Once we rate a domain, we won't look at it again for 30 days, so our system can't be used to DDOS other web sites.)
We once had a user from an Italian university who was trying to request info on a huge number of web sites. He put over 100,000 requests into the queue, and it didn't hurt performance for other users. After a few days, though, we looked at the logs, and noticed that the requests that returned "try again later" were never being followed up with requests for the actual info. So it was all wasted work. I sent a note to the department chair of the university involved, indicating that we had no objection to their using our service, but that their client program was poorly written and wasn't doing anything useful. The traffic stopped.
Obviously not, because they are going to re-build.
Manhattan barely has to rebuild anything. Building codes are tough there - everything has to be brick, concrete, or steel. Building foundations go down to bedrock. Few Manhattan buildings were damaged by the hurricane. One three-story slum had the front facade collapse; the walls and floors held and no one was injured. One construction crane had its boom broken by the wind, but the safety cables held and it didn't fall. That was about it for Manhattan.
Yes, there was about a half billion gallons of water in subway, railroad, and road tunnels. Was. The MTA has big pumps. They have pumping trains made from old subway cars which they pushed up to the water with small Diesel locomotives. Half the East River tunnels are already pumped out and some lines under the river are operating. Limited subway service between Manhattan and Brooklyn should resume tomorrow.
Power never failed for Manhattan above 34th St, and it's back on now for most of lower Manhattan. Even when flooded, underground power lines can be restored rapidly. That will speed up the remaining pumping work.
With power back on, New York City's gasoline pipeline is running again, and gas stations are reopening.
The areas that are severely damaged are single-family residential frame structures in coastal communities. Some of them are totally wiped out. People in the outer boroughs and the Jersey shore are getting cold and hungry. The first supermarket in Far Rockaway reopens at 11 AM Saturday. In Manhattan, as soon as the infrastructure came back up, the city was ready to go. Not so in the 'burbs.
The idiots who stayed on Fire Island despite a mandatory evacuation order were finally rescued, with great difficulty. The first group of rescuers had to themselves be rescued; they were cut off when water cut all the way across the island. Now the people who built expensive frame houses facing the Atlantic Ocean only a few feet above sea level are whining for Government funding to rebuild.
Insurers, including Medicare, support this. About half of Medicare expenditures occur during the last six months of life. If people can be encouraged to die faster, it will help balance the budget.
It's a useful article on keyboard mechanisms, and it's a good discussion of the tradeoffs between thin keyboards and ergonomics. The history is weak.
There's no mention of key rollover, or "can you push a key before releasing the previous key"? Modern keyboards report a key down and key up event for each key, so rollover can be unlimited. Early keyboards struggled with this. The Selectric, and Teletype machines, were mechanically interlocked against multiple key-presses. Some early keyboards wouldn't handle two keys down at the same time at all.
The feedback issue was a big one. Some keyboards clicked, some had a "clicker" inside to create the illusion that they clicked, and some beeped, an annoyance which has returned with some touch screens.
It's amusing that iPad-like devices have reverted to a 3-row keyboard with multiple shifts. That's where Teletype machines were a century ago. The keyboard layout of an iPad is very similar to that of a 1930s Teletype.
Krantz's book, "Failure is not an Option", covers much of this material, with more information about the people at those consoles and what they did.
Those Philco-Ford console systems showed up in a number of other places, including NORAD HQ in Cheyenne Mountain and the USAF Satellite Control Center in Sunnyvale. Those screens are TV screens on a cable TV system, with a TV tuner in the console. All video generation is elsewhere. Anyone on the system could tune in anyone's screen. Military command and control centers are often set up with that capability even today. It makes coordinated teamwork possible without people having to physically hang around the console where the action is.
Capability Based Security can make our systems secure.
It would help, but you have to be hard-assed about who gets what capability tickets. No, Angry Birds, you can't phone home and share high scores (and user ID info). No, you can't paste data from emails into the banking program.
There are a few things that ought to be done, at least as demos. An EAL 7 BGP server and an EAL 7 DNS server, formally verified down to the machine instructions, would be a good place to start,
Of course it's a mess. iTunes is a music player, a video player, a file manager, a sync program, and a shopping cart. Quicktime is also bundled in there somewhere.
The main function of iTunes is to create a direct connection between Apple and your bank account. So Apple is unlikely to separate the shopping cart function from the other functions.
... GPS software also generates wrong results under acceleration to discourage DIY missile systems...
You got a source for that assertion?
There's an International Trafficking in Arms Regulation which designates as weapons GPS systems "designed for producing navigation results above 60,000 feet altitude and at 1,000 knots velocity or greater" (i.e. for ICBMs) or "Designed or modified for use with unmanned air vehicle systems capable of delivering at least a 500 kg payload to a range of at least 300 km" (i.e. for cruise missiles). The second one is kind of pointless, since there's no way the GPS system can tell. The first one, though, is implemented in most GPS units. High altitude balloon experimenters and rocket guidance system companies have workarounds.
Slashdot Mirror
Auto dealer franchise laws reflect a long history of auto manufacturers screwing dealers. Auto dealers were traditionally small businesses with one supplier, which put them firmly under the thumb of the manufacturer. Many dealers still are, although some are big mufti-manufacturer chains.
After looking at the New York and Massachusetts laws, it's not clear that they prohibit a manufacturer from selling entirely through their own stores. What the laws clearly prohibit is a manufacturer competing with its own dealers. If a manufacturer doesn't have any independent dealers, the law probably doesn't apply. The dealers are trying to stretch the law by arguing that the manufacturer is unfairly competing with their dealership, but that may not work.
California prohibits a manufacturer from opening a company store within 10 miles of a dealer, so Tesla has no problem there.
USB mice are standardized self-configuring devices. The handshake when an HID device plugs in tells the host what the device can do. There's no reason that an HID device needs a custom driver.
They're right - who in the US today studies how to set up a production line? Who gets a degree in production engineering? Production engineering is a complex problem. Failure can occur in non-obvious ways. Mistakes can lead to a plant which looks great, produces good product, and costs too much per unit produced. Or a good plant which needs huge capital expenditures and a long shutdown to change to a different product. Apple, in their manufacturing days, made both those mistakes.
Must be a slow month on the ISS. There's no reason that testing this should actually involve having someone in space operate a robot on the ground.
If this guy had really been a good CIA ops officer, he would have said nothing until he knew who the leaker was.
Stupid idea. Aluminum is easily and profitably recycled. On the other side, there's more biodegradable sludge available than anyone can use.
This product doesn't quite add up.
I agree. The presentation is all hand-waving with PowerPoint. The claim is that their box uses about half the power and space of the least efficient company data center they used for comparison. However, an existing data center won't have all the latest, highest density drives - the average drive will probably be a few years old. So their average figures will be worse.
What it looks like they did is build a box where you can't replace drives without taking the box out of the rack. That yields higher density, and that's not totally unreasonable. You could operate your disks on the assumption that you replace at the box level, not the drive level. Provide enough installed spares to cover the anticipated life of the box, administer it remotely, and never open it. You'll probably have fewer maintenance-induced failures and you need fewer server monkeys. You need a storage management system like Google's, where drives are assumed to be unreliable and everything is duplicated, usually in a drive physically distant.
So this guy proposes to improve security by replacing web sites with executable applications on the user's machine? What's wrong with this picture?
The author argues that disallowing clicks on transparent objects would break too much. It would break some minor functions on a few pages at Google and Facebook, and Yahoo if anybody still cares. They can fix that; it's bad coding, not something that they needed to do. It would break thousands of annoying popups. Win. If the pixel clicked isn't at least 25% opaque, the image doesn't get the click. This enforces visual fidelity.
(I put that in my image buttons - if you click on the image rectangle, but not on the round button, nothing happens.)
What's impressive about Musk is that he's good at running manufacturing. Space-X designs and builds rockets and spacecraft in their own plants with their own employees. Same for Tesla. That's what impressed Automobile Magazine. The Tesla roadster was, in their opinion, just a Lotus Elise with an electric power plant. But the Tesla sedan is an all-new design and a well-executed one.
Apple is a design house and a marketing operation. The manufacturing is done by low-wage workers at Hon Hai Precision Industries in Shenzhen. Apple used to be a manufacturer, but they had trouble running plants efficiently and gave up.
This is the logic behind a flinch reflex. It's just enough approaching obstacle detection to avoid hitting stuff. It's good to have in a UAV that has to operate near obstacles. It's not full SLAM, but it doesn't need to be.
Nice. Now get it into the toy helicopter market.
casual discussions about a potential acquisition/hire agreement
That's a long way from an actual buyout. About half of big deals that are announced as done fall through. If you're not in a conference room at Wilson Sonsini on Page Mill talking to someone who reports to Zuckerberg, its not serious yet.
A pending patent is mentioned. That's not that big a deal. Anybody who's any good in Silicon Valley has a patent or two. I have six, two of which have produced significant revenue. It's unlikely that broad coverage can be achieved on a scheme for packing disk drives into racks. There's much prior art. You probably have a six month advantage over the competition. If that. Evtron doesn't seem to be actually shipping product. Compare what AmpliStore is actually shipping. The number of disk drives per rack is lower, but there are 40 computers in there, too. A storage farm with small ARM-based CPUs might cut the space needed for the compute power, but that's not exactly an original idea.
So take the money if you can get it.
What about the Go language as a replacement to C?
Go is primarily intended for writing server-side web applications that need to go fast. This is a huge market and a crucial part of Google's business. But it's a garbage-collected language. It's not suitable for writing the garbage collector itself, or for embedded applications. It's not a language you can use down at the bottom. Go itself is implemented in C, not Go.
Erlang can be used down at the bottom on bare metal, but is not popular outside the telecom area.
The only people who dismiss C++ templates as being in "la-la land" are those who don't understand templates.
That's not the problem. The problem is that C++ templates are a reasonably decent template system being abused as a compile-time programming language. Templates are a set of rewrite rules, and just because a term-rewriting system can be made Turing-complete doesn't mean that's a good way to code. A decade of misplaced effort has gone into making the C++ template system work as a compile-time programming language.
Compile-time programming has a long history, mostly bad. PL/1 had a compile-time language. There was interest in "extensible languages" in the 1970s and 1980s. The usual result was unreadable code. LISP macros weren't too bad; at least the same language was used at compile time and at run time. The UNIX world had a few go-rounds in that direction; UNIX still has m4, used for almost nothing except sendmail config files. C macros are extremely limited, which is probably a good thing.
In the more dynamic languages, this is less of an issue. You can usually do at run time what one struggles to do at compile time for hard-compiled languages.
C++ templates are touted as a language safety improvement. They don't fix the underlying problems in the language; they just wallpaper over them. Then the mold seeps through as raw pointers escape the collection classes.
Frustrations of being an old programmer:
Javascript is at last a decent object-oriented programming language, but much of the Javascript out there is miserably written by people who have no clue. Much of it is cut and pasted from older bad Javascript, with special cases for different browsers. Even worse are front ends to convert Java or something else into obfuscated Javascript.
C should have died decades ago. The problem is that all of the replacements were worse. Modula tanked because Wirth and DEC botched the marketing. Ada tanked because it was too verbose. All the languages with garbage collection are unsuitable for low-level work. The C++ committee went off into template la-la land and became irrelevant. So we still have buffer overflows, security breaches, and crashes all over the place because the key language of the infrastructure sucks. Treating arrays as pointers was a horrible mistake.
HTML browsers should have required, from the beginning, that the opening and closing brackets balance. Instead, we now have HTML5, with clearly defined semantics for broken HTML. Have you ever seen what has to go into an HTML 5 parser to make that work?
Machine learning is great, but the notation of the field sucks. Most of what's going on is better visualized geometrically.
Microsoft says the future of programming is adding trivial little "apps" to a Microsoft-provided core and being paid peanuts for them. Apple insists they get to monopolize anything worth doing, and others can only develop "apps" in areas Apple can't profit from. Not a good future.
System administration is a blue-collar job, like electricians. But without unions.
Check out The Animation Guild, IATSE Local 839, AFL-CIO. The Animation Guild represents artists and computer graphics workers in Southern California. Computer graphics people at Cartoon Network, Dreamworks, Fox, Hasbro, Marvel, Nicolodeon, Sony, Disney, and Warner are all in that union local.
What do they get for it? Here's a summary of current contracts. First, there's a union wage scale, but it's a minimum. Most workers are paid more than "scale". Second, hours worked and overtime pay are strictly enforced. More than 8 hours per day, overtime pay. More than 40 hours per week, overtime pay. More than 5 days per week, overtime pay. These multiply, so that if you work 14 hours on a Sunday, the hourly rate is huge. Movie projects have "crunches" too, and when they do, the employees get paid a lot of money. This is why production scheduling and budgeting are taken very seriously in Hollywood. So seriously that there are completion bond companies which, if a project gets too far behind, have the authority to fire the director and producer and put in their own people.
The Animation Guild also runs a pension fund. They point out that the Guild has been around longer than all but two animation studios. Hanna-Barbera (Flintstones, Jetsons, etc. and Walter Lanz (Woody Woodpecker), once big names in animation, are long gone; the Animation Guild is still there.
I've run into an IATSE organizer at SIGGRAPH meetings. They've tried to organize the video game industry, but so far, without success. In Redwood City, Electronic Arts and Dreamworks have adjacent buildings. Dreamworks is union; EA is not. The working conditions at EA are much worse.
Whose problem does it solve?
Google's. Their "anal probe" approach to acquiring customer information doesn't work for in-person purchases. They're trying to roll up the "affinity card" business. Payment is just something they handle to get their hooks into your transaction data.
Unless you have really crappy credit, why would you need more than one credit card? I have one personal credit card, one business credit card, and an ATM card. If you want to borrow money, credit cards are a terrible loan deal.
If they actually contacted you, report that to the FBI. They're probably contacting other people, too. A pattern will emerge.
A useful technical solution that seems not to be used much is to make web site services "fair", rather than first-in, first out. If something has a queue, and you're handling an request from source X, take the next work item from a source other than X. The result is the volume of attacks coming from an individual IP address doesn't matter. Only the number of attacking IP addresses matters. Your real users will still get through, although there will be degradation in proportion to the number of hostile IP addresses.That really should be a feature in Apache.
We use this for a free API service we offer. If you make a request, it may either be satisfied immediately if we have the data available, or the request is queued for processing (this involves examining and rating a web site) and the caller gets a "try again later" status. The processing queue is "fair", so no single source can overwhelm it. (Once we rate a domain, we won't look at it again for 30 days, so our system can't be used to DDOS other web sites.)
We once had a user from an Italian university who was trying to request info on a huge number of web sites. He put over 100,000 requests into the queue, and it didn't hurt performance for other users. After a few days, though, we looked at the logs, and noticed that the requests that returned "try again later" were never being followed up with requests for the actual info. So it was all wasted work. I sent a note to the department chair of the university involved, indicating that we had no objection to their using our service, but that their client program was poorly written and wasn't doing anything useful. The traffic stopped.
Obviously not, because they are going to re-build.
Manhattan barely has to rebuild anything. Building codes are tough there - everything has to be brick, concrete, or steel. Building foundations go down to bedrock. Few Manhattan buildings were damaged by the hurricane. One three-story slum had the front facade collapse; the walls and floors held and no one was injured. One construction crane had its boom broken by the wind, but the safety cables held and it didn't fall. That was about it for Manhattan.
Yes, there was about a half billion gallons of water in subway, railroad, and road tunnels. Was. The MTA has big pumps. They have pumping trains made from old subway cars which they pushed up to the water with small Diesel locomotives. Half the East River tunnels are already pumped out and some lines under the river are operating. Limited subway service between Manhattan and Brooklyn should resume tomorrow.
Power never failed for Manhattan above 34th St, and it's back on now for most of lower Manhattan. Even when flooded, underground power lines can be restored rapidly. That will speed up the remaining pumping work. With power back on, New York City's gasoline pipeline is running again, and gas stations are reopening.
The areas that are severely damaged are single-family residential frame structures in coastal communities. Some of them are totally wiped out. People in the outer boroughs and the Jersey shore are getting cold and hungry. The first supermarket in Far Rockaway reopens at 11 AM Saturday. In Manhattan, as soon as the infrastructure came back up, the city was ready to go. Not so in the 'burbs.
The idiots who stayed on Fire Island despite a mandatory evacuation order were finally rescued, with great difficulty. The first group of rescuers had to themselves be rescued; they were cut off when water cut all the way across the island. Now the people who built expensive frame houses facing the Atlantic Ocean only a few feet above sea level are whining for Government funding to rebuild.
Insurers, including Medicare, support this. About half of Medicare expenditures occur during the last six months of life. If people can be encouraged to die faster, it will help balance the budget.
It's a useful article on keyboard mechanisms, and it's a good discussion of the tradeoffs between thin keyboards and ergonomics. The history is weak.
There's no mention of key rollover, or "can you push a key before releasing the previous key"? Modern keyboards report a key down and key up event for each key, so rollover can be unlimited. Early keyboards struggled with this. The Selectric, and Teletype machines, were mechanically interlocked against multiple key-presses. Some early keyboards wouldn't handle two keys down at the same time at all.
The feedback issue was a big one. Some keyboards clicked, some had a "clicker" inside to create the illusion that they clicked, and some beeped, an annoyance which has returned with some touch screens.
It's amusing that iPad-like devices have reverted to a 3-row keyboard with multiple shifts. That's where Teletype machines were a century ago. The keyboard layout of an iPad is very similar to that of a 1930s Teletype.
Even in Palo Alto, the Mercury News reports that reaction was "meh". No big lines. Someone showed up at 2AM expecting a line, and no one was there.
Krantz's book, "Failure is not an Option", covers much of this material, with more information about the people at those consoles and what they did.
Those Philco-Ford console systems showed up in a number of other places, including NORAD HQ in Cheyenne Mountain and the USAF Satellite Control Center in Sunnyvale. Those screens are TV screens on a cable TV system, with a TV tuner in the console. All video generation is elsewhere. Anyone on the system could tune in anyone's screen. Military command and control centers are often set up with that capability even today. It makes coordinated teamwork possible without people having to physically hang around the console where the action is.
Capability Based Security can make our systems secure.
It would help, but you have to be hard-assed about who gets what capability tickets. No, Angry Birds, you can't phone home and share high scores (and user ID info). No, you can't paste data from emails into the banking program.
There are a few things that ought to be done, at least as demos. An EAL 7 BGP server and an EAL 7 DNS server, formally verified down to the machine instructions, would be a good place to start,
Of course it's a mess. iTunes is a music player, a video player, a file manager, a sync program, and a shopping cart. Quicktime is also bundled in there somewhere.
The main function of iTunes is to create a direct connection between Apple and your bank account. So Apple is unlikely to separate the shopping cart function from the other functions.
... GPS software also generates wrong results under acceleration to discourage DIY missile systems ...
You got a source for that assertion?
There's an International Trafficking in Arms Regulation which designates as weapons GPS systems "designed for producing navigation results above 60,000 feet altitude and at 1,000 knots velocity or greater" (i.e. for ICBMs) or "Designed or modified for use with unmanned air vehicle systems capable of delivering at least a 500 kg payload to a range of at least 300 km" (i.e. for cruise missiles). The second one is kind of pointless, since there's no way the GPS system can tell. The first one, though, is implemented in most GPS units. High altitude balloon experimenters and rocket guidance system companies have workarounds.
Yes, it is rocket science.