I thought they had negotiated a deal with the Pleiedes cluster. We would provide them with instruments of torture (Noel Edmunds, Ken Dodd, Keith Chegwin and Chris Tarrent), and they would provide us with a copy of all transmissions that had been beamed into space (other than those featuring said instruments of torture).
I've been working with computers for a while now - I started in 1978 - and have a fairly clear idea of how much it costs to build a solid server that can support a typical primary school, provide a high level of robustness, be easily maintained, and be reasonably secure. Five hundred squid for a server is very impressive and would probably require access to wholesale prices or some help - at least for most people - but I've built very reasonable desktops for a shade under that at retail cost for parts, so I know it is an achievable goal. To actually achieve it for a server - ah, now that is the impressive part. The theory of custom-built machines is often different from the practice. I salute you for managing what is actually quite a challenging task.
I think you mean "abusive, idiotic" management. And my experience has been that skilled IT workers are hired only until a code monkey can be stuffed in there to replace them - assuming anyone is thought to be needed at all. The most "amusing" situation I've faced was a prohibition on charging overtime. Overtime was mandatory, but could not be billed as time worked. It was equally required that all time in the office be accounted for. This meant that overtime was listed as negative vacation. This wasn't time anyone could actually use, but was an excellent trick for getting the hours down to 40 per week, no matter how long anyone actually worked in practice.
I have never aged so rapidly as when I've been working for some of the more truly abusive bosses. Sadly, employment in America is "by will" and there's no meaningful support for IT employees. Talk of simply going to another job assumes (a) that any such other job will be any better - in practice, the "geographic cure" is merely a good way to waste time and money, and (b) that we're still in the Dot Com era where IT jobs outnumbered employees by an order of magnitude. We're in a recession, guys. Outside a software shop, IT is merely a support role, and support roles earn no income. They're expendable.
Very likely, if you're in the areas of Bismarck or Jamestown - doubly so if you know families that moved over from Norway or Poland in the 1800s. If your last name is Wyngarden or Woychick, you probably don't just know my mother but are probably a long-lost cousin.
Thermionic valves have vastly superior tolerance to electromagnetic radiation, acceleration, shocks, and other hostile conditions. The power they can push through is astonishing. On the flip-side, their mean time between failures isn't always so great, they're bulkier, they use more power, and it can be very hard to find some of the older lines.
If you wanted to build some part of an embedded device that absolutely had to take some really ugly conditions, you could do a whole lot worse than to build that specific module using valves. Let's say you wanted to build a new module for the IIS, for example. The internal circuits can largely be protected, so conventional radiation-proof chips would be fine. However, if you wanted reliable computing elements that could be strapped to the outside of the pod, you've harsh conditions indeed. Lead-smothered rad-hardened silicon chips that can handle space tolerances and have their own heating elements would probably work. Lots of things that can go wrong, though. Complexity-wise and weight-wise you're probably not significantly better off than using thermionic valves with none of the extras.
Where else could valves be used? Easy. If the cathode and anode are deliberately mis-aligned, then one or more grids must be set to a value such that the directed power completes the circuit. If something goes wrong (too much power, something fails, whatever), then the beam is either not pushed at all or pushed far too far. In either case, you've an all-electronic circuit-breaker - ideal if you want to get rid of fuseboxes and mechanical trip-switches.
Nobody, but nobody, buys version 1 of a product - if they've any sense. It's bad enough to buy a whole-number release (those are likely to be the buggiest) but version 1 is a huge no-no. In the case of Microsoft, the first service pack has acquired a reputation for not being good either. Virtually all Windows SP1 releases have been followed rapidly by hotfixes and even other service packs. This isn't unique to Windows - the majority of brown paper bag releases of the Linux kernel that seriously impact users are also x.y.0 releases. It's a fundamental principle of software purchase that has always been true and will likely always be true.
On the other hand, Vista was under-developed, rushed, and had integral features removed. That last part is more significant than it might first appear. If you remove chunks out of the foundations of a building, you can expect the building to collapse. The same is true in software - if it's designed to be present, then removing that feature will destabilize everything depending on it. Yes, it was late. So what. The contribution Vista is making to Microsoft is negligible in terms of sales and disastrous in terms of PR in the European courts. Investing a year or two more work into the project would have been cheaper, produced a better product and generally given Microsoft a lot of plusses.
There was pressure for Vista being released. Yeah, and a company that can pay billions in daily fines without working up a sweat needs to pay attention to such pressure why? Due to lost market share? Lost to whom? Other OS' may be catching up, but it'll be five to ten years before they can capture significant marketshare. Three or four years more development would have kept Microsoft's lead and secured it with far less risk of legal retribution.
All in all, Vista's release marked very poor marketing decisions, not just very poor technical ones, although it need not have been that way.
I can't use my 25th level hamster mage. (For those interested, dragon eyes have limited shielding between the lids and the brain, and afterwards you get 1XP per GP stolen. It just takes a while to carry off enough to get the spells to nick the rest.)
Last I heard, artificial diamonds were neither large nor particularly flaw-free. But that could have changed substantially. Obviously not by enough to threaten the world markets - yet. There are interesting technical challenges in making flawless ultra-pure mono-isotopic crystals of any kind, and quite a few financial problems, but if they can be solved for one crystal then I see no reason why they could not be solved for this theoretical crystal, provided the structure is similar enough.
You can't know you have 0.1% accuracy. You can't stipulate any bounds on precision or certainty at all. For further information, see James Gleik's "Chaos" and Benoit Mandelbrot's "Fractal Geometry of Nature".
It boils down to Information Theory. Isaac Asimov describes it well in his Foundation series, where Hari Seldon is uncertain if psychohistory can exist, that maybe the simplest description of the system is the system itself. It simply cannot be reduced further without introducing so many errors (Butterfly Effect) as to make the reduced model useless.
If this is the case, then the only computer you could run a simulation of the Universe on and get results would be the Universe itself and no simulator is required or - indeed - usable.
They're non-differentiable, so the only way to calculate them is to perform each and every step. As they are also infinitely sensitive to intial conditions, you must perform an infinite number of steps at infinte precision in order to guarantee the calculations are correct.
In the end, this is where a simulator MUST break down. It will always be possible for another simulator to take the smallest sub-division of time that the first simulator is using and then divide it further. If this was the case, you would be able to take a short enough timeframe over a small enough system and demonstrate that increasing the accuracy of the calculations actually produced a decrease in the accuracy of the results. The simulator of the Universe would be subject to the Butterfly Effect before your simulator was, and so the "real" results would contradict what was expected.
If you can show an example of this, you can show the Universe can only be a simulator. It can't be anything else. If you cannot show this, if you can't find a way to gain greater precision than the Universe Simulator, then no such simulator exists.
Ultimately, it boils down to this. A computer, no matter how advanced, must be finite in scope. It may be very very large, but it cannot be infinite. It is bound by the properties of physical state, which are strictly quantized. There are simply not an infinite number of states for any given component, no place to put an infinite number of components, and no way to link an infinite number of components together in a way that would take finite time to perform any useful work and still be thought of as part of the same system. This is a constraint any simulator must have - it must be able to simulate finite time in finite time.
Physical systems are bound by other laws. It's unclear what laws of nature must hold true for all universes within the multiverse - laws so fundamental that even if the universe was simulated, the simulator must also adhere to those laws. However, some such laws almost certainly exist. Determining those would be helpful, as they place additional constraints.
This is the crux of my argument. You only need to find a constraint - any constraint - which must apply to a simulator but which need not apply to the raw essential ingredients within the Universe itself. A simulator can either be event-driven (ie: go to the next thing that happens), or it can step through in small increments of time. If you can show that time is quantized but that you cannot predict that from first principles alone, then you've an excellent case for a simulator. If you can show that causality cannot be violated at the quantum level, but QM predicts that this should happen, you've an excellent case for a simulator.
If, however, you can show that there is any trait in the Universe whatsoever that would require an infinite number of steps at infinite precision to duplicate, you've a strong case against a simulator. Likewise, if you can show from first principles alone that there is a single facet of the Universe which cannot be modeled by anything simpler than itself, you've again a strong argument against a simulator.
These last two cases are widespread throughout the Universe. Strange Attrators and The Butterfly Effect are the norm, not the exception. Quantum wormholes at any non-zero relative speed (which will be all of them) and indeed Quantum foam itself violate causality as a matter of course. Time does not appear to be quantized, and the Universe is just too complex to be event-driven.
The Universe is probably a great many things, some unprintable, but a simulator isn't one of them.
The point of a chaotic system is that it is infinitely sensitive to initial conditions (so you can never know the initial conditions well enough), that you need to use infinitesimal steps (implicit in it being infinitely sensitive, the result of which being that you can never make the system fine-enough grained to be able to step through it), and that the system is non-differentiable (you cannot interpolate, you cannot take short-cuts, you cannot infer - the only way to know the result of N steps is to take all N of them).
In consequence, it is non-predictable (the only way to know the result is to actually go through each and every step, of which there are an infinite number and each of which requires infinite precision) but wholly deterministic (there are no random elements, if the entire system was reset exactly to an infinite degree of precision to a given state, then it would produce identical results).
But here is the crux of it. You cannot store an infinitely accurate number on a computer, nor can you execute an infinite number of steps. The best a computer can ever do is a crude approximation. If you were to calculate the Mandelbrot Set at a resolution in which there is no more information, you would run out of places to store the individual floating-point numbers long before you got to the point of actually performing any arithmetic on them. Such a shape cannot be represented as a whole by anything simpler or more compact than itself.
There are some valuable ideas that can be borrowed from such lines of thinking. First off, if someone made a computer simulation of you and placed that simulation in a restricted environment (and you in an identical physical environment), how could you determine which one of you was real? Is there a way to know?
The answer is yes - chaos theory can be simulated on a computer but it is not going to be as sensitive to initial conditions as something in an analogue universe. Thus, chaotic systems on a computer can be repeated. Real-world chaotic systems never can.
The next thing that can be drawn is a better understanding of the brain. It should be obvious by now that the senses do not link directly to the conciousness but rather are used to update a mental "virtual reality" construct within the brain. Thus, everyone is living in their own virtual reality in a sense. This is easily demonstrated - there are hundreds of psychological tests that show how the brain fills in missing information, which only makes sense if there is some internal model from which such information can be obtained.
On the other hand, people on the autistic spectrum have fewer mirror neurons and show abnormal activity in the pre-frontal lobes, according to fMRI scans. They are also well-known for having an astonishing level of focus to the point where activity beyond a relatively low level is painfully overloading. This would make sense using this VR idea, as their brains' internal VR would be skewed from experience, above a certain level of input, creating intense stress and confusion. Exactly what you find with people on the autistic spectrum.
Does this internal VR model mean that all of reality is a VR model? No. If it did, then the VR models could always agree even when there is a bottleneck or information degradation. Since this is clearly not the case, it seems reasonable to conclude that the brain's VR is a crude approximation to reality and not reality itself.
Doesn't the idea of the conciousness existing within a brain-level VR contradict the notion of experimental science? No. The VR is not what you experience, the VR - or whatever you want to call it - is simply a mental construct to allow the brain to anticipate and to act in advance of actual data, or act where actual data is too noisy to directly use. Processing sensory data is hard work and can't possibly be done in real-time all of the time. However, measurements are not made in real-time. You observe a system as it exists in a snapshot of time, and you can continue observing that snapshot all you like. Since that is the case, any momentary disparity between the internal VR and the external world should be eliminated.
I believe it is testable. All computers ultimately reduce to the Turing Machine. This includes neural networks and at least some classes of quantum computer. (Heresy, I know. Terrible. Now go find a medium-rare steak to burn me on.) However, not all problems reduce to computable problems. If there is a non-computable system that exists in the real world, then it cannot be the product of a simulation, no matter how advanced the computer is.
Do such problems exist? Well, chaos theory is full of them. You cannot have a system that is truly chaotic and computable at the same time - the two are mutually exclusive. Both are deterministic, but only one is predictable.
Oolite is good, but the chief maintainer quit and 1.65 is getting long in the tooth. I couldn't get the hang of Vega Strike. Nice background stills, but nothing of the feel that the Elite family of games gives. There's nothing quite like mashing up a tech 1 anarchy in a beat-up spaceship with a trusty military beam by your side. And front. And back.
True, but whatever potential skills a child has are useless without (a) information, (b) an opportunity to experiment with that information, and (c) exchange information with other people who have similar interests.
This isn't a guaranteed way to learn, but history suggests that groups tend to learn efficiently when they have not only the right information but also the right conditions. What constitutes the right conditions is not something that is really known, but it's a fair bet that "Classical" schooling is superior in many ways to modern schooling when it comes to diversity and understanding.
Well, Open Office, any Open Office derivative (and there do seem to be a few), a variety of other office products, etc. Enforced obsolescence is nothing new, almost by definition. It is more of a problem when Microsoft pulls such a stunt, as opposed to say graphics manipulation packages that drop GIF support, simply because Microsoft is bigger. Oh, and yes, I know why GIF support was pulled - damn software patents - but the principle is the same: abolition of an old, largely redundant format so as to force users to adopt the format of the developer's choice, be that a benign or malevolent choice. In the end, it is the Will of the Maintainers that rules supreme, a techno-theocracy whether the deity is a freeing or enslaving one.
Of course, it is impossible for one program to support all formats. You end up with an exponential growth in the number of translation mechanisms needed, if you do that. Ideally, you'd have everything translated to/from a universal common meta-format, which is the premise of the next-generation file formats. Of course, such formats are still sufficiently flat that they cannot be truly universal. I'd hate to use any such meta-formats to represent structured data stored in an extension for NetCDF 4 over HDF 5 over ROMIO, the parallel I/O library. But then if meta-formats were any good at being universal, why would anyone want to have such a horribly complex layer-over-layer-over-layer mishmash in order to store/retrieve data? Such formats might be good for processing, because that's what they're good at, but a universal meta-format would be superior for storage as it would interact with more software, right?
Wrong. Specialist formats really are needed for some things, making universal translation unlikely to happen any time soon, in turn making some formats, well, doomed to be dropped. Although there are better ways for this to happen.
Things like Libraries of Congress, Libraries of Alexandria, Spams per Square Inch. You know, the units that people have become familiar with. Besides which, are they power-two gigagytes or SI gigabytes? Also, how much bandwidth is needed to shift all that data? In the standard Imperial units of Clay Tablets per German Juggernaut per unit of French motorway, naturally.
Well, attempts by Open Source developers to borrow from BBC Elite to produce a comprehensive open-ended gaming environment have so far not achieved a whole lot. Partly through legal complications, but also through lack of developers. I can't remember the last time the TORC group actually produced a release. Development on Empire seems limited to non-existant. The Netrek genre seems to have died. XTank was interesting, but died through licensing complications. Very few MUD or MUSH servers are under any kind of development, and it's limited at best. LambdaMOO has faded into oblivion. Omega offered an interesting twist to the Rouge/Nethack family, but the entire family seems to have been vanquished by a dragon.
In other words, the projects exist. People have been interested. But for whatever reason - my bet is a mix of it being hard and developers making lousy publicists - the efforts have struggled to maintain sufficient interest and have eventually collapsed through brain drains and burnout taking out original developers with nobody to replace them.
Slashdot Mirror
I thought they had negotiated a deal with the Pleiedes cluster. We would provide them with instruments of torture (Noel Edmunds, Ken Dodd, Keith Chegwin and Chris Tarrent), and they would provide us with a copy of all transmissions that had been beamed into space (other than those featuring said instruments of torture).
I've been working with computers for a while now - I started in 1978 - and have a fairly clear idea of how much it costs to build a solid server that can support a typical primary school, provide a high level of robustness, be easily maintained, and be reasonably secure. Five hundred squid for a server is very impressive and would probably require access to wholesale prices or some help - at least for most people - but I've built very reasonable desktops for a shade under that at retail cost for parts, so I know it is an achievable goal. To actually achieve it for a server - ah, now that is the impressive part. The theory of custom-built machines is often different from the practice. I salute you for managing what is actually quite a challenging task.
I have never aged so rapidly as when I've been working for some of the more truly abusive bosses. Sadly, employment in America is "by will" and there's no meaningful support for IT employees. Talk of simply going to another job assumes (a) that any such other job will be any better - in practice, the "geographic cure" is merely a good way to waste time and money, and (b) that we're still in the Dot Com era where IT jobs outnumbered employees by an order of magnitude. We're in a recession, guys. Outside a software shop, IT is merely a support role, and support roles earn no income. They're expendable.
Very likely, if you're in the areas of Bismarck or Jamestown - doubly so if you know families that moved over from Norway or Poland in the 1800s. If your last name is Wyngarden or Woychick, you probably don't just know my mother but are probably a long-lost cousin.
My mother is from North Dakota, so naturally I deny it exists. :)
If you wanted to build some part of an embedded device that absolutely had to take some really ugly conditions, you could do a whole lot worse than to build that specific module using valves. Let's say you wanted to build a new module for the IIS, for example. The internal circuits can largely be protected, so conventional radiation-proof chips would be fine. However, if you wanted reliable computing elements that could be strapped to the outside of the pod, you've harsh conditions indeed. Lead-smothered rad-hardened silicon chips that can handle space tolerances and have their own heating elements would probably work. Lots of things that can go wrong, though. Complexity-wise and weight-wise you're probably not significantly better off than using thermionic valves with none of the extras.
Where else could valves be used? Easy. If the cathode and anode are deliberately mis-aligned, then one or more grids must be set to a value such that the directed power completes the circuit. If something goes wrong (too much power, something fails, whatever), then the beam is either not pushed at all or pushed far too far. In either case, you've an all-electronic circuit-breaker - ideal if you want to get rid of fuseboxes and mechanical trip-switches.
Nature's techno-science hacks are all Bohr-ing. However, shouldn't naked vaccuum tubes be in Pentode Magazine?
I think the plan is to use The Blinkenlights Project and every hall they can find in a one mile radius.
Plan 9 is technically old, but because of the way it was handled prior to open-sourcing, I'd argue that it's really effectively young.
....that the chips were actually salvaged from a fleet of BBC television detector vans?
On the other hand, Vista was under-developed, rushed, and had integral features removed. That last part is more significant than it might first appear. If you remove chunks out of the foundations of a building, you can expect the building to collapse. The same is true in software - if it's designed to be present, then removing that feature will destabilize everything depending on it. Yes, it was late. So what. The contribution Vista is making to Microsoft is negligible in terms of sales and disastrous in terms of PR in the European courts. Investing a year or two more work into the project would have been cheaper, produced a better product and generally given Microsoft a lot of plusses.
There was pressure for Vista being released. Yeah, and a company that can pay billions in daily fines without working up a sweat needs to pay attention to such pressure why? Due to lost market share? Lost to whom? Other OS' may be catching up, but it'll be five to ten years before they can capture significant marketshare. Three or four years more development would have kept Microsoft's lead and secured it with far less risk of legal retribution.
All in all, Vista's release marked very poor marketing decisions, not just very poor technical ones, although it need not have been that way.
I can't use my 25th level hamster mage. (For those interested, dragon eyes have limited shielding between the lids and the brain, and afterwards you get 1XP per GP stolen. It just takes a while to carry off enough to get the spells to nick the rest.)
Last I heard, artificial diamonds were neither large nor particularly flaw-free. But that could have changed substantially. Obviously not by enough to threaten the world markets - yet. There are interesting technical challenges in making flawless ultra-pure mono-isotopic crystals of any kind, and quite a few financial problems, but if they can be solved for one crystal then I see no reason why they could not be solved for this theoretical crystal, provided the structure is similar enough.
You can't know you have 0.1% accuracy. You can't stipulate any bounds on precision or certainty at all. For further information, see James Gleik's "Chaos" and Benoit Mandelbrot's "Fractal Geometry of Nature".
If this is the case, then the only computer you could run a simulation of the Universe on and get results would be the Universe itself and no simulator is required or - indeed - usable.
In the end, this is where a simulator MUST break down. It will always be possible for another simulator to take the smallest sub-division of time that the first simulator is using and then divide it further. If this was the case, you would be able to take a short enough timeframe over a small enough system and demonstrate that increasing the accuracy of the calculations actually produced a decrease in the accuracy of the results. The simulator of the Universe would be subject to the Butterfly Effect before your simulator was, and so the "real" results would contradict what was expected.
If you can show an example of this, you can show the Universe can only be a simulator. It can't be anything else. If you cannot show this, if you can't find a way to gain greater precision than the Universe Simulator, then no such simulator exists.
Physical systems are bound by other laws. It's unclear what laws of nature must hold true for all universes within the multiverse - laws so fundamental that even if the universe was simulated, the simulator must also adhere to those laws. However, some such laws almost certainly exist. Determining those would be helpful, as they place additional constraints.
This is the crux of my argument. You only need to find a constraint - any constraint - which must apply to a simulator but which need not apply to the raw essential ingredients within the Universe itself. A simulator can either be event-driven (ie: go to the next thing that happens), or it can step through in small increments of time. If you can show that time is quantized but that you cannot predict that from first principles alone, then you've an excellent case for a simulator. If you can show that causality cannot be violated at the quantum level, but QM predicts that this should happen, you've an excellent case for a simulator.
If, however, you can show that there is any trait in the Universe whatsoever that would require an infinite number of steps at infinite precision to duplicate, you've a strong case against a simulator. Likewise, if you can show from first principles alone that there is a single facet of the Universe which cannot be modeled by anything simpler than itself, you've again a strong argument against a simulator.
These last two cases are widespread throughout the Universe. Strange Attrators and The Butterfly Effect are the norm, not the exception. Quantum wormholes at any non-zero relative speed (which will be all of them) and indeed Quantum foam itself violate causality as a matter of course. Time does not appear to be quantized, and the Universe is just too complex to be event-driven.
The Universe is probably a great many things, some unprintable, but a simulator isn't one of them.
In consequence, it is non-predictable (the only way to know the result is to actually go through each and every step, of which there are an infinite number and each of which requires infinite precision) but wholly deterministic (there are no random elements, if the entire system was reset exactly to an infinite degree of precision to a given state, then it would produce identical results).
But here is the crux of it. You cannot store an infinitely accurate number on a computer, nor can you execute an infinite number of steps. The best a computer can ever do is a crude approximation. If you were to calculate the Mandelbrot Set at a resolution in which there is no more information, you would run out of places to store the individual floating-point numbers long before you got to the point of actually performing any arithmetic on them. Such a shape cannot be represented as a whole by anything simpler or more compact than itself.
The answer is yes - chaos theory can be simulated on a computer but it is not going to be as sensitive to initial conditions as something in an analogue universe. Thus, chaotic systems on a computer can be repeated. Real-world chaotic systems never can.
The next thing that can be drawn is a better understanding of the brain. It should be obvious by now that the senses do not link directly to the conciousness but rather are used to update a mental "virtual reality" construct within the brain. Thus, everyone is living in their own virtual reality in a sense. This is easily demonstrated - there are hundreds of psychological tests that show how the brain fills in missing information, which only makes sense if there is some internal model from which such information can be obtained.
On the other hand, people on the autistic spectrum have fewer mirror neurons and show abnormal activity in the pre-frontal lobes, according to fMRI scans. They are also well-known for having an astonishing level of focus to the point where activity beyond a relatively low level is painfully overloading. This would make sense using this VR idea, as their brains' internal VR would be skewed from experience, above a certain level of input, creating intense stress and confusion. Exactly what you find with people on the autistic spectrum.
Does this internal VR model mean that all of reality is a VR model? No. If it did, then the VR models could always agree even when there is a bottleneck or information degradation. Since this is clearly not the case, it seems reasonable to conclude that the brain's VR is a crude approximation to reality and not reality itself.
Doesn't the idea of the conciousness existing within a brain-level VR contradict the notion of experimental science? No. The VR is not what you experience, the VR - or whatever you want to call it - is simply a mental construct to allow the brain to anticipate and to act in advance of actual data, or act where actual data is too noisy to directly use. Processing sensory data is hard work and can't possibly be done in real-time all of the time. However, measurements are not made in real-time. You observe a system as it exists in a snapshot of time, and you can continue observing that snapshot all you like. Since that is the case, any momentary disparity between the internal VR and the external world should be eliminated.
Do such problems exist? Well, chaos theory is full of them. You cannot have a system that is truly chaotic and computable at the same time - the two are mutually exclusive. Both are deterministic, but only one is predictable.
Oolite is good, but the chief maintainer quit and 1.65 is getting long in the tooth. I couldn't get the hang of Vega Strike. Nice background stills, but nothing of the feel that the Elite family of games gives. There's nothing quite like mashing up a tech 1 anarchy in a beat-up spaceship with a trusty military beam by your side. And front. And back.
This isn't a guaranteed way to learn, but history suggests that groups tend to learn efficiently when they have not only the right information but also the right conditions. What constitutes the right conditions is not something that is really known, but it's a fair bet that "Classical" schooling is superior in many ways to modern schooling when it comes to diversity and understanding.
Of course, it is impossible for one program to support all formats. You end up with an exponential growth in the number of translation mechanisms needed, if you do that. Ideally, you'd have everything translated to/from a universal common meta-format, which is the premise of the next-generation file formats. Of course, such formats are still sufficiently flat that they cannot be truly universal. I'd hate to use any such meta-formats to represent structured data stored in an extension for NetCDF 4 over HDF 5 over ROMIO, the parallel I/O library. But then if meta-formats were any good at being universal, why would anyone want to have such a horribly complex layer-over-layer-over-layer mishmash in order to store/retrieve data? Such formats might be good for processing, because that's what they're good at, but a universal meta-format would be superior for storage as it would interact with more software, right?
Wrong. Specialist formats really are needed for some things, making universal translation unlikely to happen any time soon, in turn making some formats, well, doomed to be dropped. Although there are better ways for this to happen.
Things like Libraries of Congress, Libraries of Alexandria, Spams per Square Inch. You know, the units that people have become familiar with. Besides which, are they power-two gigagytes or SI gigabytes? Also, how much bandwidth is needed to shift all that data? In the standard Imperial units of Clay Tablets per German Juggernaut per unit of French motorway, naturally.
In other words, the projects exist. People have been interested. But for whatever reason - my bet is a mix of it being hard and developers making lousy publicists - the efforts have struggled to maintain sufficient interest and have eventually collapsed through brain drains and burnout taking out original developers with nobody to replace them.