In civil suits, the plaintiff's attorneys normally take 33% of the winnings. In this case, they're only asking 25%. Why do judges allow this? It's because of the substantial risk involved in taking a case like this. If the plaintiff loses, they get nothing. When representing a defendant, they are paid by the hour, and make damn sure they get that retainer up front. The risk is quite substantial, because a lawyer who works for 2 years on a big case and gets nothing is going to have a hard time eating. Allowing large contingency fees increases the likelyhood that these cases ever see the inside of a courtroom. Since contingency fees tend to be used when the little guy is suing the big guy, this tends to help the people who need it most and hurt the people who feel it least. Whether $3,000/hour is appropriate to further the interest of justice is left to the discretion of the judge handling the case, but it isn't an inherently outrageous fee under the circumstances.
Now, if you're wondering why lawyers often charge on the order of $200/hour with a straight face, it's because they have to pay their secretaries, paralegals, bookkeepers, phone bill, LEXIS-NEXIS subscription, malpractice insurance, rent, and, of course, Windows licensing fees. My parents are both attorneys with excellent professional reputations, and fairly thrifty people, but I still have college loans, having already spent many thousands of dollars on tuition out of my own savings. The savings didn't come from gifts or anything like that, they came from working since I was 15. It would be far worse if I had gone to college out of state, but we simply couldn't afford that at all. I don't blame my parents for any of this, because it's not like they've been neglecting me. They're doing the best they can. There's a fairly decent chance that at age 21 I'll have a higher income as a software developer than they do as (very good) attorneys.
There are certainly lawyers who become quite wealthy from their profession, but most of them end up somewhere in the middle class. If you can think of a way to streamline the legal system to significantly reduce those costs, your lawyer will surely pass the savings on to you. Unlike Microsoft, your lawyer has to compete.
Of course, if you're getting a Mac, the best time is always right after they announce an upgrade. If it's been a while since an upgrade, you run the significant risk of your order being obsolete by the time the package arrives. Personally, I hold off on buying computers until I hit the point where I realize that the inconvenience of dealing with my current system's inadequacy is measurably impacting the amount of time I could be spending hanging out with hot women.
The great thing about wavelets is how they work at arbitrary resolution without much of a performance hit. Edges look like edges. Since you can basically make a general description of an image and just keep adding more detailed wavelets until you've got the compression/quality ratio you're looking for, and you can define quality however you'd like. One of the ideas for JPEG2000 is to have a field in image tags to specify how much of the image a browser should download, so you'd only have to keep one copy on the server. (By the way, where the hell is JPEG2000?)
The above just takes advantage of spatial similarity (if a pixel is one color, it's neighbors are probably similar), but you can also take advantage of temporal similarity (if a pixel is one color in this frame, it's probably a similar color in the next one). You can also do motion compression, though when you get to that level of optimization you generally lose the symmetry between sender and receiver resource consumption. Of course, that might just be another CS dissertation away.
I was on the U.Va. team a few years ago. I distinctly remember you guys helping us out in some way or other, but I can't exactly remember what (but thanks!) I haven't really kept up with it lately, but congratulations on your win.
Anyway, if you're in the stock class, I'd imagine you're using 12% cells, at the max. Just imagine if this kind of thing gets cheap. They'll be running the 2011 ASC at the Indianapolis Motor Speedway.
That's because many codecs have performance tuning parameters built into the encoding standard, like with MPEG. Wavelet-based methods don't need to do this, so their performance tends to scale quite smoothly. More traditional compression techniques may still beat them out at their "sweet spots", but the wavelet methods are very general.
It would be REALLY nice if that were true, but it's not. The US patent system has been going downhill for a while, and while it may be intended to promote research, it's definitely not doing that nearly as well as it used to.
The product of this reaction is Helium-4. No neutrons necessary. While you might expect some neutrons to escape in vacuum reactions and create some other byproducts, it's entirely possible that it works some other way in a metal lattice.
For portability reasons, implementing vector with a non-integer multiplier is a bad idea, and for efficiency reasons, using a multiplier larger than 3 is a bad idea. This doesn't stop you from tweaking things yourself, but this is once again something that you're better off letting a VM do for you, since it will almost always guess at least as well as you at little expense, and will save you lots of time and debugging. Some of the tricks you described sound almost like reimplementing the VM inside your application, in which case you're better off using one that's already written and heavily optimized.
As for the containers and adapters, deque requires amortized constant time push_front and push_back as well as random access. While you could reimplement page tables in your deque class, that's certainly an ugly way of doing it. The standard is written loosely enough to describe the scheme you describe, but it's typically implemented with something resembling a vector under the hood. List may have been intended to be implemented as a traditional C-style linked list, and again the standard leaves a lot of freedom to implementors, but it's typically implemented with array-like memory under the hood. The standard only requires that inserts take amortized constant time, and random access is not allowed, so there are a lot of tricks you can do with this. By doing it with arrays/vectors, you reduce memory needs (at least for doubly-linked lists) and most importantly, preserve spatial locality that keeps your cache hit ratio extremely high. I'm sure there are exceptions, but the vector-like list implementation is common and fast. Stack and queue generally use whatever deque uses. Sure, you can tweak this yourself, but that doesn't make it a good idea.
That's the point of the end of the sentence you quoted "unless you know enough to tell it to do otherwise". Usually you don't, and most of the time when you do, the performance gain is meager compared to a smart compiler, VM, and JIT.
I completely agree with you about C++.NET. I was contrasting VM languages with non-VM languages, and I still think of C++ as a non-VM language. I think of the VM as providing a huge win for 99% of all application development.
Remember the Unix philosophy, "Worse is Better"? The idea was that by writing lots of small things that do one thing each and do them well, the performance hit that comes from linking them all together is less than the speedup from hardware advance in the time it takes for the monolithic developer to get everything right. As a pleasant side effect, the bugs are a hell of a lot easier to fix, and if some component is horribly broken and the maintainer won't fix it, you can rip it out and replace it with something that has similar functionality.
The reason you need a 3GHz P4 with 2GB of RAM is that so many developers have taken to linking together things that are much more like entire monolithic applications just to integrate very simple functionality.
The introduction of modern programming languages like C# and Java allows us to effectively turn back the clock 30 years on code bloat, since the important stuff is now part of the language and can be handled intelligently by the compiler without everyone writing their own differently broken implementation. If you read his code, you'll note that Rijndael is a part of the standard libraries. Because of that, he managed to write the whole thing in 210 lines of whitespace-heavy code, and the bytecode compiler will link in the bare minimum to run the program, and the JIT will cache of that only what is actually running, incurring a significant performance hit if and only if something exceptional happens, in which case performance is probably not a serious concern anyway.
That said, it is most certainly possible to write bloated C# code, but in my experience, you always win when you let the language/compiler do something for you, instead of writing your own or using external libraries.
Remember that research that found that emulating the underlying hardware with a sufficiently intelligent userland dynamic profiler was usually faster than running directly on the underlying hardware? The dynamic profiler can optimize like no compiler will ever be able to do with static analysis. It's a similar principle to what Transmeta does with their x86 emulation. Modern Just-In-Time Compilers use dynamic profiling to accelerate things, and they're getting quite good. It's certainly quite possible to design a C# vector class that's both more memory and processor efficient in most cases than C++. Here's how:
1) Record in the virtual machine/JIT every time a vector gets resized. 2) Based on the pattern of resizing, speculatively allocate for new vectors/resizes as much memory as they'll ever need, or at least as much as they'll need any time soon. 3) When you guess wrong about a speculative allocation, adjust your speculation.
C++ doubles the amount of space allocated for a vector (or queue, or list, or stack, or dequeue, or binary heap, etc) whenever a resize exceeds the amount already allocated, unless you know enough to tell it to do otherwise. This keeps the amortized cost of increasing size by one constant. C++ doesn't benefit from profiling like C# does because there's no virtual machine that can change what binary code is actually sent to the processor. You could hack vector profiling together yourself, but it would be slow. Of course, this doesn't really help C# if you're never resizing your vectors, but that doesn't mean C# can't do better than C, even if C++ will have it beat. If you've ever done much benchmarking of the C++ STL, you know that it's usually faster than otherwise identical code written with arrays, which shouldn't be possible, since the array access code can be done fairly easily in assembly without virtual function table lookups and such, but nonetheless is quite real.
The trick to this whole scheme is doing the speculation quickly and accurately. We may not be to the point yet where JIT code reliably outperforms statically compiled code in less space, but there are an army of extraordinarily intelligent grad students out there writing dissertations on the topic, and I assure you they'll make it happen.
We should put these guys to work doing human spam filtering. Sure, it sounds like a recipe for disaster, but just tie their pay to their performance. Let through an ad for grey-market software at huge discounts? Looks like you're gonna have to get friendly with Bubba if you want your cigarettes this week. Of course you also spot-check their rejects. Drop the email from the ex asking if you wanna get drinks some time? TWO WEEKS IN SOLITARY!
In all seriousness, 5 years ago I would have said that multi-year prison sentences for spamming would be extreme, at least in cases without other crimes involved. On its face, it's still extreme, but these guys now hold an entire communication system hostage. If sending several of them to prison for their transgressions (which ARE transgressions) can be a deterrent, then I'm for it. I think it really will be a deterrent if we can get some convictions. It's not like people spam in a brief burst of anger. These people generally have some business or technical skills that could find them legitimate employment (perhaps somewhat less lucrative, but above the poverty line) even in the lousy tech economy. I hear the porn industry does well when the economy is lousy. I'm sure my mom would much rather I manage a (legitimate) porn server farm than a spam server farm anyway.
In theory, you're right. Sadly, Congress and the Courts have been of the mindset that if a little property is good, then more must be better. These days you can't sneeze without tortiously infringing on a copyright, because you might have seen a movie with some particularly distinctive sneeze. In fact, even if you didn't see the movie, if it played in your town, you had access to it, and you have the burden of proving you didn't see it.
That's funny, because Sprint's residential DSL, which partners with Earthlink under an arrangement whose terms are not known to me...
wait for it...
HAS SPYWARE BUILT INTO THE SETUP SOFTWARE!
Or at least it did when my parents subscribed. Nothing that a disk crash and a reinstall with RASPPPOE couldn't solve, but wow.
I guess that's not quite as bad as their new Compaq desktop that came with spyware PREINSTALLED.
There's a general philosophy amount consumer software and hardware distributors that people don't want to know what goes on underneath, and give their tacit permission for them to put whatever the hell they want on there. We keep blaming code bloat for making computers run slower than they used to, but maybe the fact that people have a couple dozen completely unnecessary processes running, each using just a few megs of RAM that nobody would notice missing by itself, has something to do with it.
Granted, I could probably cut down on a few things running on my fairly stock Debian/KDE workstation, but they use about 1/100 the resources as the useless crap on my parents' machine. Of course, none of the things on my machine that I'm not entirely sure I need are designed to scan web pages I read for key words and deliver pop-up ads for competitors.
If the Nullsoft release was unauthorized (what constitutes unauthorized is not as clear-cut as AOL would have us believe) then the fact that the code was GPL'd is irrelevant. Go roll your own people. Don't even look at the WASTE source. You'll be tainted.
This is why precedents don't extend beyond the jurisdiction of the court that sets them, and why appeals courts accept amicus curiae briefs from other parties, which may actually be more influential than the arguments of the attorneys of the parties directly involved in the suit.
Generally speaking, if an employee of a corporation does something on behalf of a corporation in a manner consistent with an official act of the corporation (like, posting it on their main website), regardless of whether or not they were supposed to, the corporation's recourse is fire the employee. The exceptions to this would be if the employee knew that the release was unauthorized and did it anyway, making them no longer an agent of the company. As long as they did it in good faith, they were acting as an agent of the company, however poor their judgement may have been.
Of course, there are other exceptions to this as well, more obscure and probably even more open to subjective interpretation in court. If you wanna use something like this, consult a lawyer, or roll your own.
This kind of thing would never be an issue with postal mail, because a tangible copy is made. By convention this is assumed to be true of email. Instant messaging blurs the lines. Thanks to the wonders of internet technology, we have modes of communication inherently unlike anything contemplated by legislative bodies at the time of the writing of these laws.
What do appeals courts generally do to convictions based on laws that weren't written with the circumstances of the alleged crime in mind? They generally throw them out. Lets hope this holds, and also badger our legislators to allow recording of communications by those known by the speaker to be receiving them. Two-party consent lets the powerful screw over the little people and not be held accountable for it. It's arguably a first amendment violation, though apparently nobody has argued that lately, or at least not convincingly.
The unix world has the advantage of starting with a lot of software designed to be used in a multi-user environment by non-administrative users. The Windows world is rife with the artifacts of DOS/Win16/Win9x/WinME software development which makes no such assumptions, with the result that it is considered normal by both users and developers to have and require administrative access. While this is slowly changing, it is a problem that unix-based environments won't have to face.
Of course, people will come out with things like Lindows and face them anyway, but it should be pretty easy for them to adopt the standard linux security model if they start having significant problems.
They've known for a long time that LEAP is inherently flawed, and no patch can fix it. That said, it's a hell of a lot simpler to deploy than more secure things like EAP-TLS. This attack still requires an offline brute force decryption attempt. Granted, it may be a highly accelerated brute force decryption attempt, but if you don't allow your users to use passwords that are vulnerable to dictionary attacks, LEAP is Good Enough for many purposes.
The Deep Space Nine episode you're describing sounds like "Favor the Bold" or "Sacrifice of Angels". Wow, I never thought of myself as a Trek geek until now.
I read this as meaning that they won't go messing with their backups to purge your data, which is perfectly reasonable, especially for a free service. Backups are supposed to be write-once recordings. If you've got a problem with that, don't transmit sensitive data over a free webmail service. Nothing spooky here. In fact, I imagine every webmail service works this way. Gmail is just kind enough to warn you about this.
Yes, it really is about an order of magnitude better. 0.95/0.38 = 2.5. Now, keeping in mind that we're dealing with a 2-D surface, 2.5^2 = 6.25. Log(6.25) = 0.796, so this is about 4/5 of an order of magnitude improvement.
It's my understanding that a lot of the organization for the coup that took down the Soviet Union was done by email. Of course, at the time, the internet was under the radar. More recently, massive demonstrations that brought down a government were organized by text messaging. The lesson here is that communication technology may indeed be a threat in places that actually allow the people to use it. The Chinese government is right to be afraid, because internet usage is quite widespread there. North Korea probably has nothing to fear.
Slashdot Mirror
In civil suits, the plaintiff's attorneys normally take 33% of the winnings. In this case, they're only asking 25%. Why do judges allow this? It's because of the substantial risk involved in taking a case like this. If the plaintiff loses, they get nothing. When representing a defendant, they are paid by the hour, and make damn sure they get that retainer up front. The risk is quite substantial, because a lawyer who works for 2 years on a big case and gets nothing is going to have a hard time eating. Allowing large contingency fees increases the likelyhood that these cases ever see the inside of a courtroom. Since contingency fees tend to be used when the little guy is suing the big guy, this tends to help the people who need it most and hurt the people who feel it least. Whether $3,000/hour is appropriate to further the interest of justice is left to the discretion of the judge handling the case, but it isn't an inherently outrageous fee under the circumstances.
Now, if you're wondering why lawyers often charge on the order of $200/hour with a straight face, it's because they have to pay their secretaries, paralegals, bookkeepers, phone bill, LEXIS-NEXIS subscription, malpractice insurance, rent, and, of course, Windows licensing fees. My parents are both attorneys with excellent professional reputations, and fairly thrifty people, but I still have college loans, having already spent many thousands of dollars on tuition out of my own savings. The savings didn't come from gifts or anything like that, they came from working since I was 15. It would be far worse if I had gone to college out of state, but we simply couldn't afford that at all. I don't blame my parents for any of this, because it's not like they've been neglecting me. They're doing the best they can. There's a fairly decent chance that at age 21 I'll have a higher income as a software developer than they do as (very good) attorneys.
There are certainly lawyers who become quite wealthy from their profession, but most of them end up somewhere in the middle class. If you can think of a way to streamline the legal system to significantly reduce those costs, your lawyer will surely pass the savings on to you. Unlike Microsoft, your lawyer has to compete.
Of course, if you're getting a Mac, the best time is always right after they announce an upgrade. If it's been a while since an upgrade, you run the significant risk of your order being obsolete by the time the package arrives. Personally, I hold off on buying computers until I hit the point where I realize that the inconvenience of dealing with my current system's inadequacy is measurably impacting the amount of time I could be spending hanging out with hot women.
The great thing about wavelets is how they work at arbitrary resolution without much of a performance hit. Edges look like edges. Since you can basically make a general description of an image and just keep adding more detailed wavelets until you've got the compression/quality ratio you're looking for, and you can define quality however you'd like. One of the ideas for JPEG2000 is to have a field in image tags to specify how much of the image a browser should download, so you'd only have to keep one copy on the server. (By the way, where the hell is JPEG2000?)
The above just takes advantage of spatial similarity (if a pixel is one color, it's neighbors are probably similar), but you can also take advantage of temporal similarity (if a pixel is one color in this frame, it's probably a similar color in the next one). You can also do motion compression, though when you get to that level of optimization you generally lose the symmetry between sender and receiver resource consumption. Of course, that might just be another CS dissertation away.
I was on the U.Va. team a few years ago. I distinctly remember you guys helping us out in some way or other, but I can't exactly remember what (but thanks!) I haven't really kept up with it lately, but congratulations on your win.
Anyway, if you're in the stock class, I'd imagine you're using 12% cells, at the max. Just imagine if this kind of thing gets cheap. They'll be running the 2011 ASC at the Indianapolis Motor Speedway.
That's because many codecs have performance tuning parameters built into the encoding standard, like with MPEG. Wavelet-based methods don't need to do this, so their performance tends to scale quite smoothly. More traditional compression techniques may still beat them out at their "sweet spots", but the wavelet methods are very general.
It would be REALLY nice if that were true, but it's not. The US patent system has been going downhill for a while, and while it may be intended to promote research, it's definitely not doing that nearly as well as it used to.
Britain. India. Pakistan. Kashmir.
The product of this reaction is Helium-4. No neutrons necessary. While you might expect some neutrons to escape in vacuum reactions and create some other byproducts, it's entirely possible that it works some other way in a metal lattice.
For portability reasons, implementing vector with a non-integer multiplier is a bad idea, and for efficiency reasons, using a multiplier larger than 3 is a bad idea. This doesn't stop you from tweaking things yourself, but this is once again something that you're better off letting a VM do for you, since it will almost always guess at least as well as you at little expense, and will save you lots of time and debugging. Some of the tricks you described sound almost like reimplementing the VM inside your application, in which case you're better off using one that's already written and heavily optimized.
As for the containers and adapters, deque requires amortized constant time push_front and push_back as well as random access. While you could reimplement page tables in your deque class, that's certainly an ugly way of doing it. The standard is written loosely enough to describe the scheme you describe, but it's typically implemented with something resembling a vector under the hood. List may have been intended to be implemented as a traditional C-style linked list, and again the standard leaves a lot of freedom to implementors, but it's typically implemented with array-like memory under the hood. The standard only requires that inserts take amortized constant time, and random access is not allowed, so there are a lot of tricks you can do with this. By doing it with arrays/vectors, you reduce memory needs (at least for doubly-linked lists) and most importantly, preserve spatial locality that keeps your cache hit ratio extremely high. I'm sure there are exceptions, but the vector-like list implementation is common and fast. Stack and queue generally use whatever deque uses. Sure, you can tweak this yourself, but that doesn't make it a good idea.
That's the point of the end of the sentence you quoted "unless you know enough to tell it to do otherwise". Usually you don't, and most of the time when you do, the performance gain is meager compared to a smart compiler, VM, and JIT.
I completely agree with you about C++.NET. I was contrasting VM languages with non-VM languages, and I still think of C++ as a non-VM language. I think of the VM as providing a huge win for 99% of all application development.
Remember the Unix philosophy, "Worse is Better"? The idea was that by writing lots of small things that do one thing each and do them well, the performance hit that comes from linking them all together is less than the speedup from hardware advance in the time it takes for the monolithic developer to get everything right. As a pleasant side effect, the bugs are a hell of a lot easier to fix, and if some component is horribly broken and the maintainer won't fix it, you can rip it out and replace it with something that has similar functionality.
The reason you need a 3GHz P4 with 2GB of RAM is that so many developers have taken to linking together things that are much more like entire monolithic applications just to integrate very simple functionality.
The introduction of modern programming languages like C# and Java allows us to effectively turn back the clock 30 years on code bloat, since the important stuff is now part of the language and can be handled intelligently by the compiler without everyone writing their own differently broken implementation. If you read his code, you'll note that Rijndael is a part of the standard libraries. Because of that, he managed to write the whole thing in 210 lines of whitespace-heavy code, and the bytecode compiler will link in the bare minimum to run the program, and the JIT will cache of that only what is actually running, incurring a significant performance hit if and only if something exceptional happens, in which case performance is probably not a serious concern anyway.
That said, it is most certainly possible to write bloated C# code, but in my experience, you always win when you let the language/compiler do something for you, instead of writing your own or using external libraries.
Remember that research that found that emulating the underlying hardware with a sufficiently intelligent userland dynamic profiler was usually faster than running directly on the underlying hardware? The dynamic profiler can optimize like no compiler will ever be able to do with static analysis. It's a similar principle to what Transmeta does with their x86 emulation. Modern Just-In-Time Compilers use dynamic profiling to accelerate things, and they're getting quite good. It's certainly quite possible to design a C# vector class that's both more memory and processor efficient in most cases than C++. Here's how:
1) Record in the virtual machine/JIT every time a vector gets resized.
2) Based on the pattern of resizing, speculatively allocate for new vectors/resizes as much memory as they'll ever need, or at least as much as they'll need any time soon.
3) When you guess wrong about a speculative allocation, adjust your speculation.
C++ doubles the amount of space allocated for a vector (or queue, or list, or stack, or dequeue, or binary heap, etc) whenever a resize exceeds the amount already allocated, unless you know enough to tell it to do otherwise. This keeps the amortized cost of increasing size by one constant. C++ doesn't benefit from profiling like C# does because there's no virtual machine that can change what binary code is actually sent to the processor. You could hack vector profiling together yourself, but it would be slow. Of course, this doesn't really help C# if you're never resizing your vectors, but that doesn't mean C# can't do better than C, even if C++ will have it beat. If you've ever done much benchmarking of the C++ STL, you know that it's usually faster than otherwise identical code written with arrays, which shouldn't be possible, since the array access code can be done fairly easily in assembly without virtual function table lookups and such, but nonetheless is quite real.
The trick to this whole scheme is doing the speculation quickly and accurately. We may not be to the point yet where JIT code reliably outperforms statically compiled code in less space, but there are an army of extraordinarily intelligent grad students out there writing dissertations on the topic, and I assure you they'll make it happen.
We should put these guys to work doing human spam filtering. Sure, it sounds like a recipe for disaster, but just tie their pay to their performance. Let through an ad for grey-market software at huge discounts? Looks like you're gonna have to get friendly with Bubba if you want your cigarettes this week. Of course you also spot-check their rejects. Drop the email from the ex asking if you wanna get drinks some time? TWO WEEKS IN SOLITARY!
In all seriousness, 5 years ago I would have said that multi-year prison sentences for spamming would be extreme, at least in cases without other crimes involved. On its face, it's still extreme, but these guys now hold an entire communication system hostage. If sending several of them to prison for their transgressions (which ARE transgressions) can be a deterrent, then I'm for it. I think it really will be a deterrent if we can get some convictions. It's not like people spam in a brief burst of anger. These people generally have some business or technical skills that could find them legitimate employment (perhaps somewhat less lucrative, but above the poverty line) even in the lousy tech economy. I hear the porn industry does well when the economy is lousy. I'm sure my mom would much rather I manage a (legitimate) porn server farm than a spam server farm anyway.
In theory, you're right. Sadly, Congress and the Courts have been of the mindset that if a little property is good, then more must be better. These days you can't sneeze without tortiously infringing on a copyright, because you might have seen a movie with some particularly distinctive sneeze. In fact, even if you didn't see the movie, if it played in your town, you had access to it, and you have the burden of proving you didn't see it.
That's funny, because Sprint's residential DSL, which partners with Earthlink under an arrangement whose terms are not known to me...
wait for it...
HAS SPYWARE BUILT INTO THE SETUP SOFTWARE!
Or at least it did when my parents subscribed. Nothing that a disk crash and a reinstall with RASPPPOE couldn't solve, but wow.
I guess that's not quite as bad as their new Compaq desktop that came with spyware PREINSTALLED.
There's a general philosophy amount consumer software and hardware distributors that people don't want to know what goes on underneath, and give their tacit permission for them to put whatever the hell they want on there. We keep blaming code bloat for making computers run slower than they used to, but maybe the fact that people have a couple dozen completely unnecessary processes running, each using just a few megs of RAM that nobody would notice missing by itself, has something to do with it.
Granted, I could probably cut down on a few things running on my fairly stock Debian/KDE workstation, but they use about 1/100 the resources as the useless crap on my parents' machine. Of course, none of the things on my machine that I'm not entirely sure I need are designed to scan web pages I read for key words and deliver pop-up ads for competitors.
If the Nullsoft release was unauthorized (what constitutes unauthorized is not as clear-cut as AOL would have us believe) then the fact that the code was GPL'd is irrelevant. Go roll your own people. Don't even look at the WASTE source. You'll be tainted.
This is why precedents don't extend beyond the jurisdiction of the court that sets them, and why appeals courts accept amicus curiae briefs from other parties, which may actually be more influential than the arguments of the attorneys of the parties directly involved in the suit.
Generally speaking, if an employee of a corporation does something on behalf of a corporation in a manner consistent with an official act of the corporation (like, posting it on their main website), regardless of whether or not they were supposed to, the corporation's recourse is fire the employee. The exceptions to this would be if the employee knew that the release was unauthorized and did it anyway, making them no longer an agent of the company. As long as they did it in good faith, they were acting as an agent of the company, however poor their judgement may have been.
Of course, there are other exceptions to this as well, more obscure and probably even more open to subjective interpretation in court. If you wanna use something like this, consult a lawyer, or roll your own.
This kind of thing would never be an issue with postal mail, because a tangible copy is made. By convention this is assumed to be true of email. Instant messaging blurs the lines. Thanks to the wonders of internet technology, we have modes of communication inherently unlike anything contemplated by legislative bodies at the time of the writing of these laws.
What do appeals courts generally do to convictions based on laws that weren't written with the circumstances of the alleged crime in mind? They generally throw them out. Lets hope this holds, and also badger our legislators to allow recording of communications by those known by the speaker to be receiving them. Two-party consent lets the powerful screw over the little people and not be held accountable for it. It's arguably a first amendment violation, though apparently nobody has argued that lately, or at least not convincingly.
The unix world has the advantage of starting with a lot of software designed to be used in a multi-user environment by non-administrative users. The Windows world is rife with the artifacts of DOS/Win16/Win9x/WinME software development which makes no such assumptions, with the result that it is considered normal by both users and developers to have and require administrative access. While this is slowly changing, it is a problem that unix-based environments won't have to face.
Of course, people will come out with things like Lindows and face them anyway, but it should be pretty easy for them to adopt the standard linux security model if they start having significant problems.
They've known for a long time that LEAP is inherently flawed, and no patch can fix it. That said, it's a hell of a lot simpler to deploy than more secure things like EAP-TLS. This attack still requires an offline brute force decryption attempt. Granted, it may be a highly accelerated brute force decryption attempt, but if you don't allow your users to use passwords that are vulnerable to dictionary attacks, LEAP is Good Enough for many purposes.
I don't even have to say it...
The Deep Space Nine episode you're describing sounds like "Favor the Bold" or "Sacrifice of Angels". Wow, I never thought of myself as a Trek geek until now.
I read this as meaning that they won't go messing with their backups to purge your data, which is perfectly reasonable, especially for a free service. Backups are supposed to be write-once recordings. If you've got a problem with that, don't transmit sensitive data over a free webmail service. Nothing spooky here. In fact, I imagine every webmail service works this way. Gmail is just kind enough to warn you about this.
Yes, it really is about an order of magnitude better. 0.95/0.38 = 2.5. Now, keeping in mind that we're dealing with a 2-D surface, 2.5^2 = 6.25. Log(6.25) = 0.796, so this is about 4/5 of an order of magnitude improvement.
It's my understanding that a lot of the organization for the coup that took down the Soviet Union was done by email. Of course, at the time, the internet was under the radar. More recently, massive demonstrations that brought down a government were organized by text messaging. The lesson here is that communication technology may indeed be a threat in places that actually allow the people to use it. The Chinese government is right to be afraid, because internet usage is quite widespread there. North Korea probably has nothing to fear.