I don't know the exact dimensions of A4, but it's similar to 8.5"x11" (at . Suppose each spot of color stores 24 bits of information -- this is really a stretch, for reasons I'll get to below. Normal Xerox paper supports something like 300 dpi of resolution (that is the scale of the fibers of the paper). Then you can store at most 3*8.5*11*300*300 bytes of information on a piece of paper with color coding. That is just 0.025 GB.
Now, one might argue that nice photo paper could do better -- but it couldn't do better than about 1200x1200 -- the resolution of inkjet photopaper. That would yield a factor-of-16 increase, for 0.4 GB.
In practice, you can't do that well at all, because each location on the paper doesn't store 24 bits, it generally only stores one bit for each type of ink you put down. With an RGB ink system you get only 3 bits per location, even with ideal placement. That yields 0.003 GB on normal paper or 0.05 GB on photo paper.
The only way to do better is to increase the number of symbols on the paper (writing smaller) or to increase the information content of each symbol (using more subtle gradations of color). If it were easy, we'd all be doing it already.
BBC Radio 1 blankets the U.K. with so much energy that the Soviets used to use little dipole antennas to rectify the RF and power small audio bugs -- ubiquitous broadcast power on demand!
Now, that is ~100MHz rather than ~2GHz -- but then again, microwave ovens tend to leak in the 2 GHz band at the 10 milliwatt level.
Even if "benign" license control is all this is for, it ain't so benign. Having to ask permission before acting is the hallmark of totalitarianism. Even if the license were free of monetary charge, giving up that much control is too high a price to watch "X-Men 7" or "Police Academy 32".
More conventional thermoelectric junctions use dissimilar metals with different work functions -- the canonical example is aluminum and copper. The electrons are more tightly bound in the copper than in the aluminum, so it costs a little bit of energy to move them from the copper to the aluminum. Generally that energy comes from the thermal field around the junction. Likewise, moving electrons from alluminum to copper gives off a little bit of heat. You can buy Peltier coolers that have hundreds of short aluminum and copper pillars holding two thin plates apart. All the copper->aluminum junctions are on one side, and all the aluminum->copper junctions are on the other. You can use the assembly to move heat from one side to the other (by passing an electric current through it) or to generate electricity (by heating one side with, say, a blowtorch and cooling the other with, say, water). The infamous RTGs on deep spacecraft like Cassini are exactly the same technology, only they use a block of plutonium as the warm side -- the plutonium stays hot as it decays.
The limitation of conventional Peltier piles is that you want to thermally isolate the junctions while maintaining good electrical conductivity -- but aluminum and copper are both good thermal conductors! In fact those two effects are related -- the free electrons in aluminum and copper carry the heat through the metal, so improving thermal isolation (by alloying the metal, f'rinstance) also ruins the electrical conductivity and what you save by not leaking heat you lose by adding electrical resistance.
Semiconductors have controllable electron binding characteristics depending on how you dope them, so it "should" be a simple matter of junction design to make a dynamite thermopile with 'em. Further, the band gap (the amount of energy carried by a conduction electron) can be quite large, so you wouldn't have to carry so many electrons through the junctions to move a LOT of heat around.
I worked at D3D 'way back in the 1980s, when people thought breakeven would be achieved before the turn of the millennium. If as much effort were put into electrostatic confinement (the Farnsworth fusor we keep hearing so much about) that might have actually happened. The advantage of the Farnsworth fusor is that it uses a confinement field with a divergence term!
The magnetic field has no divergence (there are no magnetic monopoles) so it is extremely difficult to confine anything -- you can only slow down the leakage. That comes with some problems -- for example, it's very hard to get anything into or out of a magnetic bottle (as in a Tokamak) unless it is electrically neutral. Accelerating and heating the plasma are hard because the energy sources you can use (manipulation of the magnetic field itself, either at radiofrequency (RF heating) or near DC (betatron heating), themselves destabilize the confinement.
D3D used the innovation of firing neutral atoms in through the magnetic bottle, which provides material and heat into the plasma (the atoms generally ionize once they get in -- and then they're trapped like the rest of the plasma). The problem there is that we have no technology to accelerate neutral particles -- so they had these little tiny particle accelerators that fired their beams through GIANT TANKS of reactant that was intended to neutralize the input beams on-the-fly. Some small percentage of the particles got neutralized, and the rest bounced off the outside of the magnetic bottle into a beam dump. Seeing the size of the equipment made me realize that tokamak fusion is probably a dead end for power generation -- if it can be made to work at all (in the sense of achieving, say, 10x heat gain), the ancillary equipment is HUGE and it's not at all clear that economies of scale are enough to make it worthwhile.
The Farnsworth-Hirsch type fusors have the advantage that you can fire in charged particles -- they rattle around and lose some of their kinetic energy, and after that they're trapped in a normal potential well. Like muon-catalyzed fusion machines, the Farnsworth fusor is in a race to get the energy out of a fusible nucleus before it leaks away -- but fresh hydrogen or deuterium ions are much, much cheaper than muons, and it seems to have a better chance of working.
(Remember muon-catalyzed fusion? Muons act like electrons, only more massive -- so atoms that have an electron replaced with a muon get smaller [it's a quantum thing], bringing the nuclei closer together and boosting the fusion rate. You can get a pretty high fusion rate (a few fusions per muon per microsecond) at close to room temperature in pretty tame materials. The problem is that muons only last about two microseconds before decaying into energy, neutrinos, and electrons -- so you have to make several hundred fusions per microsecond, to make the energy worth the effort of making a muon in the first place. Nobody was able to make it pay off.)
Oh, I think we're in agreement. Centralized control, in the right hands, works great -- the problem is that the "right hands" are so rare. In practice, hierarchies are not selfless and smart -- they made of selfish, stupid individuals. Worse, once the hierarchy is large enough the individuals' interests don't align well with the organization's. Managers throughout the hierarchy can better their careers more effectively (in the short term) by managing the information supply to their bosses, than by actually managing. This effect snowballs as it trickles down the hierarchy: it is one reason that whistleblowers are so often punished, and it leads to a sort of institutional psychosis as feedback and decision-making ability both deteriorate.
That is a general problem in any large organization (as evidenced by the popularity of books like "The Peter Principle" or the old chestnut about the fictional element Administratium).
The problem is that an effectively-led organization has a decay time associated with it: once the initial population of competent leaders move up or on, or retire, decay sets in. Large, long-lived, functional organizations do exist but they are the exception rather than the norm.
No, the problem is the ".*" -- it matches "..", which is the containing directory, so it blows away every user's files just like "rm -r *" would in/usr/users.
That 'painful period of adjustment? How many of those periods will we get, and how long will they last? Decades?
A lot. Forever.
The problem is high expectation. Quasi-free-market capitalism simply doesn't work very well when organizations get large -- the individual peoples' incentives always tilt toward value hoarding (which translates to corporate behaviors like rent-seeking and personal behaviors like deception) rather than value creation.
The only advantage capitalism has is that everything else seems to be worse in the long run -- other economic systems all seem to be unstable to the same incentive problems that capitalism has. They tend to collapse under their own weight, and either become horrible, stifling dictatorships where nothing works or else develop capitalist black markets that grow to dominate the economy anyway.
Systems like health care or oil extraction, where entry costs are high and there is a long delay between action and market feedback, tend to be worse than systems with better feedback and lower entry costs. We simply don't have the appropriate technology yet to manage those long-term infrastructural industries well. (I'm not referring to transistors here, but rather to systems of management and institutional organization)
The efficiency is only over the long haul. The advantage of a free-market system is that, when large organizations get sufficiently bad, they fail and are replaced by other, presumably more efficient, ones. It is painful and takes years to happen, but they do. In a centralized economy, large, stultified, inefficient organizations are coded into law and can't fail until there is a revolution.
Polya's magnum opus is a practical manual on how to train yourself to have good hunches. I have many more authoritative books on specific subjects but this one is my all-around favorite.
I don't have a good recommendation for an applications book -- but both discrete and continuous group theories help develop good thinking pathways about discrete sets. Rather than wave my hands yet more, I'll give you two examples that are (slightly) more specific: having to trudge through proofs in Galois group theory will give you an unerring nose for corner cases in UI/API design; and thinking about subgroups and symmetry classes will give you a nice sense of how to generalize algorithms that you might otherwise implement with copy/paste.
Corner cases are a big deal -- considering what happens at the extreme limits of allowed input (in an N-dimensional input space) is (IMAO) the single most neglected part of code design, and it yields fuggliness and bugs at every turn. Well-handled corner cases are a delight to encounter, code, and use, and make life good. Poorly-handled or ignored corner cases increase complexity and cause endless headaches later. Typical corner cases are things like (in language design) zero-length arrays or the scalar-vs-array-of-length-1 wart, or (in GUI window management) windows of zero or very large size or very large nesting depth, or (in analytic geometry applications) lines that happen (in 2-D) to be parallel or (in 3-D) to be coplanar.
Generalization has to do with deeper subtleties of algorithm design, but if you are used to thinkin in terms of symmetry and generalization, you will more frequently and more readily be constructively Lazy rather than implementing specifics that must later be painstakingly hacked to accept new features or be adapted to new applications.
So, er, think of it as "deep background" with no specific applications to particular problems -- just good training in a style of thinking that will serve you in good stead later.
...but I have yet to use n-dimensional calculus in non-cartesian space for practical business applications.
Funny you should mention that: one of my buddies in graduate school was courted heavily by Wall Street to do exactly that. They wanted him to apply n-dimensional non-Euclidean analysis to stock market prediction to see if there was some additional arbitrage to be had that way. At a time when physics post-docs were earning ~$35k per annum they offered him something like $180k per annum with incentive bonuses on top of that.
When I finished my own post-doc and was scouting around for a position, I was courted by several IT firms precisely because I was familiar with n-dimensional calculus and non-Euclidean spaces. So that kind of knowledge does appear to be in demand for business applications -- perhaps just not in your particular sector.
Admittedly I'm not a CS -- just a computer-minded astrophysicist -- but one of the two or three most useful courses I had in graduate school was discrete signal processing. Anyone who's doing original work in data mining, signal compression (video, sound, text), communications (packet routing, protocol design), or signal recognition (image interpretation, sound recognition, handwriting recognition) absolutely needs to have a good overview of basic DSP techniques. Anyone who uses those things seriously, which is pretty much anyone who reads Slashdot, ought to at least have an overview of the field.
Abstract algebra is important because it opens up large pathways of logical thinking -- even if you're "just" monkeying around with basic SQL queries, you'll be much more effective if you have had a course in linear algebra or (preferably) abstract algebra: it will make you think differently about the data you are querying. Even if you can write elegant, fast code at a basic language-usage level, algorithmic design is also subject to a bajillion types of boneheadedness and stupidity. Knowing abstract algebra will help you write algorithms that are faster (both to write and to run) and cleaner.
So -- at least one upper division proof-oriented math course (linear algebra, abstract algebra, advanced calculus) and at least one DSP course (Z-transforms, DSP, data vector spaces, Hanning codes, the channel capacity theorem, and compression theory) should be required!
Ionizing radiation is just awful on chemical bonds and crystal structures. After all, it works by knocking electrons or whole atoms loose from the nice, bound states they were in. That's how the radiation damages you, too -- it's just that biological systems are a whole lot more sensitive to being scrambled, than are bulk objects like bricks or (to pick a not-so-random example) bundles of carbon nanotubes.
High doses of radiation do strange things to materials -- increase cross-links, damage coherent structure, add skillions of crystal defects. If you lower a nice flexible, white piece of polyethelene plastic into a nuclear reactor for a while, you are liable to pull out a yellow, harder, brittle, fragile piece that has the same overall shape.
If I understand the nature of the space elevator right, each particle "hit" would tear apart a carbon nanotube, gradually shortening the average tube length and weakening the whole bulk structure. I'm sure someone has thought of this effect, but we haven't seen much of it in the space elevator press packets.
Noise canceling headphones do protect your hearing but most civilian models don't afford serious protection at those dB levels. The reason is that, if the headset is capable of generating antinoise that can prevent ear damage, it si equally capable (perhaps more capable) of generating *noise* that can damage your ears. That is a huge liability problem for the mass market.
Given the ease of making consumer models feed back (cup your hands just right over the earphones of a Sharper Image, Panasonic, or even Bose set, and you'll be rewarded with an ear-piercing squeal), I'd rather that the drivers simply not be capable of working with instant-deafness levels of sound pressure.
Military models are different -- the ANC headphones military chopper pilots wear reduce the bone-shaking noise of an open cockpit Huey, to a noise level comparable to a civilian sedan on the freeway (I had the pleasure of experiencing this in the early 1990s, flying recovery for a sounding rocket at White Sands). I imagine that the very best of the civilian models can do that too, but you won't find them in the catalog in your airplane seat-back or at the mall, more like Sporty's Pilot Shop.
Of course it's the Republicans who are nominally Toryoids (conservatives) and the Democrats who are nominally Whigoids/Progressives (liberals). Too much posting, too little coffee -- sorry to get that backwards.
Over the past few years, the Republicans (normally Whigoids, to UKers) have turned the political process into a tribalist scrum rather than an informed debate. The Democrats (normally Toryoids, to UKers) have been completely locked out of the process -- many bills are forced to a vote by the leading party, without even having been debated in the full chamber, and many bills are drafted and revised in closed chambers that Democrats have literally not been allowed to enter. The Republicans have made use of their power over the process, to pass any damnned bill that crosses their minds, whether or not it agrees with their nominal party platform.
Republicans "traditionally" stand for limited government and fiscal responsibility, but over the past few years we have seen drastic increases in the legislated power of the executive branch over the people, including measures that can only be described as fascist. You may not pay much attention to our federal budget from the U.K. but our federal budget deficit is the worst in history due to extremely imprudent spend-but-don't-tax policies passed by the current round of Republicans, to the point that many economists have predicted third-world style hyperinflation of the U.S. dollar.
Meanwhile, the Democrats "traditionally" stand for more socialist policies and larger government, and have a long-term reputation for deficit spending -- but the Clintonian democrats of the 1990s were able to rein in spending to the point (with the help of moderate Republicans then in office) they were able to balance the federal budget. That was a task nobody had accomplished since World War 2.
The Republicans position themselves as supporting traditional moral values, particularly as regards sexual issues, while the Democrats position themselves as supporting individual choice of lifestyle --- but lately the Republicans seem to be the ones getting mired in scandal and having problems with personal debauchery (wife-strangling, homosexual page-fucking, and the like).
So if you're a bit confused over there, well, so are we! There's a major shift happening in U.S. politics, due in large part to the wilful self-destruction of the Republican platform and the moderation of the Democrat platform.
The now-venerable annual International Nethack Tournament starts on 10/31 as always at http://nethack.devnull.net. These guys are the original online NH tourney, accept no imitations:-)
If you click over there you can volunteer to run a game server, which is kind of fun. Nethack probably has the highest ratio of player brain resources to server CPU resources of any current computer game -- I ran a game server a few years in a row and was always bemused to find that 20 people were playing nethack on my [random ancient piece of hardware] without noticeably impacting its load average or even clogging up my DSL line.
DAC and ADC circuits are really good these days. By really good I mean that a $100 sound card is better than a high-end tape deck from the 1980s, or even than most audiophile turntables playing brand-new vinyl. The built-in soundcard on your motherboard probably "sucks", which means it's only as good as that really nice component tape deck your older brother bought in the 1990s and you drooled over until you discovered mp3s. The suckiness is probably digital noise from the motherboard, leaking it at the -50 or -60 db level (about the same as the noise floor for a cassette tape w/o Dolby or DBX). Harmonic distortion is probably buried in the digital leakage, even on cost-engineered, sucky on-board sound.
A few years ago I did audio comparisons between a cheap-ass I-Opener computer playing mp3s ripped from a record and a midrange Technics tape deck playing the same tracks recorded from the same record, and the I-Opener did better.
So if you've paid any attention at all to your sound card, you probably won't hear any distortion from passing the sound through it. You're much more likely to notice the fuzz and tinkle-bells from the initial low-quality Rhapsody encoding.
Imagine being able to have your music collection centralized at home on ONE MACHINE. NOT files strewn about all over the place, but one centralized location.
To keep me buying CDs (or, rather, get me started again) the industry would have to lower prices drastically. When the CD of the "Bride and Prejudice" soundtrack costs twice as much as the movie itself, there is a serious problem with pricing.
Most of these are owned by private entities, making it quite difficult to access the information -- for example, a copy of the California building codes costs close to $500 in three-ring binder form. Most jurisdictions incorporate the copyrighted documents into law by reference only, trying to sidestep the problem that the law of the land is not copyrightable.
Slashdot Mirror
I don't know the exact dimensions of A4, but it's similar to 8.5"x11" (at . Suppose each spot of color stores 24 bits of information -- this is really a stretch, for reasons I'll get to below. Normal Xerox paper supports something like 300 dpi of resolution (that is the scale of the fibers of the paper). Then you can store at most 3*8.5*11*300*300 bytes of information on a piece of paper with color coding. That is just 0.025 GB.
Now, one might argue that nice photo paper could do better -- but it couldn't do better than about 1200x1200 -- the resolution of inkjet photopaper. That would yield a factor-of-16 increase, for 0.4 GB.
In practice, you can't do that well at all, because each location on the paper doesn't store 24 bits, it generally only stores one bit for each type of ink you put down. With an RGB ink system you get only 3 bits per location, even with ideal placement. That yields 0.003 GB on normal paper or 0.05 GB on photo paper.
The only way to do better is to increase the number of symbols on the paper (writing smaller) or to increase the information content of each symbol (using more subtle gradations of color). If it were easy, we'd all be doing it already.
BBC Radio 1 blankets the U.K. with so much energy that the Soviets used to use little dipole antennas to rectify the RF and power small audio bugs -- ubiquitous broadcast power on demand!
Now, that is ~100MHz rather than ~2GHz -- but then again, microwave ovens tend to leak in the 2 GHz band at the 10 milliwatt level.
Wifi is the very least of the RF "problem".
Even if "benign" license control is all this is for, it ain't so benign. Having to ask permission before acting is the hallmark of totalitarianism. Even if the license were free of monetary charge, giving up that much control is too high a price to watch "X-Men 7" or "Police Academy 32".
More conventional thermoelectric junctions use dissimilar metals with different work functions -- the canonical example is aluminum and copper. The electrons are more tightly bound in the copper than in the aluminum, so it costs a little bit of energy to move them from the copper to the aluminum. Generally that energy comes from the thermal field around the junction. Likewise, moving electrons from alluminum to copper gives off a little bit of heat. You can buy Peltier coolers that have hundreds of short aluminum and copper pillars holding two thin plates apart. All the copper->aluminum junctions are on one side, and all the aluminum->copper junctions are on the other. You can use the assembly to move heat from one side to the other (by passing an electric current through it) or to generate electricity (by heating one side with, say, a blowtorch and cooling the other with, say, water). The infamous RTGs on deep spacecraft like Cassini are exactly the same technology, only they use a block of plutonium as the warm side -- the plutonium stays hot as it decays.
The limitation of conventional Peltier piles is that you want to thermally isolate the junctions while maintaining good electrical conductivity -- but aluminum and copper are both good thermal conductors! In fact those two effects are related -- the free electrons in aluminum and copper carry the heat through the metal, so improving thermal isolation (by alloying the metal, f'rinstance) also ruins the electrical conductivity and what you save by not leaking heat you lose by adding electrical resistance.
Semiconductors have controllable electron binding characteristics depending on how you dope them, so it "should" be a simple matter of junction design to make a dynamite thermopile with 'em. Further, the band gap (the amount of energy carried by a conduction electron) can be quite large, so you wouldn't have to carry so many electrons through the junctions to move a LOT of heat around.
I wonder why nobody has done this before?
I worked at D3D 'way back in the 1980s, when people thought breakeven would be achieved before the turn of the millennium. If as much effort were put into electrostatic confinement (the Farnsworth fusor we keep hearing so much about) that might have actually happened. The advantage of the Farnsworth fusor is that it uses a confinement field with a divergence term!
The magnetic field has no divergence (there are no magnetic monopoles) so it is extremely difficult to confine anything -- you can only slow down the leakage. That comes with some problems -- for example, it's very hard to get anything into or out of a magnetic bottle (as in a Tokamak) unless it is electrically neutral. Accelerating and heating the plasma are hard because the energy sources you can use (manipulation of the magnetic field itself, either at radiofrequency (RF heating) or near DC (betatron heating), themselves destabilize the confinement.
D3D used the innovation of firing neutral atoms in through the magnetic bottle, which provides material and heat into the plasma (the atoms generally ionize once they get in -- and then they're trapped like the rest of the plasma). The problem there is that we have no technology to accelerate neutral particles -- so they had these little tiny particle accelerators that fired their beams through GIANT TANKS of reactant that was intended to neutralize the input beams on-the-fly. Some small percentage of the particles got neutralized, and the rest bounced off the outside of the magnetic bottle into a beam dump. Seeing the size of the equipment made me realize that tokamak fusion is probably a dead end for power generation -- if it can be made to work at all (in the sense of achieving, say, 10x heat gain), the ancillary equipment is HUGE and it's not at all clear that economies of scale are enough to make it worthwhile.
The Farnsworth-Hirsch type fusors have the advantage that you can fire in charged particles -- they rattle around and lose some of their kinetic energy, and after that they're trapped in a normal potential well. Like muon-catalyzed fusion machines, the Farnsworth fusor is in a race to get the energy out of a fusible nucleus before it leaks away -- but fresh hydrogen or deuterium ions are much, much cheaper than muons, and it seems to have a better chance of working.
(Remember muon-catalyzed fusion? Muons act like electrons, only more massive -- so atoms that have an electron replaced with a muon get smaller [it's a quantum thing], bringing the nuclei closer together and boosting the fusion rate. You can get a pretty high fusion rate (a few fusions per muon per microsecond) at close to room temperature in pretty tame materials. The problem is that muons only last about two microseconds before decaying into energy, neutrinos, and electrons -- so you have to make several hundred fusions per microsecond, to make the energy worth the effort of making a muon in the first place. Nobody was able to make it pay off.)
Oh, I think we're in agreement. Centralized control, in the right hands, works great -- the problem is that the "right hands" are so rare. In practice, hierarchies are not selfless and smart -- they made of selfish, stupid individuals. Worse, once the hierarchy is large enough the individuals' interests don't align well with the organization's. Managers throughout the hierarchy can better their careers more effectively (in the short term) by managing the information supply to their bosses, than by actually managing. This effect snowballs as it trickles down the hierarchy: it is one reason that whistleblowers are so often punished, and it leads to a sort of institutional psychosis as feedback and decision-making ability both deteriorate.
That is a general problem in any large organization (as evidenced by the popularity of books like "The Peter Principle" or the old chestnut about the fictional element Administratium).
The problem is that an effectively-led organization has a decay time associated with it: once the initial population of competent leaders move up or on, or retire, decay sets in. Large, long-lived, functional organizations do exist but they are the exception rather than the norm.
Er, rant mode off.
No, the problem is the ".*" -- it matches "..", which is the containing directory, so it blows away every user's files just like "rm -r *" would in /usr/users.
A lot. Forever.
The problem is high expectation. Quasi-free-market capitalism simply doesn't work very well when organizations get large -- the individual peoples' incentives always tilt toward value hoarding (which translates to corporate behaviors like rent-seeking and personal behaviors like deception) rather than value creation.
The only advantage capitalism has is that everything else seems to be worse in the long run -- other economic systems all seem to be unstable to the same incentive problems that capitalism has. They tend to collapse under their own weight, and either become horrible, stifling dictatorships where nothing works or else develop capitalist black markets that grow to dominate the economy anyway.
Systems like health care or oil extraction, where entry costs are high and there is a long delay between action and market feedback, tend to be worse than systems with better feedback and lower entry costs. We simply don't have the appropriate technology yet to manage those long-term infrastructural industries well. (I'm not referring to transistors here, but rather to systems of management and institutional organization)
The efficiency is only over the long haul. The advantage of a free-market system is that, when large organizations get sufficiently bad, they fail and are replaced by other, presumably more efficient, ones. It is painful and takes years to happen, but they do. In a centralized economy, large, stultified, inefficient organizations are coded into law and can't fail until there is a revolution.
From the 1980s -- I watched a new sysadmin blowing away an old user directory on the college VAX (which had about 600 user accounts):
/usr/users/olduser/* /usr/users/olduser
/usr/users/olduser: directory not empty /usr/users/olduser/.*
# rm -r
# rmdir
# rm -r
Shortly thereafter, the phone started ringing...
(think about it...)
Polya's magnum opus is a practical manual on how to train yourself to have good hunches. I have many more authoritative books on specific subjects but this one is my all-around favorite.
I don't have a good recommendation for an applications book -- but both discrete and continuous group theories help develop good thinking pathways about discrete sets. Rather than wave my hands yet more, I'll give you two examples that are (slightly) more specific: having to trudge through proofs in Galois group theory will give you an unerring nose for corner cases in UI/API design; and thinking about subgroups and symmetry classes will give you a nice sense of how to generalize algorithms that you might otherwise implement with copy/paste.
Corner cases are a big deal -- considering what happens at the extreme limits of allowed input (in an N-dimensional input space) is (IMAO) the single most neglected part of code design, and it yields fuggliness and bugs at every turn. Well-handled corner cases are a delight to encounter, code, and use, and make life good. Poorly-handled or ignored corner cases increase complexity and cause endless headaches later. Typical corner cases are things like (in language design) zero-length arrays or the scalar-vs-array-of-length-1 wart, or (in GUI window management) windows of zero or very large size or very large nesting depth, or (in analytic geometry applications) lines that happen (in 2-D) to be parallel or (in 3-D) to be coplanar.
Generalization has to do with deeper subtleties of algorithm design, but if you are used to thinkin in terms of symmetry and generalization, you will more frequently and more readily be constructively Lazy rather than implementing specifics that must later be painstakingly hacked to accept new features or be adapted to new applications.
So, er, think of it as "deep background" with no specific applications to particular problems -- just good training in a style of thinking that will serve you in good stead later.
Funny you should mention that: one of my buddies in graduate school was courted heavily by Wall Street to do exactly that. They wanted him to apply n-dimensional non-Euclidean analysis to stock market prediction to see if there was some additional arbitrage to be had that way. At a time when physics post-docs were earning ~$35k per annum they offered him something like $180k per annum with incentive bonuses on top of that.
When I finished my own post-doc and was scouting around for a position, I was courted by several IT firms precisely because I was familiar with n-dimensional calculus and non-Euclidean spaces. So that kind of knowledge does appear to be in demand for business applications -- perhaps just not in your particular sector.
Admittedly I'm not a CS -- just a computer-minded astrophysicist -- but one of the two or three most useful courses I had in graduate school was discrete signal processing. Anyone who's doing original work in data mining, signal compression (video, sound, text), communications (packet routing, protocol design), or signal recognition (image interpretation, sound recognition, handwriting recognition) absolutely needs to have a good overview of basic DSP techniques. Anyone who uses those things seriously, which is pretty much anyone who reads Slashdot, ought to at least have an overview of the field.
Abstract algebra is important because it opens up large pathways of logical thinking -- even if you're "just" monkeying around with basic SQL queries, you'll be much more effective if you have had a course in linear algebra or (preferably) abstract algebra: it will make you think differently about the data you are querying. Even if you can write elegant, fast code at a basic language-usage level, algorithmic design is also subject to a bajillion types of boneheadedness and stupidity. Knowing abstract algebra will help you write algorithms that are faster (both to write and to run) and cleaner.
So -- at least one upper division proof-oriented math course (linear algebra, abstract algebra, advanced calculus) and at least one DSP course (Z-transforms, DSP, data vector spaces, Hanning codes, the channel capacity theorem, and compression theory) should be required!
Let me get this straight...
You want to take the tallest, most dangerous structure ever built, and fire terawatt laser beams at it from all over the continent?
Hmmm... what could possibly go wrong here?
Ionizing radiation is just awful on chemical bonds and crystal structures. After all, it works by knocking electrons or whole atoms loose from the nice, bound states they were in. That's how the radiation damages you, too -- it's just that biological systems are a whole lot more sensitive to being scrambled, than are bulk objects like bricks or (to pick a not-so-random example) bundles of carbon nanotubes.
High doses of radiation do strange things to materials -- increase cross-links, damage coherent structure, add skillions of crystal defects. If you lower a nice flexible, white piece of polyethelene plastic into a nuclear reactor for a while, you are liable to pull out a yellow, harder, brittle, fragile piece that has the same overall shape.
If I understand the nature of the space elevator right, each particle "hit" would tear apart a carbon nanotube, gradually shortening the average tube length and weakening the whole bulk structure. I'm sure someone has thought of this effect, but we haven't seen much of it in the space elevator press packets.
Noise canceling headphones do protect your hearing but most civilian models don't afford serious protection at those dB levels. The reason is that, if the headset is capable of generating antinoise that can prevent ear damage, it si equally capable (perhaps more capable) of generating *noise* that can damage your ears. That is a huge liability problem for the mass market.
Given the ease of making consumer models feed back (cup your hands just right over the earphones of a Sharper Image, Panasonic, or even Bose set, and you'll be rewarded with an ear-piercing squeal), I'd rather that the drivers simply not be capable of working with instant-deafness levels of sound pressure.
Military models are different -- the ANC headphones military chopper pilots wear reduce the bone-shaking noise of an open cockpit Huey, to a noise level comparable to a civilian sedan on the freeway (I had the pleasure of experiencing this in the early 1990s, flying recovery for a sounding rocket at White Sands). I imagine that the very best of the civilian models can do that too, but you won't find them in the catalog in your airplane seat-back or at the mall, more like Sporty's Pilot Shop.
Of course it's the Republicans who are nominally Toryoids (conservatives) and the Democrats who are nominally Whigoids/Progressives (liberals). Too much posting, too little coffee -- sorry to get that backwards.
Over the past few years, the Republicans (normally Whigoids, to UKers) have turned the political process into a tribalist scrum rather than an informed debate. The Democrats (normally Toryoids, to UKers) have been completely locked out of the process -- many bills are forced to a vote by the leading party, without even having been debated in the full chamber, and many bills are drafted and revised in closed chambers that Democrats have literally not been allowed to enter. The Republicans have made use of their power over the process, to pass any damnned bill that crosses their minds, whether or not it agrees with their nominal party platform.
Republicans "traditionally" stand for limited government and fiscal responsibility, but over the past few years we have seen drastic increases in the legislated power of the executive branch over the people, including measures that can only be described as fascist. You may not pay much attention to our federal budget from the U.K. but our federal budget deficit is the worst in history due to extremely imprudent spend-but-don't-tax policies passed by the current round of Republicans, to the point that many economists have predicted third-world style hyperinflation of the U.S. dollar.
Meanwhile, the Democrats "traditionally" stand for more socialist policies and larger government, and have a long-term reputation for deficit spending -- but the Clintonian democrats of the 1990s were able to rein in spending to the point (with the help of moderate Republicans then in office) they were able to balance the federal budget. That was a task nobody had accomplished since World War 2.
The Republicans position themselves as supporting traditional moral values, particularly as regards sexual issues, while the Democrats position themselves as supporting individual choice of lifestyle --- but lately the Republicans seem to be the ones getting mired in scandal and having problems with personal debauchery (wife-strangling, homosexual page-fucking, and the like).
So if you're a bit confused over there, well, so are we! There's a major shift happening in U.S. politics, due in large part to the wilful self-destruction of the Republican platform and the moderation of the Democrat platform.
So go now, while you're still thinking about it, and register. You're too late for this election but you're nice and early for 2008.
The now-venerable annual International Nethack Tournament starts on 10/31 as always at http://nethack.devnull.net. These guys are the original online NH tourney, accept no imitations :-)
If you click over there you can volunteer to run a game server, which is kind of fun. Nethack probably has the highest ratio of player brain resources to server CPU resources of any current computer game -- I ran a game server a few years in a row and was always bemused to find that 20 people were playing nethack on my [random ancient piece of hardware] without noticeably impacting its load average or even clogging up my DSL line.
DAC and ADC circuits are really good these days. By really good I mean that a $100 sound card is better than a high-end tape deck from the 1980s, or even than most audiophile turntables playing brand-new vinyl. The built-in soundcard on your motherboard probably "sucks", which means it's only as good as that really nice component tape deck your older brother bought in the 1990s and you drooled over until you discovered mp3s. The suckiness is probably digital noise from the motherboard, leaking it at the -50 or -60 db level (about the same as the noise floor for a cassette tape w/o Dolby or DBX). Harmonic distortion is probably buried in the digital leakage, even on cost-engineered, sucky on-board sound.
A few years ago I did audio comparisons between a cheap-ass I-Opener computer playing mp3s ripped from a record and a midrange Technics tape deck playing the same tracks recorded from the same record, and the I-Opener did better.
So if you've paid any attention at all to your sound card, you probably won't hear any distortion from passing the sound through it. You're much more likely to notice the fuzz and tinkle-bells from the initial low-quality Rhapsody encoding.
You can do that now. I do. BSD Samba server + SONOS + iTunes = happy Zowie.
To keep me buying CDs (or, rather, get me started again) the industry would have to lower prices drastically. When the CD of the "Bride and Prejudice" soundtrack costs twice as much as the movie itself, there is a serious problem with pricing.
Most of these are owned by private entities, making it quite difficult to access the information -- for example, a copy of the California building codes costs close to $500 in three-ring binder form. Most jurisdictions incorporate the copyrighted documents into law by reference only, trying to sidestep the problem that the law of the land is not copyrightable.