No space elevator designs that are even vaguely plausible include a moving cable. To understand why, consider the mass of such a cable: the energy required to accelerate it and then decelerate it for the cars to start and stop would be phenomenal. You could potentially have a loop that would continuously move in a circle, but you'd still have problems starting it. Just dropping things from the top wouldn't be enough, because you'd need to get them a fair way down before they'd stop orbiting and actually provide force in the correct direction. I don't even want to think about the lateral forces that such a structure would have to endure.
I bought a new fridge about 5 years ago. I moved house and worked out that the difference in power consumption between the old fridge I had and the new one that I bought meant that the new fridge paid for itself in 2-3 years. Newer utilities are significantly lower power than ones from even the '80s and '90s. I bet that the next set of low-operating-cost white goods will all have some kind of Internet-related insecurity as standard.
i.e. with less warning than the revalidation time for a lot of safety-critical systems.
Anybody who knows about problems with leap days?
Well, aside from the Zune infinite looping...
Leap days (which we call leap years, because consistency is hard) are predictable. Software written 40 years ago will have the extra days at exactly the same times and with exactly the same frequency that the designers thought that they would. You never have problems where some parts of a distributed system got the update and others didn't. Either the code is working, or it's broken. It's also really easy to test.
You do understand that the navigation is ALSO intrinsically tied to the astronomical positioning of things, right?
Today? Mostly (for anything where accuracy matters to the degree that leap seconds will make a difference in under a few hundred years) it depends on the GPS position, or some equivalent. GPS time, unlike UTC, does not have leap seconds.
If we don't bother with leap seconds, then the distance that the sun will be off from being directly overhead at the equinox is about the same as it is now from being a couple of hundred miles away from the meridian. A simpler solution to the problem would be to, every couple of thousand years, have a one-hour reset. There is basically nothing that depends on the position of the sun in the sky to that level of accuracy, but there are a huge number of things (including all air-traffic control systems) that depend on keeping time in sync to sub-second accuracy and are safety critical. These things all need some special handling for every adjustment and an extra hour would be no more difficult for them than an extra second, so doing one big correction every couple of thousand years would be far, far cheaper. That's of course assuming that we still care much for a time system that's predicated on a single planet's relationship to its star in 2,000 years. It seems likely that we'll either be sufficiently disbursed that we don't, or that we'll have damaged our civilisation enough that we will have far bigger problems to deal with.
Are those numbers by currency unit or by transaction? I probably do about as many transactions in cash as by card (though contactless is reducing that quickly), but they're all very low value. I'll easily spend more in one card transaction than I will in a couple of weeks of cash transactions (I'm in the UK).
That's almost as silly as Red Mercury. Do you know why no terrorist plots have actually detonated a dirty bomb? It's not because radioactive materials are hard to get hold of, it's because building an effective dirty bomb is really hard. You have to find something that is sufficiently radioactive to be a problem, that is easy to disburse over a wide area, but which won't disburse so far / quickly that it will simply drop to background radioactivity levels.
Well, nothing aside from convincing the USSR to spend so much on defence that their economy collapsed, shortly followed by their political system.
Oh, and those intelligence agencies that are now crying for a greater ability to spy on us so that they can protect us from terrorists? They were completely surprised and thought that the collapse of the USSR was a hoax.
It's aimed at the same niche as Fortran, but comes with memory safety, a REPL environment and a JIT, and good interfacing with other languages (including Fortran) to reuse library code. Fortran is probably still a better choice if you're writing a library that lots of people will use.
I suspect that most people doing scientific computation have heard of Julia. It's a pretty neat language (and has a lot shallower a learning curve than Fortran, which is the big player in the space. If you're not in a field that uses Fortran a lot, Julia probably isn't too relevant to you). I had a student last year work on hoisting bounds checking out of loops to expose better optimisation opportunities for autovectorisation. In combination with Polly, this got a factor of 4-8 speedup for a lot of workloads and also paves the way for things like GPU offloading for critical loops. There are definitely a lot of places where some of that $600K could be spent on small improvements that would give huge improvements to end users.
To give a simple example: If you want to share a document between two apps on Windows Phone, then one passes it directly to the other. To do the same on Android, one writes it to the (typically emulated) SD card and now that you've completely bypassed the security model, everything else can open it. It would be easy to designate a directory on Windows Phone that would act as the SD card, but only Android apps would be allowed to use it, you wouldn't have the same user model for document sharing between Android and Windows apps as you do between Windows apps. There are lots of other issues with differences in standard widget implementations and so on. Even moving between iOS and Android (whose UIs are a lot more similar to each other than either is to Windows Phone), the difference in behaviour of the back button is confusing.
They are already popular despite that not happening yet. Wrong guess. Maybe try something other than a guess next time?
Where are you getting your numbers from? That's why we bought them, and it's why the companies that I talk to who buy them in lots of a thousand buy them. In the P4 days, we were buying Opterons almost exclusively. I think it's been five years since we last bought one.
It's really easy to reimplement it in that way if you want to - you can reuse all of the AOSP code directly and just plug in your own back ends for OpenGL, Skia, and so on. The problem is that you'll now have an Android app running on a Windows phone, looking and behaving like an Android app. If you've ever tried to use WINE for anything other than games (which try not to use the platform-native UI, as it breaks immersion) on OS X then you'll know how badly that sucks. You won't have an app that integrates nicely with the underlying system services (even things like copy and paste won't work correctly) and you'll have UI patterns that don't quite fit and feel weird and wrong, and a load of frustrated users. Even UI ports where a human is involved are very hard. Automated ones suck - remember Java's AWT?
So please explain why Transmeta didn't take off despite aiming directly for that metric
and good performance, there was no reason to buy Transmeta processors. Note that they're not actually dead: nVidia bought Transmeta and used their ideas in their Project Denver ARM SoCs, which are selling pretty well now, in a different market where performance-per-Watt does matter.
and why there are so many power hungry Xeons out there.
There aren't. Xeons are so popular precisely because they give you the best performance within a given power envelope that you can currently buy (unless you're willing to go with custom accelerators or less general cores such as GPUs). Xeons became really popular once they started beating Opterons in performance per Watt. The new Cavium ThunderX parts claim to be in the same ballparks (better on some workloads, worse on others), but there hasn't been much competition for Xeons for a while.
FLOPS/Watt matters a lot to the customers of this kind of thing. When you're spending a few tens of millions on the supercomputer, you really don't care what the CPUs or accelerator cores cost. You do care about power consumption though, because that translates to cooling requirements and directly limits how big a system you can build.
So what exactly is the real world application of such a beast?
All of the things where you really, really wish that you could do GPU offloading, but can't because you have diverging flow control and the GPU version ends up coming nowhere near to the theoretical peak performance of the hardware. The Xeon Phi cores are pretty simple, but there are loads of them, they have real branch prediction and caches (so handle the same kind of workloads as normal CPU cores, just a bit slower) and have fairly beefy vector units (so when they're running in a straight line they're actually pretty fast). Anything that you can make run on multiple threads should work nicely on them. That includes a load of HPC code that uses OpenMP, but which doesn't map particularly well to a CPU programming model without significant rewriting and redesigning of the core algorithms.
And that's why no one is deploying general homomorphic encryption. There are a few approaches that run almost as fast as doing the unencrypted work, but only support a very limited set of operations and have space requirements that scale linearly (or worse) with the number of operations that you want to support. If it's targeted at a specific application then it's conceivable that they've managed to create something that's fast, but only useful for a very specific use case.
Red tape, presumably. Anyway, because this is all handed off to private enterprise. Paid for with public money, most of which does not come from taxes off other corporations. So the net result is another stream of public funds into private hands. Guess who lobbied for that, and guess whose interests this does not serve.
Add to that two words: regulatory capture. With a revolving door between the large corporations and the regulators, a lot of that 'expensive' regulation is lobbied for by the established companies and has a disproportionate impact on smaller businesses trying to compete in that market.
It happens in the west, but it doesn't employ many people. Part of the reason that it's starting to come back, aside from QA concerns, is that cheap labour can't compete with machines (which work round the clock and produce exactly the same quality for all of that time). Even if the up-front investment is more, it's going to be offset over the lifetime of the factory and being closer to your consumers makes the supply chains easier to manage. That doesn't really help, if the capital is still controlled by very few people.
College councilors don't try to scare people away from this major (by, say, giving stats on how competitive the industry is), because it brings in too much money for the school.
This is the second comment I've seen along this line in the thread. Isn't this the job of school careers advisors, before the child applies to universities?
Slashdot Mirror
No space elevator designs that are even vaguely plausible include a moving cable. To understand why, consider the mass of such a cable: the energy required to accelerate it and then decelerate it for the cars to start and stop would be phenomenal. You could potentially have a loop that would continuously move in a circle, but you'd still have problems starting it. Just dropping things from the top wouldn't be enough, because you'd need to get them a fair way down before they'd stop orbiting and actually provide force in the correct direction. I don't even want to think about the lateral forces that such a structure would have to endure.
I bought a new fridge about 5 years ago. I moved house and worked out that the difference in power consumption between the old fridge I had and the new one that I bought meant that the new fridge paid for itself in 2-3 years. Newer utilities are significantly lower power than ones from even the '80s and '90s. I bet that the next set of low-operating-cost white goods will all have some kind of Internet-related insecurity as standard.
Your link doesn't go anywhere, which makes me think that it's a retracted story based on the initial announcement, which turned out to be incorrect.
Leap seconds are announced months in advance
i.e. with less warning than the revalidation time for a lot of safety-critical systems.
Anybody who knows about problems with leap days?
Well, aside from the Zune infinite looping...
Leap days (which we call leap years, because consistency is hard) are predictable. Software written 40 years ago will have the extra days at exactly the same times and with exactly the same frequency that the designers thought that they would. You never have problems where some parts of a distributed system got the update and others didn't. Either the code is working, or it's broken. It's also really easy to test.
You do understand that the navigation is ALSO intrinsically tied to the astronomical positioning of things, right?
Today? Mostly (for anything where accuracy matters to the degree that leap seconds will make a difference in under a few hundred years) it depends on the GPS position, or some equivalent. GPS time, unlike UTC, does not have leap seconds.
If we don't bother with leap seconds, then the distance that the sun will be off from being directly overhead at the equinox is about the same as it is now from being a couple of hundred miles away from the meridian. A simpler solution to the problem would be to, every couple of thousand years, have a one-hour reset. There is basically nothing that depends on the position of the sun in the sky to that level of accuracy, but there are a huge number of things (including all air-traffic control systems) that depend on keeping time in sync to sub-second accuracy and are safety critical. These things all need some special handling for every adjustment and an extra hour would be no more difficult for them than an extra second, so doing one big correction every couple of thousand years would be far, far cheaper. That's of course assuming that we still care much for a time system that's predicated on a single planet's relationship to its star in 2,000 years. It seems likely that we'll either be sufficiently disbursed that we don't, or that we'll have damaged our civilisation enough that we will have far bigger problems to deal with.
And anyway... "News flash, giant multinational project sees schedule slip - details at 9!"
Multinational? But the headline says that it's French! I thought it was only a multinational project in stories with a positive spin.
Are those numbers by currency unit or by transaction? I probably do about as many transactions in cash as by card (though contactless is reducing that quickly), but they're all very low value. I'll easily spend more in one card transaction than I will in a couple of weeks of cash transactions (I'm in the UK).
That's almost as silly as Red Mercury. Do you know why no terrorist plots have actually detonated a dirty bomb? It's not because radioactive materials are hard to get hold of, it's because building an effective dirty bomb is really hard. You have to find something that is sufficiently radioactive to be a problem, that is easy to disburse over a wide area, but which won't disburse so far / quickly that it will simply drop to background radioactivity levels.
Well, nothing aside from convincing the USSR to spend so much on defence that their economy collapsed, shortly followed by their political system.
Oh, and those intelligence agencies that are now crying for a greater ability to spy on us so that they can protect us from terrorists? They were completely surprised and thought that the collapse of the USSR was a hoax.
It's aimed at the same niche as Fortran, but comes with memory safety, a REPL environment and a JIT, and good interfacing with other languages (including Fortran) to reuse library code. Fortran is probably still a better choice if you're writing a library that lots of people will use.
Unfortunately it has not chance of dislodging R,
The places I've seen Julia make the most inroads, it's been displacing Fortran, not R. And anything that displaces Fortran makes me happy.
Your enumeration of requirements omits scientific computing, which is the niche where Fortran is currently dominant and Julia is growing.
I suspect that most people doing scientific computation have heard of Julia. It's a pretty neat language (and has a lot shallower a learning curve than Fortran, which is the big player in the space. If you're not in a field that uses Fortran a lot, Julia probably isn't too relevant to you). I had a student last year work on hoisting bounds checking out of loops to expose better optimisation opportunities for autovectorisation. In combination with Polly, this got a factor of 4-8 speedup for a lot of workloads and also paves the way for things like GPU offloading for critical loops. There are definitely a lot of places where some of that $600K could be spent on small improvements that would give huge improvements to end users.
To give a simple example: If you want to share a document between two apps on Windows Phone, then one passes it directly to the other. To do the same on Android, one writes it to the (typically emulated) SD card and now that you've completely bypassed the security model, everything else can open it. It would be easy to designate a directory on Windows Phone that would act as the SD card, but only Android apps would be allowed to use it, you wouldn't have the same user model for document sharing between Android and Windows apps as you do between Windows apps. There are lots of other issues with differences in standard widget implementations and so on. Even moving between iOS and Android (whose UIs are a lot more similar to each other than either is to Windows Phone), the difference in behaviour of the back button is confusing.
They are already popular despite that not happening yet. Wrong guess. Maybe try something other than a guess next time?
Where are you getting your numbers from? That's why we bought them, and it's why the companies that I talk to who buy them in lots of a thousand buy them. In the P4 days, we were buying Opterons almost exclusively. I think it's been five years since we last bought one.
It's really easy to reimplement it in that way if you want to - you can reuse all of the AOSP code directly and just plug in your own back ends for OpenGL, Skia, and so on. The problem is that you'll now have an Android app running on a Windows phone, looking and behaving like an Android app. If you've ever tried to use WINE for anything other than games (which try not to use the platform-native UI, as it breaks immersion) on OS X then you'll know how badly that sucks. You won't have an app that integrates nicely with the underlying system services (even things like copy and paste won't work correctly) and you'll have UI patterns that don't quite fit and feel weird and wrong, and a load of frustrated users. Even UI ports where a human is involved are very hard. Automated ones suck - remember Java's AWT?
So please explain why Transmeta didn't take off despite aiming directly for that metric
and good performance, there was no reason to buy Transmeta processors. Note that they're not actually dead: nVidia bought Transmeta and used their ideas in their Project Denver ARM SoCs, which are selling pretty well now, in a different market where performance-per-Watt does matter.
and why there are so many power hungry Xeons out there.
There aren't. Xeons are so popular precisely because they give you the best performance within a given power envelope that you can currently buy (unless you're willing to go with custom accelerators or less general cores such as GPUs). Xeons became really popular once they started beating Opterons in performance per Watt. The new Cavium ThunderX parts claim to be in the same ballparks (better on some workloads, worse on others), but there hasn't been much competition for Xeons for a while.
It's 20 years since 1989?
FLOPS/Watt matters a lot to the customers of this kind of thing. When you're spending a few tens of millions on the supercomputer, you really don't care what the CPUs or accelerator cores cost. You do care about power consumption though, because that translates to cooling requirements and directly limits how big a system you can build.
So what exactly is the real world application of such a beast?
All of the things where you really, really wish that you could do GPU offloading, but can't because you have diverging flow control and the GPU version ends up coming nowhere near to the theoretical peak performance of the hardware. The Xeon Phi cores are pretty simple, but there are loads of them, they have real branch prediction and caches (so handle the same kind of workloads as normal CPU cores, just a bit slower) and have fairly beefy vector units (so when they're running in a straight line they're actually pretty fast). Anything that you can make run on multiple threads should work nicely on them. That includes a load of HPC code that uses OpenMP, but which doesn't map particularly well to a CPU programming model without significant rewriting and redesigning of the core algorithms.
And that's why no one is deploying general homomorphic encryption. There are a few approaches that run almost as fast as doing the unencrypted work, but only support a very limited set of operations and have space requirements that scale linearly (or worse) with the number of operations that you want to support. If it's targeted at a specific application then it's conceivable that they've managed to create something that's fast, but only useful for a very specific use case.
Red tape, presumably. Anyway, because this is all handed off to private enterprise. Paid for with public money, most of which does not come from taxes off other corporations. So the net result is another stream of public funds into private hands. Guess who lobbied for that, and guess whose interests this does not serve.
Add to that two words: regulatory capture. With a revolving door between the large corporations and the regulators, a lot of that 'expensive' regulation is lobbied for by the established companies and has a disproportionate impact on smaller businesses trying to compete in that market.
It happens in the west, but it doesn't employ many people. Part of the reason that it's starting to come back, aside from QA concerns, is that cheap labour can't compete with machines (which work round the clock and produce exactly the same quality for all of that time). Even if the up-front investment is more, it's going to be offset over the lifetime of the factory and being closer to your consumers makes the supply chains easier to manage. That doesn't really help, if the capital is still controlled by very few people.
College councilors don't try to scare people away from this major (by, say, giving stats on how competitive the industry is), because it brings in too much money for the school.
This is the second comment I've seen along this line in the thread. Isn't this the job of school careers advisors, before the child applies to universities?