NIST required a reference implementation in C and two optimized implementations in C and old Java (JDK1.1). Reasonably good software benchmarks (even on 8-bit microprocessors) and probable hardware complexity were part of what they selected for.
That most definitely does not follow. In a free, pluralistic society with a democratic government, the only laws that are valid and deserve to exist are necessarily based on ethical norms that are widely shared and have very strong consensus behind them. IOW, there are lots of things I personally think people shouldn't do, but they shouldn't be banned because my ethical sense doesn't trump anyone else's and the government doesn't belong solely to me. I can't seriously support my own pet laws for the same reason I wouldn't support yours (though I can wish the world would get a clue). Every law is an infringement on freedom (hopefully in an honest effort to strike a balance between conflicting rights), and should be there not merely to make society a better place (that's what voluntary acts are for), but to maximize our ability to exercise our rights.
I can't speak for Stallman, but that's my argument (shaped by the ideals of the US, I know) for how people can advocate their ethics without supporting laws mandating them.
do you support a law making it illegal to sell any computer that isn't built from readily-available spare parts?
Not at all. Every box in our apartment is beige (and I have a scarred knuckle to prove it), and I think buying proprietary hardware is awfully short-sighted, but there's no consensus that it's inherently evil.
Our claim to fame is being tool-users that learn. IMHO a wholly consumer society that doesn't let people learn from and improve their tools has failed, and is probably a threat to our entire civilization. I hate the idea that we're punishing tool buyers for doing things that tool sellers don't approve of, though I don't know how to avoid this without gutting contract law (probably by a crude hack like declaring certain contract terms unenforceable). But then, "free" negotiation between unequals is my one big problem with libertarianism.
To clarify, if I patent the Frobnicator and you patent the Super Frobnicator (based on the Frobnicator), nobody (including you) can make a Frobnicator without a license from me, and nobody (including me!) can make a Super Frobnicator without licenses from each of us.
Corel (or whoever they bought WordPerfect from) suffered precisely because Microsoft had no liability for problems caused by changes they made, so they could break their customers' third-party apps with impunity. If Windows had a warranty with teeth, they couldn't have gotten away with doing it.
I'm glad they've worked on it, but I have to wonder which session is "current" for a service none of them started, and whether every service will have a security token that grants access to every session.
Yeah, it's mostly about thinking of the interface contracts at your level and the next, and knowing how failures are indicated. You have to be a lot more paranoid without exception support. But you probably wanted
since stdout to a non-tty could be block buffered, and stdio might not see the error until it flushes your file buffers at exit. And I don't think non-POSIX systems mandate that exit(0) is success and exit(1) is failure (I think VMS errors are all negative, but the C library special-cased 0 and 1).
(And you need to write "<" to get "<" in sample code with HTML.)
They probably can't sue you unless your own negligence contributed somehow. But if they somehow can, you and the other hundred software vendors (who each got sued for $X) can bring a class action against Microsoft for $100X and come out even--and the threat of that will ensure the whole mess doesn't happen.
Writing correct software isn't impossible, just more difficult. For now the payoff is so small that hardly anyone bothers.
Does everyone who wants to take the exam need a degree, or just whoever first created the exam?
Anyone with a degree from an accredited college already passed a bunch of tests covering all that material. Wouldn't this exam only be useful to qualify candidates who (for whatever reason) don't have degrees? Or is this a "have they forgotten everything" requalification everyone must pass periodically?
An OS written in C++ is okay with me (though I think the efficiency advantages of allowing "undefined behavior" are overblown). What worries me is that more people are going to think it's reasonable to make system interfaces (like the AtheOS GUI) that are only usable by C++ applications, making life harder for less hidebound hackers.
Dependency walker won't find calls to LoadLibrary (which some apps call because a plugin is declared in the registry or because the runtime linker's feedback about missing DLLs is so useless) or CoGetClassObject (which is how every COM client loads server or proxy DLLs). And I can't expect it to be there on an end user's system.
I can't see why you would want to [list all other packages that depend on a given package] though.
You can't ever safely remove a package without answering the question "does anything (that I'm not removing) depend on this?"
Terminal services is really only good for remote admin. X's model (any app on any box can ask for a window on your display, and all windows are equal) is much cleaner. When you put an entire virtual display (with yet another window manager) in a window, there are bizzare rules about which events are interpreted as commands by which window manager, and you can't really fit apps on different machines on your real display at once because all your real estate is wasted on remote desktops. Then there are the implementation quirks, like services blocking on modal dialogs you can only dismiss (or even see) by using the main console (VNC is a big win here, though it's awfully high-bandwidth and slow from Win32).
GNU and BSD are so huge it's difficult to contribute to them, so it's reassuring to see that it's still possible for one person to write a credible kernel. But with all the recent progress in CORBA, JVM, and CLR (the pseudo-JVM for.NET), it's a worrisome trend to see new interfaces that can be used from C++ but not from more robust languages (Eiffel, Python, Perl, Ruby, Ada...). And pressure on the GCC team against improving their ABI (because of all the underspecified interfaces that'll break) can't be good.
GIF can't encode images that use more than 256 colors--there's some debate whether you can validly compose areas with several palettes in one image, but certainly you can't expect every implementation to display that correctly. But once you quantize the image you can keep decoding, manipulating, and re-encoding it forever without losing information. PNG, MNG, TIFF, and GIF are suitable editing formats, while JPEG and MPEG aren't because the output continually degrades.
Hardly anyone knew LZW was patented, partly because nobody reads patents in general, partly because Terry Welch neglected to even mention the patent when he described LZW in the June 1984 issue of IEEE Computer.
It's impossible, even in principle, to be certain no US patent applies to technology you're using. The patent database is a mountain of gibberish (just reading them is a skill ordinary developers don't have), and nobody can afford to keep doing searches so thorough they'll definitely find every relevant one. But even if you could, you wouldn't be allowed to see the pending applications that might be approved years later.
US patent applications have never been public record; it's always been viewed as important that the invention can be a trade secret if the patent isn't issued.
Certainly examiners are underpaid (thus it's very unlikely the USPTO can attract many skilled software developers) and aren't suitably rewarded for diligently seeking prior art (in fact doing their job correctly would reduce their throughput so severly they'd probably be dismissed), but fairly involving the public would be very difficult. IMHO the USPTO ought to line up a few thoroughly nondisclosed consulting firms--give them a problem statement and a small budget (say one week of a senior developer's time), and if they can reproduce the invention it's presumed obvious and ineligible for a patent.
The GPL can't impose restrictions beyond copyright law. However, copyright law restrictions do apply to derived works. It's not entirely clear just when reusing others' software does or does not make your software a derived work (there hasn't been much precedent), but unlike most of us the FSF hired a lawyer to tell them what courts will probably (or at least ought to) decide.
I don't think that all software will ever be free simply because that will require force against those who disagree
Not necessarily. All we have to do is refuse to penalize people who reverse-engineer, modify, and redistribute proprietary software.
As corporations gradually realize the inevitability of any product being remade to do its new owner's rightful bidding, they're demanding ever greater legal powers to control their customers' behavior. We need to establish once and for all that learning from and improving our tools are some of the most important of our inalienable rights as sentient tool-using beings.
Minaturization costs money, at least in the macroscopic world. If you have a square foot of motherboard, relatively straightforward designs and big cheap commodity components become feasible. Smaller systems save a little power and make higher clock rates attainable (in a 1GHz clock cycle of 1ns, an electronic signal only travels about a foot), but until commodity desktop/minitower hits such a performance wall (or simply goes out of mass-market fashion) it'll always be cheaper than embeddable versions or laptops.
IMHO assert() is showing its age. It ought to throw an exception (and the runtime should provide a stack-trace facility with module names and line numbers for exceptions that aren't specially handled; Java does a good job here), and should be selectively disabled or not at all. Disabling them all is almost always the Wrong Thing.
Language bloat isn't a measure of how big any one implementation is, but a measure of how many gratuitous (because any user could implement them) features are built in. Libraries can offer all the domain-specific data types you could want, and all users really need is some convenient way to write an expression that produces such a value--you don't have to build a privileged few into the fundamental syntax of the language!
There are filesystems that intentionally make it difficult to distinguish between encrypted data and cover noise, so They know you can never prove that you've revealed all your information.
Stroustrup advocated the Resource Acquisition Is Initialization pattern in the canonical C++ manual. It's practically what destructors are for. Anyone who didn't understand that book just isn't qualified to use the language.
Commercial OSes have to adopt the same official stance towards errors as Communist countries had towards poverty. For doctrinal reasons it was not possible to admit that poverty was a serious problem in Communist countries, because the whole point of Communism was to eradicate poverty. Likewise, commercial OS companies like Apple and Microsoft can't go around admitting that their software has bugs and that it crashes all the time, any more than Disney can issue press releases stating that Mickey Mouse is an actor in a suit.
Slashdot Mirror
NIST required a reference implementation in C and two optimized implementations in C and old Java (JDK1.1). Reasonably good software benchmarks (even on 8-bit microprocessors) and probable hardware complexity were part of what they selected for.
That most definitely does not follow. In a free, pluralistic society with a democratic government, the only laws that are valid and deserve to exist are necessarily based on ethical norms that are widely shared and have very strong consensus behind them. IOW, there are lots of things I personally think people shouldn't do, but they shouldn't be banned because my ethical sense doesn't trump anyone else's and the government doesn't belong solely to me. I can't seriously support my own pet laws for the same reason I wouldn't support yours (though I can wish the world would get a clue). Every law is an infringement on freedom (hopefully in an honest effort to strike a balance between conflicting rights), and should be there not merely to make society a better place (that's what voluntary acts are for), but to maximize our ability to exercise our rights.
I can't speak for Stallman, but that's my argument (shaped by the ideals of the US, I know) for how people can advocate their ethics without supporting laws mandating them.
Not at all. Every box in our apartment is beige (and I have a scarred knuckle to prove it), and I think buying proprietary hardware is awfully short-sighted, but there's no consensus that it's inherently evil.
Our claim to fame is being tool-users that learn. IMHO a wholly consumer society that doesn't let people learn from and improve their tools has failed, and is probably a threat to our entire civilization. I hate the idea that we're punishing tool buyers for doing things that tool sellers don't approve of, though I don't know how to avoid this without gutting contract law (probably by a crude hack like declaring certain contract terms unenforceable). But then, "free" negotiation between unequals is my one big problem with libertarianism.
To clarify, if I patent the Frobnicator and you patent the Super Frobnicator (based on the Frobnicator), nobody (including you) can make a Frobnicator without a license from me, and nobody (including me!) can make a Super Frobnicator without licenses from each of us.
Corel (or whoever they bought WordPerfect from) suffered precisely because Microsoft had no liability for problems caused by changes they made, so they could break their customers' third-party apps with impunity. If Windows had a warranty with teeth, they couldn't have gotten away with doing it.
I'm glad they've worked on it, but I have to wonder which session is "current" for a service none of them started, and whether every service will have a security token that grants access to every session.
Yeah, it's mostly about thinking of the interface contracts at your level and the next, and knowing how failures are indicated. You have to be a lot more paranoid without exception support. But you probably wanted
since stdout to a non-tty could be block buffered, and stdio might not see the error until it flushes your file buffers at exit. And I don't think non-POSIX systems mandate that exit(0) is success and exit(1) is failure (I think VMS errors are all negative, but the C library special-cased 0 and 1).(And you need to write "<" to get "<" in sample code with HTML.)
Writing correct software isn't impossible, just more difficult. For now the payoff is so small that hardly anyone bothers.
Does everyone who wants to take the exam need a degree, or just whoever first created the exam?
Anyone with a degree from an accredited college already passed a bunch of tests covering all that material. Wouldn't this exam only be useful to qualify candidates who (for whatever reason) don't have degrees? Or is this a "have they forgotten everything" requalification everyone must pass periodically?
An OS written in C++ is okay with me (though I think the efficiency advantages of allowing "undefined behavior" are overblown). What worries me is that more people are going to think it's reasonable to make system interfaces (like the AtheOS GUI) that are only usable by C++ applications, making life harder for less hidebound hackers.
Dependency walker won't find calls to LoadLibrary (which some apps call because a plugin is declared in the registry or because the runtime linker's feedback about missing DLLs is so useless) or CoGetClassObject (which is how every COM client loads server or proxy DLLs). And I can't expect it to be there on an end user's system.
You can't ever safely remove a package without answering the question "does anything (that I'm not removing) depend on this?"
Traditionally, /bin/ls only defaults to multi-column output if stdout is a tty. Does "ls|cat" look just like "ls" on your box?
Terminal services is really only good for remote admin. X's model (any app on any box can ask for a window on your display, and all windows are equal) is much cleaner. When you put an entire virtual display (with yet another window manager) in a window, there are bizzare rules about which events are interpreted as commands by which window manager, and you can't really fit apps on different machines on your real display at once because all your real estate is wasted on remote desktops. Then there are the implementation quirks, like services blocking on modal dialogs you can only dismiss (or even see) by using the main console (VNC is a big win here, though it's awfully high-bandwidth and slow from Win32).
GNU and BSD are so huge it's difficult to contribute to them, so it's reassuring to see that it's still possible for one person to write a credible kernel. But with all the recent progress in CORBA, JVM, and CLR (the pseudo-JVM for .NET), it's a worrisome trend to see new interfaces that can be used from C++ but not from more robust languages (Eiffel, Python, Perl, Ruby, Ada...). And pressure on the GCC team against improving their ABI (because of all the underspecified interfaces that'll break) can't be good.
GIF can't encode images that use more than 256 colors--there's some debate whether you can validly compose areas with several palettes in one image, but certainly you can't expect every implementation to display that correctly. But once you quantize the image you can keep decoding, manipulating, and re-encoding it forever without losing information. PNG, MNG, TIFF, and GIF are suitable editing formats, while JPEG and MPEG aren't because the output continually degrades.
Hardly anyone knew LZW was patented, partly because nobody reads patents in general, partly because Terry Welch neglected to even mention the patent when he described LZW in the June 1984 issue of IEEE Computer.
It's impossible, even in principle, to be certain no US patent applies to technology you're using. The patent database is a mountain of gibberish (just reading them is a skill ordinary developers don't have), and nobody can afford to keep doing searches so thorough they'll definitely find every relevant one. But even if you could, you wouldn't be allowed to see the pending applications that might be approved years later.
US patent applications have never been public record; it's always been viewed as important that the invention can be a trade secret if the patent isn't issued.
Certainly examiners are underpaid (thus it's very unlikely the USPTO can attract many skilled software developers) and aren't suitably rewarded for diligently seeking prior art (in fact doing their job correctly would reduce their throughput so severly they'd probably be dismissed), but fairly involving the public would be very difficult. IMHO the USPTO ought to line up a few thoroughly nondisclosed consulting firms--give them a problem statement and a small budget (say one week of a senior developer's time), and if they can reproduce the invention it's presumed obvious and ineligible for a patent.
The GPL can't impose restrictions beyond copyright law. However, copyright law restrictions do apply to derived works. It's not entirely clear just when reusing others' software does or does not make your software a derived work (there hasn't been much precedent), but unlike most of us the FSF hired a lawyer to tell them what courts will probably (or at least ought to) decide.
Not necessarily. All we have to do is refuse to penalize people who reverse-engineer, modify, and redistribute proprietary software.
As corporations gradually realize the inevitability of any product being remade to do its new owner's rightful bidding, they're demanding ever greater legal powers to control their customers' behavior. We need to establish once and for all that learning from and improving our tools are some of the most important of our inalienable rights as sentient tool-using beings.
Minaturization costs money, at least in the macroscopic world. If you have a square foot of motherboard, relatively straightforward designs and big cheap commodity components become feasible. Smaller systems save a little power and make higher clock rates attainable (in a 1GHz clock cycle of 1ns, an electronic signal only travels about a foot), but until commodity desktop/minitower hits such a performance wall (or simply goes out of mass-market fashion) it'll always be cheaper than embeddable versions or laptops.
IMHO assert() is showing its age. It ought to throw an exception (and the runtime should provide a stack-trace facility with module names and line numbers for exceptions that aren't specially handled; Java does a good job here), and should be selectively disabled or not at all. Disabling them all is almost always the Wrong Thing.
Language bloat isn't a measure of how big any one implementation is, but a measure of how many gratuitous (because any user could implement them) features are built in. Libraries can offer all the domain-specific data types you could want, and all users really need is some convenient way to write an expression that produces such a value--you don't have to build a privileged few into the fundamental syntax of the language!
There are filesystems that intentionally make it difficult to distinguish between encrypted data and cover noise, so They know you can never prove that you've revealed all your information.
Stroustrup advocated the Resource Acquisition Is Initialization pattern in the canonical C++ manual. It's practically what destructors are for. Anyone who didn't understand that book just isn't qualified to use the language.