Surely this is just further confirmation of what gestalt pyshcologists determined in the 19th century ? that we only pay attention to a small part of our visual field at any given time. The only new aspect seems to be the neurological confirmation of what was alread known through psychological experiment.
Also, its not the image processing that is serial. We've known for some time that that is parallel - large parts of visual cortex recognise lines at different angles, changes in color, etc, using what are quite close to standard image processing algorithms.
It would be interesting if Red Hat (or some other free software related company) started registering software patents and allowing free use of them for free software development.
They'd then be able to enter into the kind of "patent swap" deals that big companies do with one another on behalf of the free software world.
One of the Manchester machines (I cannot remember if it was the Mark I or the SSEM) was a stored program computer, and therefore the first electronic stored program computer. There were probably mechanical predecessors though.
Looking back, there is a continuity of calculating machines all the way back to the abacus, each of which was the first to implement some notable feature of modern computers.
I agree completely about the risks of strong bad government. Overthrowing such beasts is almost impossible - look at Iraq.
I agree with what you say about most effort in many liberal regimes (and most especially the US) being directed at keeping government weak. I think this has been an unspoken bu longstanding principle for some time - not only must we forbid that the government take certain powers, we must ensure it remains weak so if it does take them, we can overthrow it.
My concern with this is that weak bad governments, being unscrupulous, have a very easy time turning themselves into strong bad governments, especially if people are angry or not particularly aware of what is going on.
The situation the US is in now, where it has a moderately strong government with a widely questionable record, raises the question of what to do to make it at least a good government, if not a strong good government. If people set out to weaken governments on the ground that they are not particularly good, I think its quite possible large portions of the electorate will get upset because certain tasks they think the the responsibility of the state are not being filled. That increases the risk of unscrupulous, and probably bad, governments being elected by promising to clean up the mess. This is what looks like it might happen in Russia.
I appreciate the point about the nature of US administrations, and to a lesser extent all western governments since around 1950.
Nonetheless, I do believe they are *relatively* well-behaved, especially when you look at the alternatives. I don't live in the US, so maybe I underestimate the degree to which are behaving badly, but I do still think that limiting the informational resources of even only slightly benign government might increase the chances of the public supporting one that will turn out not to be benign at all.
There's something I don't get here. I know its a long standing conviction of Anglo-American civil libertarians that this kind of thing is Bad News. I am more or less in agreement with that, but thinking about it a bit more, what exactly is wrong with it ?
It seems to me that measures like this increase the ability of the government to enforce the law, regardless of what that law actually is. That makes the assertion that this will help to catch drug smugglers/terrorists/child molersters/people with green hair/tomorrow's public enemy number one, reallistic. It also, of course, means it can be used to implement arbitary and stupid laws, such as, for instance drug prohibition (just to pick a nice uncontroversial example).
However the problem is not the databases and so on themselves, but the laws they are being used to implement. If the US had the sort of utterly minimal libertarian code of laws that many people here probably favour, would measures like this still be a problem ?
I'm concerned about this, because I do believe that if we stop relatively (note thats relatively) well-behaved governments from doing their jobs effectively, we may well be increasing our chances of getting larger, more unwieldly, more tyranical and more expensive governments in their place.
Oh, and please don't reply with some stupid platitude and privacy and freedom. I know that's what is believed, I want to know why it is necessarily true.
I wasn't trying to make a fair comparison. I was trying, apparently unsuccessfully, to illustrate the stupidity of making unfair comparisons.
The person to whom I was replying was trying to argue that Linux won't "fail" like Java because its better designed. Firstly either Java hasn't failed or I've halucinated every project I've worked on for the last 2 years.
Secondly, the comparison is completely spurious, and this is what I was trying to illustrate. Linux does something fairly well understood and quite commonly implemented - its a Unix-like kernel. He was probably refering to the entire "operating system", not the kernel. If so, you can't even really argue its universally well designed and implemented. Linux is not just the Linux kernel.
Java's AWT tries to be a cross platform GUI framework. This is hard, infrequently attempted, and when it is attempted in invariably has only marginal success. It did is to within the state of the art, which is not very good, and then swung (pun intended) back to the other known solution to the problem. And Java is not just the AWT.
You can't tar the whole OS with the brush of the kernel, nor the whole class library with the brush of the hardest part. Not can you fairly compare the engineering in a fairly well-understood task with that in one that is universally seen to be difficult.
In spite of the flurry of bile directed at Sun for Java that this posting willing no doubt generate, the comparison between Linux and Java here is actually not bad.
Java did not take over the world, but its a pretty good language and is being very widely used. In fact its the fastest uptake of any programming language I've ever seen, in spite of its problems.
Similarly Linux will never take over the world. It might get a 30 or even 70 percent share, but it will never completely obscure the possibility of using anything else the way Windows has in some places. Its not really within anyone's interest or ability to do that.
The problem with the article is that it assumes, as many people here do, and MS's and Sun's marketing machines pretend to, that the purpose of some technologies is to take over the world. It isn't, and by and large they can't. You wouldn't want them to anyway. Java hasn't failed because we've not all started to use Solaris (god forbid, in fact), and likewise Linux won't have failed if some people keep using NT. Articles like this are based on a false perception that there is some winner and some loser in technology battles. It never turns out that way, even if thats what the participants want.
Fine if you are programming bear metal on your own. Not so fine if you're in a team. Especially not fine if you want to reuse code. Most especially not fine if the language is bearly usable without it (templates in C++).
Interfaces distinguish between the subtype hierarchy, which should be publicly visible so that users of a package know what objects they can treat in the same way, and the inheritance hierarchy, which is all about things being implemented the same way, which you should not need to know.
Multiple inheritance of implementation is generally thought to be bad because its actually very rare to find classes you can mix together sensibly. If you want to know that two classes with the same interface fulfill certain rules, then I would say that Eiffel-style conditions are the right way to go. If you want to share implementation, delegate to another object.
I'm not convinced namespace protection matters. If your classes are that big, I suggest they're too big.
Smalltalk development environments are still the best ever created, and most of the work was done more than 20 years ago now. The language is OK. The lack of typing makes it hard to have much confidence in Smalltalk programs, and the syntax has its good and bad points, but I'd probably still choose Smalltalk over Java.
Regrettably there's not much choice these days. ObjectShare have gone even further into VB-land and lost the last few really serious projects to use it. Java is good in that it keeps many of the good things about Smalltalk, although I find it hard to understand why so little work has been done bringing the runtimes up to Smalltalk standards. The worst feature of Java is there distinction between reference and primitive types. You simply don't need to do this, the supposed "inefficiency" can be removed by the compiler before it even gets to the runtime.
Ummm... Java removes most of the widely-reputed problem areas of C++ (excessive syntax, badly implemented generics, multiple inheritance, operator overloading and pass by value).
Its a shame this article doesn't say what "real world programming concepts" people need to master.
Its my experience that many otherwise very competent programmers have huge holes in their knowledge, and the primary one seems to be ability to do decent abstract design. Teaching people C emphatically won't help with that. Even if you don't like OO, something like Ada or Standard ML is going to be a lot more use that C.
If the above is a reference to a lack of knowledge about real algorithm design and implementation, and an overreliance on libraries, then I agree. People need to learn that stuff, or they'll get stuck when the library of choice is missing the function they want. Nonetheless, if you have libraries, you should use them. There's nothing more irritating than programmers who faff around reimplementing stuff thats already in the standard libraries for "efficiency reasons" (not that they've profiled it or anything).
ML is semantically nice, but syntactically its evil. "An undistinguished ALGOL-family syntax" is how one of the standard's author's put it. Trust me on this, I've written an interpreter. The syntax is evil.
More seriously, ML is one of my favourite programming languages, but you and I were taught ML as a first or second language because we were CS students. It gets you out of nasty imperitive ways of thinking, and gives you a different way of seeing problems.
Most of the professional computer programmers I know would have trouble understanding the recursive pattern matching style which is typical of ML. It generally takes a good few weeks of teaching before CS students get their heads around it, and some never do.
The goal of a project like this is to teach people how to do a bit of simple programming, so they can get more out of their hardware, not to start them down the road to a CS PhD.
You have never read 1984 etc I must assume. Privacy is a fundamental basis of humanity. Take away that, and all the rest will follow. Why not individuality next? who needs that? we are all good at heart, so lets take down our walls and drone for the greater good, mindlessly tending the machines of society. No thanks, I'll continue to do my legal activities in the privacy of my own home.
This is the wrong way round. Tyrannical governments will inevitably violate people's privacy to enforce their tyranny. A government can have spy sattelites and refrain from trying to control what you do in bed just as easily as it can have an army and refrain from trying to block the streets.
Do you really imagine that if the government really did become despotic, they would refrain from surveilance because you didn't like it ? I really cannot see the logic here.
Resources that can be beneficial in the hands of a a broadly liberal state (look it up, OK, it doesn't mean what you think) are the tools of oppression under a tyranny. You don't avoid tyrrany by binding the hands of the liberal state. The more you make it impossible for necessary government to do its job, the more likely it is your moronic fellow countrymen will support unnecessary government.
I don't know very much about MIPS in particular, except that it started out as an especially radical approach to RISC (I think they tried to eliminate all pipeline interlocks - I don't know exactly what happened to that).
But I must argue with the idea that adding SIMD to MIPS, was neceaarily against the spirit of the project just on the grounds that "its not RISC".
RISC is to microprocessor design what OO is to software engineering. Everyone believes in it. Everyone thinks its a Good Thing, but noone can ever quite agree about what it entails doing in practice.
The original RISC projects at Stanford and Cambridge, took the (huge and crufty) instruction sets of the day, and analysed the instruction use frequency of compilers for certain languages (C mainly, but I think the Stanford group also looked at LISP and Smalltalk). The conclusion was that for C at least, a lot of silicon real estate could be freed by losing lots of instructions and addressing modes, concentrating only on those that were frequently used and making them fast by using more space in their implementation. It worked, but the optimisations made were pretty much the same in all RISC processors. They all tended to have a load store architecture, a large register file, relatively deep pipelines, and instruction and data caches.
It is this common feature set of RISC processors, which most people ended up understanding as the essence of RISC, rather than the careful empirical understanding of instruction usage that lay behind them.
In the design of the 486, Intel realise that you can apply these ideas to the design of CISC processors to. They separated the memory parts of the CISC instructions from the rest, added more registers (but hid them from the programmer), pipelines the system and added a cache. Hence the claim at that time that "RISC is dead/irrelevant/pointless". From that point on, no microprocessor designer has needed to go back to the principle of reducing the instruction set to gain speed. Mainly, this is because people have been much more careful about adding new instructions. Nonetheless the Alpha, and the modern MIPS processors have more instructions that anything that existed when the idea of RISC was first floated.
A lot has happened since then. Mainly the number of transistors on a chip has continued to double every 18 months, and inevitably that has led to a quest for ways to use this space to get more speed. Increasing the number of functional units and the depth of pipelines is one way, but beyond a certain number of FUs, the amount of parallelism that can be extracted by the instruction scheduler in the processor drops of. To gain any speed beyond that point, you need a smarter compiler, and one way to help the compiler to be smart is to add vector processing instructions. The problem then becomes getting enough data into the CPU, which is what the Cray boys were always good at, as I understand it.
This says nothing about the specific case of MIPS. Maybe MIPS really had architectural features that were incompatible with vector instructions. Maybe they just bodged it. Maybe the political infighting stopped it from working. Nonetheless, the statement "RISC processors cannot do vectors" is wrong, since many now do, and anyway the term RISC has been made meaningless through the degree to which it is now universally accecpted. Underneath the instruction set architecture - which still is now almost always very complex - there are no CISC processors any more.
Following these research findings the Council of Ministers has decided, on the advice of the vetinary subcommittee, and the subcommittee in charge of outlawing things that might scare some people, to ban mice from all EU member states.
Dr Klaus van Hoopklinger, chairman of the investigative subcommitee for researching spurious reasons to outlaw foreign products announced "Given the possibility that some mice will undoubtedly be genetically engineered for nefarious commercial reasons by our American allies, and given that European member states are currently carrying out a 10 year extensive yet innevitably inconclusive effort on the effects of superintelligent mice on the ecosystem and dietray health of EU citizens, we see no alternative but to ban all mice".
The decision is expected to be ratified by the European Parliament and made law by the governments of the member states with little opposition. A massive, Europe wide, mouse extermination program, using orgnically grown poisons, is expected to begin early in 2001, and customs barriers, including specially trained, selectively bred, sniffer cats, should be in place by early next year.
Funniest thing I've read for ages. Glad to see someone else has noticed the amazing resemblance between certain libertarians and the people they seem to hate most (Marxists and Fascists).
This article seems to contain just as much prejudice and ill-judgement as the rantings it criticises. To quote:
When the system was moved to Red Hat, the scripts broke. Thousands of others had problems when the move to glibc was rushed by Red Hat while other distributions remained cautious.
In many ways, Linux is a FreeBSD clone
But many are simply curious about why a new user would choose Linux over FreeBSD, despite FreeBSD's technical superiority.
Unlike Linux advocates, FreeBSD advocates do not believe FreeBSD should be running on every microchip.
Most negative talk about FreeBSD is baseless and intended to destroy, scare, or subvert potential users
All of these statements say or imply things about Linux or Linux users that are not generally true. Most ranting by Linux zealots about FreeBSD is based on prejudice and misinformation. Repeating the same mistake in the other direction doesn't help.
The BSDs and Linux are all excellent, and very similar, operating systems. The only way to compare them is with strict technical or legal comparisons. Rhetoric doesn't help anyone to make decisions.
To the extent that Unisys is a single entity, yes, you have the right to communicate your feelings to it. You can even swear at it, or insult its parentage, if thats the kind of thing you enjoy.
However it is impossible to communicate with the entity called Unisys, you have to talk to its employees instead. Unfortunately mail from ordinary people sent to large companies does not get read by the people responsible for their decisions (and in this case, I might add, Unisys isn't doing much wrong). It gets read by other ordinary people, who really cannot do very much about said decisions. If you insult them, then frankly, thats just rude, and you should apologise.
They aren't OPEN sourcing it, they are giving it the boneheaded license they've used for their other "no where to go but into the ground" projects. Who wants to work for free to increase Sun's profit margins? The use of the term "community" in the license is laughable.
Where do you get off with this stuff ? They get money, they also get to enforce compatibility. For some technologies, they let you ship non-commercial releases. Big deal - its their code, and with Java at least I can see where they are coming from.
Since when do the second fastest growing programming language in the world, one of only two Unix office suites of any quality, and one of a very few genuine efforts at universal plug and play have nowhere to go but the ground ? You mean they can't make money ? Do you really think they'd make money if they made them open source ?
People who would rather pay a little or look at some adverts to write a letter to their gran than shell out $x00 for MS Office. I barely use a word processor, and fond as I am of LaTeX, I don't like it for certain purposes. I *might* use such services, depending on how they are set up.
Seriously, there are a lot of people betting serious money on this "application publishing" stuff, and there does seem to be a market.
Slashdot Mirror
Surely this is just further confirmation of what gestalt pyshcologists determined in the 19th century ? that we only pay attention to a small part of our visual field at any given time. The only new aspect seems to be the neurological confirmation of what was alread known through psychological experiment.
Also, its not the image processing that is serial. We've known for some time that that is parallel - large parts of visual cortex recognise lines at different angles, changes in color, etc, using what are quite close to standard image processing algorithms.
It would be interesting if Red Hat (or some other free software related company) started registering software patents and allowing free use of them for free software development.
They'd then be able to enter into the kind of "patent swap" deals that big companies do with one another on behalf of the free software world.
One of the Manchester machines (I cannot remember if it was the Mark I or the SSEM) was a stored program computer, and therefore the first electronic stored program computer. There were probably mechanical predecessors though.
Looking back, there is a continuity of calculating machines all the way back to the abacus, each of which was the first to implement some notable feature of modern computers.
I agree completely about the risks of strong bad government. Overthrowing such beasts is almost impossible - look at Iraq.
I agree with what you say about most effort in many liberal regimes (and most especially the US) being directed at keeping government weak. I think this has been an unspoken bu longstanding principle for some time - not only must we forbid that the government take certain powers, we must ensure it remains weak so if it does take them, we can overthrow it.
My concern with this is that weak bad governments, being unscrupulous, have a very easy time turning themselves into strong bad governments, especially if people are angry or not particularly aware of what is going on.
The situation the US is in now, where it has a moderately strong government with a widely questionable record, raises the question of what to do to make it at least a good government, if not a strong good government. If people set out to weaken governments on the ground that they are not particularly good, I think its quite possible large portions of the electorate will get upset because certain tasks they think the the responsibility of the state are not being filled. That increases the risk of unscrupulous, and probably bad, governments being elected by promising to clean up the mess. This is what looks like it might happen in Russia.
I appreciate the point about the nature of US administrations, and to a lesser extent all western governments since around 1950.
Nonetheless, I do believe they are *relatively* well-behaved, especially when you look at the alternatives. I don't live in the US, so maybe I underestimate the degree to which are behaving badly, but I do still think that limiting the informational resources of even only slightly benign government might increase the chances of the public supporting one that will turn out not to be benign at all.
There's something I don't get here. I know its a long standing conviction of Anglo-American civil libertarians that this kind of thing is Bad News. I am more or less in agreement with that, but thinking about it a bit more, what exactly is wrong with it ?
It seems to me that measures like this increase the ability of the government to enforce the law, regardless of what that law actually is. That makes the assertion that this will help to catch drug smugglers/terrorists/child molersters/people with green hair/tomorrow's public enemy number one, reallistic. It also, of course, means it can be used to implement arbitary and stupid laws, such as, for instance drug prohibition (just to pick a nice uncontroversial example).
However the problem is not the databases and so on themselves, but the laws they are being used to implement. If the US had the sort of utterly minimal libertarian code of laws that many people here probably favour, would measures like this still be a problem ?
I'm concerned about this, because I do believe that if we stop relatively (note thats relatively) well-behaved governments from doing their jobs effectively, we may well be increasing our chances of getting larger, more unwieldly, more tyranical and more expensive governments in their place.
Oh, and please don't reply with some stupid platitude and privacy and freedom. I know that's what is believed, I want to know why it is necessarily true.
I wasn't trying to make a fair comparison. I was trying, apparently unsuccessfully, to illustrate the stupidity of making unfair comparisons.
The person to whom I was replying was trying to argue that Linux won't "fail" like Java because its better designed. Firstly either Java hasn't failed or I've halucinated every project I've worked on for the last 2 years.
Secondly, the comparison is completely spurious, and this is what I was trying to illustrate. Linux does something fairly well understood and quite commonly implemented - its a Unix-like kernel. He was probably refering to the entire "operating system", not the kernel. If so, you can't even really argue its universally well designed and implemented. Linux is not just the Linux kernel.
Java's AWT tries to be a cross platform GUI framework. This is hard, infrequently attempted, and when it is attempted in invariably has only marginal success. It did is to within the state of the art, which is not very good, and then swung (pun intended) back to the other known solution to the problem. And Java is not just the AWT.
You can't tar the whole OS with the brush of the kernel, nor the whole class library with the brush of the hardest part. Not can you fairly compare the engineering in a fairly well-understood task with that in one that is universally seen to be difficult.
If you want it covered, submit it, rather than posting offtopic garbage. If you want to winge, go somewhere else.
Oh yeah, Xlib is just so much easier to program in than the AWT.
In spite of the flurry of bile directed at Sun for Java that this posting willing no doubt generate, the comparison between Linux and Java here is actually not bad.
Java did not take over the world, but its a pretty good language and is being very widely used. In fact its the fastest uptake of any programming language I've ever seen, in spite of its problems.
Similarly Linux will never take over the world. It might get a 30 or even 70 percent share, but it will never completely obscure the possibility of using anything else the way Windows has in some places. Its not really within anyone's interest or ability to do that.
The problem with the article is that it assumes, as many people here do, and MS's and Sun's marketing machines pretend to, that the purpose of some technologies is to take over the world. It isn't, and by and large they can't. You wouldn't want them to anyway. Java hasn't failed because we've not all started to use Solaris (god forbid, in fact), and likewise Linux won't have failed if some people keep using NT. Articles like this are based on a false perception that there is some winner and some loser in technology battles. It never turns out that way, even if thats what the participants want.
Fine if you are programming bear metal on your own. Not so fine if you're in a team. Especially not fine if you want to reuse code. Most especially not fine if the language is bearly usable without it (templates in C++).
Interfaces distinguish between the subtype hierarchy, which should be publicly visible so that users of a package know what objects they can treat in the same way, and the inheritance hierarchy, which is all about things being implemented the same way, which you should not need to know.
Multiple inheritance of implementation is generally thought to be bad because its actually very rare to find classes you can mix together sensibly. If you want to know that two classes with the same interface fulfill certain rules, then I would say that Eiffel-style conditions are the right way to go. If you want to share implementation, delegate to another object.
I'm not convinced namespace protection matters. If your classes are that big, I suggest they're too big.
Nice quote, by the way.
Smalltalk development environments are still the best ever created, and most of the work was done more than 20 years ago now. The language is OK. The lack of typing makes it hard to have much confidence in Smalltalk programs, and the syntax has its good and bad points, but I'd probably still choose Smalltalk over Java.
Regrettably there's not much choice these days. ObjectShare have gone even further into VB-land and lost the last few really serious projects to use it. Java is good in that it keeps many of the good things about Smalltalk, although I find it hard to understand why so little work has been done bringing the runtimes up to Smalltalk standards. The worst feature of Java is there distinction between reference and primitive types. You simply don't need to do this, the supposed "inefficiency" can be removed by the compiler before it even gets to the runtime.
Ummm ... Java removes most of the widely-reputed problem areas of C++ (excessive syntax, badly implemented generics, multiple inheritance, operator overloading and pass by value).
If you know of others, please enlighten us.
Its a shame this article doesn't say what "real world programming concepts" people need to master.
Its my experience that many otherwise very competent programmers have huge holes in their knowledge, and the primary one seems to be ability to do decent abstract design. Teaching people C emphatically won't help with that. Even if you don't like OO, something like Ada or Standard ML is going to be a lot more use that C.
If the above is a reference to a lack of knowledge about real algorithm design and implementation, and an overreliance on libraries, then I agree. People need to learn that stuff, or they'll get stuck when the library of choice is missing the function they want. Nonetheless, if you have libraries, you should use them. There's nothing more irritating than programmers who faff around reimplementing stuff thats already in the standard libraries for "efficiency reasons" (not that they've profiled it or anything).
ML is semantically nice, but syntactically its evil. "An undistinguished ALGOL-family syntax" is how one of the standard's author's put it. Trust me on this, I've written an interpreter. The syntax is evil.
More seriously, ML is one of my favourite programming languages, but you and I were taught ML as a first or second language because we were CS students. It gets you out of nasty imperitive ways of thinking, and gives you a different way of seeing problems.
Most of the professional computer programmers I know would have trouble understanding the recursive pattern matching style which is typical of ML. It generally takes a good few weeks of teaching before CS students get their heads around it, and some never do.
The goal of a project like this is to teach people how to do a bit of simple programming, so they can get more out of their hardware, not to start them down the road to a CS PhD.
You have never read 1984 etc I must assume. Privacy is a fundamental basis of humanity. Take away that, and all the rest will follow. Why not individuality next? who needs that? we are all good at heart, so lets take down our walls and drone for the greater good, mindlessly tending the machines of society. No thanks, I'll continue to do my legal activities in the privacy of my own home.
This is the wrong way round. Tyrannical governments will inevitably violate people's privacy to enforce their tyranny. A government can have spy sattelites and refrain from trying to control what you do in bed just as easily as it can have an army and refrain from trying to block the streets.
Do you really imagine that if the government really did become despotic, they would refrain from surveilance because you didn't like it ? I really cannot see the logic here.
Resources that can be beneficial in the hands of a a broadly liberal state (look it up, OK, it doesn't mean what you think) are the tools of oppression under a tyranny. You don't avoid tyrrany by binding the hands of the liberal state. The more you make it impossible for necessary government to do its job, the more likely it is your moronic fellow countrymen will support unnecessary government.
I don't know very much about MIPS in particular, except that it started out as an especially radical approach to RISC (I think they tried to eliminate all pipeline interlocks - I don't know exactly what happened to that).
But I must argue with the idea that adding SIMD to MIPS, was neceaarily against the spirit of the project just on the grounds that "its not RISC".
RISC is to microprocessor design what OO is to software engineering. Everyone believes in it. Everyone thinks its a Good Thing, but noone can ever quite agree about what it entails doing in practice.
The original RISC projects at Stanford and Cambridge, took the (huge and crufty) instruction sets of the day, and analysed the instruction use frequency of compilers for certain languages (C mainly, but I think the Stanford group also looked at LISP and Smalltalk). The conclusion was that for C at least, a lot of silicon real estate could be freed by losing lots of instructions and addressing modes, concentrating only on those that were frequently used and making them fast by using more space in their implementation. It worked, but the optimisations made were pretty much the same in all RISC processors. They all tended to have a load store architecture, a large register file, relatively deep pipelines, and instruction and data caches.
It is this common feature set of RISC processors, which most people ended up understanding as the essence of RISC, rather than the careful empirical understanding of instruction usage that lay behind them.
In the design of the 486, Intel realise that you can apply these ideas to the design of CISC processors to. They separated the memory parts of the CISC instructions from the rest, added more registers (but hid them from the programmer), pipelines the system and added a cache. Hence the claim at that time that "RISC is dead/irrelevant/pointless". From that point on, no microprocessor designer has needed to go back to the principle of reducing the instruction set to gain speed. Mainly, this is because people have been much more careful about adding new instructions. Nonetheless the Alpha, and the modern MIPS processors have more instructions that anything that existed when the idea of RISC was first floated.
A lot has happened since then. Mainly the number of transistors on a chip has continued to double every 18 months, and inevitably that has led to a quest for ways to use this space to get more speed. Increasing the number of functional units and the depth of pipelines is one way, but beyond a certain number of FUs, the amount of parallelism that can be extracted by the instruction scheduler in the processor drops of. To gain any speed beyond that point, you need a smarter compiler, and one way to help the compiler to be smart is to add vector processing instructions. The problem then becomes getting enough data into the CPU, which is what the Cray boys were always good at, as I understand it.
This says nothing about the specific case of MIPS. Maybe MIPS really had architectural features that were incompatible with vector instructions. Maybe they just bodged it. Maybe the political infighting stopped it from working. Nonetheless, the statement "RISC processors cannot do vectors" is wrong, since many now do, and anyway the term RISC has been made meaningless through the degree to which it is now universally accecpted. Underneath the instruction set architecture - which still is now almost always very complex - there are no CISC processors any more.
Following these research findings the Council of Ministers has decided, on the advice of the vetinary subcommittee, and the subcommittee in charge of outlawing things that might scare some people, to ban mice from all EU member states.
Dr Klaus van Hoopklinger, chairman of the investigative subcommitee for researching spurious reasons to outlaw foreign products announced "Given the possibility that some mice will undoubtedly be genetically engineered for nefarious commercial reasons by our American allies, and given that European member states are currently carrying out a 10 year extensive yet innevitably inconclusive effort on the effects of superintelligent mice on the ecosystem and dietray health of EU citizens, we see no alternative but to ban all mice".
The decision is expected to be ratified by the European Parliament and made law by the governments of the member states with little opposition. A massive, Europe wide, mouse extermination program, using orgnically grown poisons, is expected to begin early in 2001, and customs barriers, including specially trained, selectively bred, sniffer cats, should be in place by early next year.
Funniest thing I've read for ages. Glad to see someone else has noticed the amazing resemblance between certain libertarians and the people they seem to hate most (Marxists and Fascists).
This article seems to contain just as much prejudice and ill-judgement as the rantings it criticises. To quote:
When the system was moved to Red Hat, the scripts broke. Thousands of others had problems when the move to glibc was rushed by Red Hat while other distributions remained cautious.
In many ways, Linux is a FreeBSD clone
But many are simply curious about why a new user would choose Linux over FreeBSD, despite FreeBSD's technical superiority.
Unlike Linux advocates, FreeBSD advocates do not believe FreeBSD should be running on every microchip.
Most negative talk about FreeBSD is baseless and intended to destroy, scare, or subvert potential users
All of these statements say or imply things about Linux or Linux users that are not generally true. Most ranting by Linux zealots about FreeBSD is based on prejudice and misinformation. Repeating the same mistake in the other direction doesn't help.
The BSDs and Linux are all excellent, and very similar, operating systems. The only way to compare them is with strict technical or legal comparisons. Rhetoric doesn't help anyone to make decisions.
To the extent that Unisys is a single entity, yes, you have the right to communicate your feelings to it. You can even swear at it, or insult its parentage, if thats the kind of thing you enjoy.
However it is impossible to communicate with the entity called Unisys, you have to talk to its employees instead. Unfortunately mail from ordinary people sent to large companies does not get read by the people responsible for their decisions (and in this case, I might add, Unisys isn't doing much wrong). It gets read by other ordinary people, who really cannot do very much about said decisions. If you insult them, then frankly, thats just rude, and you should apologise.
They aren't OPEN sourcing it, they are giving it the boneheaded license they've used for their other "no where to go but into the ground" projects. Who wants to work for free to increase Sun's profit margins? The use of the term "community" in the license is laughable.
Where do you get off with this stuff ? They get money, they also get to enforce compatibility. For some technologies, they let you ship non-commercial releases. Big deal - its their code, and with Java at least I can see where they are coming from.
Since when do the second fastest growing programming language in the world, one of only two Unix office suites of any quality, and one of a very few genuine efforts at universal plug and play have nowhere to go but the ground ? You mean they can't make money ? Do you really think they'd make money if they made them open source ?
People who would rather pay a little or look at some adverts to write a letter to their gran than shell out $x00 for MS Office. I barely use a word processor, and fond as I am of LaTeX, I don't like it for certain purposes. I *might* use such services, depending on how they are set up.
Seriously, there are a lot of people betting serious money on this "application publishing" stuff, and there does seem to be a market.