"If Linus calls his kernel Linux and not gnuLinux then its called Linux."
RMS doesn't care what Linus names his kernel. RMS does care what name people use to describe the operating system that fundamentally matches the project he started twenty years ago.
It bothers me to use "Linux" as the name of a kernel *and* the name of a class of operating systems. That's just plain confusing.
Doesn't it also seem strange to name an operating system after its kernel, which was named after a single kernel developer? And when looking for a name for all Linux-based operating systems, it's striking that they all (AFAIK) are based on the GNU project. Nobody is bundling Linux with FreeBSD tools, are they?
Calling a large class of operating systems "Linux" just seems strange to me, since Linus had very little to do with any of them. He wrote a kernel, and (with the "help" of a friend) name it Linux. And that's the only part of operating systems that Torvald's really seems to care much about (it's not even clear that he cares about every subsystem in "his" kernel). So why call any complete operating system Linux? As far as I know, Linus does not have his own distrobution.
Since things have only gotten worse since fvwm2, I don't know why a productive person would waste much time on switching to "modern" application environments. I often conteplate going back to fvwm2, because 1) it worked, 2) had an understandable and reliable config system, 3) does everything I need.
Can anyone suggest to me what Sawfish or Metacity offers that fvwm2 doesn't? I'm sure there has to be many such things, but I can't think of any right now. What about useful things?
I think people are underestimating what Emacs can do. It had it's own "windowing system" before X existed. It worked then, and it worked now. For writing email and code, Emacs is a great environment with or without X. Within his domain of interest, I bet RMS has a very good sense for what is good and what is bad.
On the other hand, the video games that come with Emacs aren't nearly as cool-looking as those you can run under X. But then, I'll bet RMS doesn't play too many video games.
And talking about the "Open Source bits and pieces" probably won't attract RMS. He has strongly-held convictions about Free Software, and just isn't very interested in Open Source as a category.
Did you know that most accidents happen w/in 5 miles of home? OTOH, how much driving occurs w/in 5 miles of home? Until you answer the second, the first is meaningless.
72 helicopter accidents is total disaster if there were 72 helicopter flights. If there were 150,000 helicopter flights per day, then 72 might be a good number. But then, we should probably decide whether time-in-flight is important, or distance covered, etc.
FWIW, I agree that cruise missiles are probably a bad idea for traffic info. The efficiency seems poor. Blimps would do better. And in most places you could just put fixed cameras or sensors. Isn't this what traffic.com is doing?
Humans have really bad inuition when it comes to math and statistics. For a fun example, look up the Monty Hall problem.
It's gonna move a whole lot faster, and I'd guess straighter/mile, than a bird. I'd think you could pick it out if you had a regular blip. I'm guessing, of course.
"the notion of a file system should be hidden from a user"
I disagree, very strongly. Windows has tried this, and the result is that their users are (collectively) stupider than ever. If you have a file system, expose it and get people to spend the time it takes to learn about it. Then the users are smart and powerful, and don't need tech support.
This applies for any type of file system, including fancy relational extensions. Heck, I think it goes for just about everything in computing. If you want simplicity, make/buy appliances. If you want a *computer*, make people learn how to use it.
Computers are only "too complicated" for people who don't actually need a computer.
Are you sure that Windows never had a Presentation Manager? I'm glad you knew what I was getting at. But I thought Windows (maybe 3.0, 3.1, or 3.11) had a pm process running somewhere.
I played with Xmouse when I was using Windows, and I agree with your assesment. My wife has messed with various virtual-screen switchers in Windows, and all of them seem to have problems with some "too clever" apps, just like Xmouse. I've seen a lot of folks at my university play with various X "servers" (well, mostly Hummingbird) for Windows, but that doesn't seem worth the effort in the long run given the troubles they've had.
I was once told that a productive user accepts the default settings. Messing with Windows to make it like UNIX, and vice-versa, seems a silly waste of time. Especially since you can have GNU/Linux for free on a virtually zero-cost PC.
In my experience, GUIs apply fundamental restrictions to data manipulation due to the graphically-dominated input and output systems. There are lots of neat programs like SAS for doing data manipulation through a gui, but every worthwhile tool I've ever used on Windows also offered a CLI. When I've encountered experts with these applications, they've always preferred the CLI Perhaps they prefer the CLI because that's what the learned with, but I expect it is because it works better for them. Matlab, SAS, Mathematica, and IDL are some examples.
Once you are working with a CLI, things are easier if the other apps and the OS are dominated by CLI mentality, in my opinion. For instance, cutting and pasting between text apps is more difficult in the Windows Presentation Manger (or Explorer, or whatever it is now) than in X. Focus issues in Windows (can these be changed yet?) don't allow separation of focus and stacking order (focused window is always on top). Furthermore, CLI apps generally use stdin and stdout, while GUI apps don't really have the same standarized I/O mechanisms. Anyone who has written GUI apps in Windows knows has asked themselves at least one "so where in heck does fprintf(stdout,"foo") show up?".
Standard CLI tools on a GNU-ish system (which means most UNIX-ish installations, since GNU text and file tools are nearly standard) include many apps for parseing and rebuilding text files and streams. There are many regex tools available by default, whether CLI, programming languages, or GUIs. You can get a lot of work done with just bash+grep. Furthermore, modern UNIX shells have a mostly-reasonable programming language.
None of these things are part of Windows as MS configures it. Even the standard libs that come with Windows don't support regex (I guess the new MSDEV.NET has some sort of regex support now). Heck, Windows doesn't even come with programming support of any kind, except for pathetic batch files and, er, is there anything else with default Windows? My point is that MS built Windows with saleability in mind, and makes it attractive to people who are in charge of purchasing software. Contrast this with UNIX, which was created without permission on company time by researchers for themselves. UNIX is meant for research, and that is why an arbitrary precision stack-based RPN CLI calculator (dc) has been part of UNIX distributions since the mid seventies (just guessing).
These are the reasons that I believe support my opinion. Graphical communication systems are a pain the butt, which is why we ditched heiroglyphics thousands of years ago. Graphical Interfaces may be useful for communicating with Chimpanzees, but they are a lousy way to data manipuulation tasks in detail.
Maybe you can answer a Fed Ex tracking question for me. =-) The online Fed Ex tracking info often has high latency (long time between event, and posting of event on website) until the local delivery area is reached. I've never cared much, since it's only the local delivery that involves me directly.
All the same, do you have any idea where this latency comes from? Sometimes it is more than 12 hours.
There seems to be major differences on UPS and Fed Ex at various geographic locations. For example, every computer I've shipped to Pittsburgh has suffered case damage. One had a sheared corner strut (that bit of angle-iron, except it's not iron =-).
On the other hand, Fed Ex has not damaged a single shipment. And this is despite the fact that probably 75% of my computer shipements (including cases) are Fed Ex.
I've also watched the UPS tracking system track my packages into and right back out of Pittsburgh, and had friends see the same. Then you have to wait for the package to get turned around and head back to Pittsburgh. In contrast, I once had Fed Ex forget to put my package on a truck for local delivery, and they gave me a choice to pick it up myself the same day. In fact, the Fed Ex office here is open later than the truck drivers work, so I can always pick up my package the same day if I'm not available for a signature. The UPS office here won't do that for me.
The long and short of my point is that UPS sucks in Pittsburgh. I don't care to claim more than that, unless we start talking international shipping (in which case DHL is the champion). However, I've heard a few stories that went the opposite way (except for international shipping).
-Paul Komarek
Re:Who's that Copernicus guy, anyway?
on
Time to Face the Music
·
· Score: 2, Insightful
I'm glad I'm not the only one offended (or at least surprised) by this tasteless comparison. The RIAA can fall of the earth without any real loss of music , history, or works of human creativity. Heck, we might even find a way to recover works that RIAA publishers don't publish anymore.
That is, the RIAA just is *NOT* that important in the whole scheme of things. We had music before the RIAA, we'll have music after. And we can hope the music will get better without so much corporate "assistance".
On the other hand, burning ancient manuscripts and destruction of some of the oldest known examples of writing in Western civilization cannot be undone. I suppose we won't see the stolen artifacts for some time, either.
The comparison of music piracy (by which I mean large scale copying and redistribution in exchange for money) is a trivial, meaningless problem compared to the looting happening in Baghdad. Furthermore, informal and not-for-profit copying of songs though the interent is on an even lower tier of importance.
Actually, I had both. Thanks for the clarification, since I screwed up the title. "Programming Python" is indeed the book I was referring to by Lutz. I remember little about "Learning Python", and didn't want to say anything about it. It was my first Python book, and I guess I learned a fair bit from it. Hard to say at this point.
I cannot agree with the reviewer that M. Lutz's O'Reilly "Python" book is good. I disliked this book nearly all the way through. It jumped around too much, used too many words, and had insufficient detail on more advanced topics. Given that the book is about three inches thick (from memory -- I gave away my copy), there's enough room for details on everything.
I concur with other posters that the "Essential Reference" (white and red from New Riders, written by D. Beazley who did SWIG) is an awesome book. It is concise, making it a good reference. I wouldn't think it was a good book for those who have never programmed. If you have some programming experience, though, I expect you'll appreciate this book.
The "Quick Python" book from Manning is nice, too. This was my wife's preferred book for learning Python. I've looked through it a bit, and it seems decently concise but with more explanation than "Essential Reference". I used its section on extending Python through C, and found it very useful. That section didn't have everything bit of reference that I needed for conversion specifiers, but their examples were dead-on what I needed to get started.
I recently finished reading through "How to Think Like a Computer Scientist: Learning With Python" from Green Tea Press. This book is not a complete reference or guide for Python, nor will it be particularly 'useful' for people who have taken university-level programming and data structures classes. However, it seems to be an AWESOME book for people who don't yet program, or whose only experience is web programming or VB or Perl programming (I'm not saying those things are bad, but very often they don't encourage reasonable programming discipline and methodology). I write "it seems" only because I'm not a beginner or an instructor for intro to programming courses.
This book ("How to Think...") is aimed at classroom use, so it doesn't include info about installing Python or starting the Python interpreter under Windows, etc. It does preach solid computer science. Parts of their approach seemed a bit unusual to me, compared to my more classical training, but after a few gripes I always was forced to conclude that their approach was valid and as concise and clear as it could be. The authors are aware of the book being used in high schools and community colleges. I expect that mature students, or any adult intrested in learning proper programming, would benefit from starting with this book.
I disagree that nobody has ever put remote display tech and 802.11b together before. I have a tablet next to me (from frontpath/progear) running a GNU/Linux variant with rdp and wireless and sound and a touchscreen and usb. It lasts as long on battery, and probably weighs less than 2.6 kilos (I haven't weighed it, and don't recall). Anyway, I've had this thing for many months, probably half a year.
Furthermore, there are lots of other devices that have done this even longer. My iPaq (running a variant of Debian, more or less) does all of these things, weighs less than a pound (guessing), and is plenty bright (you can use it as a flashlight or reading lamp). I've had it for over a year, and it's a consumer device.
So the v150 may be new, but the idea isn't novel. Heck, the first thing people seemed to do with iPaq's running GNU/Linux was start up remote X apps and VNC. And debugging graphical apps on the iPaq is a heck of a lot easier with the ability to push to it's display, or pull from it.
Incidently, I am thinking of using the tablet as a remote control for a trunk-mounted ogg player in my car, similar to what you were suggesting. =-)
Measure theory is relevant here. The issue is that typical probability measures do not account for "small" events. Those events are called "events of zero measure".
Suppose you are measuring 2D area of the set of points in the disc with center (0,0) and radius 1. Suppose now that we remove the center point (0,0), and measure the area again. Should the area change? In general, we probably want the area to be the same either way. That is, we expect that the set { (0,0) } has zero 2D area.
Now, instead of saying area, let's use the term "measure". Probabilities, at least in some contexts, are measures of (usually) infinite sets. Therefore probabilities (in these contexts) are insensitive to finite differences between two infinite sets. Saying an event has a probability of 1.0 (in these contexts) means that the event has the same measure as the set of all possibilities. It does not mean that the event is identically the same as the set of all possibilities.
The result is that an event with probability 1 might not "occur", and an event with probability 0 might "occur". There is no contradiction here, just some unintuitive definitions.
I'm getting a bit sloppy with my words, and should probably quit. Just keep in mind that the last 200 years of mathematics have been dominated by examples of inadequate and failed human intuition. Words like probability, event, and measure probably don't mean what one would guess they mean, because these concepts turned out to be much more complicated than we imagined.
The difficulty of computing surface area is a great example of why we need measure theory. Maybe there's a good webpage out there on the surface area and the "crinkled cylinder" (or some name like that). Also interesting are 1D lines in 2D space which are soooo long that they have 2D area. =-)
Heck, old beetle's have great acceleration without any mods. They'll beat anyone across the near-side crosswalk. =-)
It almost made me sad to see that picture. I really loved my '72 beetle. It crossed Snoqualmie Pass (not huge, but big enough) on 3 cylinders -- maybe two, but that seems unrealistic. When I took the distributor cap and spark plug wires off to see what was the matter, one wire burst into a cloud of blue powder and another was coated with white.
My '72 beetle always got me where I was going, at which point I had to spend a day fixing it. =-)
That's a fine idea. I might just start doing that. I'm often very parnoid about accidentally hitting Enter too soon, especially since I use multiple keyboard layouts.
You just might have helped at least one person. =-)
It's pretty clear that PC power supplies are all about cheap and not at all about quality. Linear supplies were used for a while, but that was a *long* time ago.
You may care to read through this article for what appears to be a careful and reasoned assessment of Al Gore's lifetime activities with respect to technology and nationwide data networks.
You might also want to read Vinton Cerf's email about Al Gore and the internet, which was prepared by Cerf and Kahn. Here's a quote: "Bob and I believe that the vice president deserves significant credit for his early recognition of the importance of what has become the Internet." This may not answer your question directly, but provides useful information all the same.
I'm not a Gore fan. These links come from a very quick search on google. I'm guessing that anyone not educated on this subject must not care very much.
Ah, so you're a Bayesian. I should have realized this sooner.;-) Just kidding.
There are some cross-overs between your descriptions of the comp sci camp and the math/stats camp. My advisor, Andrew Moore, is one of them. The support for interdisciplinary work at Carnegie Mellon is the primary reason I came here.
Your comment about understanding or emulating intelligence reminded me of a funny quote from Andrew. I'd asked him if he had picked up any algorithm ideas from watching his son grow up. His reply:
"I try not to be biologically inspired. We can either write algorithms that work the way the human brain works, or we can write algorithms that work the way the brain *should* have worked."
Not only is it unclear whether neural nets have provided any insight into human thought, but it's unclear whether we should even care about human thought. Our brains are neat, but they might not be the best role model for building software.
I agree about the uselessness of the name "AI". However, I must admit that "Artificial Intelligence" sounds a whole lot better than "The study of how to keep robots from bumping into things." (FWIW, my work has very little to do with robotics, but robots are cool so I use them in examples)
I am a doctoral student working under the broad umbrella of AI. Because my background is mathematics, I've struggled for years to figure out just what these computer scientists are talking about. My conclusion is that AI means just about anything the speaker wishes.
One rather crude saying claims that "AI can be broken into two part, statistics and bullshit". I don't care for this assessment, but it makes it clear that even rather pedestrian stuff like decision trees and clever application of Bayes' Rule are considered part of AI by some people.
In fact, my work (and much of my lab's work) is all about making traditional statistics go fast. When you want a computer to (help) make decisions based on data, statistical methods provide a reasonable way to inform judgement.
My personal take on AI is that it covers just about anything which can be used to keep robots from stupidly running into walls.
Now, that said, my problems with characterizing MASSIVE as "AI on steroids" are:
1) The techniques used in MASSIVE are, at best, a subset of AI techniques. AI by most definitions is not a single entity that can be put on steroids.
2) Within the field I work in (call it computational statistics if you like), we use datasets with "hundreds and thousands" of data points as test cases for debugging. Our real data has hundreds *of* thousands of datapoints, each having hundreds of thousands of attributes. So in what way is a coarse simulation of an army 10,000 strong "massive"?
3) The goal of MASSIVE is simulation. If we're going to label stuff "AI", I'd prefer that the goal included "real-world", uncontrolled environments. Simulations can be useful when working on "real" tasks, but the simulation should not be the stopping point of any "AI" endeavor. I'd like to see that MASSIVE army of 10,000 deal with bad roads, monsoons, disease, insect migrations, faulty equipment, incorrect maps, politically-motivated leaders, etc. Dealing with noisy or otherwise incorrect and incomplete data is what separates AI from computational geometry, numerical analysis, and operations research.
In the end, MASSIVE is a simulation system lacking real-world data input, real-world interaction with uncontrolled environments, and real-world outcomes. To me, that's not AI.
However, I'm happy to compliment MASSIVE as a darn good simulation w/r/t appearance on-screen. Writing good simulations requires a good programmer. Sometimes, getting a simulation to resemble anything even remotely real is just as hard as building software that deals with the real world.
Slashdot Mirror
"If Linus calls his kernel Linux and not gnuLinux then its called Linux."
RMS doesn't care what Linus names his kernel. RMS does care what name people use to describe the operating system that fundamentally matches the project he started twenty years ago.
It bothers me to use "Linux" as the name of a kernel *and* the name of a class of operating systems. That's just plain confusing.
Doesn't it also seem strange to name an operating system after its kernel, which was named after a single kernel developer? And when looking for a name for all Linux-based operating systems, it's striking that they all (AFAIK) are based on the GNU project. Nobody is bundling Linux with FreeBSD tools, are they?
Calling a large class of operating systems "Linux" just seems strange to me, since Linus had very little to do with any of them. He wrote a kernel, and (with the "help" of a friend) name it Linux. And that's the only part of operating systems that Torvald's really seems to care much about (it's not even clear that he cares about every subsystem in "his" kernel). So why call any complete operating system Linux? As far as I know, Linus does not have his own distrobution.
-Paul Komarek
Since things have only gotten worse since fvwm2, I don't know why a productive person would waste much time on switching to "modern" application environments. I often conteplate going back to fvwm2, because 1) it worked, 2) had an understandable and reliable config system, 3) does everything I need.
Can anyone suggest to me what Sawfish or Metacity offers that fvwm2 doesn't? I'm sure there has to be many such things, but I can't think of any right now. What about useful things?
-Paul Komarek
I think people are underestimating what Emacs can do. It had it's own "windowing system" before X existed. It worked then, and it worked now. For writing email and code, Emacs is a great environment with or without X. Within his domain of interest, I bet RMS has a very good sense for what is good and what is bad.
On the other hand, the video games that come with Emacs aren't nearly as cool-looking as those you can run under X. But then, I'll bet RMS doesn't play too many video games.
And talking about the "Open Source bits and pieces" probably won't attract RMS. He has strongly-held convictions about Free Software, and just isn't very interested in Open Source as a category.
-Paul Komarek
Then again, if the country's defenses are fairly week, perhaps he's planning to take it over himself! Mwa ha ha ha!
-Paul Komarek
Did you know that most accidents happen w/in 5 miles of home? OTOH, how much driving occurs w/in 5 miles of home? Until you answer the second, the first is meaningless.
72 helicopter accidents is total disaster if there were 72 helicopter flights. If there were 150,000 helicopter flights per day, then 72 might be a good number. But then, we should probably decide whether time-in-flight is important, or distance covered, etc.
FWIW, I agree that cruise missiles are probably a bad idea for traffic info. The efficiency seems poor. Blimps would do better. And in most places you could just put fixed cameras or sensors. Isn't this what traffic.com is doing?
Humans have really bad inuition when it comes to math and statistics. For a fun example, look up the Monty Hall problem.
-Paul Komarek
It's gonna move a whole lot faster, and I'd guess straighter/mile, than a bird. I'd think you could pick it out if you had a regular blip. I'm guessing, of course.
-Paul Komarek
"the notion of a file system should be hidden from a user"
I disagree, very strongly. Windows has tried this, and the result is that their users are (collectively) stupider than ever. If you have a file system, expose it and get people to spend the time it takes to learn about it. Then the users are smart and powerful, and don't need tech support.
This applies for any type of file system, including fancy relational extensions. Heck, I think it goes for just about everything in computing. If you want simplicity, make/buy appliances. If you want a *computer*, make people learn how to use it.
Computers are only "too complicated" for people who don't actually need a computer.
-Paul Komarek
Try cooking oil. It worked for me, and it's safe enough to eat. Furthermore, it doesn't dry up your hands or have inhalation warnings.
-Paul Komarek
Are you sure that Windows never had a Presentation Manager? I'm glad you knew what I was getting at. But I thought Windows (maybe 3.0, 3.1, or 3.11) had a pm process running somewhere.
I played with Xmouse when I was using Windows, and I agree with your assesment. My wife has messed with various virtual-screen switchers in Windows, and all of them seem to have problems with some "too clever" apps, just like Xmouse. I've seen a lot of folks at my university play with various X "servers" (well, mostly Hummingbird) for Windows, but that doesn't seem worth the effort in the long run given the troubles they've had.
I was once told that a productive user accepts the default settings. Messing with Windows to make it like UNIX, and vice-versa, seems a silly waste of time. Especially since you can have GNU/Linux for free on a virtually zero-cost PC.
-Paul Komarek
In my experience, GUIs apply fundamental restrictions to data manipulation due to the graphically-dominated input and output systems. There are lots of neat programs like SAS for doing data manipulation through a gui, but every worthwhile tool I've ever used on Windows also offered a CLI. When I've encountered experts with these applications, they've always preferred the CLI Perhaps they prefer the CLI because that's what the learned with, but I expect it is because it works better for them. Matlab, SAS, Mathematica, and IDL are some examples.
.NET has some sort of regex support now). Heck, Windows doesn't even come with programming support of any kind, except for pathetic batch files and, er, is there anything else with default Windows? My point is that MS built Windows with saleability in mind, and makes it attractive to people who are in charge of purchasing software. Contrast this with UNIX, which was created without permission on company time by researchers for themselves. UNIX is meant for research, and that is why an arbitrary precision stack-based RPN CLI calculator (dc) has been part of UNIX distributions since the mid seventies (just guessing).
Once you are working with a CLI, things are easier if the other apps and the OS are dominated by CLI mentality, in my opinion. For instance, cutting and pasting between text apps is more difficult in the Windows Presentation Manger (or Explorer, or whatever it is now) than in X. Focus issues in Windows (can these be changed yet?) don't allow separation of focus and stacking order (focused window is always on top). Furthermore, CLI apps generally use stdin and stdout, while GUI apps don't really have the same standarized I/O mechanisms. Anyone who has written GUI apps in Windows knows has asked themselves at least one "so where in heck does fprintf(stdout,"foo") show up?".
Standard CLI tools on a GNU-ish system (which means most UNIX-ish installations, since GNU text and file tools are nearly standard) include many apps for parseing and rebuilding text files and streams. There are many regex tools available by default, whether CLI, programming languages, or GUIs. You can get a lot of work done with just bash+grep. Furthermore, modern UNIX shells have a mostly-reasonable programming language.
None of these things are part of Windows as MS configures it. Even the standard libs that come with Windows don't support regex (I guess the new MSDEV
These are the reasons that I believe support my opinion. Graphical communication systems are a pain the butt, which is why we ditched heiroglyphics thousands of years ago. Graphical Interfaces may be useful for communicating with Chimpanzees, but they are a lousy way to data manipuulation tasks in detail.
-Paul Komarek
You got to choose the license for your thesis? Doesn't your university own your thesis?
-Paul Komarek
Maybe you can answer a Fed Ex tracking question for me. =-) The online Fed Ex tracking info often has high latency (long time between event, and posting of event on website) until the local delivery area is reached. I've never cared much, since it's only the local delivery that involves me directly.
All the same, do you have any idea where this latency comes from? Sometimes it is more than 12 hours.
-Paul Komarek
There seems to be major differences on UPS and Fed Ex at various geographic locations. For example, every computer I've shipped to Pittsburgh has suffered case damage. One had a sheared corner strut (that bit of angle-iron, except it's not iron =-).
On the other hand, Fed Ex has not damaged a single shipment. And this is despite the fact that probably 75% of my computer shipements (including cases) are Fed Ex.
I've also watched the UPS tracking system track my packages into and right back out of Pittsburgh, and had friends see the same. Then you have to wait for the package to get turned around and head back to Pittsburgh. In contrast, I once had Fed Ex forget to put my package on a truck for local delivery, and they gave me a choice to pick it up myself the same day. In fact, the Fed Ex office here is open later than the truck drivers work, so I can always pick up my package the same day if I'm not available for a signature. The UPS office here won't do that for me.
The long and short of my point is that UPS sucks in Pittsburgh. I don't care to claim more than that, unless we start talking international shipping (in which case DHL is the champion). However, I've heard a few stories that went the opposite way (except for international shipping).
-Paul Komarek
I'm glad I'm not the only one offended (or at least surprised) by this tasteless comparison. The RIAA can fall of the earth without any real loss of music , history, or works of human creativity. Heck, we might even find a way to recover works that RIAA publishers don't publish anymore.
That is, the RIAA just is *NOT* that important in the whole scheme of things. We had music before the RIAA, we'll have music after. And we can hope the music will get better without so much corporate "assistance".
On the other hand, burning ancient manuscripts and destruction of some of the oldest known examples of writing in Western civilization cannot be undone. I suppose we won't see the stolen artifacts for some time, either.
The comparison of music piracy (by which I mean large scale copying and redistribution in exchange for money) is a trivial, meaningless problem compared to the looting happening in Baghdad. Furthermore, informal and not-for-profit copying of songs though the interent is on an even lower tier of importance.
Grrr.
-Paul Komarek
Actually, I had both. Thanks for the clarification, since I screwed up the title. "Programming Python" is indeed the book I was referring to by Lutz. I remember little about "Learning Python", and didn't want to say anything about it. It was my first Python book, and I guess I learned a fair bit from it. Hard to say at this point.
-Paul
I cannot agree with the reviewer that M. Lutz's O'Reilly "Python" book is good. I disliked this book nearly all the way through. It jumped around too much, used too many words, and had insufficient detail on more advanced topics. Given that the book is about three inches thick (from memory -- I gave away my copy), there's enough room for details on everything.
I concur with other posters that the "Essential Reference" (white and red from New Riders, written by D. Beazley who did SWIG) is an awesome book. It is concise, making it a good reference. I wouldn't think it was a good book for those who have never programmed. If you have some programming experience, though, I expect you'll appreciate this book.
The "Quick Python" book from Manning is nice, too. This was my wife's preferred book for learning Python. I've looked through it a bit, and it seems decently concise but with more explanation than "Essential Reference". I used its section on extending Python through C, and found it very useful. That section didn't have everything bit of reference that I needed for conversion specifiers, but their examples were dead-on what I needed to get started.
I recently finished reading through "How to Think Like a Computer Scientist: Learning With Python" from Green Tea Press. This book is not a complete reference or guide for Python, nor will it be particularly 'useful' for people who have taken university-level programming and data structures classes. However, it seems to be an AWESOME book for people who don't yet program, or whose only experience is web programming or VB or Perl programming (I'm not saying those things are bad, but very often they don't encourage reasonable programming discipline and methodology). I write "it seems" only because I'm not a beginner or an instructor for intro to programming courses.
This book ("How to Think...") is aimed at classroom use, so it doesn't include info about installing Python or starting the Python interpreter under Windows, etc. It does preach solid computer science. Parts of their approach seemed a bit unusual to me, compared to my more classical training, but after a few gripes I always was forced to conclude that their approach was valid and as concise and clear as it could be. The authors are aware of the book being used in high schools and community colleges. I expect that mature students, or any adult intrested in learning proper programming, would benefit from starting with this book.
-Paul Komarek
I disagree that nobody has ever put remote display tech and 802.11b together before. I have a tablet next to me (from frontpath/progear) running a GNU/Linux variant with rdp and wireless and sound and a touchscreen and usb. It lasts as long on battery, and probably weighs less than 2.6 kilos (I haven't weighed it, and don't recall). Anyway, I've had this thing for many months, probably half a year.
Furthermore, there are lots of other devices that have done this even longer. My iPaq (running a variant of Debian, more or less) does all of these things, weighs less than a pound (guessing), and is plenty bright (you can use it as a flashlight or reading lamp). I've had it for over a year, and it's a consumer device.
So the v150 may be new, but the idea isn't novel. Heck, the first thing people seemed to do with iPaq's running GNU/Linux was start up remote X apps and VNC. And debugging graphical apps on the iPaq is a heck of a lot easier with the ability to push to it's display, or pull from it.
Incidently, I am thinking of using the tablet as a remote control for a trunk-mounted ogg player in my car, similar to what you were suggesting. =-)
-Paul Komarek
Measure theory is relevant here. The issue is that typical probability measures do not account for "small" events. Those events are called "events of zero measure".
Suppose you are measuring 2D area of the set of points in the disc with center (0,0) and radius 1. Suppose now that we remove the center point (0,0), and measure the area again. Should the area change? In general, we probably want the area to be the same either way. That is, we expect that the set { (0,0) } has zero 2D area.
Now, instead of saying area, let's use the term "measure". Probabilities, at least in some contexts, are measures of (usually) infinite sets. Therefore probabilities (in these contexts) are insensitive to finite differences between two infinite sets. Saying an event has a probability of 1.0 (in these contexts) means that the event has the same measure as the set of all possibilities. It does not mean that the event is identically the same as the set of all possibilities.
The result is that an event with probability 1 might not "occur", and an event with probability 0 might "occur". There is no contradiction here, just some unintuitive definitions.
I'm getting a bit sloppy with my words, and should probably quit. Just keep in mind that the last 200 years of mathematics have been dominated by examples of inadequate and failed human intuition. Words like probability, event, and measure probably don't mean what one would guess they mean, because these concepts turned out to be much more complicated than we imagined.
The difficulty of computing surface area is a great example of why we need measure theory. Maybe there's a good webpage out there on the surface area and the "crinkled cylinder" (or some name like that). Also interesting are 1D lines in 2D space which are soooo long that they have 2D area. =-)
-Paul Komarek
"It has always sucked."
;-)
/., back when Rob was learning to make mugs in ceramics class.
That's simply not true, Mr. uid>10,000.
I remember when you could get good super-computing advice on
-Paul Komarek
Heck, old beetle's have great acceleration without any mods. They'll beat anyone across the near-side crosswalk. =-)
It almost made me sad to see that picture. I really loved my '72 beetle. It crossed Snoqualmie Pass (not huge, but big enough) on 3 cylinders -- maybe two, but that seems unrealistic. When I took the distributor cap and spark plug wires off to see what was the matter, one wire burst into a cloud of blue powder and another was coated with white.
My '72 beetle always got me where I was going, at which point I had to spend a day fixing it. =-)
-Paul Komarek
That's a fine idea. I might just start doing that. I'm often very parnoid about accidentally hitting Enter too soon, especially since I use multiple keyboard layouts.
You just might have helped at least one person. =-)
-Paul Komarek
It's pretty clear that PC power supplies are all about cheap and not at all about quality. Linear supplies were used for a while, but that was a *long* time ago.
-Paul Komarek
You may care to read through this article for what appears to be a careful and reasoned assessment of Al Gore's lifetime activities with respect to technology and nationwide data networks.
You might also want to read Vinton Cerf's email about Al Gore and the internet, which was prepared by Cerf and Kahn. Here's a quote: "Bob and I believe that the vice president deserves significant credit for his early recognition of the importance of what has become the Internet." This may not answer your question directly, but provides useful information all the same.
I'm not a Gore fan. These links come from a very quick search on google. I'm guessing that anyone not educated on this subject must not care very much.
-Paul Komarek
Ah, so you're a Bayesian. I should have realized this sooner. ;-) Just kidding.
There are some cross-overs between your descriptions of the comp sci camp and the math/stats camp. My advisor, Andrew Moore, is one of them. The support for interdisciplinary work at Carnegie Mellon is the primary reason I came here.
Your comment about understanding or emulating intelligence reminded me of a funny quote from Andrew. I'd asked him if he had picked up any algorithm ideas from watching his son grow up. His reply:
"I try not to be biologically inspired. We can either write algorithms that work the way the human brain works, or we can write algorithms that work the way the brain *should* have worked."
Not only is it unclear whether neural nets have provided any insight into human thought, but it's unclear whether we should even care about human thought. Our brains are neat, but they might not be the best role model for building software.
I agree about the uselessness of the name "AI".
However, I must admit that "Artificial Intelligence" sounds a whole lot better than "The study of how to keep robots from bumping into things." (FWIW, my work has very little to do with robotics, but robots are cool so I use them in examples)
-Paul Komarek
I am a doctoral student working under the broad umbrella of AI. Because my background is mathematics, I've struggled for years to figure out just what these computer scientists are talking about. My conclusion is that AI means just about anything the speaker wishes.
One rather crude saying claims that "AI can be broken into two part, statistics and bullshit". I don't care for this assessment, but it makes it clear that even rather pedestrian stuff like decision trees and clever application of Bayes' Rule are considered part of AI by some people.
In fact, my work (and much of my lab's work) is all about making traditional statistics go fast. When you want a computer to (help) make decisions based on data, statistical methods provide a reasonable way to inform judgement.
My personal take on AI is that it covers just about anything which can be used to keep robots from stupidly running into walls.
Now, that said, my problems with characterizing MASSIVE as "AI on steroids" are:
1) The techniques used in MASSIVE are, at best, a subset of AI techniques. AI by most definitions is not a single entity that can be put on steroids.
2) Within the field I work in (call it computational statistics if you like), we use datasets with "hundreds and thousands" of data points as test cases for debugging. Our real data has hundreds *of* thousands of datapoints, each having hundreds of thousands of attributes. So in what way is a coarse simulation of an army 10,000 strong "massive"?
3) The goal of MASSIVE is simulation. If we're going to label stuff "AI", I'd prefer that the goal included "real-world", uncontrolled environments. Simulations can be useful when working on "real" tasks, but the simulation should not be the stopping point of any "AI" endeavor. I'd like to see that MASSIVE army of 10,000 deal with bad roads, monsoons, disease, insect migrations, faulty equipment, incorrect maps, politically-motivated leaders, etc. Dealing with noisy or otherwise incorrect and incomplete data is what separates AI from computational geometry, numerical analysis, and operations research.
In the end, MASSIVE is a simulation system lacking real-world data input, real-world interaction with uncontrolled environments, and real-world outcomes. To me, that's not AI.
However, I'm happy to compliment MASSIVE as a darn good simulation w/r/t appearance on-screen. Writing good simulations requires a good programmer. Sometimes, getting a simulation to resemble anything even remotely real is just as hard as building software that deals with the real world.
-Paul Komarek