Seriously though, there are no new technology on the horizon that would make silicon run cooler, and the speed of core-voltage drop does not keep up with frequency bumps (heat is square of frequency for CMOS gates).
Sure there are: asynchronous logic, reconfigurable computing, reversible computation, and many other technologies. Of course, that requires significant changes to the way we design software and hardware...
The interconnects in most Origins and Altix systems are 3.2 gigaBYTE per second with extremely low latency.
I think SGI is playing with the numbers there: while that is the limit for how much data you can push through a link, it seems unlikely that you get that kind of performance for arbitrary processor-to-processor communications or as aggregate bandwidth.
In any case, very fast interconnects are not cost-efficient when they are not needed. 3.2GB is the memory bandwidth of a 400MHz bus motherboard. Most numerical applications do not spend all their time moving stuff around between different processors, and it just isn't worth paying a premium in order to get that kind of bandwidth if you don't need it.
Also, the Altix can handle up to 64 processors per single machine / single node (or 128 with a very beta set of patches). The cluster in the article is actually four Altix systems, each with 64 processors. The Origin 3800/3900 can handle 512 processors per node (or 1024 with a special "XXL" IRIX kernel).
Again, you pay a premium for a product that just doesn't make a lot of sense (at least to me): an abstraction of parallel computation in which 64 or 1024 processors all read and write a gigantic shared memory just isn't a good way of thinking about parallel computations, even if you could build it cheaply. And because it's NUMA, it only looks like a shared memory, but it really isn't anyway.
I think the SGIs have a niche: you can take dusty FORTRAN and multi-threaded C applications and scale them up to many processors with a minimum of fuss. That's very valuable, even if you pay a premium for the convenience. But I sure hope that the future of parallel programming won't look like that.
I think violence in movies or games probably can occasionally cause people to become murderers. And, yes, by that I mean that if those people had not played violent video games, they would not have become killers.
But just because A occasionally causes B doesn't mean that A by itself is sufficient to cause B, or that reducing A is the best way of reducing B. If you take happy, healthy kids and have them play violent video games or show them standard Hollywood movies dripping with violence, they won't turn into killers--something else needs to go wrong. Conversely, US courts have established that Twinkies sometimes cause people to become killers--should we ban Twinkies as well?
The fact that we have had very high levels of violence in the US (compared to other nations) far longer than video games have been around shows us that video games are not a major cause of violence in this culture. Let's find out what really turns so many Americans into killers and fix it; I suspect it's more related to poverty, education, and deep-rooted social problems than the depiction of violence in video games or movies.
It seems unlikely that the SGI is the first Linux cluster with global shared memory. There are plenty of distributed shared memory systems in software, some of them open source. You can find a list here. For most computations and most hardware, you are probably still better off with MPI or PVM rather than shared memory.
Note also that there are several high speed interconnects for Linux clusters available from many different vendors, including InfiniBand, Gigabit Ethernet, FireWire, and Myrinet.
Just tune into traffic info radio or pick up your cell phone and call one of the many free traffic info services. Or, have the information sent to your cell phone. But most people realize pretty quickly that that kind of knowledge is pretty useless: even if 101 or 405 are stop-and-go, taking alternate routes probably still takes longer than just living with it.
The only thing you can do is to stay a little longer at the office until traffic has died down. And to see when that has happened, you don't need a wireless gadget, you just point your desktop web browser at a traffic site.
For the same things I would use Perl, Java, or Matlab for: CGI scripts, system administration, text processing, numerical processing, GUI application development, and many other applications.
What are its strengths and its weaknesses?
Strengths: easy to learn, easy to read, much better error checking than Perl, much more concise than Java, lots of libraries, lots of GUI toolkits (Gtk+, wxWindows, Qt, Tk, FLTK, others).
Weaknesses: library modules are somewhat haphazard, no equivalent to CPAN, C implementation is rather ugly under the covers, C language interface is not very good (but still better than Perl). For short scripts, it's not as concise as Perl. While Numerical Python is "fast" in the same way Matlab is, it's still an interpreted language, and Java or C++ are a lot faster for low-level numerics and system stuff.
Why is it worth learning another programming language?
Python is a great language for learning how to program and for learning new concepts. It supports functional programming styles and object-oriented styles, but at the same time, it is also a very practical language.
It's good to see that Linksys is complying with the GPL. It's unfortunate that they are not making their platform more compelling by releasing a complete source tree.
In the end, Linksys APs are just not very good. I have had two, and both of them have had serious bugs. They are now stuck in a closet. Furthermore, their functionality was pretty limited as well.
So, if you want something hackable or powerful, don't bother with the Linksys APs. You are better off with a Mini-ITX board running Linux or BSD and some dumb wireless hardware (USB, wireless bridge, etc.). It's a little more expensive up-front, but much higher quality and much more flexible and functional.
That argument falls flat because that's not how Linksys is selling their gateways. So far, they have had one model in each category, they have provided updates for it, and they have come out with new models only when a new networking standard has come out.
And their development group is barely capable of producing a working wireless router as is. If they released the sources, people could fix their bugs and build new, interesting services with it.
That would mean selling a lot more boxes, instead of what is happening now: my Linksys AP is stuffed into a closet because its software was too flaky.
Where have you been hiding? People patent concepts all the time. No working implementation is required anymore. And a lot of SciFi writing is actually quite a bit more detailed than just a "concept".
I wonder whether science fiction can count as prior art for patent purposes: a lot of science fiction writers seem to specify their ideas in about as much detail as a lot of patents.
Conversely, when are science fiction writers going to start taking out patents prior to publishing their writings?
It doesn't have to. POP3 already has authentication.
Neither SSL nor IPSEC automatically make your POP or IMAP services "secure", e.g. one can still perform buffer overflow attacks over an encrypted channel.
It's naive to think that anything makes software "automatically" secure. Application level encryption like that used in SupraSphere can have security bugs as well.
I'd rather trust the combination of well-debugged POP/IMAP servers combined with standard IPsec implementations than trust some closed source software based on new, untested, application-level protocols.
Re:No Knee-jerk Privacy responses please...
on
Twist on DNA Privacy
·
· Score: 2, Insightful
The wrongful conviction of these men cannot be assigned to anything but the insincerity of the police and prosecuters in seeking out the real perpetrator.
Prosecutors and judges in Britain have little incentive to be "insincere" in order to achieve convictions--they aren't up for elections every few years.
Miscarriages of justice and mistakes can occur anywhere; but overall, I would have much more confidence in the British system than in the US system. Furthermore, if there is a mistake, people don't get executed and they aren't subjected to a virtual death penalty (which imprisonment in many US prisons amounts to), so there is time to correct the mistake.
Re:DNA profiling is an inevitable step
on
Twist on DNA Privacy
·
· Score: 2, Insightful
DNA profiling is probably the single most important anti-crime tool of the future.
DNA evidence is much easier to fake than, say, fingerprints. All you need is a little blood, skin, saliva, or other sample from someone, anyone. As a criminal, you only care about having someone else's DNA show up more prominently than your own.
And if you want to implicate someone in particular, getting the tiniest biological sample (e.g., hair from a comb), you can synthesize as much incriminating DNA for that person as you like, using standard techniques of molecular biology.
Furthermore, the possibility of laboratory errors are much worse with DNA evidence than with other kinds of physical evidence. I do believe that Simpson was guilty, but the objections his defense team raised to the handling of the DNA evidence were valid, and the lab work was shoddy. Handling DNA evidence correctly is much harder than anything that forensic labs are used to.
DNA will probably be useful for "crimes of passion", but for anything that is planned ahead of time, it will provide criminals with lots of opportunities for obfuscation and misdirection. And DNA may well result in many false convictions as juries become too confident in it. Because, when all is said and done, DNA only tells you that there is a bit of biological material from a person at a particular location; it doesn't tell you what actually happened or how it got there.
I was burglared last month, robbed of about $30,000. The thieves left a cigarette stub on the carpet.
Yes, and if DNA evidence were widely used, that "cigarette stub" may well have been something the burglar picked up outside on the street and dropped on your carpet to misdirect the police.
One to allow access to acesspoint and a second second one - dynamically assigned to individual clients(probably recognized by unique mac address) for all data communication between that unique client and accesspoint.
But that means that the access point needs to be able to store one key per client. Furthermore, in order to be reasonably convenient, there needs to be a protocol to do the key exchange without user intervention. Adding this on to 802.11 looks like a major headache. (I think the new Bluetooth standard has provisions for this sort of thing, however.)
However, why couldn't there be a RSA or symmetric encryption for 802.11[x]?
Bluetooth seems to address this: its encryption does not have the weaknesses of 802.11x, and newer versions apparently allow 128bit encrypted open/ad-hoc connections.
I'll take the chance that someone sees my penis-enlargement spam.
The problem is that people also see your POP3 password, which means that they may be removing both your penis-enlargement spam and your real mail from your mailbox after getting your password.
I'm amazed that people still use unencrypted anything over the Internet
What choice do people have? For example, my ISP only offers unencrypted POP3 access, and that ISP is the only ISP that offers broadband access in my area.
If you have some suggestions for third party mail boxes that offer encrypted IMAP4 access, well, please share them.
This only verifies the importance of application level encryption. Every socket communication should be encrypted so that security doesn't rely on the network connection itself.
And one very easy way of encrypting "every socket communication" is via IPsec. And, guess what, you don't need to hack every application to do it. Nor, for that matter, do you need
Suprasphere encrypts all socket communication using a dynamically generated Diffie-Hellman key exchange. This is much better than SSL because it does not require using a CA so you can set it all up without any administrative overhead.
OK, so you are putting in a blatant plug for a commercial product. Why not give a URL? Are you afrad the ridiculous web design at suprasphere.com would scare people away?
As for not requiring a CA, ssh and SSL do not require a CA either. The ability to use a CA just gives you additional functionality that you simply cannot achieve without a CA.
identical twins, clones, other factors
on
Twist on DNA Privacy
·
· Score: 4, Interesting
This reminds us of a problem with DNA testing: it can't distinguish identical twins or clones. Also, inbreeding might increase the odds for a false positive match, and there may be many other real-world factors we don't know about that increase the probability of a false positive beyond what common estimates would lead you to guess.
These problems are compounded by the widespread misapplication of statistical prodecures in the biomedical sciences (most of the FDA drug testing is based on outdated and basically faulty statistical procedures, and it's probably the same in the forensic sciences).
The problem you guys have is "why didn't I think of that back then".
The problem you guys have is that you don't have a clue. I mean, what could you possibly be referring to by "why didn't I think of that back then"?
When Gates came out with MS-DOS, people already had workstations, Smalltalk-80, and multitasking PC operating systems.
And heaven forbid he donates money to charities, research, funds scholarships and hosts parties at his lakeside house thingy.
Monarchs did all those things as well, that doesn't make monarchies a desirable form of government.
And the truth of the matter is he played by the rules and won.
In a free market economy, winning itself is against the rules: without dozens of competitors in the game, markets fail to be efficient. And that's, of course, what we are seeing in the PC software industry.
Yes, and in both cases, the/.-centric summary would be valid.
In any case, the greed of Bill Gates doesn't start with dividend payouts, it starts with wanting to own the entire computer industry and crushing every competitor to dust through unfair competition and sleazy tricks.
In the way of background, note that Gold is the same Gold of the Bondi-Holye-Gold steady state cosmological model, proposed in the 1940s and 1950s as an attempt to "fix problems" with the big bang model, and has long held non-conventional views on light. Gold and others invoked "tired light" -- photons which became redder from their point of emission, even though doing so contradicted momentum-energy conservation.
It's perfectly fine for a physics outsider not to bother understanding, say, standard thermodynamics, but if he wants to use standard thermodynamics to draw conclusions, he has to understand it and use it correctly.
Frankly, standard thermodynamics is simple enough that anybody who has been in the business since the 1940's should have an understanding of it. It just makes me shudder to think what kind of nonsense this guy teaches in his physics classes.
You could use the same argument to show that you couldn't play billiards, which is obviously nonsense
The fallacy in his argument is that Carnot's rule applies to thermal energy. Thermal energy is microscopic motion with an overall total zero momentum for any macroscopic chunk you look at. Carnot's rule tells us nothing about what you can do with macroscopic momentum: it doesn't tell us about billiard balls, for example. But the total momentum of the light hitting a solar sail does not have a total zero momentum. The solar sail manages to carve out a macroscopic chunk of all the radiation leaving the sun whose total momentum is non-zero. Since it has a non-zero total macroscopic momentum, considerations of heat engines don't come into play, and it can be used for propelling a solar sail just like you can propel one billiard ball by hitting it with another one.
This is the kind of patent all big computer companies file a lot of: trivial technology and lots of related prior art, but nobody else has patented exactly this thing. Why is Microsoft doing this? To be able to achieve cross-licensing with other companies that have big patent portfolios. The effect is to keep small competitors from being able to enter any of their markets (because they will be stepping on some of Microsoft's patents), and to be able to have leverage against open source projects.
I think this is ultimately only fixable legislatively. It's important that the EU do not pass software patents--by having at least one large market where open source software can be developed without this nonsense, people will keep creating software even for functionality that's patented in the US. But in the long run, we really need to get patent reform in the US.
The effects of these kinds of patents are so hostile to business and competition that sooner or later, legislators must see the light.
Slashdot Mirror
Seriously though, there are no new technology on the horizon that would make silicon run cooler, and the speed of core-voltage drop does not keep up with frequency bumps (heat is square of frequency for CMOS gates).
Sure there are: asynchronous logic, reconfigurable computing, reversible computation, and many other technologies. Of course, that requires significant changes to the way we design software and hardware...
The interconnects in most Origins and Altix systems are 3.2 gigaBYTE per second with extremely low latency.
I think SGI is playing with the numbers there: while that is the limit for how much data you can push through a link, it seems unlikely that you get that kind of performance for arbitrary processor-to-processor communications or as aggregate bandwidth.
In any case, very fast interconnects are not cost-efficient when they are not needed. 3.2GB is the memory bandwidth of a 400MHz bus motherboard. Most numerical applications do not spend all their time moving stuff around between different processors, and it just isn't worth paying a premium in order to get that kind of bandwidth if you don't need it.
Also, the Altix can handle up to 64 processors per single machine / single node (or 128 with a very beta set of patches). The cluster in the article is actually four Altix systems, each with 64 processors. The Origin 3800/3900 can handle 512 processors per node (or 1024 with a special "XXL" IRIX kernel).
Again, you pay a premium for a product that just doesn't make a lot of sense (at least to me): an abstraction of parallel computation in which 64 or 1024 processors all read and write a gigantic shared memory just isn't a good way of thinking about parallel computations, even if you could build it cheaply. And because it's NUMA, it only looks like a shared memory, but it really isn't anyway.
I think the SGIs have a niche: you can take dusty FORTRAN and multi-threaded C applications and scale them up to many processors with a minimum of fuss. That's very valuable, even if you pay a premium for the convenience. But I sure hope that the future of parallel programming won't look like that.
I think violence in movies or games probably can occasionally cause people to become murderers. And, yes, by that I mean that if those people had not played violent video games, they would not have become killers.
But just because A occasionally causes B doesn't mean that A by itself is sufficient to cause B, or that reducing A is the best way of reducing B. If you take happy, healthy kids and have them play violent video games or show them standard Hollywood movies dripping with violence, they won't turn into killers--something else needs to go wrong. Conversely, US courts have established that Twinkies sometimes cause people to become killers--should we ban Twinkies as well?
The fact that we have had very high levels of violence in the US (compared to other nations) far longer than video games have been around shows us that video games are not a major cause of violence in this culture. Let's find out what really turns so many Americans into killers and fix it; I suspect it's more related to poverty, education, and deep-rooted social problems than the depiction of violence in video games or movies.
It seems unlikely that the SGI is the first Linux cluster with global shared memory. There are plenty of distributed shared memory systems in software, some of them open source. You can find a list here. For most computations and most hardware, you are probably still better off with MPI or PVM rather than shared memory.
Note also that there are several high speed interconnects for Linux clusters available from many different vendors, including InfiniBand, Gigabit Ethernet, FireWire, and Myrinet.
Just tune into traffic info radio or pick up your cell phone and call one of the many free traffic info services. Or, have the information sent to your cell phone. But most people realize pretty quickly that that kind of knowledge is pretty useless: even if 101 or 405 are stop-and-go, taking alternate routes probably still takes longer than just living with it.
The only thing you can do is to stay a little longer at the office until traffic has died down. And to see when that has happened, you don't need a wireless gadget, you just point your desktop web browser at a traffic site.
What do Slashdotters use python for?
For the same things I would use Perl, Java, or Matlab for: CGI scripts, system administration, text processing, numerical processing, GUI application development, and many other applications.
What are its strengths and its weaknesses?
Strengths: easy to learn, easy to read, much better error checking than Perl, much more concise than Java, lots of libraries, lots of GUI toolkits (Gtk+, wxWindows, Qt, Tk, FLTK, others).
Weaknesses: library modules are somewhat haphazard, no equivalent to CPAN, C implementation is rather ugly under the covers, C language interface is not very good (but still better than Perl). For short scripts, it's not as concise as Perl. While Numerical Python is "fast" in the same way Matlab is, it's still an interpreted language, and Java or C++ are a lot faster for low-level numerics and system stuff.
Why is it worth learning another programming language?
Python is a great language for learning how to program and for learning new concepts. It supports functional programming styles and object-oriented styles, but at the same time, it is also a very practical language.
It's good to see that Linksys is complying with the GPL. It's unfortunate that they are not making their platform more compelling by releasing a complete source tree.
In the end, Linksys APs are just not very good. I have had two, and both of them have had serious bugs. They are now stuck in a closet. Furthermore, their functionality was pretty limited as well.
So, if you want something hackable or powerful, don't bother with the Linksys APs. You are better off with a Mini-ITX board running Linux or BSD and some dumb wireless hardware (USB, wireless bridge, etc.). It's a little more expensive up-front, but much higher quality and much more flexible and functional.
That argument falls flat because that's not how Linksys is selling their gateways. So far, they have had one model in each category, they have provided updates for it, and they have come out with new models only when a new networking standard has come out.
And their development group is barely capable of producing a working wireless router as is. If they released the sources, people could fix their bugs and build new, interesting services with it.
That would mean selling a lot more boxes, instead of what is happening now: my Linksys AP is stuffed into a closet because its software was too flaky.
Where have you been hiding? People patent concepts all the time. No working implementation is required anymore. And a lot of SciFi writing is actually quite a bit more detailed than just a "concept".
"GTK/Gnome finally catches up by implementing usless feature copied from OSX"
OSX did not pioneer the use of shadows in user interfaces; it is actually decades old.
I wonder whether science fiction can count as prior art for patent purposes: a lot of science fiction writers seem to specify their ideas in about as much detail as a lot of patents.
Conversely, when are science fiction writers going to start taking out patents prior to publishing their writings?
IPSEC cannot authenticate users to a service
It doesn't have to. POP3 already has authentication.
Neither SSL nor IPSEC automatically make your POP or IMAP services "secure", e.g. one can still perform buffer overflow attacks over an encrypted channel.
It's naive to think that anything makes software "automatically" secure. Application level encryption like that used in SupraSphere can have security bugs as well.
I'd rather trust the combination of well-debugged POP/IMAP servers combined with standard IPsec implementations than trust some closed source software based on new, untested, application-level protocols.
The wrongful conviction of these men cannot be assigned to anything but the insincerity of the police and prosecuters in seeking out the real perpetrator.
Prosecutors and judges in Britain have little incentive to be "insincere" in order to achieve convictions--they aren't up for elections every few years.
Miscarriages of justice and mistakes can occur anywhere; but overall, I would have much more confidence in the British system than in the US system. Furthermore, if there is a mistake, people don't get executed and they aren't subjected to a virtual death penalty (which imprisonment in many US prisons amounts to), so there is time to correct the mistake.
DNA profiling is probably the single most important anti-crime tool of the future.
DNA evidence is much easier to fake than, say, fingerprints. All you need is a little blood, skin, saliva, or other sample from someone, anyone. As a criminal, you only care about having someone else's DNA show up more prominently than your own.
And if you want to implicate someone in particular, getting the tiniest biological sample (e.g., hair from a comb), you can synthesize as much incriminating DNA for that person as you like, using standard techniques of molecular biology.
Furthermore, the possibility of laboratory errors are much worse with DNA evidence than with other kinds of physical evidence. I do believe that Simpson was guilty, but the objections his defense team raised to the handling of the DNA evidence were valid, and the lab work was shoddy. Handling DNA evidence correctly is much harder than anything that forensic labs are used to.
DNA will probably be useful for "crimes of passion", but for anything that is planned ahead of time, it will provide criminals with lots of opportunities for obfuscation and misdirection. And DNA may well result in many false convictions as juries become too confident in it. Because, when all is said and done, DNA only tells you that there is a bit of biological material from a person at a particular location; it doesn't tell you what actually happened or how it got there.
I was burglared last month, robbed of about $30,000. The thieves left a cigarette stub on the carpet.
Yes, and if DNA evidence were widely used, that "cigarette stub" may well have been something the burglar picked up outside on the street and dropped on your carpet to misdirect the police.
One to allow access to acesspoint and a second second one - dynamically assigned to individual clients(probably recognized by unique mac address) for all data communication between that unique client and accesspoint.
But that means that the access point needs to be able to store one key per client. Furthermore, in order to be reasonably convenient, there needs to be a protocol to do the key exchange without user intervention. Adding this on to 802.11 looks like a major headache. (I think the new Bluetooth standard has provisions for this sort of thing, however.)
However, why couldn't there be a RSA or symmetric encryption for 802.11[x]?
Bluetooth seems to address this: its encryption does not have the weaknesses of 802.11x, and newer versions apparently allow 128bit encrypted open/ad-hoc connections.
I'll take the chance that someone sees my penis-enlargement spam.
The problem is that people also see your POP3 password, which means that they may be removing both your penis-enlargement spam and your real mail from your mailbox after getting your password.
I'm amazed that people still use unencrypted anything over the Internet
What choice do people have? For example, my ISP only offers unencrypted POP3 access, and that ISP is the only ISP that offers broadband access in my area.
If you have some suggestions for third party mail boxes that offer encrypted IMAP4 access, well, please share them.
This only verifies the importance of application level encryption. Every socket communication should be encrypted so that security doesn't rely on the network connection itself.
And one very easy way of encrypting "every socket communication" is via IPsec. And, guess what, you don't need to hack every application to do it. Nor, for that matter, do you need
Suprasphere encrypts all socket communication using a dynamically generated Diffie-Hellman key exchange. This is much better than SSL because it does not require using a CA so you can set it all up without any administrative overhead.
OK, so you are putting in a blatant plug for a commercial product. Why not give a URL? Are you afrad the ridiculous web design at suprasphere.com would scare people away?
As for not requiring a CA, ssh and SSL do not require a CA either. The ability to use a CA just gives you additional functionality that you simply cannot achieve without a CA.
This reminds us of a problem with DNA testing: it can't distinguish identical twins or clones. Also, inbreeding might increase the odds for a false positive match, and there may be many other real-world factors we don't know about that increase the probability of a false positive beyond what common estimates would lead you to guess.
These problems are compounded by the widespread misapplication of statistical prodecures in the biomedical sciences (most of the FDA drug testing is based on outdated and basically faulty statistical procedures, and it's probably the same in the forensic sciences).
And, as a fallback plan, they make products people want.
So did Standard Oil and Ma Bell, and we still broke them up because they were ultimately harmful to the economy and not in the interest of consumers.
And so do cigarette makers and drug dealers, for that matter.
The problem you guys have is "why didn't I think of that back then".
The problem you guys have is that you don't have a clue. I mean, what could you possibly be referring to by "why didn't I think of that back then"?
When Gates came out with MS-DOS, people already had workstations, Smalltalk-80, and multitasking PC operating systems.
And heaven forbid he donates money to charities, research, funds scholarships and hosts parties at his lakeside house thingy.
Monarchs did all those things as well, that doesn't make monarchies a desirable form of government.
And the truth of the matter is he played by the rules and won.
In a free market economy, winning itself is against the rules: without dozens of competitors in the game, markets fail to be efficient. And that's, of course, what we are seeing in the PC software industry.
Yes, and in both cases, the /.-centric summary would be valid.
In any case, the greed of Bill Gates doesn't start with dividend payouts, it starts with wanting to own the entire computer industry and crushing every competitor to dust through unfair competition and sleazy tricks.
In the way of background, note that Gold is the same Gold of the Bondi-Holye-Gold steady state cosmological model, proposed in the 1940s and 1950s as an attempt to "fix problems" with the big bang model, and has long held non-conventional views on light. Gold and others invoked "tired light" -- photons which became redder from their point of emission, even though doing so contradicted momentum-energy conservation.
It's perfectly fine for a physics outsider not to bother understanding, say, standard thermodynamics, but if he wants to use standard thermodynamics to draw conclusions, he has to understand it and use it correctly.
Frankly, standard thermodynamics is simple enough that anybody who has been in the business since the 1940's should have an understanding of it. It just makes me shudder to think what kind of nonsense this guy teaches in his physics classes.
You could use the same argument to show that you couldn't play billiards, which is obviously nonsense
The fallacy in his argument is that Carnot's rule applies to thermal energy. Thermal energy is microscopic motion with an overall total zero momentum for any macroscopic chunk you look at. Carnot's rule tells us nothing about what you can do with macroscopic momentum: it doesn't tell us about billiard balls, for example. But the total momentum of the light hitting a solar sail does not have a total zero momentum. The solar sail manages to carve out a macroscopic chunk of all the radiation leaving the sun whose total momentum is non-zero. Since it has a non-zero total macroscopic momentum, considerations of heat engines don't come into play, and it can be used for propelling a solar sail just like you can propel one billiard ball by hitting it with another one.
This is the kind of patent all big computer companies file a lot of: trivial technology and lots of related prior art, but nobody else has patented exactly this thing. Why is Microsoft doing this? To be able to achieve cross-licensing with other companies that have big patent portfolios. The effect is to keep small competitors from being able to enter any of their markets (because they will be stepping on some of Microsoft's patents), and to be able to have leverage against open source projects.
I think this is ultimately only fixable legislatively. It's important that the EU do not pass software patents--by having at least one large market where open source software can be developed without this nonsense, people will keep creating software even for functionality that's patented in the US. But in the long run, we really need to get patent reform in the US.
The effects of these kinds of patents are so hostile to business and competition that sooner or later, legislators must see the light.