Rights are worthless if you are unable to exercise them. Rights are only as strong as their enforcement. If someone violates your rights, who will enforce penalties against them? Who will set up incentives to prevent others from doing so? In most civilised countries, we delegate responsibility for this to an organisation that we call a government. It doesn't give you the rights, but it does try to stop anyone else from taking them away. The government is accountable to the people and the paragraph that you quote attempts, in the USA, to prevent the government from attempting to take away certain rights.
If someone wants to have fun, they should set up an automated system that allows people to send Facebook a right to be forgotten notice under the GDPR and provides a record of who sent it. They might not delete the data, but they're liable for very large fines if the EU discovers that they didn't comply.
(I hear in Britain the correct strategy is to claim the homeowner shot a burglar, then their version of SWAT will show up. Might've just been a joke, though.)
The British equivalent of SWAT is Specialist Firearms Command (usually referred to as SO19 in films and TV). They train specialist firearms officers, who receive extra training and must pass a battery of psychological tests before they are even accepted for training. There is an automatic investigation in the case of any firearm discharge by one of these officers.
SFC is almost never the first response. An unarmed officer will attempt to judge whether they need to be deployed. When they are deployed, they have been trained to avoid firing unless all other options have been exhausted. This training sometimes fails, but it seems to work a lot better than the US model of arming all of the police and giving them appallingly bad training.
Some do. OpenStreetMap was instrumental in persuading the UK government to include some of the government-maintained maps under an open license. Historically, high-quality mapping data has been regarded as a military asset, because if you want to invade a country then having decent maps is essential. That's less important in an age of satellite photography (any foreign power that could plan in invasion almost certainly has satellites that can provide them with accurate mapping data) and so there's a slow shift to seeing it as a commercial asset.
You seem to be implying that Google Maps has consistently high quality data. My house doesn't exist on Google Maps (it does, and is correctly numbered on OSM) and the street that I live on is now half there, as of a couple of months ago, on Google Maps, in spite of the part that they've documented being finished almost three years ago and the half that they haven't being finished over a year ago.
A lot of the things that use Google Maps (or, increasingly, Bing Maps) do so via the APIs as well. If you go to a hotel web site, for example, odds are that they'll have either a Google or Bing map embedded in their 'how to find us' page. It is fairly easy to do the same with OSM, but there is a quite large selection of web front ends and there's no real attempt to curate them and many are poorly documented (it took me ages, for example, to realise that the JavaScript OSM overlay library that I used for a conference was putting my waypoints in the wrong place because of a poorly documented projection option having a default that wasn't the same as the other tool that I was using). This kind of thing prevents OSM being the default choice for embedding elsewhere, which means that it doesn't get the kind of visibility that Google Maps provides.
Clang supports all of the features of Objective-C on all *NIX platforms and Windows. The GNUstep Objective-C runtime, in combination with clang, supports a superset of the language features of Apple's implementation and is used by Microsoft in their Windows bridge for iOS. You might be surprised at some of the places Objective-C ends up - there are quite a few larger (mostly non-Web, though there are also things like SOGo) server systems that use Objective-C - I've consulted for a couple of companies that maintain them.
Apple really likes to play up Objective-C as the competitor to Swift, but the elephant in the room is Objective-C++, which (since C++11) is a far better language than either. You can have the late binding of Objective-C objects and the compile-time specialisation of C++ templates and switch between them easily. With ARC, you can store Objective-C object pointers inside C++ collections and have the memory management work properly (which was the biggest irritation of pre-ARC Objective-C++). If you use C++ objects as instance variables in Objective-C, then their destructors run automatically after -dealloc, so the memory management works both ways (you can, for example, use C++11 unique_ptr or shared_ptr to manage ownership of non-Objective-C resources by Objective-C++ objects). Most importantly, it has a trivial way of using C++ libraries, something that Swift can achieve only if you wrap them in [Objective-]C.
That was different. The Althon and Opteron were both better designs than Intel, but were made using an older process technology. The first Opterons were made on a 130nm process in 2003, moving to 90nm in 2005. The Pentium 4 was using 130nm 2002 and moved to 90nm in 2004 - Intel had a clear one-year lead on process technology. AMD only kept up by having much better processor designs.
On the desktop, power usage isn't nearly so important - it's plugged in. Your budget isn't a power budget on the desktop, it's a dollar budget
At a rough approximation, 1W for one year costs me £1. A 20W power difference over the lifetime of a device is £20/year if it's always on. Now that a 3-year upgrade cycle has extended to a 5-year cycle for most people, a 20W power increase is a £100 difference to TCO. For a corporate budget, it's worse because it's a £20/year operating expense rather than a £100 capital write-down.
NetBurst wasn't a failure of microarchitecture, it was a failure of process technology. NetBurst was designed on the assumption that Intel's fab people would be able to produce 10GHz versions in 2005, in the same power budget as the 2GHz version in 2000. This failed dismally and was one of the first indications that Intel's dominance over fabrication techniques was not going to last forever. If Intel in 2005 had been able to produce the process technology that they were forecasting in 1996, when the NetBurst design was started, then AMD would be long dead. They had absolutely nothing that could have competed with a 10GHz Pentium 4. Unfortunately for Intel, AMD had quite a lot that could compete with a 3GHz Pentium 4.
Cyrix's performance failure was really one of marketing. I had a 6x86 P166+. It was a 133MHz core, with Cyrix's stupid 'P-rating' thing claiming it was faster than a Pentium 166. It actually was faster than a 133MHz Pentium for anything I benchmarked, but quite a bit slower than a P166. It was also a lot cheaper than a P133. If they'd marketed it against the P133, it would have been quite compelling (faster and cheaper), and they wouldn't have got such a reputation for exaggerating their performance.
While their market share dropped over the years as people discovered that AMD CPUs were faster, more reliable, ran cooler and cost less all at the same time
I'll accept faster and cooler, but for a long time Intel was shipping thermal throttling and AMD wasn't. In some cases, this was a problem for Intel. We had an Opteron cluster and a P4 Xeon cluster. When the cooling failed in a node on the AMD cluster, the CPU burned and the node died. The cluster management system noted the failure and redistributed the work. With the P4 cluster, the CPU would clock down to something tiny like 200MHz. It would still be responsive to command messages and so work wouldn't be redistributed, but jobs would take ten times as long to complete (and the tail latency of the slowest node would cause the entire job to slow down).
Lower power, as I recall, was only really true for desktop and up. The Pentium M, once it shipped, was lower power than anything AMD had to offer and gave Intel some breathing room while the laptop was the fastest-growing market segment and AMD's laptop offerings were lacklustre. The Opteron was wiping the floor with anything Intel had to offer in the server space, largely because of HyperTransport (basically the Alpha CPU interconnect) on multi-socket systems and moving the memory controller on die, while Intel chips were stuck with the memory controller on a separate north bridge limiting memory bandwidth (I seem to recall that some Xeons were actually faster than AMD cores if you had a workload that fitted in cache). The Althon64 then came and outperformed anything Intel had on the desktop for even approximately the same price. The only advantage the mobile versions had over the Pentium M was 64-bit support, and no one wanted 64-bit Windows on mobile (and the market for non-Windows laptops was tiny).
Developing their own processor will be ridiculously expensive. Even their i devices are mostly ARM design.
Apple's in-house cores use an ARM ISA, but they are categorically not an ARM design. They are a complete reimplementation and currently outperform all of their competitors. In the same power envelope, they also outperform Intel. Given that they're already spending the money designing these, it's not too crazy to imagine that they'd switch to them.
Also, you think they can do better at securing cpus than Intel and amd - both of which have larger market shares?
Yes. Apple's secure element is not vulnerable to any of the speculation-based attacks against SGX enclaves.
You missed an important one. The NT in Windows NT originally referred to the Intel i860's code name (N-Ten). The NT was developed on the i860 and then ported to x86, to ensure that the codebase didn't include any x86isms.
It's a bit more subtle than that. Since about 2007, we've not been seeing Dennard scaling, so although transistor counts have gone up, the number that you can power at any given time has not increased by nearly as much. This is what people mean when they talk about 'dark silicon'. The problem with the x86 decoder is that it needs to be powered most of the time, because any time you're executing instructions that are not in the trace cache, it's necessary. The same is true for caches (though you can turn off some of the ways in a set-associative cache when you're seeing that they're not used - not sure if anyone does this in commercial processors) and the register rename engine. All of the execution pipelines can be power gated so you can turn some of them off depending on the instruction mix (for example, powering down the AVX pipes in scalar workloads).
It doesn't matter much to Intel at the high end, because it isn't a huge part of the total power consumption, but it is a problem at the low end when reducing the power consumption of everything else makes it start to dominate again.
What if Google has most of the market because they make a better product for less money?
That depends. Are they making it for less money because they're subsidising it with income from other products? If so, it may count as dumping and so is illegal (for good reason: it means that no company that doesn't have an independent revenue stream can compete).
How much market share do the Fire phones have? Is it enough that Google isn't an effective monopoly? How many apps are not available via the Amazon store that are available via Google Play?
Manufacturers aren't prohibited from installing their own software, including software which serves the same functions as Google's options
Except that if you want to install any Google apps, then you must install the entire suite and if you want to install competing apps then there are different licensing terms that cost more. And if you don't then it's impossible for your customers to install most third-party software because Google has managed to achieve an effective monopoly on distribution of most Android apps. And if you do install the Play store then you also need Play services, which run with insane permissions and hook into almost every app installed from Play.
I doubt you can get it very far in the time that it takes to send a security guard or the police out. The thing is several tons and you can't just drag it through the water, you need to lift it up onto a boat, during which point you'll trigger all sorts of alarms (vibration, motion, shock).
I wondered about this. If the heat were spread throughout the ocean, it would be a negligible impact, but raising the temperature in a bay by one degree can have quite significant ecological impacts. It depends a lot on how much tidal flow there is and how much heat the thing is dissipating.
Google stopped doing that over a decade ago and I presume Azure has a similar policy. A disk fails, they offline the machine. Enough machines in a rack fail, and they replace the rack. This datacentre is specifically designed not to be accessible by humans (the gas inside has had oxygen and water vapour removed, to reduce the damage that these do to components) and is operating under the same model: you keep using it until enough hardware is dead that it's worth pulling it up and doing a complete refit.
It looks pretty heavy. It's also not very far out to sea, so I doubt you could steal it without anyone seeing and it wouldn't take MS very long to notice an entire datacentre dropping off the Internet and sending someone out to take a look...
I've read a few articles over the last couple of months that indicate that millennials are saving more, inflation adjusted, than their parents were at the same age, but they're still less able to buy houses.
I did the calculation here for what we pay postdocs a couple of years back. The exact numbers don't matter too much: Wage increases are n% a year, house prices are increasing by m% per year. As long as m is greater than n, eventually you will reach the point where no matter what percentage of your income you save, you will never be able to afford a house. With some fairly conservative estimates of n and m, I found that within a decade the increase on the deposit required to buy a house each year will be more than a postdoc makes in a year, before tax. Postdocs make around 2-3 times the median wage here. Someone working an unskilled job will never be able to own a house here already.
No, it is far simpler than looking for some mythical "political" issues. It is simply that hackers - especially amateur ones, who write code as a hobby - dislike trying to work out how old stuff works. They like writing new stuff, instead.
I suspect it's partly that, but also that companies like Red Hat make money from training, so like obsoleting old tools and the developers who care about the Principle of Least Astonishment have long since moved on from Linux to other systems.
Slashdot Mirror
Rights are worthless if you are unable to exercise them. Rights are only as strong as their enforcement. If someone violates your rights, who will enforce penalties against them? Who will set up incentives to prevent others from doing so? In most civilised countries, we delegate responsibility for this to an organisation that we call a government. It doesn't give you the rights, but it does try to stop anyone else from taking them away. The government is accountable to the people and the paragraph that you quote attempts, in the USA, to prevent the government from attempting to take away certain rights.
If someone wants to have fun, they should set up an automated system that allows people to send Facebook a right to be forgotten notice under the GDPR and provides a record of who sent it. They might not delete the data, but they're liable for very large fines if the EU discovers that they didn't comply.
(I hear in Britain the correct strategy is to claim the homeowner shot a burglar, then their version of SWAT will show up. Might've just been a joke, though.)
The British equivalent of SWAT is Specialist Firearms Command (usually referred to as SO19 in films and TV). They train specialist firearms officers, who receive extra training and must pass a battery of psychological tests before they are even accepted for training. There is an automatic investigation in the case of any firearm discharge by one of these officers.
SFC is almost never the first response. An unarmed officer will attempt to judge whether they need to be deployed. When they are deployed, they have been trained to avoid firing unless all other options have been exhausted. This training sometimes fails, but it seems to work a lot better than the US model of arming all of the police and giving them appallingly bad training.
Some do. OpenStreetMap was instrumental in persuading the UK government to include some of the government-maintained maps under an open license. Historically, high-quality mapping data has been regarded as a military asset, because if you want to invade a country then having decent maps is essential. That's less important in an age of satellite photography (any foreign power that could plan in invasion almost certainly has satellites that can provide them with accurate mapping data) and so there's a slow shift to seeing it as a commercial asset.
You seem to be implying that Google Maps has consistently high quality data. My house doesn't exist on Google Maps (it does, and is correctly numbered on OSM) and the street that I live on is now half there, as of a couple of months ago, on Google Maps, in spite of the part that they've documented being finished almost three years ago and the half that they haven't being finished over a year ago.
A lot of the things that use Google Maps (or, increasingly, Bing Maps) do so via the APIs as well. If you go to a hotel web site, for example, odds are that they'll have either a Google or Bing map embedded in their 'how to find us' page. It is fairly easy to do the same with OSM, but there is a quite large selection of web front ends and there's no real attempt to curate them and many are poorly documented (it took me ages, for example, to realise that the JavaScript OSM overlay library that I used for a conference was putting my waypoints in the wrong place because of a poorly documented projection option having a default that wasn't the same as the other tool that I was using). This kind of thing prevents OSM being the default choice for embedding elsewhere, which means that it doesn't get the kind of visibility that Google Maps provides.
Clang supports all of the features of Objective-C on all *NIX platforms and Windows. The GNUstep Objective-C runtime, in combination with clang, supports a superset of the language features of Apple's implementation and is used by Microsoft in their Windows bridge for iOS. You might be surprised at some of the places Objective-C ends up - there are quite a few larger (mostly non-Web, though there are also things like SOGo) server systems that use Objective-C - I've consulted for a couple of companies that maintain them.
Apple really likes to play up Objective-C as the competitor to Swift, but the elephant in the room is Objective-C++, which (since C++11) is a far better language than either. You can have the late binding of Objective-C objects and the compile-time specialisation of C++ templates and switch between them easily. With ARC, you can store Objective-C object pointers inside C++ collections and have the memory management work properly (which was the biggest irritation of pre-ARC Objective-C++). If you use C++ objects as instance variables in Objective-C, then their destructors run automatically after -dealloc, so the memory management works both ways (you can, for example, use C++11 unique_ptr or shared_ptr to manage ownership of non-Objective-C resources by Objective-C++ objects). Most importantly, it has a trivial way of using C++ libraries, something that Swift can achieve only if you wrap them in [Objective-]C.
That was different. The Althon and Opteron were both better designs than Intel, but were made using an older process technology. The first Opterons were made on a 130nm process in 2003, moving to 90nm in 2005. The Pentium 4 was using 130nm 2002 and moved to 90nm in 2004 - Intel had a clear one-year lead on process technology. AMD only kept up by having much better processor designs.
On the desktop, power usage isn't nearly so important - it's plugged in. Your budget isn't a power budget on the desktop, it's a dollar budget
At a rough approximation, 1W for one year costs me £1. A 20W power difference over the lifetime of a device is £20/year if it's always on. Now that a 3-year upgrade cycle has extended to a 5-year cycle for most people, a 20W power increase is a £100 difference to TCO. For a corporate budget, it's worse because it's a £20/year operating expense rather than a £100 capital write-down.
NetBurst wasn't a failure of microarchitecture, it was a failure of process technology. NetBurst was designed on the assumption that Intel's fab people would be able to produce 10GHz versions in 2005, in the same power budget as the 2GHz version in 2000. This failed dismally and was one of the first indications that Intel's dominance over fabrication techniques was not going to last forever. If Intel in 2005 had been able to produce the process technology that they were forecasting in 1996, when the NetBurst design was started, then AMD would be long dead. They had absolutely nothing that could have competed with a 10GHz Pentium 4. Unfortunately for Intel, AMD had quite a lot that could compete with a 3GHz Pentium 4.
Cyrix's performance failure was really one of marketing. I had a 6x86 P166+. It was a 133MHz core, with Cyrix's stupid 'P-rating' thing claiming it was faster than a Pentium 166. It actually was faster than a 133MHz Pentium for anything I benchmarked, but quite a bit slower than a P166. It was also a lot cheaper than a P133. If they'd marketed it against the P133, it would have been quite compelling (faster and cheaper), and they wouldn't have got such a reputation for exaggerating their performance.
While their market share dropped over the years as people discovered that AMD CPUs were faster, more reliable, ran cooler and cost less all at the same time
I'll accept faster and cooler, but for a long time Intel was shipping thermal throttling and AMD wasn't. In some cases, this was a problem for Intel. We had an Opteron cluster and a P4 Xeon cluster. When the cooling failed in a node on the AMD cluster, the CPU burned and the node died. The cluster management system noted the failure and redistributed the work. With the P4 cluster, the CPU would clock down to something tiny like 200MHz. It would still be responsive to command messages and so work wouldn't be redistributed, but jobs would take ten times as long to complete (and the tail latency of the slowest node would cause the entire job to slow down).
Lower power, as I recall, was only really true for desktop and up. The Pentium M, once it shipped, was lower power than anything AMD had to offer and gave Intel some breathing room while the laptop was the fastest-growing market segment and AMD's laptop offerings were lacklustre. The Opteron was wiping the floor with anything Intel had to offer in the server space, largely because of HyperTransport (basically the Alpha CPU interconnect) on multi-socket systems and moving the memory controller on die, while Intel chips were stuck with the memory controller on a separate north bridge limiting memory bandwidth (I seem to recall that some Xeons were actually faster than AMD cores if you had a workload that fitted in cache). The Althon64 then came and outperformed anything Intel had on the desktop for even approximately the same price. The only advantage the mobile versions had over the Pentium M was 64-bit support, and no one wanted 64-bit Windows on mobile (and the market for non-Windows laptops was tiny).
Developing their own processor will be ridiculously expensive. Even their i devices are mostly ARM design.
Apple's in-house cores use an ARM ISA, but they are categorically not an ARM design. They are a complete reimplementation and currently outperform all of their competitors. In the same power envelope, they also outperform Intel. Given that they're already spending the money designing these, it's not too crazy to imagine that they'd switch to them.
Also, you think they can do better at securing cpus than Intel and amd - both of which have larger market shares?
Yes. Apple's secure element is not vulnerable to any of the speculation-based attacks against SGX enclaves.
You missed an important one. The NT in Windows NT originally referred to the Intel i860's code name (N-Ten). The NT was developed on the i860 and then ported to x86, to ensure that the codebase didn't include any x86isms.
It's a bit more subtle than that. Since about 2007, we've not been seeing Dennard scaling, so although transistor counts have gone up, the number that you can power at any given time has not increased by nearly as much. This is what people mean when they talk about 'dark silicon'. The problem with the x86 decoder is that it needs to be powered most of the time, because any time you're executing instructions that are not in the trace cache, it's necessary. The same is true for caches (though you can turn off some of the ways in a set-associative cache when you're seeing that they're not used - not sure if anyone does this in commercial processors) and the register rename engine. All of the execution pipelines can be power gated so you can turn some of them off depending on the instruction mix (for example, powering down the AVX pipes in scalar workloads).
It doesn't matter much to Intel at the high end, because it isn't a huge part of the total power consumption, but it is a problem at the low end when reducing the power consumption of everything else makes it start to dominate again.
What if Google has most of the market because they make a better product for less money?
That depends. Are they making it for less money because they're subsidising it with income from other products? If so, it may count as dumping and so is illegal (for good reason: it means that no company that doesn't have an independent revenue stream can compete).
How much market share do the Fire phones have? Is it enough that Google isn't an effective monopoly? How many apps are not available via the Amazon store that are available via Google Play?
Manufacturers aren't prohibited from installing their own software, including software which serves the same functions as Google's options
Except that if you want to install any Google apps, then you must install the entire suite and if you want to install competing apps then there are different licensing terms that cost more. And if you don't then it's impossible for your customers to install most third-party software because Google has managed to achieve an effective monopoly on distribution of most Android apps. And if you do install the Play store then you also need Play services, which run with insane permissions and hook into almost every app installed from Play.
I doubt you can get it very far in the time that it takes to send a security guard or the police out. The thing is several tons and you can't just drag it through the water, you need to lift it up onto a boat, during which point you'll trigger all sorts of alarms (vibration, motion, shock).
I wondered about this. If the heat were spread throughout the ocean, it would be a negligible impact, but raising the temperature in a bay by one degree can have quite significant ecological impacts. It depends a lot on how much tidal flow there is and how much heat the thing is dissipating.
Google stopped doing that over a decade ago and I presume Azure has a similar policy. A disk fails, they offline the machine. Enough machines in a rack fail, and they replace the rack. This datacentre is specifically designed not to be accessible by humans (the gas inside has had oxygen and water vapour removed, to reduce the damage that these do to components) and is operating under the same model: you keep using it until enough hardware is dead that it's worth pulling it up and doing a complete refit.
It looks pretty heavy. It's also not very far out to sea, so I doubt you could steal it without anyone seeing and it wouldn't take MS very long to notice an entire datacentre dropping off the Internet and sending someone out to take a look...
I've read a few articles over the last couple of months that indicate that millennials are saving more, inflation adjusted, than their parents were at the same age, but they're still less able to buy houses.
I did the calculation here for what we pay postdocs a couple of years back. The exact numbers don't matter too much: Wage increases are n% a year, house prices are increasing by m% per year. As long as m is greater than n, eventually you will reach the point where no matter what percentage of your income you save, you will never be able to afford a house. With some fairly conservative estimates of n and m, I found that within a decade the increase on the deposit required to buy a house each year will be more than a postdoc makes in a year, before tax. Postdocs make around 2-3 times the median wage here. Someone working an unskilled job will never be able to own a house here already.
No, it is far simpler than looking for some mythical "political" issues. It is simply that hackers - especially amateur ones, who write code as a hobby - dislike trying to work out how old stuff works. They like writing new stuff, instead.
I suspect it's partly that, but also that companies like Red Hat make money from training, so like obsoleting old tools and the developers who care about the Principle of Least Astonishment have long since moved on from Linux to other systems.