We've tried to use WebSphere in our site (we're currently working on PHP3 and wnated to switch to a more professional architecture). So I sort of know this problem. I won't comment about the rest, but WebSphere has the wrong model. We wanted to implement an ecommerce site with our main bussines and some e-stores. That's suposedly the use, right? The fact is that we couldn't. We used some base classes and that's it. We ended up taking some of the product data model. But even that was brain dead. Let's take for example the attributes tables. That's a great idea (which everybody also had) with a horrid implementation. Instead of using a separate table for attribute names, are use as PK! Soy you kill any usefulness to put intelligence in your software. Or you end up setting up a new table with the attributes names (doing what should have been done in first place). You have some sort of object inheritance, But instead of going the whole mile (that's an expression?) they simply left it as a product/item relationship with nothing else shared. I don't see why te names are all 8 char. That's unacceptable. Particularly when your supported plataforms do support longer names (Oracle and DB2). You also couldn't keep the same name for a FK row and the pointed row? Come on guys? I will simply say the Catalog Archtech is such an ill concieved application that I spend se least productive day in my life convincing myslef that the're wasn't anything useful in there. Not to mention the fact that precludes te use of the little inheritance that you have. And the whole model is ridiculous! You only have e-stores and the products are linked to them. You can't have a single catalog. You don't have stocks management. The cart is a the least friendly thing I've ever seen. And there's no way of mixing data from different estores. But I have to admit that VisualAge is une fine editor. And the JVM is the very best thing ever. We actually ended up buing the database and Java suporting (and enabling) components. And as usual the whole enchilada wasn't availableunder Linux. At least the instalation scripts weren't broken (ejem Oracle, Informix) and hasn't hardcoded path to libraries and applications (Ejem again! Oracle).
The reason is that is a national effort to achieve a locally interconnected Internet. And it has a GREAT disadvantage: it hurts local telephone monopolies. In Argentina it took years just to have a local NAP (CABASE). But when ISP start pricing different from national and international connection (for bussiness, of course) that flourish. For example, here, the price of an international E1 (that's a 2Mbps for Americans) is 30 times a national E1.
The Montecarlo method is conceptually very simple: when you want to estimate the integral of some set that you can't integrate directly, but: a) is contained by some easy to handle and well know set. b) you have a function that tells you whether an element is inside the set or outside the set. Then all you have to do is to sample the known set with randomly distributed points and ask wether they belong or they don't. Then you simply take the proportion of points inside and multiply it by the size of the container. It's actually a bit more general and complicated. But that's basically the idea.
Thoy wouldn't be able to sue you because it's your original work. You can only patent original work. So not only would they spoil their patent but you could counter sue them for misapropiation of your work or plagiarism. That's why I like BSD licences. Though for the most part LGPL is de definitive patent for libraries. About other work, well GPL is good enough. At least that's what I think.
I believe you're quite wrong. All that they usually do is to put./configure --config-dir=/etc or something like this. You can actually install a bare system and download the latest sources, compile them and get everything under/usr/local were things are supposed to go. And don't you ever forget that Red Hat is funding the core Linux development team (Alan Cox & company) plus GNOME (no flames about this, please) and are an active member of the group which is working to standarize Linux. Besides almost all of their work is GPL. I've downloaded ISO images of their distribution since 5.1. And now I've actually bought 6.2 with the electronic commerce pakage. And I already have almost all the software downloaded from the Inet, but it seemed right to pay for their service. I will only add that I couldn't get to know linux before because Slackware 2.0 was simply impossible to install for a neophite like me. Thanks to Red Hat I was able to learn enough to download a bare distribution and install the packages I like.
Obviously You should get the authors permission in any case. But on principle if you made a game and used a GPL dll which you put a separate download to enable sound I think it would be fair. As usual, I'm not a lawyer so check before. And get a signed authorization faxed from the authors. /*Rant I've read the filosofical posture of the FSF and I don't really get it. I mean, the have this axiom: "Software should be free, as in freedom" (strange a language that don't distiguish gratis from libre.) But after a geek I'm also an economist (I know, but at least we have to learn dynamic optimization, chaotic systems and topology, which is a lot better than most other social sciences, besides I've been redeemed working as a CTO in a.com company) so I don't take such thing as an axiom. I believe that we should maximize the welfare of the society which is a function of the goods (a concept which I won't define yet) consumed (tip: axiom 1). I also believe in the greedyness of the human beings, not as an absolute value, but to a certain degree (tip bis: axiom 2). So I believe that we should enforce the property rights (tip: hypothesis 1. If you want to know why read Microeconomis, by Hal Varian.) This means that there should be property. I don't believe much in intellectual property, much less in the brain dead American law that let's you patent "anything manmade under the sun." Why? Ok, lets get to it. If we want to maximize the welfare we want to use the biggest technology set available in the know universe. So ideally anybody should get free access to the most advanced technics in each fields (this doesn't mean that AMD Dresden fab should be open to anyone, but it's plans and underlaying technology should, at least until this point). So, from this point of view, if we consider source code as a technology it should be on the public domain. But being the greedy animals we are, making huge investemnts of resources on investigating technology for the sake of it just doesn't seem compelling. Evidently we need some incentive to make research and development more interesting. So if we allow the developer of a new technique or technology to reap some of the benefits that his work has brought to society, we may have a more people willing to do R & D. Let's get to the software side of things. We can see a program as any other product, so let's think of it for a while (I wil try to discuss later a bit about the portions of a program, a tracendental point on this discussion.) It need a certain amount of resources to be designed, and another to be replicated. The former may be huge while the latter is insignificant. So what we have is that the cost of one extra unit of a program is basically costless. From microeconomics 101 we know that in order to maximize welfare the price of a good should equal the cost of the last unit produced (not exaclty, but close enough for a rant.) So from this point of view software should be costless. The problem comes from the huge development cost. If every all source code were in the public domain, then few programmers would work and programs which would increase the welfare would not be written even though the benefits to society are less than their cost. But if all software were commercial we would end up paying way more than the development cost (cough*Windows*cough). Even worst, a lot of consumers of the software wouldn't pay it, even though the cost asociated with their consumption is in practice zero. Worst yet, all programmers should reinvent the wheel for each new program (not to mention the added storage and debugging costs). But let's add pattents that allow to get property on trivial code. So you should license a printf() function! Developers wouldn't even have the chance to reinvent the wheel, they would have to pay for it every single time. Evidently, basic functions and algorithms should be almost free because all the other programs are based on those. But let's say that we have a huge company A that's competing with an even bigger company B. And a manager in Company A decides that they need a given software that's unavailable. That program means a lot of money to the, so they are willing to pay a lot for it. If they couldn't enforce their property on that software they wouldn't want to develop it because they would be investing money to get company B the same benefits without the costs. Or worst yet, some people simply like to own what they buy (not so crazy if you think about it). But let's suppose that company B want's to pay for that program too. If there's a patent they won't be able to use it. So company A would have an unfair advantage to company B. Which hurts competition and don't maximizes welfare. I'm letting the comments discuss my ideas while I continue to work. Don't bother to go my home page because I'm not posting this yet.
What's the performance difference on these engines? I mean. If Cal wants low emission they could simply enforce manual tranmissions and get a 30% improvement in efficiency overnight. But yo should add to that if Americans wouldn't use automatic gears they wouldn't need 4 liters. My 2 cents
I don't feel bad about using their services. tonic.to in fact was fast, easy to use and a great experience (albeit a bit expensive one). On the plus side there are words in Spanish or other latin languages that end in to so you can have a domain like juan.sos.recho.to which is a sentence neatly separated by dots. Quite nice if you ask me.
The difference between Type I and TT lays in the quality of the designer. Actually the TT has the hinting built in while the Type I has "some". So no solution is better than the other. BOTH are open standards? That would be the solution. But the important fact here is that if you define all the vector graphics in logical units you could do so for raster images. The scaling may not be the most beautyful, but is a solution. Or in places with dynamic layout you could define a range of DPI where the picture is shown at 1:1, then at 2:1 then at 3:1, etc. A third solution would be to use fractal compression of images. It saves space and is resolution independent (Problem: there's no free fractal compression standard). Though the images lose sharpness. OK, my 2 cents. Baldusi
First and foremost Via has a joint venture with S3 that can make the chips under cross license agreement. But it's obvious that Intel is using strong arm tactics which are highlly illigal under US antitrust laws. Now I'm going to be a bit paranoic . Intel and AMD are US companies, so Intel can't really be that bad to AMD. But VIA is a taiwanese company. So it's possible that when the FTC makes the cost-benefit analysis they find that the cost to consumers in US is less than the monopolistic benefits that Intel gets from the rest of the world (and on which it pay's income tax). Ok, may be it's a bit extreme. But I think you should think about it.
You are forgetting that the user experiences the whole computer, not just the processor. The sad true story is that while AMD processors are just as good as Intel's (if not better, just look at Intel's errata) the companion chipsets are really crappy. And you have to add that the cheaper systems have crappy motherboards (Gateway anyone?). So while the processor is robust and compatible, if you have crappy chipsets (which is all the Super 7's) which renders the whole machine unstable and incompatible. So don't let your AMD fan feelings blind you from the truth: it's more difficult to get a cheap, stable AMD computer than an Intel's one (that's the vantage of Intel's chipsets).
While it maight be true that some instruction implementations are not the best possible, I would like to focus on the quality of it's chipsets. It's very difficult to check against SGI, Sun or Alpha chipsets, so I guess we are constrained to x86 comparison. And in that market, at least until BX they had the very best top of the line. If nothing else thay are the only ones who implement protocols flawlesslly. I mean, who could get an AGP to work perfectly on a VIA or ALi without disabling features (like X2 mode)? I had to switch to a Celeron when my k6 couldn't talk to my USB camera (BTW the Intel TX used it perfectly, but I needed 100Mhz FSB). No only that but their performance used to be stellar. Simply unreachable by the competitors. If you've seen benchmarks through Internet you'll see that the Apollo Pro with PC133 can't beat a BX with PC100. That's simply amazing. Until Intel launched the i740 they had the fastest, most compatible and flawless products. But then... i740 (horrid performance, no local memory texturing, though it was a high quality product), the first Celeron (16 Kb cache, horrid performante), the i810 (horrid performante and reliability issues), Rambus (questionable performance, outrageous cost, intent to corner the RAM market, the reason of all current problems), i810e (horrid performance), i820 (late, buggy, slow, expensive), Coppermine (late, buggy), i840 (expensive), SpeedStep (late), 800Mhz (vaporware) and the Itanium (brain dead, clumbsy, bloated, slow, expensive, late, late, late, etc.). So it may be true that the current offering is crap or dated (the venerable BX). But they have a history of delivering good products. Though I doubt they can keep up any longer
I've been thinking that if it's true that the opcodes could be implemented faster at relatively no cost then it's quite possible that they implemented them "badly" just to force compiler to use simpler instructions. Sort of trying to RISCify the x86 by having a bad implementation of the "messy" instructions.
Slashdot Mirror
Have you heard of the KAME project? It's a project to develop an IPv6 subsystema for de *BSDs. And it has aparently been quite successful.
We've tried to use WebSphere in our site (we're currently working on PHP3 and wnated to switch to a more professional architecture). So I sort of know this problem. I won't comment about the rest, but WebSphere has the wrong model.
We wanted to implement an ecommerce site with our main bussines and some e-stores. That's suposedly the use, right? The fact is that we couldn't. We used some base classes and that's it. We ended up taking some of the product data model. But even that was brain dead.
Let's take for example the attributes tables. That's a great idea (which everybody also had) with a horrid implementation. Instead of using a separate table for attribute names, are use as PK! Soy you kill any usefulness to put intelligence in your software. Or you end up setting up a new table with the attributes names (doing what should have been done in first place). You have some sort of object inheritance, But instead of going the whole mile (that's an expression?) they simply left it as a product/item relationship with nothing else shared.
I don't see why te names are all 8 char. That's unacceptable. Particularly when your supported plataforms do support longer names (Oracle and DB2). You also couldn't keep the same name for a FK row and the pointed row? Come on guys?
I will simply say the Catalog Archtech is such an ill concieved application that I spend se least productive day in my life convincing myslef that the're wasn't anything useful in there. Not to mention the fact that precludes te use of the little inheritance that you have.
And the whole model is ridiculous! You only have e-stores and the products are linked to them. You can't have a single catalog. You don't have stocks management. The cart is a the least friendly thing I've ever seen. And there's no way of mixing data from different estores.
But I have to admit that VisualAge is une fine editor. And the JVM is the very best thing ever. We actually ended up buing the database and Java suporting (and enabling) components. And as usual the whole enchilada wasn't availableunder Linux. At least the instalation scripts weren't broken (ejem Oracle, Informix) and hasn't hardcoded path to libraries and applications (Ejem again! Oracle).
The reason is that is a national effort to achieve a locally interconnected Internet. And it has a GREAT disadvantage: it hurts local telephone monopolies.
In Argentina it took years just to have a local NAP (CABASE). But when ISP start pricing different from national and international connection (for bussiness, of course) that flourish. For example, here, the price of an international E1 (that's a 2Mbps for Americans) is 30 times a national E1.
The Montecarlo method is conceptually very simple: when you want to estimate the integral of some set that you can't integrate directly, but:
a) is contained by some easy to handle and well know set.
b) you have a function that tells you whether an element is inside the set or outside the set.
Then all you have to do is to sample the known set with randomly distributed points and ask wether they belong or they don't. Then you simply take the proportion of points inside and multiply it by the size of the container.
It's actually a bit more general and complicated. But that's basically the idea.
Thoy wouldn't be able to sue you because it's your original work. You can only patent original work. So not only would they spoil their patent but you could counter sue them for misapropiation of your work or plagiarism. That's why I like BSD licences. Though for the most part LGPL is de definitive patent for libraries. About other work, well GPL is good enough. At least that's what I think.
I believe you're quite wrong. All that they usually do is to put ./configure --config-dir=/etc or something like this. You can actually install a bare system and download the latest sources, compile them and get everything under /usr/local were things are supposed to go.
And don't you ever forget that Red Hat is funding the core Linux development team (Alan Cox & company) plus GNOME (no flames about this, please) and are an active member of the group which is working to standarize Linux. Besides almost all of their work is GPL.
I've downloaded ISO images of their distribution since 5.1. And now I've actually bought 6.2 with the electronic commerce pakage. And I already have almost all the software downloaded from the Inet, but it seemed right to pay for their service.
I will only add that I couldn't get to know linux before because Slackware 2.0 was simply impossible to install for a neophite like me. Thanks to Red Hat I was able to learn enough to download a bare distribution and install the packages I like.
Obviously You should get the authors permission in any case. But on principle if you made a game and used a GPL dll which you put a separate download to enable sound I think it would be fair. As usual, I'm not a lawyer so check before. And get a signed authorization faxed from the authors.
/*Rant .com company) so I don't take such thing as an axiom.
I've read the filosofical posture of the FSF and I don't really get it. I mean, the have this axiom: "Software should be free, as in freedom" (strange a language that don't distiguish gratis from libre.) But after a geek I'm also an economist (I know, but at least we have to learn dynamic optimization, chaotic systems and topology, which is a lot better than most other social sciences, besides I've been redeemed working as a CTO in a
I believe that we should maximize the welfare of the society which is a function of the goods (a concept which I won't define yet) consumed (tip: axiom 1). I also believe in the greedyness of the human beings, not as an absolute value, but to a certain degree (tip bis: axiom 2). So I believe that we should enforce the property rights (tip: hypothesis 1. If you want to know why read Microeconomis, by Hal Varian.) This means that there should be property. I don't believe much in intellectual property, much less in the brain dead American law that let's you patent "anything manmade under the sun." Why? Ok, lets get to it.
If we want to maximize the welfare we want to use the biggest technology set available in the know universe. So ideally anybody should get free access to the most advanced technics in each fields (this doesn't mean that AMD Dresden fab should be open to anyone, but it's plans and underlaying technology should, at least until this point). So, from this point of view, if we consider source code as a technology it should be on the public domain. But being the greedy animals we are, making huge investemnts of resources on investigating technology for the sake of it just doesn't seem compelling. Evidently we need some incentive to make research and development more interesting. So if we allow the developer of a new technique or technology to reap some of the benefits that his work has brought to society, we may have a more people willing to do R & D.
Let's get to the software side of things. We can see a program as any other product, so let's think of it for a while (I wil try to discuss later a bit about the portions of a program, a tracendental point on this discussion.) It need a certain amount of resources to be designed, and another to be replicated. The former may be huge while the latter is insignificant. So what we have is that the cost of one extra unit of a program is basically costless. From microeconomics 101 we know that in order to maximize welfare the price of a good should equal the cost of the last unit produced (not exaclty, but close enough for a rant.) So from this point of view software should be costless. The problem comes from the huge development cost. If every all source code were in the public domain, then few programmers would work and programs which would increase the welfare would not be written even though the benefits to society are less than their cost.
But if all software were commercial we would end up paying way more than the development cost (cough*Windows*cough). Even worst, a lot of consumers of the software wouldn't pay it, even though the cost asociated with their consumption is in practice zero. Worst yet, all programmers should reinvent the wheel for each new program (not to mention the added storage and debugging costs). But let's add pattents that allow to get property on trivial code. So you should license a printf() function! Developers wouldn't even have the chance to reinvent the wheel, they would have to pay for it every single time.
Evidently, basic functions and algorithms should be almost free because all the other programs are based on those. But let's say that we have a huge company A that's competing with an even bigger company B. And a manager in Company A decides that they need a given software that's unavailable. That program means a lot of money to the, so they are willing to pay a lot for it. If they couldn't enforce their property on that software they wouldn't want to develop it because they would be investing money to get company B the same benefits without the costs. Or worst yet, some people simply like to own what they buy (not so crazy if you think about it).
But let's suppose that company B want's to pay for that program too. If there's a patent they won't be able to use it. So company A would have an unfair advantage to company B. Which hurts competition and don't maximizes welfare.
I'm letting the comments discuss my ideas while I continue to work. Don't bother to go my home page because I'm not posting this yet.
Regards,
Alejandro Belluscio
What's the performance difference on these engines? I mean. If Cal wants low emission they could simply enforce manual tranmissions and get a 30% improvement in efficiency overnight.
But yo should add to that if Americans wouldn't use automatic gears they wouldn't need 4 liters. My 2 cents
I don't feel bad about using their services. tonic.to in fact was fast, easy to use and a great experience (albeit a bit expensive one). On the plus side there are words in Spanish or other latin languages that end in to so you can have a domain like juan.sos.recho.to which is a sentence neatly separated by dots. Quite nice if you ask me.
Pedazo de pelotudo! Por qué mierda ponés esas cosas? Nos hacés quedar mal a todos. Escribí eso en el forum de M$. baldusi
The difference between Type I and TT lays in the quality of the designer. Actually the TT has the hinting built in while the Type I has "some". So no solution is better than the other. BOTH are open standards? That would be the solution. But the important fact here is that if you define all the vector graphics in logical units you could do so for raster images. The scaling may not be the most beautyful, but is a solution. Or in places with dynamic layout you could define a range of DPI where the picture is shown at 1:1, then at 2:1 then at 3:1, etc. A third solution would be to use fractal compression of images. It saves space and is resolution independent (Problem: there's no free fractal compression standard). Though the images lose sharpness. OK, my 2 cents. Baldusi
First and foremost Via has a joint venture with S3 that can make the chips under cross license agreement. But it's obvious that Intel is using strong arm tactics which are highlly illigal under US antitrust laws. Now I'm going to be a bit paranoic . Intel and AMD are US companies, so Intel can't really be that bad to AMD. But VIA is a taiwanese company. So it's possible that when the FTC makes the cost-benefit analysis they find that the cost to consumers in US is less than the monopolistic benefits that Intel gets from the rest of the world (and on which it pay's income tax). Ok, may be it's a bit extreme. But I think you should think about it.
You are forgetting that the user experiences the whole computer, not just the processor. The sad true story is that while AMD processors are just as good as Intel's (if not better, just look at Intel's errata) the companion chipsets are really crappy. And you have to add that the cheaper systems have crappy motherboards (Gateway anyone?). So while the processor is robust and compatible, if you have crappy chipsets (which is all the Super 7's) which renders the whole machine unstable and incompatible. So don't let your AMD fan feelings blind you from the truth: it's more difficult to get a cheap, stable AMD computer than an Intel's one (that's the vantage of Intel's chipsets).
While it maight be true that some instruction implementations are not the best possible, I would like to focus on the quality of it's chipsets. It's very difficult to check against SGI, Sun or Alpha chipsets, so I guess we are constrained to x86 comparison. And in that market, at least until BX they had the very best top of the line. If nothing else thay are the only ones who implement protocols flawlesslly. I mean, who could get an AGP to work perfectly on a VIA or ALi without disabling features (like X2 mode)? I had to switch to a Celeron when my k6 couldn't talk to my USB camera (BTW the Intel TX used it perfectly, but I needed 100Mhz FSB). No only that but their performance used to be stellar. Simply unreachable by the competitors. If you've seen benchmarks through Internet you'll see that the Apollo Pro with PC133 can't beat a BX with PC100. That's simply amazing. Until Intel launched the i740 they had the fastest, most compatible and flawless products. But then... i740 (horrid performance, no local memory texturing, though it was a high quality product), the first Celeron (16 Kb cache, horrid performante), the i810 (horrid performante and reliability issues), Rambus (questionable performance, outrageous cost, intent to corner the RAM market, the reason of all current problems), i810e (horrid performance), i820 (late, buggy, slow, expensive), Coppermine (late, buggy), i840 (expensive), SpeedStep (late), 800Mhz (vaporware) and the Itanium (brain dead, clumbsy, bloated, slow, expensive, late, late, late, etc.). So it may be true that the current offering is crap or dated (the venerable BX). But they have a history of delivering good products. Though I doubt they can keep up any longer
I've been thinking that if it's true that the opcodes could be implemented faster at relatively no cost then it's quite possible that they implemented them "badly" just to force compiler to use simpler instructions. Sort of trying to RISCify the x86 by having a bad implementation of the "messy" instructions.