You actually mean Andrew Morton and the rest of Linus's trusted lieutenants. Alan Cox is off doing other stuff currently.
Re:A great success story of Linux on the desktop..
on
Rome Moving to Linux
·
· Score: 4, Insightful
Because they installed Debian, which is not commercial. Like it or not, software procurement goes in the "business" section of the newspaper, and that section is mainly information about the successes and failures of companies in doing business. Munich gets reported because it is a deal with IBM and SuSE, so we hear about how these companies are being successful in the sector. Extremadura isn't a big success for any corporate entity (as far as I can tell), nor a particularly big failure for MicroSoft; it's a great success story for the region, but that's only regional news.
I personally think that it's a much better example of adoption of OSS; Extremadura is actually using the freedoms that RMS goes on about to change the things they don't find right about the software, rather than treating the software as proprietary software sold on reasonable terms. But this won't be interesting to the newspapers until somebody in Madrid notices that Extremadura has better IT on a smaller budget.
All systems require training. The real question is when the documentation will be sufficient training. Linux documentation has gotten to be quite good, but it is not clearly organized, nor does it start at the level of a complete newbie. Ideally, the training requirement will become a short-term loss of productivity (as people look things up the first time on the new system) rather than something where you need people to come in and teach you stuff.
My experience is that sound never works the first time you try to set it up, but the second time you try, you find that it already works. 2.6 has been, for me, the exception; it worked the first time. Of course, I'm still using the OSS emulation in ALSA, not the ALSA API at all. But it seems to me to be better than the real OSS; it will play two sounds at the same time (the second through the second PCM) even if programs are only trying to use the first device. I'm even setting the volume levels with a program configured for OSS (which seems to me to actually work better than the ALSA programs).
They're already planning a major change. This is not actually a commitment to use any OSS, but rather a decision to evaluate OSS solutions as part of the change. Furthermore, they run relatively little software on their desktops; most of the work is done on their mainframes. So the OSS portion of this is not really a massive undertaking. For that matter, if they start running Linux on their mainframes (side-by-side with what they're currently running), they can start a tighter integration between their mainframe and desktop environment.
If your phone, which is hooked to the network without any paid-for service, can reach 911, what do you need the VoIP provider for? The old POTS phone line coming into your house is a much more direct, reliable, and logical way of reaching the local 911 center than anything the VoIP provider could possibly do. And the way to make sure everybody has 911 access isn't to require VoIP providers to have it, but rather to cover it with property taxes; if you have a building, your taxes should supply you with a 911 phone line, and potentially a phone to connect to it.
(In point of fact, my phone book does have a list of emergency numbers in the front, on the page titled "911 emergency numbers"; this allows you to directly call the poison control center or the secret service, but it doesn't have any town-level services)
It seems to me that a VoIP provider is entirely the wrong place to route 911 calls. How is the provider supposed to know who to connect you to? Since it's IP from you to the provider, you could be anywhere when you make your call, regardless of what your home address is. If you make a 911 call from a cell phone, it should connect you to the police for the cell you're in, not your home police department, who will be useless if you're 100 miles from home.
The right way to set things up is to have the physical phone require a phone line (which need not be in service), which it can use to make 911 calls over POTS. This system should be funded out of property taxes (since it's local infrastructure).
Just recently, I was doing some math in which 1+1=0, because + in this case is the same as xor for binary representations of integers. There's a lot of useful math you can do based only on having a small set of principles for how your operations behave, and letting all of the other properties go. In this case, it gives you +, -, *, and / (except for/0), with a total of 256 different numbers. You can do algebra perfectly well with this system, and the error correction on CDs actually uses this system.
The arithmetic you're used to is integer arithmetic. But it is only one of a large set of possible arithmetics that each have some useful properties while not having other properties. These other arithmetics are quite different, but mathematicians reuse the usual symbols for them to show the properties they have. So "0" is the thing that doesn't affect +, "1" is the thing that doesn't affect *, "0" * anything is "0", (a+b)*c = a*c+b*c, and so forth.
Your usual arithmetic is based on counting; you state how you use the numbers to count, and you say that every number except 0 is after some other number, that after no two different numbers come the same number, and that, if you start counting from 0, you'll eventually get to every number. Addition is defined by saying that 0+x=x and if you have a sum and you count down on the left and up on the right, the total is the same. 1*x=x, and y*x+x=(y+1)*x. From that, and the ordering of the digits, you get the math you're familiar with.
I think a cross between R2-D2 and an Ewok would be great, if the lines were all delivered in a properly depressed tone of voice, and he moved appropriately. He is supposed to be "your plastic friend who's fun to be with", except for software problems and faulty hardware. Having him in a body totally inappropriate for his behavior would be just the thing, I think.
I think a more significant response would be to put pressure on the DoJ to file criminal charges against the RIAA for their tactics. If this woman's case has any merits, and voters seem to care enough to make it an campaign issue in the presidential race, it's possible. At that point, the RIAA would clearly quit, because they're practically certain to have their civil cases dismissed with contempt of court if those cases are the subject of criminal charges, and they wouldn't be able to get any lawyers (a.k.a. co-conspirators, who lose privilege) for the cases anyway.
Sure, it's not as fitting an end to the RIAA as being gunned down by a rival street gang in LA (or arrested by the LAPD), but it's something. I wonder if they're due for an audit...
Suing Munich would be a bit tricky, because they'd actually have to make some sort of claim. And, last I heard, Munich is in Germany, and there's an injunction against SCO making claims about Linux in Germany. Perhaps Darl will send Munich a blank threatening letter.
He's talking about an IP network in any case, so the voice and TV streams only come over your connection if you're actually using them.
The interesting thing, of course, is how much upstream bandwidth you'd have. If you and each of your neighbors have a 10M connection, will your block have enough upstream to support all of you using the full 10M at the same time? (The TV streams are easy, because they're broadcast, so they can enter the neighborhood as a single stream).
The tricky thing is really that you don't want to provide enough bandwidth for everybody at the same time, because it would never all be used, but you also want to make sure that, regardless of what is going on, a call to 911 gets through reliably.
He is firm on proprietary software; he won't use anything that doesn't given him the freedoms he considers important. But the GPL is designed not merely to provide those freedoms, but to aid in getting those freedoms to other works. The issue with this kind of license is that it doesn't scale well to a large number of combined projects, but XFree86 is a self-contained project anyway. The main issue is that using such a license encourages other people to use a similar license for other sorts of project, which would cause problems in the long run.
As far as I can tell from studying linguistics and computer science, formal grammars are not particularly good ways of representing either sort of language. Chomsky's main founding point was that formal methods could be used to study natural language; he proposed context-free grammars, and then quickly abandoned them, because no language is actually context-free.
Computer language syntax picked up context-free grammars, because computer languages are generally context-free, at least to a certain extent. Of course, you can't actually implement an arbitrary context-free grammar efficiently, so they turned to a restricted subset which is sufficient for the important cases. Of course, the grammar is (as natural languages discovered millenia ago) insufficient for anything useful, so they developed interesting semantics behind the overly-strict grammar.
At this point, the interesting work in linguistics (which relies heavily on obscure languages to test the boundaries of what the human language faculty produces) is in the ways that language goes beyond what is feasible to define and use in an unambiguous way; this is stuff which is unsuitable for programming languages, because it is, by definition, impossible to interpret predictably. Compiler and computer language design has not informed linguistics significantly, because natural language uses an entirely different set of tricks for an entirely different set of goals.
The research in computer languages, on the other hand, is in bits of semantics which are entirely unlike any semantics used by natural languages, but are understandable by other faculties. It is focused on the formal representation of data structures and processes, two things that natural language is entirely inadaquate for and relies entirely on extra-linguistic methods (such as demonstration) to convey.
Consider, for example, the addition of a simple bit of natural language to a computer language. Say there were an "it" keyword, which referred to the most recently used variable which type checks in the context in which it is used, except that in the arguments of a method, it cannot refer to the object on which the method is called. Such a keyword would be practically impossible to use reasonably, since it would be extremely fragile and hard to interpret. However, such a keyword is present and its use is required in almost all natural languages. Natural language is really more like a machine language than a high-level programming language; the machine it is for is to be found about your left ear, and it has only been partially reverse engineered.
Actually, this would be an interesting question to pose to RMS: the problem is that this software, while Free by itself, doesn't play well with others. You can't repackage it in some ways and distribute the result. You may not be able to link a GPL program against the library and distribute the result.
But all of the freedoms the user is supposed to demand are available when considering the program in isolation. It seems to me that RMS would say that Mandrake is right not to offer packages of XFree86 4.4 due to the license. But I don't think RMS would have any problem with getting and using XFree86 4.4 yourself if you needed it. The only caveat is that, if you distribute something, it must work with 4.3's libraries, because 4.4 is not free for that sort of use.
The girlfriend can tell the difference, it's not the diamond that gives it away. She can't tell if the diamond was mined or manufactured, but the 30% discount is quite obvious.
I'm unclear on why a new C# compiler written in C# is necessary; wouldn't it be easier to add a C# front end and a CLR back end to gcc, which not only already has a good intermediate language, but already has a Java front end, which would be close already. This would also provide native executable output for a huge number of architectures.
I think the "free" argument should actually be an "unencumbered" argument. With open source licensed software, you can actually own your copy of the software, and use it without needing a license, much as you can read a book you own without needing a license. It is necessary to have a license in order to distribute, copy, or modify the software, but only IT needs this license (to install software for end users); the end users don't need licenses at all. If you are using exclusively open source software, you can legally have a 60000-person company using one license, so you have a vastly smaller license management problem. The benefit of "free" isn't that the price is zero; the benefit is that accounting isn't needed.
Technology also creates jobs doing things that wouldn't be possible without it, because tasks that wouldn't be possible before become possible, and people are then needed to perform those tasks. For any particular task, technology (ideally) makes it so you don't need so many people to do it, but that is outweighed by the "next step" tasks that become possible.
So technology generally creates jobs at about the same rate, overall, than it eliminates them. Which is why American cities aren't more overrun with unneeded, out-of-work, former workers than ancient Rome was.
Open source doesn't need marketing, because it doesn't matter all that much to open source whether end users use it. The point of open source is to solve your local problem better with the help of like-minded people.
On the other hand, companies selling solutions based on open source projects do need marketing, because they do actually have to get otherwise uninterested end users interested. Of course, they seem to be doing pretty well at just that.
Certainly if you want to be more than a faceless minion you can't be anonymous. But you can be anonymous and say something that comes to the attention of known people. If your point is good enough, it will stand on its own without your reputation to back it up, or it will be backed by the reputations of people who find and promote it.
Furthermore, there are ideas that nobody wants to officially back, but which people will bring up if they have been mentioned anonymously. People can say, without lending their support to the idea, that a good argument is needed to refute it, and this can lead to the idea being adopted if nobody can come up with a good response.
So there is value to anonymous comments in the democratic process, even if there is also more weight given to the concerns of known individuals. Furthermore, with well-protected pseudonyms (e.g., anonymous posts signed with a certificate with no name but a recognizable fingerprint), it is possible to identify look for comments from the same source, if the source wants to associate the comments with each other. And it should be possible to sign comments, if desired, if you want your identity to be known.
I think that would be great. I think that would give potential employers a much better idea of what a person is like, particularly because you'd probably have a mix of good and bad reviews, which would tell you what sort of situations the person does well in. (And I'd like my employers to see that, despite sleeping through all of my classes, I still learned the material.)
Unfortunately, having been a TA for a couple of classes, I know that no professor is actually going to have time to write that much detail about all of the students. Much more likely is that a professor would manage to review a few of the most noteworthy students ("She came to office hours every week for extra help on material she clearly understood perfectly." "He kept correcting my sign errors before I'd finished writing the equations."), and they'd forget entirely about the existence of Mr. Smith. Besides, it would be hard for Mr. Smith to not know what professor had given that review, unless he's taking multiple courses which teach square roots, so anonymity isn't very helpful.
Slashdot Mirror
Remember, "professional" means that you get paid for it, not that you're any good at it.
You actually mean Andrew Morton and the rest of Linus's trusted lieutenants. Alan Cox is off doing other stuff currently.
Because they installed Debian, which is not commercial. Like it or not, software procurement goes in the "business" section of the newspaper, and that section is mainly information about the successes and failures of companies in doing business. Munich gets reported because it is a deal with IBM and SuSE, so we hear about how these companies are being successful in the sector. Extremadura isn't a big success for any corporate entity (as far as I can tell), nor a particularly big failure for MicroSoft; it's a great success story for the region, but that's only regional news.
I personally think that it's a much better example of adoption of OSS; Extremadura is actually using the freedoms that RMS goes on about to change the things they don't find right about the software, rather than treating the software as proprietary software sold on reasonable terms. But this won't be interesting to the newspapers until somebody in Madrid notices that Extremadura has better IT on a smaller budget.
All systems require training. The real question is when the documentation will be sufficient training. Linux documentation has gotten to be quite good, but it is not clearly organized, nor does it start at the level of a complete newbie. Ideally, the training requirement will become a short-term loss of productivity (as people look things up the first time on the new system) rather than something where you need people to come in and teach you stuff.
My experience is that sound never works the first time you try to set it up, but the second time you try, you find that it already works. 2.6 has been, for me, the exception; it worked the first time. Of course, I'm still using the OSS emulation in ALSA, not the ALSA API at all. But it seems to me to be better than the real OSS; it will play two sounds at the same time (the second through the second PCM) even if programs are only trying to use the first device. I'm even setting the volume levels with a program configured for OSS (which seems to me to actually work better than the ALSA programs).
They're already planning a major change. This is not actually a commitment to use any OSS, but rather a decision to evaluate OSS solutions as part of the change. Furthermore, they run relatively little software on their desktops; most of the work is done on their mainframes. So the OSS portion of this is not really a massive undertaking. For that matter, if they start running Linux on their mainframes (side-by-side with what they're currently running), they can start a tighter integration between their mainframe and desktop environment.
If your phone, which is hooked to the network without any paid-for service, can reach 911, what do you need the VoIP provider for? The old POTS phone line coming into your house is a much more direct, reliable, and logical way of reaching the local 911 center than anything the VoIP provider could possibly do. And the way to make sure everybody has 911 access isn't to require VoIP providers to have it, but rather to cover it with property taxes; if you have a building, your taxes should supply you with a 911 phone line, and potentially a phone to connect to it.
(In point of fact, my phone book does have a list of emergency numbers in the front, on the page titled "911 emergency numbers"; this allows you to directly call the poison control center or the secret service, but it doesn't have any town-level services)
It seems to me that a VoIP provider is entirely the wrong place to route 911 calls. How is the provider supposed to know who to connect you to? Since it's IP from you to the provider, you could be anywhere when you make your call, regardless of what your home address is. If you make a 911 call from a cell phone, it should connect you to the police for the cell you're in, not your home police department, who will be useless if you're 100 miles from home.
The right way to set things up is to have the physical phone require a phone line (which need not be in service), which it can use to make 911 calls over POTS. This system should be funded out of property taxes (since it's local infrastructure).
Just recently, I was doing some math in which 1+1=0, because + in this case is the same as xor for binary representations of integers. There's a lot of useful math you can do based only on having a small set of principles for how your operations behave, and letting all of the other properties go. In this case, it gives you +, -, *, and / (except for /0), with a total of 256 different numbers. You can do algebra perfectly well with this system, and the error correction on CDs actually uses this system.
The arithmetic you're used to is integer arithmetic. But it is only one of a large set of possible arithmetics that each have some useful properties while not having other properties. These other arithmetics are quite different, but mathematicians reuse the usual symbols for them to show the properties they have. So "0" is the thing that doesn't affect +, "1" is the thing that doesn't affect *, "0" * anything is "0", (a+b)*c = a*c+b*c, and so forth.
Your usual arithmetic is based on counting; you state how you use the numbers to count, and you say that every number except 0 is after some other number, that after no two different numbers come the same number, and that, if you start counting from 0, you'll eventually get to every number. Addition is defined by saying that 0+x=x and if you have a sum and you count down on the left and up on the right, the total is the same. 1*x=x, and y*x+x=(y+1)*x. From that, and the ordering of the digits, you get the math you're familiar with.
I think a cross between R2-D2 and an Ewok would be great, if the lines were all delivered in a properly depressed tone of voice, and he moved appropriately. He is supposed to be "your plastic friend who's fun to be with", except for software problems and faulty hardware. Having him in a body totally inappropriate for his behavior would be just the thing, I think.
I suspect that Munich does its banking in Germany.
I think a more significant response would be to put pressure on the DoJ to file criminal charges against the RIAA for their tactics. If this woman's case has any merits, and voters seem to care enough to make it an campaign issue in the presidential race, it's possible. At that point, the RIAA would clearly quit, because they're practically certain to have their civil cases dismissed with contempt of court if those cases are the subject of criminal charges, and they wouldn't be able to get any lawyers (a.k.a. co-conspirators, who lose privilege) for the cases anyway.
Sure, it's not as fitting an end to the RIAA as being gunned down by a rival street gang in LA (or arrested by the LAPD), but it's something. I wonder if they're due for an audit...
Suing Munich would be a bit tricky, because they'd actually have to make some sort of claim. And, last I heard, Munich is in Germany, and there's an injunction against SCO making claims about Linux in Germany. Perhaps Darl will send Munich a blank threatening letter.
He's talking about an IP network in any case, so the voice and TV streams only come over your connection if you're actually using them.
The interesting thing, of course, is how much upstream bandwidth you'd have. If you and each of your neighbors have a 10M connection, will your block have enough upstream to support all of you using the full 10M at the same time? (The TV streams are easy, because they're broadcast, so they can enter the neighborhood as a single stream).
The tricky thing is really that you don't want to provide enough bandwidth for everybody at the same time, because it would never all be used, but you also want to make sure that, regardless of what is going on, a call to 911 gets through reliably.
He is firm on proprietary software; he won't use anything that doesn't given him the freedoms he considers important. But the GPL is designed not merely to provide those freedoms, but to aid in getting those freedoms to other works. The issue with this kind of license is that it doesn't scale well to a large number of combined projects, but XFree86 is a self-contained project anyway. The main issue is that using such a license encourages other people to use a similar license for other sorts of project, which would cause problems in the long run.
As far as I can tell from studying linguistics and computer science, formal grammars are not particularly good ways of representing either sort of language. Chomsky's main founding point was that formal methods could be used to study natural language; he proposed context-free grammars, and then quickly abandoned them, because no language is actually context-free.
Computer language syntax picked up context-free grammars, because computer languages are generally context-free, at least to a certain extent. Of course, you can't actually implement an arbitrary context-free grammar efficiently, so they turned to a restricted subset which is sufficient for the important cases. Of course, the grammar is (as natural languages discovered millenia ago) insufficient for anything useful, so they developed interesting semantics behind the overly-strict grammar.
At this point, the interesting work in linguistics (which relies heavily on obscure languages to test the boundaries of what the human language faculty produces) is in the ways that language goes beyond what is feasible to define and use in an unambiguous way; this is stuff which is unsuitable for programming languages, because it is, by definition, impossible to interpret predictably. Compiler and computer language design has not informed linguistics significantly, because natural language uses an entirely different set of tricks for an entirely different set of goals.
The research in computer languages, on the other hand, is in bits of semantics which are entirely unlike any semantics used by natural languages, but are understandable by other faculties. It is focused on the formal representation of data structures and processes, two things that natural language is entirely inadaquate for and relies entirely on extra-linguistic methods (such as demonstration) to convey.
Consider, for example, the addition of a simple bit of natural language to a computer language. Say there were an "it" keyword, which referred to the most recently used variable which type checks in the context in which it is used, except that in the arguments of a method, it cannot refer to the object on which the method is called. Such a keyword would be practically impossible to use reasonably, since it would be extremely fragile and hard to interpret. However, such a keyword is present and its use is required in almost all natural languages. Natural language is really more like a machine language than a high-level programming language; the machine it is for is to be found about your left ear, and it has only been partially reverse engineered.
Actually, this would be an interesting question to pose to RMS: the problem is that this software, while Free by itself, doesn't play well with others. You can't repackage it in some ways and distribute the result. You may not be able to link a GPL program against the library and distribute the result.
But all of the freedoms the user is supposed to demand are available when considering the program in isolation. It seems to me that RMS would say that Mandrake is right not to offer packages of XFree86 4.4 due to the license. But I don't think RMS would have any problem with getting and using XFree86 4.4 yourself if you needed it. The only caveat is that, if you distribute something, it must work with 4.3's libraries, because 4.4 is not free for that sort of use.
The girlfriend can tell the difference, it's not the diamond that gives it away. She can't tell if the diamond was mined or manufactured, but the 30% discount is quite obvious.
And just in time for Microsoft to pay for another license...
I'm unclear on why a new C# compiler written in C# is necessary; wouldn't it be easier to add a C# front end and a CLR back end to gcc, which not only already has a good intermediate language, but already has a Java front end, which would be close already. This would also provide native executable output for a huge number of architectures.
I think the "free" argument should actually be an "unencumbered" argument. With open source licensed software, you can actually own your copy of the software, and use it without needing a license, much as you can read a book you own without needing a license. It is necessary to have a license in order to distribute, copy, or modify the software, but only IT needs this license (to install software for end users); the end users don't need licenses at all. If you are using exclusively open source software, you can legally have a 60000-person company using one license, so you have a vastly smaller license management problem. The benefit of "free" isn't that the price is zero; the benefit is that accounting isn't needed.
Technology also creates jobs doing things that wouldn't be possible without it, because tasks that wouldn't be possible before become possible, and people are then needed to perform those tasks. For any particular task, technology (ideally) makes it so you don't need so many people to do it, but that is outweighed by the "next step" tasks that become possible.
So technology generally creates jobs at about the same rate, overall, than it eliminates them. Which is why American cities aren't more overrun with unneeded, out-of-work, former workers than ancient Rome was.
Open source doesn't need marketing, because it doesn't matter all that much to open source whether end users use it. The point of open source is to solve your local problem better with the help of like-minded people.
On the other hand, companies selling solutions based on open source projects do need marketing, because they do actually have to get otherwise uninterested end users interested. Of course, they seem to be doing pretty well at just that.
Certainly if you want to be more than a faceless minion you can't be anonymous. But you can be anonymous and say something that comes to the attention of known people. If your point is good enough, it will stand on its own without your reputation to back it up, or it will be backed by the reputations of people who find and promote it.
Furthermore, there are ideas that nobody wants to officially back, but which people will bring up if they have been mentioned anonymously. People can say, without lending their support to the idea, that a good argument is needed to refute it, and this can lead to the idea being adopted if nobody can come up with a good response.
So there is value to anonymous comments in the democratic process, even if there is also more weight given to the concerns of known individuals. Furthermore, with well-protected pseudonyms (e.g., anonymous posts signed with a certificate with no name but a recognizable fingerprint), it is possible to identify look for comments from the same source, if the source wants to associate the comments with each other. And it should be possible to sign comments, if desired, if you want your identity to be known.
I think that would be great. I think that would give potential employers a much better idea of what a person is like, particularly because you'd probably have a mix of good and bad reviews, which would tell you what sort of situations the person does well in. (And I'd like my employers to see that, despite sleeping through all of my classes, I still learned the material.)
Unfortunately, having been a TA for a couple of classes, I know that no professor is actually going to have time to write that much detail about all of the students. Much more likely is that a professor would manage to review a few of the most noteworthy students ("She came to office hours every week for extra help on material she clearly understood perfectly." "He kept correcting my sign errors before I'd finished writing the equations."), and they'd forget entirely about the existence of Mr. Smith. Besides, it would be hard for Mr. Smith to not know what professor had given that review, unless he's taking multiple courses which teach square roots, so anonymity isn't very helpful.