Doesn't work, unless you're buying direct from the manufacturer: a load of the dangerous knock-offs have well-known and trusted brand labels, in spite of having nothing to do with that company.
I have an Apple computer, but almost everything I use could run on another system (though needing something that runs *NIX dev tools and Microsoft Office would probably leave me using some form of virtualisation). Of the others on the list: Facebook? Don't use them. Google? Don't use them for search, do have an Android phone but most of the Google code is replaced by other things, wouldn't mind if they went away. Microsoft? I use MS Office, but if they went away then that wouldn't be too sad (though none of the PowerPoint competitors have an equivalent of SmartArt, which is a shame). Amazon? They're convenient, but they're just a retailer: there's nothing that they sell that I can't buy elsewhere (I suspect that some things I use run on AWS, but they could probably migrate away).
I'd expect it to be a bit lower in India, because computing-related degrees are still hyped in the same way there that they were in the US and Europe just before the dot-com bubble burst. Want to make a lot of money? Be a programmer. There's going to be a much bigger long tail of incompetence, just as there was in all of the 'I've never used a computer before, but I want to become a billionaire so I'm doing computer science' students in '98 or so.
That's surprising, given that one of the design goals for USB-C was to be more robust than USB-A and it is designed and tested to support more insert/remove cycles than USB-A.
The problem with USB-C is that there are multiple power levels and some of the cheap cables don't correctly identify their maximum current, so end up catching fire when plugged into something that can provide the higher power levels. Unfortunately, these are typically sold on sites alongside the ones that will work fine.
but in a situation like that you would presumably restrict their SMB access so that they could only interact with SMB on a single, actually supported, system you control
Unfortunately, SMB uses UDP and so it's possible to spoof the origin address unless you have a correctly configured managed switch. In a company like this, which doesn't have that much in-house computer expertise, they don't have a managed switch and probably wouldn't configure it correctly if they did have one.
Because, plain and simple, of course everyone wants free money!
Under most costed UBI proposals I've seen, I'd be slightly worse off. I'm still in favour of it.
Simply doing the math will show you that so-caled 'UBI' is not feasible on a large scale
Citation please, because last time I 'did the math', assuming UBI at the level of the current tax-free earnings allowance, the overall tax rate for people earning at my level would go up by about 2-3%, but that's hardly going to bankrupt me or remove my incentive to work. The top tax brackets would go up more, but still would not come close to the highest that we've had in my parents' lifetimes.
I absolutely agree, although it has nothing to do with UBI
Yes it does, because UBI removes the big incentive for a load of minimum-wage jobs: you need one to survive. There are a lot of jobs that are poorly paid that no one actually wants to do, but they have to because they need some kind of job. UBI would push the price of these jobs up and provide a much greater incentive to automate them.
Nope. Symbian had a nice kernel design (which, amusingly, has a lot in common with Magenta in terms of design), but several incompatible bad UIs on top. WebOS, in contrast, had a Linux kernel that was a very poor fit for the device type, but a very clean UI design. The one thing that they had in common was that they were both a pig to program for.
You seem to be conflating a stop-the-world tracing collector (a mechanism) with automatic garbage collection (a policy). Garbage collection is not non-deterministic because garbage collection specifies the objective (memory is automatically reclaimed), not the implementation. There are a number of GC designs that work on realtime systems, with guaranteed maximum pause times (look at some of David F. Bacon's work, for example). Garbage collection doesn't require mysterious pauses. There are a number of concurrent collector designs, which don't stop execution of the main program (Azul ships a very high-performance one for Java, for example).
Any garbage collector implementation is in a tradeoff space between allocation throughput, tail latency, worst-case memory overhead, and overall CPU load. It's possible to heavily optimise for any of these, but not all of them at the same time. A lot of the problems people associate with garbage collection are due to picking an implementation that optimises for the ones that are not important for their workload.
You should, however, have -all- the information and not just the subset deemed supportive of the cause by invisible people
The main reason that we have representative, rather than direct, democracy is that no one has the time to do that. Do you understand the causes of the conflict in Syria? The economic impact of NAFTA? The costs and benefits of EU membership for each member state? People who spent their day jobs don't fully understand these, so what chance does the general population have? You need some kind of filter that will highlight the parts that are relevant for you to care about, the problem is that there's no accountability. News organisations can tell outright lies without penalty and ideological spin is an expected minimum.
That's not the problem. The problem is that a company like Facebook knows:
Roughly how old you are (at least enough to tell whether you're of voting age)
Where you live (roughly if it has to guess from IP addresses, precisely, if you've ever bought anything from a company that shares data with Facebook).
What news articles you read (what issues are important to you?)
What news articles you share (what are your opinions on the issues that are important to you?)
This is enough that they can identify what ads to show you to influence your opinion (Candidate X strongly supports issue Y), but more importantly they can share this info with canvassers who can target the undecided votes in a constituency and knock on their doors and say 'have you thought about [issue that we know is your number one priority], are you aware that our candidate believes [exactly what you believe]?'.
I don't know if you're a shill or simply uninformed, but you seem to be repeating the erroneous line that network neutrality prevents QoS. It is completely acceptable under network neutrality, for example, to put SIP traffic in queues with shorter or more deterministic latency. It's even permitted to charge the customer more for doing this. What is not permitted is only putting SIP traffic to the ISP's own VoIP service in a low-latency queue and putting everyone else's SIP traffic in a normal queue. It is not permitted to charge Microsoft money to prioritise Skype traffic, at the expense of other VoIP traffic. It is not permitted to penalise Netflix traffic unless Netflix pays extra to reach your customers. Which one of these do you think would be in the consumers' interest?
The problem is not that he's right, the problem is that he is solely an idealist. He has a vision of an end goal, which is one that many people would agree is beneficial, and he has some convincing arguments as to why the status quo is bad, but he doesn't have a coherent plan for how to get from here to there. He's like the street-corner communists who correctly point out the flaws with capitalism, describe an egalitarian utopia, have no plan for how to implement it, and then complain that people are still supporting a capitalist society.
There are good reasons for preferring free software and ones that are easy to argue to a business: It frees you from vendor lock-in, makes it easy to find a second source, removes most compliance-related costs, makes it easy to pay for the features you actually want and are important to your business (rather than ones that some focus group thinks will sell the next version), and so on.
There are also lots of intermediate steps, where companies open source the parts of their software that don't give them a key competitive advantage and where self-interest can motivate them. For example, Apple doesn't make money from a compiler toolchain, and it's in their best interests to open source all of their LLVM, Clang, Swift, LLDB, and so on code. Doing so doesn't cost them anything and means that other people will contribute to these projects. Yet rather than encourage this kind of thing, Stallman castigates them for the code that they don't release and endorses a license that gives them a choice of open sourcing everything (which would kill their current business model) or nothing (which gives no benefit to anyone else).
There are some cases where firewalling is difficult. I did some work some years back for a company that specialises in transferring data between different formats (including to and from paper, microfiche, old mainframe tapes, DVD, and so on). They had some very nice scanners that cost over £30K each, and which came with Windows 98 installed running the software. For that much up-front cost, I'd imagine that they're still using them - Windows 98 was EOL even back when I consulted for them. Unfortunately, for them to be useable, they need to be able to connect to the SMB server that they'll store the scanned documents on. They can be firewalled from everything else, but they're still likely to end up with malware eventually (there are known holes in the Windows 98 SMB client code). The vendor wasn't providing any software updates. There are probably a lot of other things in a similar situation, where they need to be networked for interoperability with workflow tracking stuff, but aren't safe. The real problem for ReactOS is that the people who are in this situation are typically not the sorts of companies that would fund open source development.
ReactOS uses WINE for all of the userspace stuff, so it's unlikely to be a better option than *NIX + WINE. Its main advantage is if you are using something that has a kernel component, because it provides the same device driver KPIs as Windows, so that legacy bit of expensive hardware with Windows 2000 drivers might work with ReactOS and nothing else that's actively maintained. But for pure userspace code? I doubt it's the best option.
That's not really how it works. The money repaid will be cancelled out by the removal of the mortgaged property as an asset. The value of the money is almost certainly lower than the value of the mortgaged asset, so the bank's assets will show a decrease. This then reduces the amount that they can loan by some multiple of the difference. The act of loaning to you made them able to make more loans, the act of repayment makes them able to make fewer loans.
I can't tell if you're trolling, or if that actually does work for you. When I click on it, the radio button stays pressed for a few seconds then jumps back to the D2 option.
Amusing, but the leading cause of cancer in rats is being a rat. Rats have very high mutation rates, which is why it's so difficult to eradicate them: a few in a population will probably be immune to whatever poison you try and will pass on the immunity to all of its offspring. The down side of this is that the ones that avoid poison and predators all die of cancer soon after.
There are other reasons Apple would find it difficult to buy Qualcomm. Sure, they have around four times the amount of cash required, but if they look as if they're willing to spend $80bn to make a lawsuit go away then that's a really strong indication to the market that the lawsuit has merit and that Qualcomm is undervalued. If they're willing to pay $80bn, would they be willing to pay $160bn? There are a lot of speculators who would rush to buy Qualcomm shares and push the price up, then vote against a takeover if it didn't make them money. Even if Apple did get antitrust approval, Qualcomm is a sufficiently important supplier to other companies that Apple may well find themselves bidding against a coalition (as happened with MIPS - ImagTec bought them, but most of the IP went to a consortium of companies that didn't want to be sued by patent trolls).
That's absolutely untrue in a system with fractional reserve banking. Loaning money is the action that creates money in such a system (the theory is that you're adding other capital, for example the house, to the monetary system and so it should react by creating an amount equal to that value). Repaying a loan removes value from the system and actually makes the bank less able to loan money.
Even on the desktop, it's a horrible experience now. There's a floating user box on the right that takes up a large chunk of width (hey, idiots: I made my browser window this wide because I want to see text this wide, not because I want a quarter to be a pointless empty side bar) and on the front page you can't actually get to that box (you know, the one with messages in it) unless you scroll right to be bottom, because some idiot made it move with the page, rather than scrolling sensibly.
so Intel has already lost the market to a *cluster* of companies that includes Samsung, and it's for those companies to fight it out to see who becomes the dominant one. Intel will not be one of the winners
I'm not so sure, for two reasons:
The first is that economies of scale are hugely important in the fab business. AMD spun off Global Foundries so that they could share their fab R&D costs with other companies and whoever has the highest quantities of current-process chips (older processes are less important - you run a fab for years after it's superseded, but your margins go down a lot) has a big advantage in terms of amortising their R&D costs across multiple chips. Being smaller than 5 other companies combined is not a problem, being smaller than one other company is.
Second, people typically think of Intel as a processor vendor, but they're not. They're a chip maker that happens to design x86 cores as a way of keeping their fabs running. A lot of what they sell are RAM chips, network controllers, and so on. x86 chips make up a surprisingly small percentage of their total output. They're also an ARM licensee (you might be surprised at how many ARM cores there are on an Intel motherboard), and have just upgraded their ARM license, so there's nothing to stop them fabbing ARM cores if they decide it's a good business to be in (they do already on a bunch of their FPGA products).
Yes. The difficult thing about code reuse attacks is that you have to find a useful bit of code to reuse. With a JIT in your target process, you can conveniently persuade it to generate some useful code that's full of gadgets. Mobile Safari combines ASLR with ARMv8's execute-only mappings to harden this: not only is the JIT'd region in a random location, its writeable address isn't stored anywhere in readable memory. They map a region write-only in one place and execute-only in another place. They then JIT a memcpy function that copes a buffer into the write-only region and throw away the address of the writable region, just keeping a pointer to the memcpy implementation in the executable region. You can use that to write anywhere into the executable region, but the only way of finding the writable region is to try writing there, which in a 64-bit address space is likely to crash the process probing long before you find the right address. Because the mapping for the executable part is execute only, you can't read it to find if a write happened to land in the writeable mapping and you can't read it to find where gadgets ended up.
Slashdot Mirror
Doesn't work, unless you're buying direct from the manufacturer: a load of the dangerous knock-offs have well-known and trusted brand labels, in spite of having nothing to do with that company.
I have an Apple computer, but almost everything I use could run on another system (though needing something that runs *NIX dev tools and Microsoft Office would probably leave me using some form of virtualisation). Of the others on the list: Facebook? Don't use them. Google? Don't use them for search, do have an Android phone but most of the Google code is replaced by other things, wouldn't mind if they went away. Microsoft? I use MS Office, but if they went away then that wouldn't be too sad (though none of the PowerPoint competitors have an equivalent of SmartArt, which is a shame). Amazon? They're convenient, but they're just a retailer: there's nothing that they sell that I can't buy elsewhere (I suspect that some things I use run on AWS, but they could probably migrate away).
It would be nice if that worked. There's someone at Google who has been testing a load of cables. There isn't a correlation between price and quality.
I'd expect it to be a bit lower in India, because computing-related degrees are still hyped in the same way there that they were in the US and Europe just before the dot-com bubble burst. Want to make a lot of money? Be a programmer. There's going to be a much bigger long tail of incompetence, just as there was in all of the 'I've never used a computer before, but I want to become a billionaire so I'm doing computer science' students in '98 or so.
USB-C is not as robust as USB-A.
That's surprising, given that one of the design goals for USB-C was to be more robust than USB-A and it is designed and tested to support more insert/remove cycles than USB-A.
The problem with USB-C is that there are multiple power levels and some of the cheap cables don't correctly identify their maximum current, so end up catching fire when plugged into something that can provide the higher power levels. Unfortunately, these are typically sold on sites alongside the ones that will work fine.
but in a situation like that you would presumably restrict their SMB access so that they could only interact with SMB on a single, actually supported, system you control
Unfortunately, SMB uses UDP and so it's possible to spoof the origin address unless you have a correctly configured managed switch. In a company like this, which doesn't have that much in-house computer expertise, they don't have a managed switch and probably wouldn't configure it correctly if they did have one.
Because, plain and simple, of course everyone wants free money!
Under most costed UBI proposals I've seen, I'd be slightly worse off. I'm still in favour of it.
Simply doing the math will show you that so-caled 'UBI' is not feasible on a large scale
Citation please, because last time I 'did the math', assuming UBI at the level of the current tax-free earnings allowance, the overall tax rate for people earning at my level would go up by about 2-3%, but that's hardly going to bankrupt me or remove my incentive to work. The top tax brackets would go up more, but still would not come close to the highest that we've had in my parents' lifetimes.
I absolutely agree, although it has nothing to do with UBI
Yes it does, because UBI removes the big incentive for a load of minimum-wage jobs: you need one to survive. There are a lot of jobs that are poorly paid that no one actually wants to do, but they have to because they need some kind of job. UBI would push the price of these jobs up and provide a much greater incentive to automate them.
Nope. Symbian had a nice kernel design (which, amusingly, has a lot in common with Magenta in terms of design), but several incompatible bad UIs on top. WebOS, in contrast, had a Linux kernel that was a very poor fit for the device type, but a very clean UI design. The one thing that they had in common was that they were both a pig to program for.
You seem to be conflating a stop-the-world tracing collector (a mechanism) with automatic garbage collection (a policy). Garbage collection is not non-deterministic because garbage collection specifies the objective (memory is automatically reclaimed), not the implementation. There are a number of GC designs that work on realtime systems, with guaranteed maximum pause times (look at some of David F. Bacon's work, for example). Garbage collection doesn't require mysterious pauses. There are a number of concurrent collector designs, which don't stop execution of the main program (Azul ships a very high-performance one for Java, for example).
Any garbage collector implementation is in a tradeoff space between allocation throughput, tail latency, worst-case memory overhead, and overall CPU load. It's possible to heavily optimise for any of these, but not all of them at the same time. A lot of the problems people associate with garbage collection are due to picking an implementation that optimises for the ones that are not important for their workload.
You should, however, have -all- the information and not just the subset deemed supportive of the cause by invisible people
The main reason that we have representative, rather than direct, democracy is that no one has the time to do that. Do you understand the causes of the conflict in Syria? The economic impact of NAFTA? The costs and benefits of EU membership for each member state? People who spent their day jobs don't fully understand these, so what chance does the general population have? You need some kind of filter that will highlight the parts that are relevant for you to care about, the problem is that there's no accountability. News organisations can tell outright lies without penalty and ideological spin is an expected minimum.
This is enough that they can identify what ads to show you to influence your opinion (Candidate X strongly supports issue Y), but more importantly they can share this info with canvassers who can target the undecided votes in a constituency and knock on their doors and say 'have you thought about [issue that we know is your number one priority], are you aware that our candidate believes [exactly what you believe]?'.
I don't know if you're a shill or simply uninformed, but you seem to be repeating the erroneous line that network neutrality prevents QoS. It is completely acceptable under network neutrality, for example, to put SIP traffic in queues with shorter or more deterministic latency. It's even permitted to charge the customer more for doing this. What is not permitted is only putting SIP traffic to the ISP's own VoIP service in a low-latency queue and putting everyone else's SIP traffic in a normal queue. It is not permitted to charge Microsoft money to prioritise Skype traffic, at the expense of other VoIP traffic. It is not permitted to penalise Netflix traffic unless Netflix pays extra to reach your customers. Which one of these do you think would be in the consumers' interest?
The problem is not that he's right, the problem is that he is solely an idealist. He has a vision of an end goal, which is one that many people would agree is beneficial, and he has some convincing arguments as to why the status quo is bad, but he doesn't have a coherent plan for how to get from here to there. He's like the street-corner communists who correctly point out the flaws with capitalism, describe an egalitarian utopia, have no plan for how to implement it, and then complain that people are still supporting a capitalist society.
There are good reasons for preferring free software and ones that are easy to argue to a business: It frees you from vendor lock-in, makes it easy to find a second source, removes most compliance-related costs, makes it easy to pay for the features you actually want and are important to your business (rather than ones that some focus group thinks will sell the next version), and so on.
There are also lots of intermediate steps, where companies open source the parts of their software that don't give them a key competitive advantage and where self-interest can motivate them. For example, Apple doesn't make money from a compiler toolchain, and it's in their best interests to open source all of their LLVM, Clang, Swift, LLDB, and so on code. Doing so doesn't cost them anything and means that other people will contribute to these projects. Yet rather than encourage this kind of thing, Stallman castigates them for the code that they don't release and endorses a license that gives them a choice of open sourcing everything (which would kill their current business model) or nothing (which gives no benefit to anyone else).
There are some cases where firewalling is difficult. I did some work some years back for a company that specialises in transferring data between different formats (including to and from paper, microfiche, old mainframe tapes, DVD, and so on). They had some very nice scanners that cost over £30K each, and which came with Windows 98 installed running the software. For that much up-front cost, I'd imagine that they're still using them - Windows 98 was EOL even back when I consulted for them. Unfortunately, for them to be useable, they need to be able to connect to the SMB server that they'll store the scanned documents on. They can be firewalled from everything else, but they're still likely to end up with malware eventually (there are known holes in the Windows 98 SMB client code). The vendor wasn't providing any software updates. There are probably a lot of other things in a similar situation, where they need to be networked for interoperability with workflow tracking stuff, but aren't safe. The real problem for ReactOS is that the people who are in this situation are typically not the sorts of companies that would fund open source development.
ReactOS uses WINE for all of the userspace stuff, so it's unlikely to be a better option than *NIX + WINE. Its main advantage is if you are using something that has a kernel component, because it provides the same device driver KPIs as Windows, so that legacy bit of expensive hardware with Windows 2000 drivers might work with ReactOS and nothing else that's actively maintained. But for pure userspace code? I doubt it's the best option.
That's not really how it works. The money repaid will be cancelled out by the removal of the mortgaged property as an asset. The value of the money is almost certainly lower than the value of the mortgaged asset, so the bank's assets will show a decrease. This then reduces the amount that they can loan by some multiple of the difference. The act of loaning to you made them able to make more loans, the act of repayment makes them able to make fewer loans.
I can't tell if you're trolling, or if that actually does work for you. When I click on it, the radio button stays pressed for a few seconds then jumps back to the D2 option.
Amusing, but the leading cause of cancer in rats is being a rat. Rats have very high mutation rates, which is why it's so difficult to eradicate them: a few in a population will probably be immune to whatever poison you try and will pass on the immunity to all of its offspring. The down side of this is that the ones that avoid poison and predators all die of cancer soon after.
There are other reasons Apple would find it difficult to buy Qualcomm. Sure, they have around four times the amount of cash required, but if they look as if they're willing to spend $80bn to make a lawsuit go away then that's a really strong indication to the market that the lawsuit has merit and that Qualcomm is undervalued. If they're willing to pay $80bn, would they be willing to pay $160bn? There are a lot of speculators who would rush to buy Qualcomm shares and push the price up, then vote against a takeover if it didn't make them money. Even if Apple did get antitrust approval, Qualcomm is a sufficiently important supplier to other companies that Apple may well find themselves bidding against a coalition (as happened with MIPS - ImagTec bought them, but most of the IP went to a consortium of companies that didn't want to be sued by patent trolls).
That's absolutely untrue in a system with fractional reserve banking. Loaning money is the action that creates money in such a system (the theory is that you're adding other capital, for example the house, to the monetary system and so it should react by creating an amount equal to that value). Repaying a loan removes value from the system and actually makes the bank less able to loan money.
Even on the desktop, it's a horrible experience now. There's a floating user box on the right that takes up a large chunk of width (hey, idiots: I made my browser window this wide because I want to see text this wide, not because I want a quarter to be a pointless empty side bar) and on the front page you can't actually get to that box (you know, the one with messages in it) unless you scroll right to be bottom, because some idiot made it move with the page, rather than scrolling sensibly.
so Intel has already lost the market to a *cluster* of companies that includes Samsung, and it's for those companies to fight it out to see who becomes the dominant one. Intel will not be one of the winners
I'm not so sure, for two reasons:
The first is that economies of scale are hugely important in the fab business. AMD spun off Global Foundries so that they could share their fab R&D costs with other companies and whoever has the highest quantities of current-process chips (older processes are less important - you run a fab for years after it's superseded, but your margins go down a lot) has a big advantage in terms of amortising their R&D costs across multiple chips. Being smaller than 5 other companies combined is not a problem, being smaller than one other company is.
Second, people typically think of Intel as a processor vendor, but they're not. They're a chip maker that happens to design x86 cores as a way of keeping their fabs running. A lot of what they sell are RAM chips, network controllers, and so on. x86 chips make up a surprisingly small percentage of their total output. They're also an ARM licensee (you might be surprised at how many ARM cores there are on an Intel motherboard), and have just upgraded their ARM license, so there's nothing to stop them fabbing ARM cores if they decide it's a good business to be in (they do already on a bunch of their FPGA products).
Yes. The difficult thing about code reuse attacks is that you have to find a useful bit of code to reuse. With a JIT in your target process, you can conveniently persuade it to generate some useful code that's full of gadgets. Mobile Safari combines ASLR with ARMv8's execute-only mappings to harden this: not only is the JIT'd region in a random location, its writeable address isn't stored anywhere in readable memory. They map a region write-only in one place and execute-only in another place. They then JIT a memcpy function that copes a buffer into the write-only region and throw away the address of the writable region, just keeping a pointer to the memcpy implementation in the executable region. You can use that to write anywhere into the executable region, but the only way of finding the writable region is to try writing there, which in a 64-bit address space is likely to crash the process probing long before you find the right address. Because the mapping for the executable part is execute only, you can't read it to find if a write happened to land in the writeable mapping and you can't read it to find where gadgets ended up.