That's because radioactive waste is a horribly misleading term. Anything that is sufficiently radioactive to be a danger is also radioactive enough to be useful, even if only in radiothermal or betavoltaic generators. The 'waste' is fuel that, for political or economic reasons, it doesn't make sense to use at the moment. Most of it can be reprocessed in breeder reactors and turned into fuel useable in existing reactors. Often, storing it and using newly mined fuel is cheaper, but 'we have so much nuclear fuel that it isn't currently economic to make efficient use of it' is a really, really bad argument against nuclear power.
Lighting is less power intensive now but desktop computers aren't
Nonsense. Even if you're talking about high-end workstations, compare them to a quad Alpha or similar from 15 years ago. We've actually had issues in the building where I work (full of computer scientists, so not exactly short on power users - for normal office users, the requirements would likely be even lower). The heating in the building was designed based on the assumption that every desk would have a CRT on it and a tower next to it. The CRTs are all gone (replaced with LCDs) and the towers now all consume less power: a modern multicore i5 or i7 consumes a lot less than a Pentium 4, modern SSDs use less power than spinning rust, so the waste heat from all of these is a lot lower than the building designers expected. I can't tell from the fan noise whether the computer under my desk is on - the machine I was using a decade ago made it very obvious from the jet engine noise.
Was it a native port? WordPerfect released Linux versions too, but they used WineLib. The WINE guys now discourage the use of WineLib, because it requires recompiling everything with a different compiler (Visual Studio can't generate ELF objects with Linux calling conventions) and there tend to be a lot of issues with code containing MS extensions or relying on MS bugs in template instantiation that make this very hard - even if it compiles, it may not run quite correctly.
Depends on the license terms. Some of the cheap OEM licenses are only applicable when running bare metal on a particular machine. If you want to run them in a VM, then you may need a different license. If you get audited and are not in compliance then you can be hit with a very large fine, or you can go to court and try to get that clause in the EULA invalidated (good luck doing this for less than the cost of the fine). If you're going to run proprietary software as part of your business, then make sure that you factor in compliance audits and lawyer time reading the EULAs into your TCO calculations.
Finally they felt they did not have to starve themselves in order to impress potential mates. So they let themselves go.
There's probably some truth in this, but there are other factors. When you're alone, you tend to eat when you're hungry and, unless you're stress eating, tend not to have huge meals because meals by yourself are quite boring. When you're in a relationship, you tend to try to eat together, which means eating at a time when you're both at least vaguely interested in food, but probably not at the time when either of you would have chosen to eat. If you're cooking for two, you're also less likely to make enough by accident that it looks like another complete meal for both of you so you're more likely to just finish it all between the two of you.
The Cocoa Cola Company makes a lot of money selling bottled water. Sugar isn't important to them, they'll sell anything that they can stick in a bottle with a brand and encourage people to buy at a huge markup.
Not sure about Mozilla. Google does it by shipping an entire graphics stack that uses very little by way of system support (an OpenGL context if one's available or a frame buffer if not). Safari does the same thing.
Foxconn is increasingly using mostly automated factories, so the cost of labour is less important. I suspect that the 50,000 jobs was part of a briefing that said 'similar factories that we've built of this size employ 50,000 people' and forget to mention that this one will use fully automated production lines and will employ 50 people. The main requirements are reliable power, easy access to materials and components, good distribution routes for the final product, and lax environmental regulation (if the surroundings are uninhabitable at the end of the operating life of the factory, that's fine - they're building a new one anyway, may as well build it somewhere else...).
Well, sort of. The blind ROP stuff works by realising that exec triggers re-randomisation, but fork does not. Server processes like nginx are vulnerable because they fork children to replace any that have crashed and each child has the same layout. Even with a 64-bit address space, it's possible to probe (some things, like PLTs, are relatively easy to find and full of gadgets). Unless it's changed recently (I've not been paying attention, so it's possible), Android reduces app startup time by having a zygote process for the Java runtime environment, forking a copy of it, setting the UID / filesystem namespace, and then loading the app. This means that the address space will be different to other devices, but will have the same layout across multiple runs on the same device. BROP does require a two-way communication channel, though it only requires one-bit communication out (did this attempt crash or infinite loop the target?).
The real mitigation is not the ASLR, its that only 0.0001% of the population can write ROP exploits
Anyone can write ROP exploits. You don't need many gadgets to get a Turing-complete execution environment and there are compilers available that will take a binary, derive a Turing-complete virtual machine from them, and compile arbitrary (more or less - in some cases you may have a limited number of jumps, but it's usually more than enough to create a shell) code to run on them.
Around the time CPUs hit 1GHz, I came to the conclusion that almost everything I did on a computer fit into one of two categories:
Stuff that was already fast enough.
Stuff that would basically never be fast enough, no matter how much CPU power I throw at it.
Big compile jobs are in the second category. When I'm doing a test build of FreeBSD, I generate around 50GB of object code (for LLVM it's only about 10GB for a debug build). The LLVM build system is nicely parallelised and so could happily consume 3,000 cores (linking would still be a bottleneck, but new linkers are becoming multithreaded) and building FreeBSD for every architecture is a similar proposition. FPGA place and route uses randomised algorithms and can basically consume as much CPU power as you want to throw at it. And, of course, these things grow over time (though a clean build of LLVM has gone from about an hour and a half when I started to under 5 minutes on my current laptop).
For everyday use, an oldish tablet is more than fast enough. For a few things, I'm going to keep complaining about the time it takes for a very long time.
Intuitive is always a red herring in UI design. There are two far more important attributes:
Is it discoverable?
Is it consistent?
Control/command-x/c/v for cut/copy/paste are discoverable and consistent: if you got to the edit menu of any application (visible by the default on non-braindead UIs and the starting point for finding any command), then you will see the icons next to it telling you what the shortcut is. Now that it's discovered, it's consistent everywhere. Except in terminals if you're using control instead of command, because terminals need control-c for interrupt and so break the good UI, but on a Mac it's the same in every single application including the terminal.
In contrast, most of the Windows-key-plus-modifier combinations can only be discovered by reading the documentation. There's nothing that a user is encouraged by the UI to click on that tells them about what these modifiers are (though I vaguely remember that Windows 98 had a 'show desktop' icon in the start bar that told you about windows-d in the tooltip).
Intuitive implies that you are meant to use your psychic powers or some innate knowledge to find how the UI works. Good UIs do not work that way, they make it easy for people to learn and then they allow the user to apply the knowledge everywhere. If someone complains about a UI not being intuitive, then it's a good hint that they don't know anything about HCI.
The IE thing makes more sense. They rewrote a lot of IE to use the new 2D text and graphics compositing APIs, so they couldn't backport it without either maintaining two distinct rendering paths or backporting a load of system library code.
The Direct3D thing is more silly, as OpenGL drivers on Windows XP expose all of the functionality from the hardware that Direct3D does on newer systems.
Since OpenGL 3, GL ES has been a subset of OpenGL, so you can run a GL ES program on any compliant OpenGL implementation. The main difference in the initial GL ES release was the half-precision floating point type, which was not widely supported in hardware on desktop GPUs.
11. Much of the UI became non-discoverable. I'm not sure if the default hiding of menu bars came with 8 or 7, but it meant that unless you knew that the alt button was magic you were unable to access the menus. Similarly, there was no discoverable way of exiting the Metro apps that would occasionally pop up when you accidentally hit one of the magic key combinations - alt-F4 works, but unless you know that that's a way to quit Windows apps, you're stuck.
I didn't realise how truly bad the UI was until my mother bought a new machine that came with Windows 8 just before I visited the Christmas before last. She's been using Windows since 3.1 and, though she's not exactly an expert, she's got more than a passing familiarity with the OS. Lots of things just left her completely stuck. I've no idea where MS found the people that they put in their usability testing lab, but they don't seem representative of users. When my girlfriend bought a new laptop, I persuaded her to buy one that came with a Windows 7 downgrade. It took her about two weeks before deciding that it was worth using, and she was someone who had managed to tolerate Vista for years. She seems pretty happy with 7 (though some parts of the UI suck: anyone know how to set up an ad-hoc WiFi network with Windows Vista, 7, or 8? The network config UIs are completely different in all three and I couldn't figure it out in any of them).
It's the circle. It was a house on Android 2.x, but that was considered too obvious an icon for home, so Google changed it. As far as I can tell, from using Android and from talking to some devs on the Android team at Google, everyone in the Android UI team is a drooling moron who could be replaced by a magpie and produce better results.
The hardware assist is a very small amount of signal processing to identify sibilants and plosives (something that a 2MHz 6502 could do in real time) and flag things that have the same pattern as a certain phrase ('Okay Google', I think - not used it) to wake up another core. I think in the Moto X, they use a small ARM core for it, though they may have a two-phase process where a hardware implementation wakes up a small ARM core that does a better job and wakes up the application cores if there's something interesting. In either case, it's probably quite difficult to change the pattern of sibilants and plosives that it looks for whatever Cortana wants, either because it's hard wired or because there's no software API for changing it.
And Siri was bought by Apple from an SRI spin-off, which developed the technology as part of a DARPA project. Lots of the components already exist in open source form (as the Sirius project has shown).
Currently, no. If there's no driver to pay, maybe. My point was that, even though it may not work for him, it will for a lot of people. As I said, however, he's an outlier in terms of usage. The commute in the UK is 8.5 miles (each way), so 85 miles (around 140km) per week. Even if recreational trips double that, then 500km/week seems a bit excessive. His commute distance puts him well in around the top 5-8% of car users. They're the ones getting the most use from a car, so the last ones to want to replace it with a shared vehicle.
There are several other factors too. $50k on a car is a pretty large capital investment in something that's going to depreciate quickly. If you're not using the car as much as him, then you'd probably be better off doing something else with the capital and letting someone else own the car (or own a fraction of one). If car sales drop off because most cars are being bought as part of fleets with fewer than one car per user, then prices are also likely to go up, so that $50k car may well not stay a $50k car.
His costs are around 25% lower than the IRS quotes for mileage, so I'm slightly sceptical of his calculations.
Rama is the same sort of story as 2001. They make good books and made good cinema in the '60s when impressive visual effects were enough to wow an audience. Now that CGI is cheap, it's expected by the audience and you need a bit more story. With both, a film doesn't give you long enough to tell the story well and so you end up with something that either feels rushed or superficial. Rama would make a good TV series, because you'd have the time to focus on the character development. It would be a terrible film.
You're spending almost $1000/month on a car and don't think that it would be possible for a service to make that cheaper? I bike most places and take taxis everywhere else and spend a fraction of that - even taking a taxi with a human driver to and from work every day and would be cheaper than you're spending. 500km/week puts you well outside the normal range for drivers, even in the US where the cities are carefully designed so that places people want to be are as far away from where people are as possible.
Not Bob specifically, but there were quite a few agent-based UIs floating around. They all failed for much the same reason as Bob (if you're going to have an agent between you and the computer, it needs to be really good at natural language parsing, or you're just building an obfuscated and poorly documented CLI).
[Quick question about the word "only". English is not my first language, and I get confused about the proper use of only. For example in the subject line, the word only applies to what? To the verb happens or to the phrase "in tech"?]
Slashdot Mirror
That's because radioactive waste is a horribly misleading term. Anything that is sufficiently radioactive to be a danger is also radioactive enough to be useful, even if only in radiothermal or betavoltaic generators. The 'waste' is fuel that, for political or economic reasons, it doesn't make sense to use at the moment. Most of it can be reprocessed in breeder reactors and turned into fuel useable in existing reactors. Often, storing it and using newly mined fuel is cheaper, but 'we have so much nuclear fuel that it isn't currently economic to make efficient use of it' is a really, really bad argument against nuclear power.
Lighting is less power intensive now but desktop computers aren't
Nonsense. Even if you're talking about high-end workstations, compare them to a quad Alpha or similar from 15 years ago. We've actually had issues in the building where I work (full of computer scientists, so not exactly short on power users - for normal office users, the requirements would likely be even lower). The heating in the building was designed based on the assumption that every desk would have a CRT on it and a tower next to it. The CRTs are all gone (replaced with LCDs) and the towers now all consume less power: a modern multicore i5 or i7 consumes a lot less than a Pentium 4, modern SSDs use less power than spinning rust, so the waste heat from all of these is a lot lower than the building designers expected. I can't tell from the fan noise whether the computer under my desk is on - the machine I was using a decade ago made it very obvious from the jet engine noise.
Was it a native port? WordPerfect released Linux versions too, but they used WineLib. The WINE guys now discourage the use of WineLib, because it requires recompiling everything with a different compiler (Visual Studio can't generate ELF objects with Linux calling conventions) and there tend to be a lot of issues with code containing MS extensions or relying on MS bugs in template instantiation that make this very hard - even if it compiles, it may not run quite correctly.
Depends on the license terms. Some of the cheap OEM licenses are only applicable when running bare metal on a particular machine. If you want to run them in a VM, then you may need a different license. If you get audited and are not in compliance then you can be hit with a very large fine, or you can go to court and try to get that clause in the EULA invalidated (good luck doing this for less than the cost of the fine). If you're going to run proprietary software as part of your business, then make sure that you factor in compliance audits and lawyer time reading the EULAs into your TCO calculations.
Finally they felt they did not have to starve themselves in order to impress potential mates. So they let themselves go.
There's probably some truth in this, but there are other factors. When you're alone, you tend to eat when you're hungry and, unless you're stress eating, tend not to have huge meals because meals by yourself are quite boring. When you're in a relationship, you tend to try to eat together, which means eating at a time when you're both at least vaguely interested in food, but probably not at the time when either of you would have chosen to eat. If you're cooking for two, you're also less likely to make enough by accident that it looks like another complete meal for both of you so you're more likely to just finish it all between the two of you.
The Cocoa Cola Company makes a lot of money selling bottled water. Sugar isn't important to them, they'll sell anything that they can stick in a bottle with a brand and encourage people to buy at a huge markup.
Not sure about Mozilla. Google does it by shipping an entire graphics stack that uses very little by way of system support (an OpenGL context if one's available or a frame buffer if not). Safari does the same thing.
Foxconn is increasingly using mostly automated factories, so the cost of labour is less important. I suspect that the 50,000 jobs was part of a briefing that said 'similar factories that we've built of this size employ 50,000 people' and forget to mention that this one will use fully automated production lines and will employ 50 people. The main requirements are reliable power, easy access to materials and components, good distribution routes for the final product, and lax environmental regulation (if the surroundings are uninhabitable at the end of the operating life of the factory, that's fine - they're building a new one anyway, may as well build it somewhere else...).
ASLR on a 64 bit device is a different story.
Well, sort of. The blind ROP stuff works by realising that exec triggers re-randomisation, but fork does not. Server processes like nginx are vulnerable because they fork children to replace any that have crashed and each child has the same layout. Even with a 64-bit address space, it's possible to probe (some things, like PLTs, are relatively easy to find and full of gadgets). Unless it's changed recently (I've not been paying attention, so it's possible), Android reduces app startup time by having a zygote process for the Java runtime environment, forking a copy of it, setting the UID / filesystem namespace, and then loading the app. This means that the address space will be different to other devices, but will have the same layout across multiple runs on the same device. BROP does require a two-way communication channel, though it only requires one-bit communication out (did this attempt crash or infinite loop the target?).
The real mitigation is not the ASLR, its that only 0.0001% of the population can write ROP exploits
Anyone can write ROP exploits. You don't need many gadgets to get a Turing-complete execution environment and there are compilers available that will take a binary, derive a Turing-complete virtual machine from them, and compile arbitrary (more or less - in some cases you may have a limited number of jumps, but it's usually more than enough to create a shell) code to run on them.
Big compile jobs are in the second category. When I'm doing a test build of FreeBSD, I generate around 50GB of object code (for LLVM it's only about 10GB for a debug build). The LLVM build system is nicely parallelised and so could happily consume 3,000 cores (linking would still be a bottleneck, but new linkers are becoming multithreaded) and building FreeBSD for every architecture is a similar proposition. FPGA place and route uses randomised algorithms and can basically consume as much CPU power as you want to throw at it. And, of course, these things grow over time (though a clean build of LLVM has gone from about an hour and a half when I started to under 5 minutes on my current laptop).
For everyday use, an oldish tablet is more than fast enough. For a few things, I'm going to keep complaining about the time it takes for a very long time.
Have they finally figured out how GUIs are supposed to work? They keep completely changing shit all the time.
Are you talking about Windows or Linux?
Control/command-x/c/v for cut/copy/paste are discoverable and consistent: if you got to the edit menu of any application (visible by the default on non-braindead UIs and the starting point for finding any command), then you will see the icons next to it telling you what the shortcut is. Now that it's discovered, it's consistent everywhere. Except in terminals if you're using control instead of command, because terminals need control-c for interrupt and so break the good UI, but on a Mac it's the same in every single application including the terminal.
In contrast, most of the Windows-key-plus-modifier combinations can only be discovered by reading the documentation. There's nothing that a user is encouraged by the UI to click on that tells them about what these modifiers are (though I vaguely remember that Windows 98 had a 'show desktop' icon in the start bar that told you about windows-d in the tooltip).
Intuitive implies that you are meant to use your psychic powers or some innate knowledge to find how the UI works. Good UIs do not work that way, they make it easy for people to learn and then they allow the user to apply the knowledge everywhere. If someone complains about a UI not being intuitive, then it's a good hint that they don't know anything about HCI.
The Direct3D thing is more silly, as OpenGL drivers on Windows XP expose all of the functionality from the hardware that Direct3D does on newer systems.
Since OpenGL 3, GL ES has been a subset of OpenGL, so you can run a GL ES program on any compliant OpenGL implementation. The main difference in the initial GL ES release was the half-precision floating point type, which was not widely supported in hardware on desktop GPUs.
11. Much of the UI became non-discoverable. I'm not sure if the default hiding of menu bars came with 8 or 7, but it meant that unless you knew that the alt button was magic you were unable to access the menus. Similarly, there was no discoverable way of exiting the Metro apps that would occasionally pop up when you accidentally hit one of the magic key combinations - alt-F4 works, but unless you know that that's a way to quit Windows apps, you're stuck.
I didn't realise how truly bad the UI was until my mother bought a new machine that came with Windows 8 just before I visited the Christmas before last. She's been using Windows since 3.1 and, though she's not exactly an expert, she's got more than a passing familiarity with the OS. Lots of things just left her completely stuck. I've no idea where MS found the people that they put in their usability testing lab, but they don't seem representative of users. When my girlfriend bought a new laptop, I persuaded her to buy one that came with a Windows 7 downgrade. It took her about two weeks before deciding that it was worth using, and she was someone who had managed to tolerate Vista for years. She seems pretty happy with 7 (though some parts of the UI suck: anyone know how to set up an ad-hoc WiFi network with Windows Vista, 7, or 8? The network config UIs are completely different in all three and I couldn't figure it out in any of them).
It's the circle. It was a house on Android 2.x, but that was considered too obvious an icon for home, so Google changed it. As far as I can tell, from using Android and from talking to some devs on the Android team at Google, everyone in the Android UI team is a drooling moron who could be replaced by a magpie and produce better results.
Cortana is not available in Canada because Microsoft enjoys giving a massive middle finger to anyone who's not an American
Wow, Microsoft is really buying into the whole 'let's copy Apple's business model as closely as we can' thing!
The hardware assist is a very small amount of signal processing to identify sibilants and plosives (something that a 2MHz 6502 could do in real time) and flag things that have the same pattern as a certain phrase ('Okay Google', I think - not used it) to wake up another core. I think in the Moto X, they use a small ARM core for it, though they may have a two-phase process where a hardware implementation wakes up a small ARM core that does a better job and wakes up the application cores if there's something interesting. In either case, it's probably quite difficult to change the pattern of sibilants and plosives that it looks for whatever Cortana wants, either because it's hard wired or because there's no software API for changing it.
And Siri was bought by Apple from an SRI spin-off, which developed the technology as part of a DARPA project. Lots of the components already exist in open source form (as the Sirius project has shown).
Given that these are obviously not valid DMCA take-down notices, I wonder if you could convincingly argue that they're advertising for the film...
There are several other factors too. $50k on a car is a pretty large capital investment in something that's going to depreciate quickly. If you're not using the car as much as him, then you'd probably be better off doing something else with the capital and letting someone else own the car (or own a fraction of one). If car sales drop off because most cars are being bought as part of fleets with fewer than one car per user, then prices are also likely to go up, so that $50k car may well not stay a $50k car.
His costs are around 25% lower than the IRS quotes for mileage, so I'm slightly sceptical of his calculations.
Rama is the same sort of story as 2001. They make good books and made good cinema in the '60s when impressive visual effects were enough to wow an audience. Now that CGI is cheap, it's expected by the audience and you need a bit more story. With both, a film doesn't give you long enough to tell the story well and so you end up with something that either feels rushed or superficial. Rama would make a good TV series, because you'd have the time to focus on the character development. It would be a terrible film.
You're spending almost $1000/month on a car and don't think that it would be possible for a service to make that cheaper? I bike most places and take taxis everywhere else and spend a fraction of that - even taking a taxi with a human driver to and from work every day and would be cheaper than you're spending. 500km/week puts you well outside the normal range for drivers, even in the US where the cities are carefully designed so that places people want to be are as far away from where people are as possible.
Not Bob specifically, but there were quite a few agent-based UIs floating around. They all failed for much the same reason as Bob (if you're going to have an agent between you and the computer, it needs to be really good at natural language parsing, or you're just building an obfuscated and poorly documented CLI).
[Quick question about the word "only". English is not my first language, and I get confused about the proper use of only. For example in the subject line, the word only applies to what? To the verb happens or to the phrase "in tech"?]
The definitive essay on this subject: Only His Only Grammarian Can Only Say Only What Only He Only Means.