Think about what it takes to work on that thing. It's in a long underground tunnel of rather small diameter for what's in there. Fixing stuff in place is difficult and hazardous. Removing a magnet involves disconnecting everything (a big deal; some of the connections are welded and superconducting), lifting the magnet onto a narrow carrier that runs on the walkway (no idea how that's actually done) and inching the carrier for kilometers to one of the two big vertical shafts where it can be hoisted out vertically. As an underground maintenance job, this is not fun.
The canceled American SSC was designed with a larger tunnel diameter. The LHC was designed with the assumption that not much magnet maintenance would be required, which cut costs but turned out to be a bad assumption.
First, as is typical, the Slashdot article is three steps removed from the actual paper, which is worth reading.
It's kind of cute. What makes it work is that the indexing part of the Vuze platform, which is distributed over a few million user machines, has an 8-hour timeout. After eight hours, otherwise unused entries are purged from cache, like DNS cache expiration. So it's possible to use Vuze for unreliable short-term storage of key-value pairs.
(Normally, the Vuze hash is used as a index to BitTorrent blocks, and if there's a block on a server, the server puts it into the hash and refreshes it periodically, so the block stays indexed. But it's possible to put arbitrary key-value pairs into the distributed hash that have no relationship to BitTorrent blocks. If you put info in the hash and don't refresh it, it goes away after eight hours.)
So the sender generates a key, encrypts the message, spreads the key across some number of key-value pairs on random Vuze clients, sends a message telling what key-value pairs in Vuze contain the crypto key, and deletes the local copy of the key. The receiver gets the message, looks up the key-value pairs specified in the Vuze hash, reconstructs the key, decrypts the message, displays it, and deletes the local copy of the key. The receiving client has to do this every time the message is viewed.
I kept hearing things about some site called "Google", so I tried running it through SiteTruth. Turns out it's some shady, fly-by-night company.
Yes, Google is in the doghouse again. Google is hosting some phishing sites, which were reported to PhishTank. SiteTruth blacklists any domain with a hit in PhishTank. On any given day, about 50 to 100 well-known domains (out of the 1.5 million in OpenDirectory) are on the blacklist, generally because of sloppy security. Microsoft, Yahoo, and eBay used to be on the phishing blacklist, but after some nagging by us and The Register, they've mostly plugged the security holes involved. The blacklist is updated every 3 hours, so companies that clean up their act quickly don't stay on the list for long.
Domains on the blacklist are usually 1) free hosting services, 2) URL redirectors like TinyURL, 3) DSL providers with weak abuse departments, and 4) sites with a software bug that lets other sites use them as a redirector. Some companies in those categories are good at quickly cleaning out such abuses; others just don't seem to care. In each category, there are plenty of companies who don't have such problems, so there's no reason to give anybody a free pass.
It says something about a company's abuse department if they're on that list for more than a day or two.
Uh, concrete testing is done via filling multiple plastic cylinders with concrete as it comes off of each truck to cure for 21-28 days prior to being test for strength, atleast here in CT.
That's the problem. You know 21 days after the pour if the concrete is no good. That kind of testing only works with a strong enforcement bureaucracy behind it.
Why doesn't an attempt to legally incorporate a new business include a "do any of your officers have a background in crime, particularly white collar crime?" check?
That's a real problem. People barred from involvement in the securities industry keep slipping back in. Bar owners barred from holding a liquor license often end up doing some deal as a "silent partner".
I get complaints from "web businesses" who want to operate anonymously, because SiteTruth down-rates them for that. (It's a criminal offense to run a business anonymously in many jurisdictions.)
The big mistake OLPC made was Nicholas Negroponte. He's too much into his own self-importance. His thing was dealing at the national leader level and getting himself into the press. What was needed was somebody who knew how to get a low-cost product out the door and sell it in quantity.
The OLPC should have been in a bubble-pack in every Wal-Mart and Walgreens in America, in every souk in the Middle East, and in every market in India, selling at a small profit and dropping in price every three months. But no, Negroponte had to try to make big deals with governments. That might have happened after they actually had the product out there in volume.
There's no technical problem making a wood building that strong. It's the enforcement that's the problem. Wood has good tensile strength, but the joints usually used in wood construction don't.
A few years ago, after some hurricanes, many Florida builders were discovered not to be building to code. Hurricane-proofing for small wood structures mostly consists of putting in metal brackets at joints to give wood-to-wood joints tensile strength. Not only do the brackets have to be put in, nails have to be hammered into all the holes in the brackets. Many contractors were sloppy about that, resulting in a big loss of tensile strength and major damage (like roofs ripped off) during hurricanes.
A big problem in the Third World is bad concrete mixes. Much concrete construction goes up without enough cement in the mix, and that results in building collapses.
Here's a good project for someone - develop a low-cost hand held device for concrete testing. The existing techniques are slow, labor-intensive, and a pain to use. Tests for hardened concrete
usually involve cutting out a plug and sending it to a lab elsewhere. Small portable devices would be a big help here.
Virtualizing physical I/O devices on PC-like architectures requires code in the hypervisor to emulate the device. The driver in the operating system does stores into "device registers" as if talking the real device. Each such store or load causes a trap to the hypervisor, which has a device emulator watching the register changes and pretending to be the real peripheral. When the right registers have been loaded with the right values, and the final register store is made that would start the I/O operation, the device emulator then figures out what the OS wanted to do, and makes a call to the hypervisor's I/O system to do it.
In many cases, the device driver in the OS is doing all the optimization for the device controller of a real disk, doing angular optimization and head movement minimization. Since the real device underneath may be completely different, most of this is wasted work, and may reduce performance instead of increasing it.
So it's common to have dummy device drivers for virtual machines that just pass the OS's request through to the hypervisor, without trying to manage a real device. Such drivers don't do much, and are usually trivial, although Microsoft will probably try to complicate them somehow.
This isn't a new idea; it first appeared in IBM's VM for the System/370, where such calls were passed through using the DIAGNOSE instruction (an opcode used for hardware diagnostics only, and thus never used in ordinary programs and available as a spare opcode.)
One of the hypervisor vendors calls this "paravirtualization".
Wikipedia isn't there to make celebrities look good. From a historical perspective, an honest picture is better than a carefully posed and manipulated one.
If a celebrity has been arrested, their picture on Wikipedia is likely to be a Government mugshot.
Isn't someone running a static checker on the Linux kernel? There are commercial tools which will find code that can dereference NULL. However, there aren't free tools that will do this.
Check out Vocaloid 2, the Japanese singing synthesizer. This does a nice job. Here's a sample.
Setting up Vocaloid for a new voice is a big job; the current version requires that the performer sing a long, standardized set of training syllables. Once you have the voice configured, you feed in a MIDI file with lyrics, pitch, and timing info, and singing or talking comes out.
Clearly, the next step is the ability to train the system from unstandardized speech and singing samples. Once that technology is developed, it should be possible to automatically generated cover versions of any desired song with any desired performer. At last, the computer industry will be in a position to crush the RIAA.
(Remember, under US copyright law, anyone can make a cover version of any song and just pay the statutory royalty to the songwriter, a modest fee. The songwriter cannot refuse.)
One of the real reasons behind this is that the gear train to connect a gas turbine to a propeller is a huge pain. Turbine shaft speeds are around 20,000 RPM, and this has to be stepped down to 200 RPM or so for the propeller. There's a clutch and reversing gear, too. Often the gearbox is bigger than the engine.
Reverse with a turbine is a headache. Some marine diesels are built so they can run in either direction, but that's not an option with a turbine. So there's either a really big gearshift, or a second powerplant for reverse operation.
Thus, there's ongoing interest in going to electric transmission, like a locomotive.
It's
not a new idea; an electric transmission was tried on a ship before WWII. But it works
much better with modern power semiconductors. Locomotives do this now. Modern locomotives use AC to DC to variable frequency polyphase AC conversion to drive the motors. This takes large switching power supplies, using very large semiconductors. It's a solved problem; GE locomotives have been doing this for ten years now, and their competitors now have comparable technology. Moving the technology to ships is an obvious move at this point.
There's also a Navy effort to develop a permanent magnet motor (!) big enough for shipboard propulsion. This gets rid of the field coils and increases efficiency. I'm not sure how that's coming along.
This is routine progress being made in heavy machinery. The combination of electronics and really big gears and motors can do things neither can do alone.
This seems to be legitimate; they're in FDA Phase I human testing (safety only, not effectiveness.). That doesn't mean it will work; if it makes it through Phase II, it's real.
A better binary diffing algorithm is useful for source control. But for security?
If the code is so awful that the bandwidth required for security updates is a problem, the product is defective by design.
It sounds like Google tried "agile programming" on trusted code, and now has to deal with the consequences of debugging a pile of crap.
"It's utter bilge. I don't think anybody will ever put up enough money to do such a thing . . . What good would it do us? If we spent the same amount of money on preparing first-class astronomical equipment we would learn much more about the universe . . . It is all rather rot" -- Riet Wolley, Astronomer Royal.
He was so right.
Basic truth: space travel with chemical rockets is just too inefficient to be useful. Fuels are as good as they can get; we've been using liquid hydrogen since the 1960s. It's not getting any better. Space travel is about weight reduction, which means fragile vehicles.
(Endeavour just had some foam fall off and hit the thermal tiles during launch. Again.)
Unless and until we get something better than chemical rockets, space travel isn't going to get any better.
If we'd launched an Orion or two in the 1950s, things might have been very different. Everyone would know there's a better way.
At $5 each, there are few options. Rainbow Kits are a possibility.
The "blinking lights" and "1W audio amplifier" kits are both under $5.99. That's about as low as you can go.
The only reason it's still interesting is because there's historical data. Giving the same test to groups over many decades is interesting in that it helps to spot long-term trends.
The MMPI, the Minnesota Multiphasic Personality Inventory, is like that. The definition of "normal" in the 1930s is rather different from the results today, and it's useful to see the trends.
Here's a classic "big secret:" Blaney's Ladder Levitation. It was first performed on TV on The Tonight show in 1973. It's still considered a "big secret" by professional magicians, but if you watch the video closely, you should be able to figure it out. Few illusions survive frame-by-frame examination.
OS 2200, for UNISYS mainframes, is somewhere around Level 48 today. The original demo release was April 4, 1967. Level 23 was reached around 1970, level 32 around 1978, and level 48 in 2007. Builds have numbers like "26.51.465".
Yes, there really is an operating system over 40 years old in current production use.
Mod parent up. In the serious crypto world, this is a good thing, provided it doesn't happen too often. Sometimes you're going to lose a key, because, for security reasons, you don't keep extra copies. You have a procedure for issuing new keys when this happens, which you're routinely doing anyway.
Take it from me, a computer science degree definitely won't help you get a job.
He may be right. Which is sad.
I'm not complaining personally. I've been fortunate enough to do plenty of good computer science over the course of a long career. I've done everything from networking theory to proof of correctness to autonomous robots. I've done web stuff, too, and today write mostly Python, so I'm not stuck in the past.
But new graduates, especially if they only have a BS, are going to have trouble finding a good career. Most of the basic problems in the field have been solved, and the ones that remain require deep specialization before you can make progress. The previous poster is right; employers want a bunch of buzzwords, because much of programming at the working levels today is knowing how to interface with a huge collection of ever-changing APIs, and those take time to learn.
A century ago, the stationary engineers who set up factories to run on electricity and steam had the same problem. In 1870, that was cutting-edge technology. By 1910, factory owners just ordered motors and hooked them up. Today, it's a union job.
There are good computer science problems to work on today, but it's hard to find a job that needs them solved. There's great stuff going on in machine learning. Game AIs could get much better. Robotics is finally starting to work. Computer vision is getting very good. There's a whole world of interesting technology associated with semiconductor design automation. But 99% of working programmers won't get into those areas. They'll be stuck doing e-commerce sites forever.
The smart young people I know all seem to be going into either bio or law. A few years ago, some of them were going into hedge funds, but that seems to be over.
In the US, this is completely legal. That's been settled law since Bridgeman vs. Corel (Corel issued a CD of photos of public domain paintings) and Feist vs. Rural Telephone (phone books not copyrightable; no creativity.). In fact, in Feist, the Supreme Court held that it's a constitutional issue; Congress's right to make copyright law is limited to creative works. Nor does the US have "database copyright", despite lobbying attempts for it. There's also Meshwerks vs. Toyota, which reinforces Bridgeman at the appellate level.
UK law in this area is still iffy. Which is going to be a problem here.
The problem is that sex in games usually doesn't provide good gameplay. If you just want to watch porn, that's easily available. Besides, simulated porn doesn't look that good.
That said, the big flap about the "hot coffee" scene in GTA was sort of silly. The GTA world ought to have sex in it. In fact, it's inconsistent that a game with strippers and hookers doesn't have sex in it. There's so much unrealized potential there, for seduction, power games, devious girlfriends - all the basic male/female drama elements.
The key is integrating sex into gameplay without having the sex dominate the game. That's a design challenge. It's not impossible. Second Life has sex, but it's not primarily a sex-oriented MMORPG. What we need are R-rated video games, dramas where sex plays a role in the plot. That could be fun.
I once worked in an R&D facility for heavy hydraulic equipment. They had about fifty test
cells of different sizes, the largest of which was used for hydraulic transmissions for medium-sized locomotives. Those test setups used a big motor and a water-cooled brake; the hot water went through a cooling tower, and then to sprinklers in what appeared to be a decorative lake out front but was really a heat sink. That gear was in the 5MW range, somewhat smaller than what's being described here, but not a lot smaller.
That setup was where it belonged, near the engineers who designed the things and the machinists who built the prototypes. When the big test cell was put in, it took a few months to build. Not five years.
Slashdot Mirror
Think about what it takes to work on that thing. It's in a long underground tunnel of rather small diameter for what's in there. Fixing stuff in place is difficult and hazardous. Removing a magnet involves disconnecting everything (a big deal; some of the connections are welded and superconducting), lifting the magnet onto a narrow carrier that runs on the walkway (no idea how that's actually done) and inching the carrier for kilometers to one of the two big vertical shafts where it can be hoisted out vertically. As an underground maintenance job, this is not fun.
The canceled American SSC was designed with a larger tunnel diameter. The LHC was designed with the assumption that not much magnet maintenance would be required, which cut costs but turned out to be a bad assumption.
This page takes six seconds to load. "data.coremetrics.com" is slow today. "c.fsdn" is slow, too. "doubleclick.com" only wasted about a second.
Maybe someone at Slashdot should read the book.
First, as is typical, the Slashdot article is three steps removed from the actual paper, which is worth reading.
It's kind of cute. What makes it work is that the indexing part of the Vuze platform, which is distributed over a few million user machines, has an 8-hour timeout. After eight hours, otherwise unused entries are purged from cache, like DNS cache expiration. So it's possible to use Vuze for unreliable short-term storage of key-value pairs.
(Normally, the Vuze hash is used as a index to BitTorrent blocks, and if there's a block on a server, the server puts it into the hash and refreshes it periodically, so the block stays indexed. But it's possible to put arbitrary key-value pairs into the distributed hash that have no relationship to BitTorrent blocks. If you put info in the hash and don't refresh it, it goes away after eight hours.)
So the sender generates a key, encrypts the message, spreads the key across some number of key-value pairs on random Vuze clients, sends a message telling what key-value pairs in Vuze contain the crypto key, and deletes the local copy of the key. The receiver gets the message, looks up the key-value pairs specified in the Vuze hash, reconstructs the key, decrypts the message, displays it, and deletes the local copy of the key. The receiving client has to do this every time the message is viewed.
This violates the Vuze terms of service, incidentally.
I kept hearing things about some site called "Google", so I tried running it through SiteTruth. Turns out it's some shady, fly-by-night company.
Yes, Google is in the doghouse again. Google is hosting some phishing sites, which were reported to PhishTank. SiteTruth blacklists any domain with a hit in PhishTank. On any given day, about 50 to 100 well-known domains (out of the 1.5 million in OpenDirectory) are on the blacklist, generally because of sloppy security. Microsoft, Yahoo, and eBay used to be on the phishing blacklist, but after some nagging by us and The Register, they've mostly plugged the security holes involved. The blacklist is updated every 3 hours, so companies that clean up their act quickly don't stay on the list for long.
Domains on the blacklist are usually 1) free hosting services, 2) URL redirectors like TinyURL, 3) DSL providers with weak abuse departments, and 4) sites with a software bug that lets other sites use them as a redirector. Some companies in those categories are good at quickly cleaning out such abuses; others just don't seem to care. In each category, there are plenty of companies who don't have such problems, so there's no reason to give anybody a free pass.
It says something about a company's abuse department if they're on that list for more than a day or two.
Uh, concrete testing is done via filling multiple plastic cylinders with concrete as it comes off of each truck to cure for 21-28 days prior to being test for strength, atleast here in CT.
That's the problem. You know 21 days after the pour if the concrete is no good. That kind of testing only works with a strong enforcement bureaucracy behind it.
Why doesn't an attempt to legally incorporate a new business include a "do any of your officers have a background in crime, particularly white collar crime?" check?
That's a real problem. People barred from involvement in the securities industry keep slipping back in. Bar owners barred from holding a liquor license often end up doing some deal as a "silent partner".
I get complaints from "web businesses" who want to operate anonymously, because SiteTruth down-rates them for that. (It's a criminal offense to run a business anonymously in many jurisdictions.)
The big mistake OLPC made was Nicholas Negroponte. He's too much into his own self-importance. His thing was dealing at the national leader level and getting himself into the press. What was needed was somebody who knew how to get a low-cost product out the door and sell it in quantity.
The OLPC should have been in a bubble-pack in every Wal-Mart and Walgreens in America, in every souk in the Middle East, and in every market in India, selling at a small profit and dropping in price every three months. But no, Negroponte had to try to make big deals with governments. That might have happened after they actually had the product out there in volume.
There's no technical problem making a wood building that strong. It's the enforcement that's the problem. Wood has good tensile strength, but the joints usually used in wood construction don't.
A few years ago, after some hurricanes, many Florida builders were discovered not to be building to code. Hurricane-proofing for small wood structures mostly consists of putting in metal brackets at joints to give wood-to-wood joints tensile strength. Not only do the brackets have to be put in, nails have to be hammered into all the holes in the brackets. Many contractors were sloppy about that, resulting in a big loss of tensile strength and major damage (like roofs ripped off) during hurricanes.
A big problem in the Third World is bad concrete mixes. Much concrete construction goes up without enough cement in the mix, and that results in building collapses.
Here's a good project for someone - develop a low-cost hand held device for concrete testing. The existing techniques are slow, labor-intensive, and a pain to use. Tests for hardened concrete usually involve cutting out a plug and sending it to a lab elsewhere. Small portable devices would be a big help here.
Big deal. Such drivers are trivial.
Virtualizing physical I/O devices on PC-like architectures requires code in the hypervisor to emulate the device. The driver in the operating system does stores into "device registers" as if talking the real device. Each such store or load causes a trap to the hypervisor, which has a device emulator watching the register changes and pretending to be the real peripheral. When the right registers have been loaded with the right values, and the final register store is made that would start the I/O operation, the device emulator then figures out what the OS wanted to do, and makes a call to the hypervisor's I/O system to do it.
In many cases, the device driver in the OS is doing all the optimization for the device controller of a real disk, doing angular optimization and head movement minimization. Since the real device underneath may be completely different, most of this is wasted work, and may reduce performance instead of increasing it.
So it's common to have dummy device drivers for virtual machines that just pass the OS's request through to the hypervisor, without trying to manage a real device. Such drivers don't do much, and are usually trivial, although Microsoft will probably try to complicate them somehow.
This isn't a new idea; it first appeared in IBM's VM for the System/370, where such calls were passed through using the DIAGNOSE instruction (an opcode used for hardware diagnostics only, and thus never used in ordinary programs and available as a spare opcode.)
One of the hypervisor vendors calls this "paravirtualization".
Wikipedia isn't there to make celebrities look good. From a historical perspective, an honest picture is better than a carefully posed and manipulated one.
If a celebrity has been arrested, their picture on Wikipedia is likely to be a Government mug shot.
Isn't someone running a static checker on the Linux kernel? There are commercial tools which will find code that can dereference NULL. However, there aren't free tools that will do this.
Check out Vocaloid 2, the Japanese singing synthesizer. This does a nice job. Here's a sample.
Setting up Vocaloid for a new voice is a big job; the current version requires that the performer sing a long, standardized set of training syllables. Once you have the voice configured, you feed in a MIDI file with lyrics, pitch, and timing info, and singing or talking comes out.
Clearly, the next step is the ability to train the system from unstandardized speech and singing samples. Once that technology is developed, it should be possible to automatically generated cover versions of any desired song with any desired performer. At last, the computer industry will be in a position to crush the RIAA.
(Remember, under US copyright law, anyone can make a cover version of any song and just pay the statutory royalty to the songwriter, a modest fee. The songwriter cannot refuse.)
One of the real reasons behind this is that the gear train to connect a gas turbine to a propeller is a huge pain. Turbine shaft speeds are around 20,000 RPM, and this has to be stepped down to 200 RPM or so for the propeller. There's a clutch and reversing gear, too. Often the gearbox is bigger than the engine.
Reverse with a turbine is a headache. Some marine diesels are built so they can run in either direction, but that's not an option with a turbine. So there's either a really big gearshift, or a second powerplant for reverse operation.
Thus, there's ongoing interest in going to electric transmission, like a locomotive. It's not a new idea; an electric transmission was tried on a ship before WWII. But it works much better with modern power semiconductors. Locomotives do this now. Modern locomotives use AC to DC to variable frequency polyphase AC conversion to drive the motors. This takes large switching power supplies, using very large semiconductors. It's a solved problem; GE locomotives have been doing this for ten years now, and their competitors now have comparable technology. Moving the technology to ships is an obvious move at this point.
There's also a Navy effort to develop a permanent magnet motor (!) big enough for shipboard propulsion. This gets rid of the field coils and increases efficiency. I'm not sure how that's coming along.
This is routine progress being made in heavy machinery. The combination of electronics and really big gears and motors can do things neither can do alone.
First, this isn't new; the company issued a press release on PR Newswire in January 2007.
It has nothing to do with Israel; the work is being done at Cleveland BioLabs in Cleveland, Ohio. The researcher behind this, Andrei Gudkov, is Russian. He was at the National Cancer Research Center in Moscow until 1990, then came to the US and became a professor at the University of Illinois.
This seems to be legitimate; they're in FDA Phase I human testing (safety only, not effectiveness.). That doesn't mean it will work; if it makes it through Phase II, it's real.
A better binary diffing algorithm is useful for source control. But for security? If the code is so awful that the bandwidth required for security updates is a problem, the product is defective by design.
It sounds like Google tried "agile programming" on trusted code, and now has to deal with the consequences of debugging a pile of crap.
"It's utter bilge. I don't think anybody will ever put up enough money to do such a thing . . . What good would it do us? If we spent the same amount of money on preparing first-class astronomical equipment we would learn much more about the universe . . . It is all rather rot" -- Riet Wolley, Astronomer Royal.
He was so right.
Basic truth: space travel with chemical rockets is just too inefficient to be useful. Fuels are as good as they can get; we've been using liquid hydrogen since the 1960s. It's not getting any better. Space travel is about weight reduction, which means fragile vehicles. (Endeavour just had some foam fall off and hit the thermal tiles during launch. Again.)
Unless and until we get something better than chemical rockets, space travel isn't going to get any better.
If we'd launched an Orion or two in the 1950s, things might have been very different. Everyone would know there's a better way.
At $5 each, there are few options. Rainbow Kits are a possibility. The "blinking lights" and "1W audio amplifier" kits are both under $5.99. That's about as low as you can go.
The only reason it's still interesting is because there's historical data. Giving the same test to groups over many decades is interesting in that it helps to spot long-term trends. The MMPI, the Minnesota Multiphasic Personality Inventory, is like that. The definition of "normal" in the 1930s is rather different from the results today, and it's useful to see the trends.
Here's a classic "big secret:" Blaney's Ladder Levitation. It was first performed on TV on The Tonight show in 1973. It's still considered a "big secret" by professional magicians, but if you watch the video closely, you should be able to figure it out. Few illusions survive frame-by-frame examination.
OS 2200, for UNISYS mainframes, is somewhere around Level 48 today. The original demo release was April 4, 1967. Level 23 was reached around 1970, level 32 around 1978, and level 48 in 2007. Builds have numbers like "26.51.465".
Yes, there really is an operating system over 40 years old in current production use.
Mod parent up. In the serious crypto world, this is a good thing, provided it doesn't happen too often. Sometimes you're going to lose a key, because, for security reasons, you don't keep extra copies. You have a procedure for issuing new keys when this happens, which you're routinely doing anyway.
Take it from me, a computer science degree definitely won't help you get a job.
He may be right. Which is sad.
I'm not complaining personally. I've been fortunate enough to do plenty of good computer science over the course of a long career. I've done everything from networking theory to proof of correctness to autonomous robots. I've done web stuff, too, and today write mostly Python, so I'm not stuck in the past.
But new graduates, especially if they only have a BS, are going to have trouble finding a good career. Most of the basic problems in the field have been solved, and the ones that remain require deep specialization before you can make progress. The previous poster is right; employers want a bunch of buzzwords, because much of programming at the working levels today is knowing how to interface with a huge collection of ever-changing APIs, and those take time to learn.
A century ago, the stationary engineers who set up factories to run on electricity and steam had the same problem. In 1870, that was cutting-edge technology. By 1910, factory owners just ordered motors and hooked them up. Today, it's a union job.
There are good computer science problems to work on today, but it's hard to find a job that needs them solved. There's great stuff going on in machine learning. Game AIs could get much better. Robotics is finally starting to work. Computer vision is getting very good. There's a whole world of interesting technology associated with semiconductor design automation. But 99% of working programmers won't get into those areas. They'll be stuck doing e-commerce sites forever.
The smart young people I know all seem to be going into either bio or law. A few years ago, some of them were going into hedge funds, but that seems to be over.
In the US, this is completely legal. That's been settled law since Bridgeman vs. Corel (Corel issued a CD of photos of public domain paintings) and Feist vs. Rural Telephone (phone books not copyrightable; no creativity.). In fact, in Feist, the Supreme Court held that it's a constitutional issue; Congress's right to make copyright law is limited to creative works. Nor does the US have "database copyright", despite lobbying attempts for it. There's also Meshwerks vs. Toyota, which reinforces Bridgeman at the appellate level.
UK law in this area is still iffy. Which is going to be a problem here.
It's already Friday in most time zones. Is this happening?
The problem is that sex in games usually doesn't provide good gameplay. If you just want to watch porn, that's easily available. Besides, simulated porn doesn't look that good.
That said, the big flap about the "hot coffee" scene in GTA was sort of silly. The GTA world ought to have sex in it. In fact, it's inconsistent that a game with strippers and hookers doesn't have sex in it. There's so much unrealized potential there, for seduction, power games, devious girlfriends - all the basic male/female drama elements.
The key is integrating sex into gameplay without having the sex dominate the game. That's a design challenge. It's not impossible. Second Life has sex, but it's not primarily a sex-oriented MMORPG. What we need are R-rated video games, dramas where sex plays a role in the plot. That could be fun.
It's so government.
I once worked in an R&D facility for heavy hydraulic equipment. They had about fifty test cells of different sizes, the largest of which was used for hydraulic transmissions for medium-sized locomotives. Those test setups used a big motor and a water-cooled brake; the hot water went through a cooling tower, and then to sprinklers in what appeared to be a decorative lake out front but was really a heat sink. That gear was in the 5MW range, somewhat smaller than what's being described here, but not a lot smaller.
That setup was where it belonged, near the engineers who designed the things and the machinists who built the prototypes. When the big test cell was put in, it took a few months to build. Not five years.