In short, we would need to capture a small fraction of the mass of one of the gas giants, and transport it to Mars. It would be interesting to see a proposal on how that could be done.
systemd does things like auto-detect all of the tty devices, and automatically associate them with login prompts when the device becomes active. This sounds good, until you hit an application where the tty device should not have a login prompt. After two days of trying to work around the issue (there is a work around), I now understand what everyone was complaining about...
The biggest issue is that everything is wrapped in layers of configuration scripts, and this makes it is difficult to do something specific. The distros in an effort to "make everything easier" then have their own distro-specific scripts, and this makes the problems even worse.
The old way had one configuration script per activity, and this had the advantage that you only had to worry about one script.
Have the crime lab test a series of random samples periodically. Get random objects from the police department or the prosecutor's office. If any come back positive, something is going on. Better quality control techniques would vary the amount of drugs (or DNA) on the object and this would show how accurate and reproducible the lab's technique is at detecting it.
A big problem in these cases is that no quality control is being done whatsoever!
Those statement's don't mean the same thing. For example: consider an experiment where X was expected, however the magnitude of the X effect is unknown, relative to background noise.
1. "We find that, at 90% confidence level, there is no statistically significant evidence for X" means that the experiment background noise overwhelmed X.
2. "We found that X didn't actually happen." could be a groundbreaking result. However, if background noise was large, how would you know?
3. "We find that, at 75% confidence level, a large number of events occurred but we could not confirm they were X" means that interesting results might happen if a better experiment were run.
4. "We find that, at 99.999% confidence level, X occurred" means X occurred at a high level of statistical certainty.
The reason for the passive tense is that academic writers tend to take academic disputes personally. Suppose a Nobel prize winner's graduate student wrote a paper where he said "we did this and found this", and someone realized the error in the experiment, then wrote "you did this, missed this, and you didn't actually find this." The professor's ego's would be massively bruised, and a massive rift formed in the research community.
If the passive voice is used, then it's the experiment's fault for being wrong. Using the passive voice is the equivalent of the "always address the chair" rule in Robert's rules of order. If everyone addresses the chair, then it is more difficult to meetings degenerate into personal name-calling. It doesn't always work, but it keeps things civil.
Copyright covers pretty much any new work, even if it is a copy of an old work. An example of this, bible companies routinely copyright the bible. The crucial caveat's are:
1. If the original work is in the public-domain, then anyone can copy the original work.
2. If you publish a copy of the original work, say as part of a collection or with editor's annotations, then you can sue the daylights out of anyone that copies your new work.
In practice this means, if a third party publishes a new copy of the "public domain" work, then they must make sure they publish a new work based on the original, and not the revised collection. Also, if the copyright holder of the later work cannot prove the third-party copy is a copy of his new work as opposed to the original work, then the court will likely just decide that the original work is in the public domain and that is that.
In the case of the bible, anyone can copy any public-domain copy of the bible. Most of the copies of the bible in circulation feature the "latest updates from the dead sea scrolls translated into common English", which makes almost all of the modern copies of the bible slightly different. As such, copyright can then be claimed on the latest and best translation. Many priests, clergy, bible scholars and theologians spend time comparing different versions of the bible to see how they are different.
Instead of excellence, the modern educational system says: "These students are interested in something, so let's educate a different group on the topic!"
People should be saying: "These students are interested in computer programming, let's make them better programmers!" Demand should be created through the celebration of accomplishments.
Taking away the achievements of the interested, results in mediocrity. Yes, it would be nice to have more girls in computer programming. However, the goal of the educational system is often to make everyone the same. To make the interested boys equal to the uninterested girls. Is this the solution we want? Because that is what the school system will implement. The modern school system is very good at targeting the average (or the below average). It sucks at enabling gifted students to excel.
It is very easy to get yourself into this situation as a software engineer. The other engineers don't ask you to cheat. This happens:
1. Some engineer figures out that since "x" isn't working, they need a "test bypass" function to keep the program going forward. By itself, this is really common. The engine might be running on a test stand, and most of the car is missing. As such, the software has to have a "test bypass" function to deal with the missing gas peddle, brakes, transmission, etc.
2. The program keeps moving along, using the "test bypass" function.
3. The entire car is assembled, and the "test bypass" function is still in use, because no one ever figured out how to make the meet the emission test.
4. The "design freeze" is made, as the car is ready for production. However, the test code is still in place.
5. The car still doesn't pass the emission test, but it is ready for production and sale.
Health-care and teaching appear much more expensive today because manufactured goods cost much less. If you look at where people are employed and making a good living, from 1960 to today, it quickly becomes obvious that certain sectors, like manufacturing, have became much smaller due to productivity improvements. Other sectors, like teaching and health care have not benefited from the productivity miracle the same extent. Additionally, certain intangible sectors of the economy that have went from non-existent in 1960 to a significant expenditure today. I'm thinking of finance, including credit cards and student loans, the software industry, and cell/internet companies. These companies have much higher worker-productivity levels than manufacturing has ever achieved. As a result, they employ very few people per profit developed.
Assume that all of the good paying jobs in the economy pay roughly the same amount, and the workers are distributed based on productivity and the realistic material needs/wants levels. It quickly becomes obvious that health care and teaching must be a larger portion of the economy in terms of workers, because we want health-care and education and therefore need the workers in these occupations. On the other hand, manufacturing, software and finance must be smaller portion of the economy in terms of workers, because they need fewer workers and there is an upper limit on the output we could want from these sectors. This macro-economic model applies at both large and small scales. If the overall economy has lots of workers in health-care, then the average person must be spending lots of money in that sector. The same applies to education.
To make it to the post-industrial utopia, we need to:
a) automate health-care and teaching in much the same way we have automated manufacturing, and
b) ensure that the people that lack high-income skill-sets remain useful and contributing members of the economy.
I tend to think America will achieve (a). (b) is tougher.
Have you looked at the cost per page of the new lasers? The same companies that brought us hideously priced ink for inkjets have created even more expensive toner for laser printers.
I don't purchase a new printer until I have computed the cost per page of the toner/ink. Often, the heavier "business" class devices have a much lower cost/page on the toner cartridges, and the savings often pays for itself in very short times (on the order of 3 to 4 sets of cartridge changes.) Also, colour lasers are more expensive to run than black and white laser printers. Beyond that, inkjets might be cheaper, or they might not.
The ink-jet cartridges measure their print out volume based on the number of droplets deposited. A +/- 5% change in ink droplet diameter represents a +/- 15% change in volume. When dealing with really small feature sizes, variable temperatures, and variable viscosities, it is really tough to control droplet diameter exactly. The result is that the ink-cartridge manufacturers need to overfill their cartridges to guarantee that some customer in some corner case doesn't experience a rash of cartridges that run out early.
This tactic is kind of like the hand-soap people that sell a 1 L container of soap with a hand-pump that only works for the first 950 mL. If we can see the soap in the container, we get annoyed because of the 50 mL of waste. However, the ink-cartridge people hide the amount of ink left in the "empty" cartridge, so we don't notice the waste.
Of course, when you are dealing with professional cartridges, and print-outs that can be worth big money, the printer cartridge manufacturers have to guarantee that the ink doesn't run out. The cheapest way to do this is to add a little bit of ink.
In the case of consumer cartridges, HP, Lexmark, and Epson would be deeply upset if a bunch of the customers complained about "empty" cartridges that still said they had 5% capacity left. To prevent complaints, add a little bit of ink...
Adding a little ink makes everyone happy, until someone actually looks at what is left in the "empty" cartridges, and measures it with sufficiently accurate equipment to realize how much "extra" ink is left.
The SLS exists to give pork to established NASA contractors. SpaceX is trying to get stuff into space cheaply.
SpaceX is centralized in a few districts so it gets relatively little support. On the other hand, the SLS has pork divided up over the whole country. Thus, if you are a politician, and want pork, you want to support the SLS. The fact that the SLS makes no scientific or financial sense whatsoever, does not factor into the decision to vote against SpaceX. To bring pork to your district, SLS is the correct program.
Unfortunately, SLS has went the way of many of the more recent military purchase programs. Yes, the F-35 can be built, but why? Yes the SLS can be built, but is this really the best way? do we really need it? Given SpaceX's development trajectory, will the SLS ever be needed? Really needed?
If it can't make it's way through a junction where the drivers are following the rules, that's bad programming. If it can't make it's way through a junction where other drivers don't come to a complete halt for it, it's not fit to be on the road with other drivers.
The problem is that people don't follow rules. We follow approximations of the rules. For instance, my driver's handbook described the correct way to deal with yielding at a four-way stop as "yield to the person on the right." For a computer, that's an obvious deadlock situation, or worse - an obvious mistake. If four cars are parked at a four way stop, and each car yields to the car on the right, then (a) a situation could occur where no one goes anywhere, and (b) if the individual cars only pay attention to the person on the right, then they could hit an on-coming car turning left, or the car on the left turning left. People process the "yield to the person on the right" rule into something much more complex.
People use a number of complex behaviours at four-way stops. Firstly, the wave of the hand, or the nod of the head to indicate that you yield to the other driver is an important signal. Secondly, in my jurisdiction, 90% of the four way stops are done on a first-come first-served basis. Lastly, and this is the bit I don't understand, often people yield to the person on the left. The actual system of navigating a four-way stop is much more complex than what an initial computer implementation might be.
The article specifically states that rank in sponsored links correlates to advertising spend, which I would expect.
I would also expect a weaker correlation between page rank and advertising spend in the normal links. Firstly, a site with significant advertising spend will hopefully generate more hits, and this should translate into page-rank. Secondly, I would expect a site with significant advertising spend to spend more on its site, which hopefully results in a more informative and more useful site. In turn, this should result in a weak correlation between advertising spend and page rank. Lastly, some correlation probably exists between advertising spending, and hits from the google search spider. This may translate into improved page rank for trending topics.
In all, I would be surprised if there were not correlations between advertising spend and Google page rank. What I do like from Google is that they clearly label the sponsored versus non-sponsored links. Also, Google also has a number of non-commercial sites at the top of many search suggestions, which indicates that they treat sites without advertising spend reasonably.
The grandparent poster is talking about compressing videos. If something is known about the data being encoded, then it is trivial to show that you can exceed the performance of arithmetic coding, because arithmetic coding makes no assumptions about the underlying message.
For instance, suppose I was encoding short sequences of positions that are random integer multiples of pi. Expressed as decimal or binary numbers, the message will seem highly random, because of the multiplication by an irrational number (pi). However, if I can back out the randomness introduced by pi, then the compression of the resulting algorithm can be huge.
The same applies to video. If it is possible to bring more knowledge of the problem domain to the application, then it is possible to do better on encoding. Especially with real-life video, there are endless cheats to optimize compression. Also, Dropbox may not be limited by real-time encoding. Drop-box might not even need intermediate frames to deal with fast-forward and out-of-order viewing. Dropbox may be solely interested in creating an exact image of the original file. Knowing the application affects compression dramatically.
Lastly, application specific cheats can save real-world companies and individuals money and time. Practical improvements count as advancements too.
They could be doing a Citrix-like thing where everyone is logging on to server-housed remote instances. If each instance was one VMWare VM, then 3,000 employees and 1500 VMWare instances makes sense.
They could also be a company where each corporate customer, or groups of individual customers, require their own virtual server. For example: an SaaS accounting firm similar to FreshBooks may have a separate virtual server for every corporate customer, or Blizzard where groups of users have their own dungeon.
Otherwise, even for a IT company, they have the IT infrastructure from hell. I can't imagine 1500 different server applications in a company of 3,000 people.
RTGs are being phased out because (a) the probes need more power than ever with modern computers, and (b) because of environmental concerns. Unfortunately, most of the environmental concerns revolve around the word "Plutonium" and the much more dangerous Plutonium-239.
The best RTGs use a chemically locked Plutonium-238 Oxide that is probably one of the safest fuel sources ever invented. The stuff is non-reactive ceramic that is almost indestructible, and is readily rejected by the human body if ingested. It's not even particularly radioactive, as radio-active isotopes go, because to make the RTG last for a long period of time, it is necessary to use an isotope with a sufficiently long-lived half-life characteristic. Plutonium-238 Oxide is the polar opposite of the more typical dangers from Plutonium-239 that everyone worries about.
On error, dynamic_cast returns NULL for pointers, and an exception for references. From cppreference:
If the cast is successful, dynamic_cast returns a value of type new_type. If the cast fails and new_type is a pointer type, it returns a null pointer of that type. If the cast fails and new_type is a reference type, it throws an exception that matches a handler of type std::bad_cast.
I gave up on the concept that I would be able to write and debug programs correctly the first time. Now all the central data structures in any long-lived control system get error-checking code added to them. For example, the sorted-list code is built with a checker to ensure it stays in order. The communications code gets error-checking. The PID controllers get min/max testing, etc.
Every once in a while I come across a bugs that are not in the source code. Often they are compiler errors. Sometimes the bugs involve a rare C/C++ or operating system eccentricity. Sometimes the errors are caused by obscure library changes. Sometimes they are hardware errors.
Especially with the embedded micro-controllers, I leave the consistency checking code in, because you just can't assume the everything always works. The nature of software bugs change with time, and it is not always in the way a programmer would expect. I am frequently surprised by how obscure some of the bugs are.
Per core or Per CPU software pricing can dominate the cost calculation. We have a CFD application, and we were considering boosting the hardware. One look at the software costs discouraged us.
A costly complication is that 3-D CFD (or FEA), is an O(n^3) problem. Doubling the mesh density means 2^3=8 times the CPU time. An increase of 10 times in the mesh density requires 10^3=1000 times the CPU time. If you are pushing the extreme, small changes in the mesh density have significant cost impacts.
It makes me wonder how many research groups are paying full-cost for the commercial CFD packages. Many universities, some quasi-government labs, and many small startups will not have the money for the full-price commercial packages.
MariaDB. Google switched from MySQL to Oracle to MySQL to Google F1 for it's AdWords technology. See the wiki page on Adwords. Since then, many companies, including Google have switched their smaller MySQL databases to MariaDB.
There is an interesting account of the Google Oracle migration at the wayback machine.
Slashdot Mirror
In short, we would need to capture a small fraction of the mass of one of the gas giants, and transport it to Mars. It would be interesting to see a proposal on how that could be done.
systemd does things like auto-detect all of the tty devices, and automatically associate them with login prompts when the device becomes active. This sounds good, until you hit an application where the tty device should not have a login prompt. After two days of trying to work around the issue (there is a work around), I now understand what everyone was complaining about ...
The biggest issue is that everything is wrapped in layers of configuration scripts, and this makes it is difficult to do something specific. The distros in an effort to "make everything easier" then have their own distro-specific scripts, and this makes the problems even worse.
The old way had one configuration script per activity, and this had the advantage that you only had to worry about one script.
Have the crime lab test a series of random samples periodically. Get random objects from the police department or the prosecutor's office. If any come back positive, something is going on. Better quality control techniques would vary the amount of drugs (or DNA) on the object and this would show how accurate and reproducible the lab's technique is at detecting it.
A big problem in these cases is that no quality control is being done whatsoever!
Those statement's don't mean the same thing. For example: consider an experiment where X was expected, however the magnitude of the X effect is unknown, relative to background noise.
1. "We find that, at 90% confidence level, there is no statistically significant evidence for X" means that the experiment background noise overwhelmed X.
2. "We found that X didn't actually happen." could be a groundbreaking result. However, if background noise was large, how would you know?
3. "We find that, at 75% confidence level, a large number of events occurred but we could not confirm they were X" means that interesting results might happen if a better experiment were run.
4. "We find that, at 99.999% confidence level, X occurred" means X occurred at a high level of statistical certainty.
The reason for the passive tense is that academic writers tend to take academic disputes personally. Suppose a Nobel prize winner's graduate student wrote a paper where he said "we did this and found this", and someone realized the error in the experiment, then wrote "you did this, missed this, and you didn't actually find this." The professor's ego's would be massively bruised, and a massive rift formed in the research community.
If the passive voice is used, then it's the experiment's fault for being wrong. Using the passive voice is the equivalent of the "always address the chair" rule in Robert's rules of order. If everyone addresses the chair, then it is more difficult to meetings degenerate into personal name-calling. It doesn't always work, but it keeps things civil.
Copyright covers pretty much any new work, even if it is a copy of an old work. An example of this, bible companies routinely copyright the bible. The crucial caveat's are:
1. If the original work is in the public-domain, then anyone can copy the original work.
2. If you publish a copy of the original work, say as part of a collection or with editor's annotations, then you can sue the daylights out of anyone that copies your new work.
In practice this means, if a third party publishes a new copy of the "public domain" work, then they must make sure they publish a new work based on the original, and not the revised collection. Also, if the copyright holder of the later work cannot prove the third-party copy is a copy of his new work as opposed to the original work, then the court will likely just decide that the original work is in the public domain and that is that.
In the case of the bible, anyone can copy any public-domain copy of the bible. Most of the copies of the bible in circulation feature the "latest updates from the dead sea scrolls translated into common English", which makes almost all of the modern copies of the bible slightly different. As such, copyright can then be claimed on the latest and best translation. Many priests, clergy, bible scholars and theologians spend time comparing different versions of the bible to see how they are different.
Instead of excellence, the modern educational system says: "These students are interested in something, so let's educate a different group on the topic!"
People should be saying: "These students are interested in computer programming, let's make them better programmers!" Demand should be created through the celebration of accomplishments.
Taking away the achievements of the interested, results in mediocrity. Yes, it would be nice to have more girls in computer programming. However, the goal of the educational system is often to make everyone the same. To make the interested boys equal to the uninterested girls. Is this the solution we want? Because that is what the school system will implement. The modern school system is very good at targeting the average (or the below average). It sucks at enabling gifted students to excel.
After the Greece fiasco, cash is king.
It is very easy to get yourself into this situation as a software engineer. The other engineers don't ask you to cheat. This happens:
1. Some engineer figures out that since "x" isn't working, they need a "test bypass" function to keep the program going forward. By itself, this is really common. The engine might be running on a test stand, and most of the car is missing. As such, the software has to have a "test bypass" function to deal with the missing gas peddle, brakes, transmission, etc.
2. The program keeps moving along, using the "test bypass" function.
3. The entire car is assembled, and the "test bypass" function is still in use, because no one ever figured out how to make the meet the emission test.
4. The "design freeze" is made, as the car is ready for production. However, the test code is still in place.
5. The car still doesn't pass the emission test, but it is ready for production and sale.
6. Massive scandal/recall ensues ...
Health-care and teaching appear much more expensive today because manufactured goods cost much less. If you look at where people are employed and making a good living, from 1960 to today, it quickly becomes obvious that certain sectors, like manufacturing, have became much smaller due to productivity improvements. Other sectors, like teaching and health care have not benefited from the productivity miracle the same extent. Additionally, certain intangible sectors of the economy that have went from non-existent in 1960 to a significant expenditure today. I'm thinking of finance, including credit cards and student loans, the software industry, and cell/internet companies. These companies have much higher worker-productivity levels than manufacturing has ever achieved. As a result, they employ very few people per profit developed.
Assume that all of the good paying jobs in the economy pay roughly the same amount, and the workers are distributed based on productivity and the realistic material needs/wants levels. It quickly becomes obvious that health care and teaching must be a larger portion of the economy in terms of workers, because we want health-care and education and therefore need the workers in these occupations. On the other hand, manufacturing, software and finance must be smaller portion of the economy in terms of workers, because they need fewer workers and there is an upper limit on the output we could want from these sectors. This macro-economic model applies at both large and small scales. If the overall economy has lots of workers in health-care, then the average person must be spending lots of money in that sector. The same applies to education.
To make it to the post-industrial utopia, we need to:
a) automate health-care and teaching in much the same way we have automated manufacturing, and
b) ensure that the people that lack high-income skill-sets remain useful and contributing members of the economy.
I tend to think America will achieve (a). (b) is tougher.
It's a question without a good answer. There doesn't appear to be a "permanently prevent Windows 10 upgrade" switch anywhere.
This could be a business opportunity. Write a piece of software that automatically finds and suppresses any attempts to update to Windows 10.
Have you looked at the cost per page of the new lasers? The same companies that brought us hideously priced ink for inkjets have created even more expensive toner for laser printers.
I don't purchase a new printer until I have computed the cost per page of the toner/ink. Often, the heavier "business" class devices have a much lower cost/page on the toner cartridges, and the savings often pays for itself in very short times (on the order of 3 to 4 sets of cartridge changes.) Also, colour lasers are more expensive to run than black and white laser printers. Beyond that, inkjets might be cheaper, or they might not.
The ink-jet cartridges measure their print out volume based on the number of droplets deposited. A +/- 5% change in ink droplet diameter represents a +/- 15% change in volume. When dealing with really small feature sizes, variable temperatures, and variable viscosities, it is really tough to control droplet diameter exactly. The result is that the ink-cartridge manufacturers need to overfill their cartridges to guarantee that some customer in some corner case doesn't experience a rash of cartridges that run out early.
This tactic is kind of like the hand-soap people that sell a 1 L container of soap with a hand-pump that only works for the first 950 mL. If we can see the soap in the container, we get annoyed because of the 50 mL of waste. However, the ink-cartridge people hide the amount of ink left in the "empty" cartridge, so we don't notice the waste.
Of course, when you are dealing with professional cartridges, and print-outs that can be worth big money, the printer cartridge manufacturers have to guarantee that the ink doesn't run out. The cheapest way to do this is to add a little bit of ink.
In the case of consumer cartridges, HP, Lexmark, and Epson would be deeply upset if a bunch of the customers complained about "empty" cartridges that still said they had 5% capacity left. To prevent complaints, add a little bit of ink ...
Adding a little ink makes everyone happy, until someone actually looks at what is left in the "empty" cartridges, and measures it with sufficiently accurate equipment to realize how much "extra" ink is left.
The SLS exists to give pork to established NASA contractors. SpaceX is trying to get stuff into space cheaply.
SpaceX is centralized in a few districts so it gets relatively little support. On the other hand, the SLS has pork divided up over the whole country. Thus, if you are a politician, and want pork, you want to support the SLS. The fact that the SLS makes no scientific or financial sense whatsoever, does not factor into the decision to vote against SpaceX. To bring pork to your district, SLS is the correct program.
Unfortunately, SLS has went the way of many of the more recent military purchase programs. Yes, the F-35 can be built, but why? Yes the SLS can be built, but is this really the best way? do we really need it? Given SpaceX's development trajectory, will the SLS ever be needed? Really needed?
The problem is that people don't follow rules. We follow approximations of the rules. For instance, my driver's handbook described the correct way to deal with yielding at a four-way stop as "yield to the person on the right." For a computer, that's an obvious deadlock situation, or worse - an obvious mistake. If four cars are parked at a four way stop, and each car yields to the car on the right, then (a) a situation could occur where no one goes anywhere, and (b) if the individual cars only pay attention to the person on the right, then they could hit an on-coming car turning left, or the car on the left turning left. People process the "yield to the person on the right" rule into something much more complex.
People use a number of complex behaviours at four-way stops. Firstly, the wave of the hand, or the nod of the head to indicate that you yield to the other driver is an important signal. Secondly, in my jurisdiction, 90% of the four way stops are done on a first-come first-served basis. Lastly, and this is the bit I don't understand, often people yield to the person on the left. The actual system of navigating a four-way stop is much more complex than what an initial computer implementation might be.
The article specifically states that rank in sponsored links correlates to advertising spend, which I would expect.
I would also expect a weaker correlation between page rank and advertising spend in the normal links. Firstly, a site with significant advertising spend will hopefully generate more hits, and this should translate into page-rank. Secondly, I would expect a site with significant advertising spend to spend more on its site, which hopefully results in a more informative and more useful site. In turn, this should result in a weak correlation between advertising spend and page rank. Lastly, some correlation probably exists between advertising spending, and hits from the google search spider. This may translate into improved page rank for trending topics.
In all, I would be surprised if there were not correlations between advertising spend and Google page rank. What I do like from Google is that they clearly label the sponsored versus non-sponsored links. Also, Google also has a number of non-commercial sites at the top of many search suggestions, which indicates that they treat sites without advertising spend reasonably.
The grandparent poster is talking about compressing videos. If something is known about the data being encoded, then it is trivial to show that you can exceed the performance of arithmetic coding, because arithmetic coding makes no assumptions about the underlying message.
For instance, suppose I was encoding short sequences of positions that are random integer multiples of pi. Expressed as decimal or binary numbers, the message will seem highly random, because of the multiplication by an irrational number (pi). However, if I can back out the randomness introduced by pi, then the compression of the resulting algorithm can be huge.
The same applies to video. If it is possible to bring more knowledge of the problem domain to the application, then it is possible to do better on encoding. Especially with real-life video, there are endless cheats to optimize compression. Also, Dropbox may not be limited by real-time encoding. Drop-box might not even need intermediate frames to deal with fast-forward and out-of-order viewing. Dropbox may be solely interested in creating an exact image of the original file. Knowing the application affects compression dramatically.
Lastly, application specific cheats can save real-world companies and individuals money and time. Practical improvements count as advancements too.
They could be doing a Citrix-like thing where everyone is logging on to server-housed remote instances. If each instance was one VMWare VM, then 3,000 employees and 1500 VMWare instances makes sense.
They could also be a company where each corporate customer, or groups of individual customers, require their own virtual server. For example: an SaaS accounting firm similar to FreshBooks may have a separate virtual server for every corporate customer, or Blizzard where groups of users have their own dungeon.
Otherwise, even for a IT company, they have the IT infrastructure from hell. I can't imagine 1500 different server applications in a company of 3,000 people.
RTGs are being phased out because (a) the probes need more power than ever with modern computers, and (b) because of environmental concerns. Unfortunately, most of the environmental concerns revolve around the word "Plutonium" and the much more dangerous Plutonium-239.
The best RTGs use a chemically locked Plutonium-238 Oxide that is probably one of the safest fuel sources ever invented. The stuff is non-reactive ceramic that is almost indestructible, and is readily rejected by the human body if ingested. It's not even particularly radioactive, as radio-active isotopes go, because to make the RTG last for a long period of time, it is necessary to use an isotope with a sufficiently long-lived half-life characteristic. Plutonium-238 Oxide is the polar opposite of the more typical dangers from Plutonium-239 that everyone worries about.
See RTG generators and plutonium oxide for more information.
On error, dynamic_cast returns NULL for pointers, and an exception for references. From cppreference:
while (something) {
// do_stuff
} while (something_else);
It compiles, is legal C, and loops endlessly if something_else is true.
It can be done in a careless moment when switching a complex piece of code from a while () loop to a do-while () loop.
I gave up on the concept that I would be able to write and debug programs correctly the first time. Now all the central data structures in any long-lived control system get error-checking code added to them. For example, the sorted-list code is built with a checker to ensure it stays in order. The communications code gets error-checking. The PID controllers get min/max testing, etc.
Every once in a while I come across a bugs that are not in the source code. Often they are compiler errors. Sometimes the bugs involve a rare C/C++ or operating system eccentricity. Sometimes the errors are caused by obscure library changes. Sometimes they are hardware errors.
Especially with the embedded micro-controllers, I leave the consistency checking code in, because you just can't assume the everything always works. The nature of software bugs change with time, and it is not always in the way a programmer would expect. I am frequently surprised by how obscure some of the bugs are.
Per core or Per CPU software pricing can dominate the cost calculation. We have a CFD application, and we were considering boosting the hardware. One look at the software costs discouraged us.
A costly complication is that 3-D CFD (or FEA), is an O(n^3) problem. Doubling the mesh density means 2^3=8 times the CPU time. An increase of 10 times in the mesh density requires 10^3=1000 times the CPU time. If you are pushing the extreme, small changes in the mesh density have significant cost impacts.
It makes me wonder how many research groups are paying full-cost for the commercial CFD packages. Many universities, some quasi-government labs, and many small startups will not have the money for the full-price commercial packages.
MariaDB. Google switched from MySQL to Oracle to MySQL to Google F1 for it's AdWords technology. See the wiki page on Adwords. Since then, many companies, including Google have switched their smaller MySQL databases to MariaDB.
There is an interesting account of the Google Oracle migration at the wayback machine.
Say what?