I am very much a utilitarian myself. While Stallman's arguments do implicitly assert a premise like the one you describe, I don't know if it should really be counted as _utilitarian_ per se. Traditional utilitarianism maximizes happiness, not freedom.
Also, in Mill's utilitarianism, mathematical reasoning very much applies. The goal of his philosophy was to find a way to mathematically determine the morality of behavior.
Finally, in any case, my argument only rested on propositional calculus and premises which I thought Bongo Bill stated. I used other parts of math as examples, but they weren't the basis for what I was saying.
Well, first, what you're saying is irrelevant. You are asserting that a free market for that which is currently protected by copyrights, trade secrets, and patents is a bad idea. I was responding to an incorrect statement which asserted that the market for music was a free market. Whether free markets are good or bad in a specific instance is irrelevant to what they are.
Beyond that, what you're saying is mostly correct. There wouldn't be "no" incentive to innovate, but there would be substantially less of an incentive to innovate. Copyrights and patents create an incentive to innovate through the creation of monopolies on innovations. These monopolies impose their own inefficiencies. If you believe that copyrights and patents are good for society, you must believe that there is no alternative to them that solves the incentives problem with greater efficiency.
I think that subsidies are a better way. We already subsidize that which is protected by patents through DARPA, the NSF, and other government funded agencies. We could feasibly get rid of patents and dramatically increase funding for these agencies to compensate. In my proposal, the inefficiency of a monopoly is replaced with the inefficiency of extra taxes. If we have a reasonably efficient tax system, I think this will easily be a net win over private monopolists.
With regard to that which is copyrighted, we could do subsidize in a similar way. Get rid of copyrights, but dramatically increase funding for the National Endowment for the Arts. As long as additional taxation is less inefficient than the inefficiency of private monopolies, this is a net win.
Now, if we as a country were to do this, I'd recommend doing it gradually. Decrease copyright and patent terms over the course of 10 years while increasing government subsidies to research and innovation. There's not a snowball's chance this will actually be done anyway, though.
> Addition is an extremely sound idea. No matter how far you carry it, it still works. I would not, however, recommend it as a good way to determine the circumference of a circle, when multiplication is so much easier in that particular circumstance.
The point is that addition still works, because addition is sound. If you think free software is a sound idea, then you must accept everything that soundness implies. If you think it implies an opposition to DRM, then you must oppose DRM. If you think free software is sound, it implies opposition to DRM, and you do not oppose DRM, then you have a contradiction on your hands.
You seem to be trying to get out of this by saying, "basically sound" rather than "sound" or "not sound". You can't do that. Operations are sound or not sound. If you think a rule is sound, except not in this case, and you have to do fudge it a little here, etc., then you really think the rule is not sound. You may still want to use this unsound rule if it works most of the time, but you have to be honest about it, and if you want to say anything useful, you must go back with a sound rule to check your work. A good example of this in math is naive set theory.
I'm using your concepts here (operations and their soundness), but I think it's better to consider propositions and their truth values. The concept of free software isn't really isn't really a sound or unsound operation; it's a set of propositions that are asserted true. Do you agree?
It's not a free market because the government creates artificial barriers to entry through copyright laws. A free market for CD's would be one in which it's legal for you to rip a CD you've purchased 1000 times and sell the copies for whatever you could get.
A free market probably wouldn't work too well for CD's, though, because of the huge spillover benefit the original seller (i.e., the artist) confers on the rest of the market. Most of the time, spillover benefits are corrected through government subsidies. In this case, the government decided to create copyright law instead.
> Open Source zealots are no more credible than any other type of zealot. It's the extension of a basically sound idea to an unrealistic, harmful, and (in the worst cases) counterintuitive extreme.
Something is not true unless it is true in all cases. Thus, if you believe something is true, you must believe it is true at its "extremes". Furthermore, everything which is intuitive is not right, so it doesn't make sense to say that counterintuitive "extremes" are "the worst cases".
The natural numbers and the rationals have the same cardinality. This is counterintuitive, but, according to the "extremes" of math, it is correct. If you accept the axioms of math, you must accept that the natural numbers and the rationals are the same size. You probably won't dispute this, but you're disputing something just like it. You want to be able to take "a basically sound idea" and deny its extremes. That is a contradiction, and you won't make any sense if you do that.
Why don't you go away, purge your mind of its contradictions, and then try again.
I am familiar with the Cingular voice mail service you are describing. If you press '7' one too many times, immediately press '*' (I think; the friendly computer voice tells you if you stay on the line) to undelete the message you just deleted. Don't hang up or press any other buttons, because you only have that one shot at undeletion.
I'm sorry you weren't familiar with this at the time, and I hope this helps in the future.
By the way, I'd be suspicious if a phone company implemented a "feature" that involves routinely keeping backup copies of all its customers' deleted voice mails indefinitely. Is that really what you want?
Semantics is meaning. There is no meaning to your statement outside its value in its language, which is defined by the language's semantics. You used the term "hybrid kernel". Someone else challenged you that said term is meaningless, making your statement meaningless. You resort to hand-waving about "not complaining about the language" because you're wrong and have no better response.
Stop being so arrogant. Admit when you're wrong and make sure to say what you mean in the future.
> I don't care what they say. It is not a public place. It is a private place with an open invitation for people to come in and spend money. They can kick anyone out for any reason. Please read my article on the matter.
I read your article, and you're wrong. A store is a public place. It is private property, but you cannot invite the general public onto your property without losing some of your rights. For what reasons the store owner can kick people out of his store depends on the laws of the state, as was said earlier. Your article on public places was a well-written, Libertarian rant, but it defined the term incorrectly.
To understand this, assume for the moment that you hate black people. You don't want any black people on your property, so you never invite them over to your house. That's fine. Now, assume you own a store, and you put up a big sign saying "No colored persons allowed." That's not fine, and the reason it's not fine is because the store is considered a public place while your house is not.
Since Libertarians are property rights absolutists, you probably think that, if you opened a store, putting a "No colored persons allowed" sign up would be just fine. I'm not saying that you personally would want to do that, just that you think you should have the right to do that. If the right to property were in fact an absolute right, as Libertarian philosphy dictates it should be, you could restrict who enters your store based on anything, including race, and you'd be right in stating that a store is not a public place. However, the law doesn't match Libertarian philosophy in this respect. Property is not an absolute right, and the store is considered a "public place".
This means a number of things. For one thing, it means the property owner cannot decide to not let people in during normal business hours because of race. It also means that customers do not have an expectation of privacy, so anyone can covertly videotape them without their permission (except in the restrooms, where there is an expectation of privacy).
Yes, I made it use less memory by using a stupider algorithm; that's not the point. My point is that Java's fatness made it very difficult to implement the data structure I needed in a tractable amount of memory. Perhaps hacks to use primitives instead of objects (where objects would be preferred conceptually) would have cut down on the fat; given enough time and motivation it's possible to work around the brain damages of almost any language, even Java.
As far as algorithm research, THAT WAS THE WHOLE POINT OF THE ASSIGNMENT! I did this project for a data structures and algorithms class. We tried out many different implementations for the search engine index, and the only ones that worked with the default size of the Java VM were stupid, flat, and returned results in linear time over the number of pages, which amazingly was still fast over 40MB of data. Storing the whole page in a StringBuffer mapped to its URL (still through TreeMap) was the least stupid thing that worked. That's why I'm so fed up about this, really. We came up with this GREAT algorithm; it would have worked given reasonable (as in, C-like) memory usage for structures/objects, and we couldn't do it in Java because Java is so fat it needs 50 FREAKING BYTES for an object with no data in it.
> The java in-memory search ran faster than the Oracle search.
Unless the Oracle search was also in memory, it damn well had better run faster.
> All good fun. But still demonstrates the ability of the language to handle the issue, in a sensible and performant manner.
No, all it means is that the search engine programmer was able to work around and overcome the deficiencies in Java.
Now that I know you were doing something with databases, what you said makes more sense to me. I was searching and indexing web pages. The smart way was to use this would be to build an index of words you can search for and store them in a TreeMap. The words would be mapped to a set of records. Each record would hold a URL, and another record holding which words followed it at which places (integer) on the page. This should NOT HAVE TAKEN as much memory as it did. Think about it: for each record, we're storing a pointer to the URL (4 bytes) (only one global list of the URL's; the records just hold a pointer to it), a list or array of integers (4 bytes each), and with these a pointer to another record (4 bytes each). There will be no more of these records than there are appearances of words indexed. So, a high estimate is 12 bytes per word. (Yes, I'm taking into account the fact that there can be more than one integer per record. That's an optimization. In the worst case, every word gets its own record.) ASCII words usually take 4 or 5 bytes, so memory use should be AT MOST 3 times the data on disk. It was more like 30 or 40 times the data on disk. Guess why? Java allocates >50 BYTES PER RECORD for overhead!
If I did this in C++, it would have worked. As it is, we ended up doing things the stupid way, which is to build a list of web pages and iterate over EVERY SINGLE WEB PAGE, searching for appearances of the word or phrase. This worked, and it was fast for the small data set we had, but it wasn't scalable. The smart way was to index based on word.
> I've done a lot of professional C++ and Java programming, and I can guarantee you did something wrong.
Ha! Of course you have. And I REALLY work for the NSA as the head of their quantum computing division (of course it exists). This is just a cover.
> The String.substring comment is probably more relevant than you think.
Nope, we explicitly dealt with the issue of multiple String copies when attempting to fix the memory usage problem, and it was a quite graceful solution iirc.
This will be my last response to ANY anonymous poster on this issue. IHBT HAND.
My professor probably wouldn't look kindly on that, seeing that almost this exact assignment is given each semester. Also, the code was written in collaboration with another person, and I don't have his permission to give the code out to random AC's on Slashdot.
Remember that we solved the problem eventually by using a slower (although not noticeably on the size data set we used) and stupider method. Object fat was taking the bulk of the memory in the original design, and that is Java's fault.
I can almost guarantee that if I implemented this EXACT SAME DESIGN in C++, there wouldn't be memory issues.
> "I was once writing a serach engine in a language I didn't know how to use, and it took a gig of RAM to do something stupid that I told it to do"
Please read my replies to your fellow ignorant posters regarding this.
> [cut further examples of your idiocy] > I haven't checked, so I could accept that Treemap is suboptimal for what you were trying to use it for. There's nothing that says you have to use it.
Yes, we could have written our own replacement for a standard library class, and if we cared enough about it (it was just a class project), we would have. It still reflects poorly on Java that its standard classes are such memory hogs.
> Disclaimer: I have in my possession (propriety commercial) code for a search engine that searches across over a gig of data. It takes up about 20MB of ram more than the data takes on disk in a flat file. It was written in a couple of hours.
> Further disclaimer: I didn't write it. Although I did suggest a performance optimisation. It was returning results in under 10ms before the optimisation.
That sounds about like how ours performed after we started doing things the stupid way. Of course, you don't mention the machine, so those stats are meaningless...
Gosh, I really should stop responding to trolls like this.
I used the standard data structure classes. This was a class assignment, and everyone using the *Set and *Map classes had memory problems, so it wasn't ineptitude in my code. Apparently, the Java standard classes just use a ton of RAM.
> How come that you had allowed Java to use 1 GB in the first place...
It would crash unless I manually set [-Xmx1024mb] with an OutOfMemoryError.
More details about why it's Java that sucks and not my code: My partner in the project actually profiled the memory usage in our TreeMap implementation. Most of the memory usage in our program came from Java's class bloat; 50 bytes just for the object identification, etc. We had to do a brain-dead StringBuffer implementation to make it work.
I was once writing a search engine in Java, and it took a gig of RAM to index 40MB using a standard TreeMap. A FREAKING GIG OF RAM!! I had to switch to doing something brain-dead simple and slow to bring it within anything resembling sane memory usage.
On Slashdot, we have an informal rule accepted by most posters (even trolls):
You don't talk about the sig.
The great-grandparent can change his sig to something completely different at any time, and this change would retroactively apply to his comment. Should he do this (and he probably will sometime after the next election), your comment will look even more idiotic than it already does. The convention of not discussing the sig is reasonable, and I hope you will decide to follow it in the future.
You express the point of view of many people in your post, and those of a significant percentage of the Slashdot community. Here's why I think you're wrong.
You talk about corporations having "equal status with individuals," but this is simply not the case. As a sibling poster mentioned, corporations cannot vote. There are many other aspects of the law that apply only to "natural persons" and not to "corporate persons."
The "corporate personhood" that you are so concerned about is in a fact a legal fiction created by the courts in order to establish that corporations could be sued. When the corporate structure was first developed, corporations argued in court that they could not be sued because they were not people, and only people could (and can) be sued. The courts responded by saying that, for the purposes of lawsuits, corporations were people. The concept of corporate personhood was then extended, by both the courts and the legislature, to place further responsibilities on corporations.
The fiction of corporate personhood is a useful structure, because in many aspects of law we do want corporations to be treated like people. They should be able to sue and be sued when they wrong or are wronged by individuals and/or corporations. They should be able to hold property and should have to pay taxes.
The distinction between "corporate personhood" and "natural personhood" is of course also useful, and it is in no danger of being blurred. Corporations cannot vote. They cannot hold public office, and they cannot serve on the jury. All mainstream politicians and scholars agree that this should be the case, and I think you'll be hard pressed to find someone with views so bizarre and extreme that he thinks corporations should be able to do these things.
You also talk about corporations in the U.S. having power in other ways. While corporations may be less regulated in the U.S. than in socialist and semi-socialist states in Europe, they are by no means unregulated. Also, the larger the corporation, the more regulated it is. Equal employment opportunity laws, for example, are major regulatory burdens for large corporations. However, since many of the more burdensome portions of these laws only apply to corporations with over a certain (fairly large) number of employees, small businesses for the most part retain the power to hire whom they wish. In this way, small corporations are more powerful than large corporations. Many business regulations share this characteristic of a "graduated burden" depending on a company's size.
I got a Zaurus off eBay a few months ago. You can get an SL5500 for about $140 used.
I used PocketWorkstation to set up a full installation of Debian Sarge on my SL5600's SD card. If you expand it with a CompactFlash WiFi card, it can browse the web with Dillo (Firefox is unusable and Minimo, while usable, is slow). It can do symbolic integration and differentiation with Maxima and matrix calculations with Octave. It supports self-hosted development through GCC and Python. I ported my custom C++ scheduling program to the Zaurus quite easily. It can emulate a TI-83+ at usable speeds (with TilEm). It supposedly can act as a remote terminal using VNC and ssh, and the ssh capability is a lifesaver for me, but I haven't used the remote VNC feature, so I can't judge.
You can check webmail with it, in Minimo if nothing else, but I don't very often. I haven't tried Thunderbird, but it's probably too slow to be usable. Text-based Linux mail clients would probably work better; I've heard reports of Sylpheed being run successfully.
My Zaurus does pretty much everything I could reasonably want it to, and the Debian universe is available to you if you need other capabilities. (Did I mention that updates to most of the packages I mentioned are just an apt-get away?:) You can fit up to a 1GB SD card in the Zaurus's SD slot, so storage probably won't be a problem. You're mostly limited by the Zaurus's 400MHz Intel XScale ARM processor and its RAM (64MB for the SL5500; 32MB for my SL5600 model).
By the way, I got quite a kick out of the announcement that Windows Vista would support swapping to flash drives. I tried that on the Zaurus ONCE. Don't do it. Swapping to flash is so slow as to be useless.
Actually, the losing side often has to pay the other side's legal fees. It might also be possible to countersue a harasser for frivolous lawsuit, but only if the case couldn't possibly have had any merit. IANAL, but I might be someday.
I said in the above comment that I disagreed with you, but heck, I'm not in a consistent enough state at this point in my life to analyze the issue; I may end up actually agreeing with you. You certainly aren't acting maliciously, and you didn't deserve that. It wasn't a troll, but it certainly wasn't rational either. Mea culpa.
Okay, fair enough. To be honest, that post wasn't exactly a troll, but I was very angry when I posted it over things unrelated to Slashdot.
I had no right to insult tgrigsby like that; I don't even know the guy. I did recognize that he was joking about literally shoving his shoe up his child's butt, but I inferred from his comment that he has no problems using aggressive corporal punishment. Had I wanted to make a respectful, rational comment arguing against corporal punishment, I should have done so instead of posting... that. I apologize to both you and him.
Slashdot Mirror
I am very much a utilitarian myself. While Stallman's arguments do implicitly assert a premise like the one you describe, I don't know if it should really be counted as _utilitarian_ per se. Traditional utilitarianism maximizes happiness, not freedom.
Also, in Mill's utilitarianism, mathematical reasoning very much applies. The goal of his philosophy was to find a way to mathematically determine the morality of behavior.
Finally, in any case, my argument only rested on propositional calculus and premises which I thought Bongo Bill stated. I used other parts of math as examples, but they weren't the basis for what I was saying.
Well, first, what you're saying is irrelevant. You are asserting that a free market for that which is currently protected by copyrights, trade secrets, and patents is a bad idea. I was responding to an incorrect statement which asserted that the market for music was a free market. Whether free markets are good or bad in a specific instance is irrelevant to what they are.
Beyond that, what you're saying is mostly correct. There wouldn't be "no" incentive to innovate, but there would be substantially less of an incentive to innovate. Copyrights and patents create an incentive to innovate through the creation of monopolies on innovations. These monopolies impose their own inefficiencies. If you believe that copyrights and patents are good for society, you must believe that there is no alternative to them that solves the incentives problem with greater efficiency.
I think that subsidies are a better way. We already subsidize that which is protected by patents through DARPA, the NSF, and other government funded agencies. We could feasibly get rid of patents and dramatically increase funding for these agencies to compensate. In my proposal, the inefficiency of a monopoly is replaced with the inefficiency of extra taxes. If we have a reasonably efficient tax system, I think this will easily be a net win over private monopolists.
With regard to that which is copyrighted, we could do subsidize in a similar way. Get rid of copyrights, but dramatically increase funding for the National Endowment for the Arts. As long as additional taxation is less inefficient than the inefficiency of private monopolies, this is a net win.
Now, if we as a country were to do this, I'd recommend doing it gradually. Decrease copyright and patent terms over the course of 10 years while increasing government subsidies to research and innovation. There's not a snowball's chance this will actually be done anyway, though.
> Addition is an extremely sound idea. No matter how far you carry it, it still works. I would not, however, recommend it as a good way to determine the circumference of a circle, when multiplication is so much easier in that particular circumstance.
The point is that addition still works, because addition is sound. If you think free software is a sound idea, then you must accept everything that soundness implies. If you think it implies an opposition to DRM, then you must oppose DRM. If you think free software is sound, it implies opposition to DRM, and you do not oppose DRM, then you have a contradiction on your hands.
You seem to be trying to get out of this by saying, "basically sound" rather than "sound" or "not sound". You can't do that. Operations are sound or not sound. If you think a rule is sound, except not in this case, and you have to do fudge it a little here, etc., then you really think the rule is not sound. You may still want to use this unsound rule if it works most of the time, but you have to be honest about it, and if you want to say anything useful, you must go back with a sound rule to check your work. A good example of this in math is naive set theory.
I'm using your concepts here (operations and their soundness), but I think it's better to consider propositions and their truth values. The concept of free software isn't really isn't really a sound or unsound operation; it's a set of propositions that are asserted true. Do you agree?
There are bijections, but I don't have one with me at the moment...
I think the easiest way to see it is to use the Schroeder-Bernstein theorem. You already see Q -> N, and you obviously have N -> Q since N SUBSETOF Q.
Productive, huh?
It's not a free market because the government creates artificial barriers to entry through copyright laws. A free market for CD's would be one in which it's legal for you to rip a CD you've purchased 1000 times and sell the copies for whatever you could get.
A free market probably wouldn't work too well for CD's, though, because of the huge spillover benefit the original seller (i.e., the artist) confers on the rest of the market. Most of the time, spillover benefits are corrected through government subsidies. In this case, the government decided to create copyright law instead.
> Open Source zealots are no more credible than any other type of zealot. It's the extension of a basically sound idea to an unrealistic, harmful, and (in the worst cases) counterintuitive extreme.
Something is not true unless it is true in all cases. Thus, if you believe something is true, you must believe it is true at its "extremes". Furthermore, everything which is intuitive is not right, so it doesn't make sense to say that counterintuitive "extremes" are "the worst cases".
The natural numbers and the rationals have the same cardinality. This is counterintuitive, but, according to the "extremes" of math, it is correct. If you accept the axioms of math, you must accept that the natural numbers and the rationals are the same size. You probably won't dispute this, but you're disputing something just like it. You want to be able to take "a basically sound idea" and deny its extremes. That is a contradiction, and you won't make any sense if you do that.
Why don't you go away, purge your mind of its contradictions, and then try again.
I am familiar with the Cingular voice mail service you are describing. If you press '7' one too many times, immediately press '*' (I think; the friendly computer voice tells you if you stay on the line) to undelete the message you just deleted. Don't hang up or press any other buttons, because you only have that one shot at undeletion.
I'm sorry you weren't familiar with this at the time, and I hope this helps in the future.
By the way, I'd be suspicious if a phone company implemented a "feature" that involves routinely keeping backup copies of all its customers' deleted voice mails indefinitely. Is that really what you want?
I'm going to try to take this one: you're wrong.
Semantics is meaning. There is no meaning to your statement outside its value in its language, which is defined by the language's semantics. You used the term "hybrid kernel". Someone else challenged you that said term is meaningless, making your statement meaningless. You resort to hand-waving about "not complaining about the language" because you're wrong and have no better response.
Stop being so arrogant. Admit when you're wrong and make sure to say what you mean in the future.
> I don't care what they say. It is not a public place. It is a private place with an open invitation for people to come in and spend money. They can kick anyone out for any reason. Please read my article on the matter.
I read your article, and you're wrong. A store is a public place. It is private property, but you cannot invite the general public onto your property without losing some of your rights. For what reasons the store owner can kick people out of his store depends on the laws of the state, as was said earlier. Your article on public places was a well-written, Libertarian rant, but it defined the term incorrectly.
To understand this, assume for the moment that you hate black people. You don't want any black people on your property, so you never invite them over to your house. That's fine. Now, assume you own a store, and you put up a big sign saying "No colored persons allowed." That's not fine, and the reason it's not fine is because the store is considered a public place while your house is not.
Since Libertarians are property rights absolutists, you probably think that, if you opened a store, putting a "No colored persons allowed" sign up would be just fine. I'm not saying that you personally would want to do that, just that you think you should have the right to do that. If the right to property were in fact an absolute right, as Libertarian philosphy dictates it should be, you could restrict who enters your store based on anything, including race, and you'd be right in stating that a store is not a public place. However, the law doesn't match Libertarian philosophy in this respect. Property is not an absolute right, and the store is considered a "public place".
This means a number of things. For one thing, it means the property owner cannot decide to not let people in during normal business hours because of race. It also means that customers do not have an expectation of privacy, so anyone can covertly videotape them without their permission (except in the restrooms, where there is an expectation of privacy).
Yes, I made it use less memory by using a stupider algorithm; that's not the point. My point is that Java's fatness made it very difficult to implement the data structure I needed in a tractable amount of memory. Perhaps hacks to use primitives instead of objects (where objects would be preferred conceptually) would have cut down on the fat; given enough time and motivation it's possible to work around the brain damages of almost any language, even Java.
As far as algorithm research, THAT WAS THE WHOLE POINT OF THE ASSIGNMENT! I did this project for a data structures and algorithms class. We tried out many different implementations for the search engine index, and the only ones that worked with the default size of the Java VM were stupid, flat, and returned results in linear time over the number of pages, which amazingly was still fast over 40MB of data. Storing the whole page in a StringBuffer mapped to its URL (still through TreeMap) was the least stupid thing that worked. That's why I'm so fed up about this, really. We came up with this GREAT algorithm; it would have worked given reasonable (as in, C-like) memory usage for structures/objects, and we couldn't do it in Java because Java is so fat it needs 50 FREAKING BYTES for an object with no data in it.
> The java in-memory search ran faster than the Oracle search.
Unless the Oracle search was also in memory, it damn well had better run faster.
> All good fun. But still demonstrates the ability of the language to handle the issue, in a sensible and performant manner.
No, all it means is that the search engine programmer was able to work around and overcome the deficiencies in Java.
Now that I know you were doing something with databases, what you said makes more sense to me. I was searching and indexing web pages. The smart way was to use this would be to build an index of words you can search for and store them in a TreeMap. The words would be mapped to a set of records. Each record would hold a URL, and another record holding which words followed it at which places (integer) on the page. This should NOT HAVE TAKEN as much memory as it did. Think about it: for each record, we're storing a pointer to the URL (4 bytes) (only one global list of the URL's; the records just hold a pointer to it), a list or array of integers (4 bytes each), and with these a pointer to another record (4 bytes each). There will be no more of these records than there are appearances of words indexed. So, a high estimate is 12 bytes per word. (Yes, I'm taking into account the fact that there can be more than one integer per record. That's an optimization. In the worst case, every word gets its own record.) ASCII words usually take 4 or 5 bytes, so memory use should be AT MOST 3 times the data on disk. It was more like 30 or 40 times the data on disk. Guess why? Java allocates >50 BYTES PER RECORD for overhead!
If I did this in C++, it would have worked. As it is, we ended up doing things the stupid way, which is to build a list of web pages and iterate over EVERY SINGLE WEB PAGE, searching for appearances of the word or phrase. This worked, and it was fast for the small data set we had, but it wasn't scalable. The smart way was to index based on word.
> I've done a lot of professional C++ and Java programming, and I can guarantee you did something wrong.
Ha! Of course you have. And I REALLY work for the NSA as the head of their quantum computing division (of course it exists). This is just a cover.
> The String.substring comment is probably more relevant than you think.
Nope, we explicitly dealt with the issue of multiple String copies when attempting to fix the memory usage problem, and it was a quite graceful solution iirc.
This will be my last response to ANY anonymous poster on this issue. IHBT HAND.
My professor probably wouldn't look kindly on that, seeing that almost this exact assignment is given each semester. Also, the code was written in collaboration with another person, and I don't have his permission to give the code out to random AC's on Slashdot.
Remember that we solved the problem eventually by using a slower (although not noticeably on the size data set we used) and stupider method. Object fat was taking the bulk of the memory in the original design, and that is Java's fault.
I can almost guarantee that if I implemented this EXACT SAME DESIGN in C++, there wouldn't be memory issues.
That is, actually, completely irrelevant to what we were doing.
> "I was once writing a serach engine in a language I didn't know how to use, and it took a gig of RAM to do something stupid that I told it to do"
Please read my replies to your fellow ignorant posters regarding this.
> [cut further examples of your idiocy]
> I haven't checked, so I could accept that Treemap is suboptimal for what you were trying to use it for. There's nothing that says you have to use it.
Yes, we could have written our own replacement for a standard library class, and if we cared enough about it (it was just a class project), we would have. It still reflects poorly on Java that its standard classes are such memory hogs.
> Disclaimer: I have in my possession (propriety commercial) code for a search engine that searches across over a gig of data. It takes up about 20MB of ram more than the data takes on disk in a flat file. It was written in a couple of hours.
> Further disclaimer: I didn't write it. Although I did suggest a performance optimisation. It was returning results in under 10ms before the optimisation.
That sounds about like how ours performed after we started doing things the stupid way. Of course, you don't mention the machine, so those stats are meaningless...
Gosh, I really should stop responding to trolls like this.
Well, first, my partner in the project and I eventually made it work (with the brain-dead StringBuffer), and it performed reasonably well.
Second, see my reply to your sibling poster.
(Note to self: stop feeding the trolls.)
> ... your ineptitude in Java ...
I used the standard data structure classes. This was a class assignment, and everyone using the *Set and *Map classes had memory problems, so it wasn't ineptitude in my code. Apparently, the Java standard classes just use a ton of RAM.
> How come that you had allowed Java to use 1 GB in the first place...
It would crash unless I manually set [-Xmx1024mb] with an OutOfMemoryError.
More details about why it's Java that sucks and not my code:
My partner in the project actually profiled the memory usage in our TreeMap implementation. Most of the memory usage in our program came from Java's class bloat; 50 bytes just for the object identification, etc. We had to do a brain-dead StringBuffer implementation to make it work.
Ha ha ha!
:)
It's funny because it's true
I was once writing a search engine in Java, and it took a gig of RAM to index 40MB using a standard TreeMap. A FREAKING GIG OF RAM!! I had to switch to doing something brain-dead simple and slow to bring it within anything resembling sane memory usage.
On Slashdot, we have an informal rule accepted by most posters (even trolls):
You don't talk about the sig.
The great-grandparent can change his sig to something completely different at any time, and this change would retroactively apply to his comment. Should he do this (and he probably will sometime after the next election), your comment will look even more idiotic than it already does. The convention of not discussing the sig is reasonable, and I hope you will decide to follow it in the future.
You express the point of view of many people in your post, and those of a significant percentage of the Slashdot community. Here's why I think you're wrong.
You talk about corporations having "equal status with individuals," but this is simply not the case. As a sibling poster mentioned, corporations cannot vote. There are many other aspects of the law that apply only to "natural persons" and not to "corporate persons."
The "corporate personhood" that you are so concerned about is in a fact a legal fiction created by the courts in order to establish that corporations could be sued. When the corporate structure was first developed, corporations argued in court that they could not be sued because they were not people, and only people could (and can) be sued. The courts responded by saying that, for the purposes of lawsuits, corporations were people. The concept of corporate personhood was then extended, by both the courts and the legislature, to place further responsibilities on corporations.
The fiction of corporate personhood is a useful structure, because in many aspects of law we do want corporations to be treated like people. They should be able to sue and be sued when they wrong or are wronged by individuals and/or corporations. They should be able to hold property and should have to pay taxes.
The distinction between "corporate personhood" and "natural personhood" is of course also useful, and it is in no danger of being blurred. Corporations cannot vote. They cannot hold public office, and they cannot serve on the jury. All mainstream politicians and scholars agree that this should be the case, and I think you'll be hard pressed to find someone with views so bizarre and extreme that he thinks corporations should be able to do these things.
You also talk about corporations in the U.S. having power in other ways. While corporations may be less regulated in the U.S. than in socialist and semi-socialist states in Europe, they are by no means unregulated. Also, the larger the corporation, the more regulated it is. Equal employment opportunity laws, for example, are major regulatory burdens for large corporations. However, since many of the more burdensome portions of these laws only apply to corporations with over a certain (fairly large) number of employees, small businesses for the most part retain the power to hire whom they wish. In this way, small corporations are more powerful than large corporations. Many business regulations share this characteristic of a "graduated burden" depending on a company's size.
I got a Zaurus off eBay a few months ago. You can get an SL5500 for about $140 used.
:) You can fit up to a 1GB SD card in the Zaurus's SD slot, so storage probably won't be a problem. You're mostly limited by the Zaurus's 400MHz Intel XScale ARM processor and its RAM (64MB for the SL5500; 32MB for my SL5600 model).
I used PocketWorkstation to set up a full installation of Debian Sarge on my SL5600's SD card. If you expand it with a CompactFlash WiFi card, it can browse the web with Dillo (Firefox is unusable and Minimo, while usable, is slow). It can do symbolic integration and differentiation with Maxima and matrix calculations with Octave. It supports self-hosted development through GCC and Python. I ported my custom C++ scheduling program to the Zaurus quite easily. It can emulate a TI-83+ at usable speeds (with TilEm). It supposedly can act as a remote terminal using VNC and ssh, and the ssh capability is a lifesaver for me, but I haven't used the remote VNC feature, so I can't judge.
You can check webmail with it, in Minimo if nothing else, but I don't very often. I haven't tried Thunderbird, but it's probably too slow to be usable. Text-based Linux mail clients would probably work better; I've heard reports of Sylpheed being run successfully.
My Zaurus does pretty much everything I could reasonably want it to, and the Debian universe is available to you if you need other capabilities. (Did I mention that updates to most of the packages I mentioned are just an apt-get away?
By the way, I got quite a kick out of the announcement that Windows Vista would support swapping to flash drives. I tried that on the Zaurus ONCE. Don't do it. Swapping to flash is so slow as to be useless.
Actually, the losing side often has to pay the other side's legal fees. It might also be possible to countersue a harasser for frivolous lawsuit, but only if the case couldn't possibly have had any merit. IANAL, but I might be someday.
I shouldn't have posted that and I'm sorry for personally attacking you.
o ld=1&commentsort=0&mode=thread&pid=15073889#150791 91
http://slashdot.org/comments.pl?sid=182353&thresh
I said in the above comment that I disagreed with you, but heck, I'm not in a consistent enough state at this point in my life to analyze the issue; I may end up actually agreeing with you. You certainly aren't acting maliciously, and you didn't deserve that. It wasn't a troll, but it certainly wasn't rational either. Mea culpa.
Okay, fair enough. To be honest, that post wasn't exactly a troll, but I was very angry when I posted it over things unrelated to Slashdot.
... that. I apologize to both you and him.
I had no right to insult tgrigsby like that; I don't even know the guy. I did recognize that he was joking about literally shoving his shoe up his child's butt, but I inferred from his comment that he has no problems using aggressive corporal punishment. Had I wanted to make a respectful, rational comment arguing against corporal punishment, I should have done so instead of posting
I actually laughed out loud reading that. Someone actually mistook me for a liberal ... wow.