Agreed. The idea that deurbanization is something that should be blandly promoted is getting pretty old. I love how off-the-cuff the remark in the original post is; it's almost as if we're all convinced that the cities should be emptied out, and only details remain.
The anti-urban crowd have also popped up for a series of idiotic, tasteless suggestions about the WTC bombing (perhaps if we didn't work in these big downtowns, we'd all be much safer).
Face it, guys. There's a reason that people pay megabucks for downtown office space in big cities: promixity to other real people. We've had Internet access, videoconferencing, Kinkos, etc. in suburbia/exburbia for years and somehow the city centers refuse to empty out.
This not to mention the very real externalities imposed by deurbanization; you know, chewing up green space, the inevitable commute once it turns out that big-screen TV and DSL fail to substitute for a social life and a vibrant work environment, etc. etc.
If you want to live a life as a wired hermit, good for you, but don't expect too many people to join you any time soon.
These are pretty pictures, but in general, trying to embed this kind of information in 2-space (or 3-space, for that matter - adding a dimension doesn't really help) is mostly futile.
I learned much more from the "bowtie" representation of the Web (that study that - roughly - divided Websites into a "mainstream", sites that linked into the mainstream but were not linked to from it, sites that were linked from the mainstream but didn't link back into it, and sites that were in isolated islands). That was nice, and used some smart analysis, rather than a huge dump of complex information onto the printed page.
CmdrTaco appears to be one of those people out there who have a rather confused notion of how severe sentences actually are. This is the second posting about how 18 months in juvie or 240 hours of community service + a criminal record amounts to a slap on the wrist.
This is pretty dumb. Jail is boring, obnoxious, demeaning and occasionally dangerous, particularly for these type of people. A sentence of several months is not a slap on the wrist. Community service sounds about right.
I thought we'd seen the high-water mark for these kind of encroachments before the Skylarov case. This fresh enormity, and Abobe's little "push for prosecution, then wash their hands of it" have convinced me that fair-minded, above-board activity to oppose these idiots doesn't go far enough. Given the incredible degree to which the MPAA/RIAA and all the other corporate whores are willing to go to corrupt our basic rights, I say we're thoroughly justified in pirating their music/software.
This is a big step for me. I'm against piracy on principle, and prefer the convenience of just going out and buying the product rather than futzing around with Napster or it's sucessors. However, with every music CD I buy, or DVD I rent, some portion of the money I'm spending is being used to erode my liberties. To hell with that. I probably should boycott, but I don't feel particularly inclined to make my life uncomfortable and principles are clearly getting thrown out the window on the other side, so what the hell.
Maybe a less profitable music/movie industry would have less money to hire lawyers and congressmen.
All-assembly programing and other odd hobbies
on
MenuetOS Debuts
·
· Score: 2
Well, I suppose this guy deserves praise for having achieved something unusual, but as far as I am concerned, this is of the order of building a 50ft Statue of Liberty replica in one's back garden using beer cans. Difficult, possibly absorbing, oddly impressive, and utterly useless.
There are a whole bunch of reasons not to do this.
1. It's not efficient, even in the limited sense "producing code that runs faster than the opposition". On modern architectures, using pure handcoded assembly only tends to have a performance benefit when you know something the compiler doesn't (say, some kind of aliasing relationship among variables in a function) or when you can use instructions that a compiler can't. Rewriting a hot-spot in pure asm is one thing, but trying to beat the compiler overall is quite another. This disparity is even more pronounced in the world of RISC/EPIC... if you doubt me, try beating global register allocation for a function with 200 basic blocks and a thousand live ranges. By hand. Good luck...
2. It's not in the slightest bit portable. Not to other architectures (obviously) - and it's liable to suffer from immense performance problems even moving to other implementations of the same architecture with different latencies and numbers of functional units.
3. It's been done already, ages ago. This is how people used to build operating systems. There are good reasons why they stopped. There is nothing innovative about doing another Commodore 64 operating system almost 20 years after the fact. The constant and strange references to color and sound sort of reinforce this impression, too.
4. It doesn't illustrate any particularly interesting principle. Using a medium-level language (or even a high-level language) to build a toy OS from the ground up allows you to concentrate on OS principles, while still getting down to the bare metal for the code that has to be written in asm. Writing a boot-loader or a well-optimized string copy function is an illuminating task (well, at least the first time you have to write one, anyhow). However, once you are a competant asm programmer, writing merge sort in asm instead of C will teach you nothing about asm or merge sort that you didn't already know. And one could use the time saved to learn the basic principles of optimization, which, based on his remarks about Linux, the author of this system clearly doesn't know.
5. Asm is difficult to write and debug. No compiler to catch obvious errors. This is so obvious to require no discussion.
6. (addressed only to those people who think that this leads somewhere practical - if you merely admire it as a beer-can-Statue-of-Liberty, ignore this one) It'll Never Work. I'm sure it does just fine as a toy system, but a POSIX layer? A TCP/IP stack? Ummn, no. It's simply never going to happen. There are plenty of tiny OS's around on which lean systems could be based, but they were written and designed to be dense and elegant, not to prove some weird point about asm programming.
Re:But it gets you compact code
on
MenuetOS Debuts
·
· Score: 2
Actually, it's unlikely that by the time Moore's law slows done, that hand-coded assembly will be very common at all. Even now, most RISC compilers do a better job than hand-coders unless there are specialized instructions to be taken advantage of or some special knowledge that the coder has that the compiler doesn't.
Take a look at the hardware documentation for the 21264 pipeline or the architecture documentation for the IA-64 and think about how much fun it would be to write reams of hand-coded assembly for them.
If Moore's law slows down, I suspect that we'll see even more work on aggresive optimizing compilers and profile-driven optimization, not on trying to hand-optimize general-purpose code for 16-way issue multithreaded speculative out-of-order... architectures.
Ok, that's an overly inflammatory subject header. However, I once had a chat with a friend and we tried to work out what practical effect it would have on the world if you could solve NP problems reliably in polynomial time. I'm sure a lot of things would become slightly better, but neither of us could think of any revolutionary new applications that would become possible that weren't previously possible.
Remember that a lot of the problems that are in NP can be approximated pretty well. If you were actually routing travelling salesmen around the US, you might not mind a solution that's off by a few percent (particularly when there are going to be a whole pile of other sources of error; your salesmen getting stuck in traffic jams or dallying with housewives, etc.).
Can anyone offer some problem domains where quantum computing would offer more than an incremental improvement (discounting crypto, which seems to be a case of gain a little, lose a little)? I'm not claiming that such domains don't exist, btw. I'd be delighted if someone could point out a few.
I doubt that Microsoft will continue this course of action.
Reading that article about how Microsoft is up to 86% market share (and still rising), I wouldn't be suprised if, in the next couple years, that Microsoft itself starts trying to make life a little easier for Netscape users. Why? Same reason that they've bailed out Apple and other, safe, weak competitors.
99% market share in the browser market isn't that much better than 86% for all practical purposes, and brings the attention of the Federal Government rather sharply.
Re:Will game consoles kill family life a little mo
on
First Looks At XBox
·
· Score: 1
Huh? Sports tend to involve far more socialization (not all of it good, mind you) than games, and develop gross-body coordination and fitness.
The idea that the adults should blatantly pander to the game consoles by "hanging around" their kids in a desperate attempt at togetherness is ridiculous. It seem to stem from a peculiar, very American idea that if kids want to spend endless amounts of time on the Playstation or in front of the TV, there's nothing that can or should be done to stop them.
That being said, I don't think that the realism of the X-box is going to make any difference. Games have reached the level of sophistication that you can play them for hours - they did that a long time ago for most people. Hell, I wasted long, long hours on Elite on the C-64.
"That 400 megahertz processor operates on about 4 times more CPU instructions per clock cycle than your X86 chip."
In theory, yes, but in practice Sparcs have never been very successful at getting the theoretical issue rate out of real code. The fight between "brainiacs" (like SPARC & PPC which try for high issue rates at the expense of clock speed) and "speed demons" (like the Alpha, which tries for the highest possible clock rates) is constantly won by the speed demon camp. Look at the historical performance of the Sparcs on (say) the SPECmarks (which granted, are not perfect, but they're certainly adequate for evaluating CPU cores if not memory systems).
Well, no need to get snotty. Don't assume that if you like something complicated, anyone who doesn't like that thing obviously isn't intellectually up to that level of complexity. That style of writing is not to everyone's taste; I found it pretentious and tiresome. I don't like Douglas Coupland either, so I guess our tastes diverge.
Personally, I'd rather have heard a little more about the actual facility and the science behind it and a little less in the way of wild philosophical speculations, references to the "copulation of fusion" and other such drivel.
Interesting, if you want to join a Language Jihad
on
Linux -- Without Unix
·
· Score: 1
This is an interesting idea, and I guess we should praise Mr. Tonneau for a great deal of originality and a lot of hard work. Beyond that...
Quite frankly, I see some serious flaws with the reasoning behind this project. It seems that Mr. Tonneau believes that the reason a lot of Free Software is bad (read big, buggy, etc) is not bad design or hasty implementation, but because it was written in C or C++.
He makes some very strange remarks about the desirability of language flexibility; talking about how limited C/C++ is because you can't build your own "for" operator. This strikes me as completely bogus. Yes, I understand that it is possible to extend a language in this way.
However, your brand new language construct (say "really_cool_for") either remains private to your own code - in which case you increase the cognitive load of someone else reading your code - or it achieves some level of public use, gradually adding to the size of the language in common use. Eventually, you'll have several loosely overlapping versions of the same language - all perfectly legal of course, according to the way he's defined the language - just not necessarily intelligible.
Granted, C and C++ can feel like this at times. The first time you hit someone else's wonderful object hierarchy with loads of overloaded operators and inheritance or macros can feel like you're programming in a new language. However, you're not.
He also points out that while C++ can define new classes (eg. complex), it doesn't really understand how to optimize them. Instead, he offers, we should tell the compiler what optimizations are possible.
His example, as it happens (concerning complex numbers - implementing x*2 as x+x) is simply wrong, suggesting that he doesn't really understand what's happening in modern compilers very well (constant propagation and inlining would take care of this example quite well). However, let's assume that we're talking about more complicated classes, which the optimizer cannot handle so easily.
Even in this case, such a feature is broken. The "how to optimize this code" module becomes part of the code; yet another thing that can be done wrong, or that can get out of step with the actual code. This will wind up as a glorified #pragma feature, with all the same problems.
In fact, I'd say the the language extensibility will make it very hard to do any complicated optimizations on the code at all. About the only thing that makes interprocedural optimizations doable in C/C++ is the fact that procedure semantics are almost always preserved in C/C++; when they aren't, it's obvious (exceptions, setjmp/longjump, a call to exit). If you can magic up your own funky interprocedural control structures (maybe because you _really_ miss continuations), then either the optimizer punts, or you have to build a nice little language so that you can tell the optimizer just what the hell is going on with your gnarly new control structure.
There are serious interactions between new language features and optimizations; garbage collection interacts with practically everything, for example. Dealing with these interactions in conventional compilers (where you know what language features exist) is hard work for smart compiler guys. Doing it automatically is probably going to be just too hard.
Another thing: I find it a little odd that the guy can write several paragraphs on how hard it is to write code across multiple platforms, a few more on some of the more obvious flaws and weaknesses of C/C++, and not mention Java. Yes, I know his language does rather more things than Java (most of them bizarre). It's just a bit disingenuous to go from talking about more obvious C/C++ lossage (that's addressed, however well or poorly, by Java) to "you need to use my language", without mentioning that C/C++ are hardly the only game in town.
And finally, what I thought was the funniest quote by far:
"For big-project managers, Pliant resembles an ideal linker since you can write various
pieces of a project in various programming
styles, with each part using the optimum style
for its functional needs..."
How's your blood pressure, "big-project managers"?
Hmmn. I can't wait to see the People's Own Chip, designed entirely without worry about whether or not existing fab technology can actually build the chip at a price that makes it worth building.
I hate to break this to you. You better sit down. Ok, good. There is no Santa Claus, you're adopted, and Intel is in the business of making money of its chips, not designing the coolest chip possible whatever the expense.
you have just implemented a really cool application for your next generation (Inferno) web phone or PDA and you want your investors, or
potential customers to have a look at it.
How do you go about it? Screen shots? PAH! Put it on your web site and let them play with the real thing.
Ummn, this seems like a bit of a fantasy.
If the Web phone or the PDA isn't actually built yet, you'll be probably faking up the interface in Flash (pick your tool), for marketing and practical reasons. After all, all of the rest of the system probably isn't built either - the hardware, the server if it uses on-line resources, etc. You probably built a nice demo for the program to sell the program to management and do early user testing - that's probably a lot better suited to this kind of application than is getting the application itself running like a fish out of water on a PC screen.
If the hardware does exist, and the potential customer exists, why not give people a demo version, or let them play with the device in a store? That way they can actually experience the whole interface experience, not a simulation of what it would look like to fetch movie times on a tiny, unreadable screen for more than the cost of a newspaper in connect time (or whatever people are doing with next-gen web phones these days).
I just found Populous II (in my mind, better than I) on www.theunderdog.org. This is a PC port and works under Win 98 if you enable expanded memory.
Populous III is a botch. The idiots who designed it essentially threw away the part that made Populous I & II interesting - namely the autonomy of your little worshippers (you can't just give each one orders). They turned it into a Command & Conquer clone, and seemed offensively proud of having done so (I saw a review where they basically said "screw you" to Populous II purists). It's pretty sad when the gaming community understands why a game worked better than the developers of a sequel.
Actually, like you suggested, the chances are pretty low. Most of China's landmass is very sparsely inhabited. Amazes me, but its true. Ah well, if you can fit 300 million into Indonesia, nothing should suprise me.
Whaa? "the real problem in the cache world is miss rate and branch prediction, not capacity" indeed.
How do you suppose things get kicked out of cache? Little elves? While conflicts are still a problem (particularly if you don't do I-cache optimizations and/or have low associativity), increasing the capacity of the cache makes a lots of problems go away. There's a reason Intel gets to charge more for the Xeon, you know, and it's not just the groovy name.
Well, actually, given the number of crypto-freaks that I've met that seem to enjoy using encryption whenever possible, I suspect that a lot of what would get revealed would be pretty picayune.
One could even make a case for totally gratuitous use of such features (anonymizers for e-mail and IP, crypto, etc.) as a means of helping to conceal genuine uses of same. You never know when you yourself might need them for real.
I agree with you here. I myself connect to my home server through Port 25 - why should I have to change my e-mail affiliation (or set up some dodgy forwarding thingy) every time I change an ISP? If my cable modem company said otherwise, I'd say bye-bye.
This seems like an area where a faster, tougher response to spam would greatly reduce their problems with it. If UUnet were to have a working group that spanned the legal, sales and abuse departments that pretty much responded to spam within 12 hours (or some similar short time) and expedited dealing with it, sure, that would cost them a lot more money.
On the other hand, they'd ultimately have less work to do; how many spammers would use the service if they knew that they'd get about 24 hours of use.
A further twist would be placing some sort of brake on large amounts of outgoing mail - perhaps every 10 complaints received reduces the # of messages per hour by 10% that UUnet will handle from these people (or further, artificially choking off the bandwidth of outgoing packets that are directed at port 25 - although that might be infeasible technically). If it turns out the complaints are not well-founded, then the brake could be removed.
Of course, if the ISPs are colluding with the spammers, there's not a whole lot one can do.
If I've got a magnetic field, then presumably something is generating the magnetic field, right? If this is working off Earth's magnetic field, then the "reaction" is that it's giving the planet Earth a tiny push - similarly for any other magnetic field that this operates in.
This sounds like a competent job, at the least. Am I the only one feeling faintly dizzy at the idea of people taking the source material of science fiction movies somewhat seriously?
I'd be happy to see an end to the days where fantasy and science fiction books were pillaged for a few cool gimmmicks, while having their plots and characters completely eviscerated. Witness the Lynch Dune (I guess I'm one of those "few" who hold some real animosity towards it - see the CNN sidebar) and Bladerunner.
[ Before I get flamed to a crisp, I must point out that I enjoyed Bladerunner, but the "androids are people too" sentimentalism of the plot was pretty much a Scott creation]
It will be interesting to see if non-Herbert readers can make any sense of the mini-series, though. It's a astonishingly complicated book, even for a six-hour mini-series. I guess I'll have to wait for it to make it to Australia...
"I'll grant that Unix may not have been what it was without Ken, but was the institutional culture such at Bell Labs that Bell Labs would have come out with something similar to Unix even if he had never been involved?"
In a word, no. I'm sure you think you're being very clever as you ask this rhetorical question. However, if you had even the faintest idea about the early history of Unix, you'd realize how off-track you are being. Without Ken Thompson and Dennis Ritchie, there would have been no UNIX or UNIX-like operating system at Bell Labs at all.
Bell Labs management didn't ask for a simple OS coded in a HLL at all. KT and DR just went ahead and did it for their own reasons. I do attibute the success of Unix to "institutional culture" (Bell Labs providing a happy little sandbox for people to play in, Bell Labs not being able to commercialize Unix immediately because of the consent decree). However, no such project would have ever happened without the individual contributions of two very unique and talented personalities.
By all means criticize Slashdot heroes, if you like. I don't worship everything that came out of the Unix Room at Bell Labs or idolize KT et al. But get your facts correct, first.
This is simply incorrect. Ken Thompson wrote the first C compiler; the design of C (as far as I remember) was shared between him and Dennis Ritchie.
Ken is a fantastically talented person with a talent for seeing in one step what would take a normal person several plodding steps. Wonderfully weird sense of humor, too.
I've heard about these kind of free wireless networks before, and it sounds pretty cool. One thing I don't understand is this: how do these wireless nets connect to the mainstream internet? More specifically, who pays for the link?
I have a always-on connection, but my ISP would crucify me on the telephone pole outside my house if I started routing random wireless user's packets through their network.
For an experimental setup, there are plenty of universities and the like willing to donate bandwidth. I don't see this scaling, though.
Slashdot Mirror
Agreed. The idea that deurbanization is something that should be blandly promoted is getting pretty old. I love how off-the-cuff the remark in the original post is; it's almost as if we're all convinced that the cities should be emptied out, and only details remain.
The anti-urban crowd have also popped up for a series of idiotic, tasteless suggestions about the WTC bombing (perhaps if we didn't work in these big downtowns, we'd all be much safer).
Face it, guys. There's a reason that people pay megabucks for downtown office space in big cities: promixity to other real people. We've had Internet access, videoconferencing, Kinkos, etc. in suburbia/exburbia for years and somehow the city centers refuse to empty out.
This not to mention the very real externalities imposed by deurbanization; you know, chewing up green space, the inevitable commute once it turns out that big-screen TV and DSL fail to substitute for a social life and a vibrant work environment, etc. etc.
If you want to live a life as a wired hermit, good for you, but don't expect too many people to join you any time soon.
These are pretty pictures, but in general, trying to embed this kind of information in 2-space (or 3-space, for that matter - adding a dimension doesn't really help) is mostly futile.
I learned much more from the "bowtie" representation of the Web (that study that - roughly - divided Websites into a "mainstream", sites that linked into the mainstream but were not linked to from it, sites that were linked from the mainstream but didn't link back into it, and sites that were in isolated islands). That was nice, and used some smart analysis, rather than a huge dump of complex information onto the printed page.
CmdrTaco appears to be one of those people out there who have a rather confused notion of how severe sentences actually are. This is the second posting about how 18 months in juvie or 240 hours of community service + a criminal record amounts to a slap on the wrist.
This is pretty dumb. Jail is boring, obnoxious, demeaning and occasionally dangerous, particularly for these type of people. A sentence of several months is not a slap on the wrist. Community service sounds about right.
I thought we'd seen the high-water mark for these kind of encroachments before the Skylarov case. This fresh enormity, and Abobe's little "push for prosecution, then wash their hands of it" have convinced me that fair-minded, above-board activity to oppose these idiots doesn't go far enough. Given the incredible degree to which the MPAA/RIAA and all the other corporate whores are willing to go to corrupt our basic rights, I say we're thoroughly justified in pirating their music/software.
This is a big step for me. I'm against piracy on principle, and prefer the convenience of just going out and buying the product rather than futzing around with Napster or it's sucessors. However, with every music CD I buy, or DVD I rent, some portion of the money I'm spending is being used to erode my liberties. To hell with that. I probably should boycott, but I don't feel particularly inclined to make my life uncomfortable and principles are clearly getting thrown out the window on the other side, so what the hell.
Maybe a less profitable music/movie industry would have less money to hire lawyers and congressmen.
Well, I suppose this guy deserves praise for having achieved something unusual, but as far as I am concerned, this is of the order of building a 50ft Statue of Liberty replica in one's back garden using beer cans. Difficult, possibly absorbing, oddly impressive, and utterly useless.
There are a whole bunch of reasons not to do this.
1. It's not efficient, even in the limited sense "producing code that runs faster than the opposition". On modern architectures, using pure handcoded assembly only tends to have a performance benefit when you know something the compiler doesn't (say, some kind of aliasing relationship among variables in a function) or when you can use instructions that a compiler can't. Rewriting a hot-spot in pure asm is one thing, but trying to beat the compiler overall is quite another. This disparity is even more pronounced in the world of RISC/EPIC... if you doubt me, try beating global register allocation for a function with 200 basic blocks and a thousand live ranges. By hand. Good luck...
2. It's not in the slightest bit portable. Not to other architectures (obviously) - and it's liable to suffer from immense performance problems even moving to other implementations of the same architecture with different latencies and numbers of functional units.
3. It's been done already, ages ago. This is how people used to build operating systems. There are good reasons why they stopped. There is nothing innovative about doing another Commodore 64 operating system almost 20 years after the fact. The constant and strange references to color and sound sort of reinforce this impression, too.
4. It doesn't illustrate any particularly interesting principle. Using a medium-level language (or even a high-level language) to build a toy OS from the ground up allows you to concentrate on OS principles, while still getting down to the bare metal for the code that has to be written in asm. Writing a boot-loader or a well-optimized string copy function is an illuminating task (well, at least the first time you have to write one, anyhow). However, once you are a competant asm programmer, writing merge sort in asm instead of C will teach you nothing about asm or merge sort that you didn't already know. And one could use the time saved to learn the basic principles of optimization, which, based on his remarks about Linux, the author of this system clearly doesn't know.
5. Asm is difficult to write and debug. No compiler to catch obvious errors. This is so obvious to require no discussion.
6. (addressed only to those people who think that this leads somewhere practical - if you merely admire it as a beer-can-Statue-of-Liberty, ignore this one) It'll Never Work. I'm sure it does just fine as a toy system, but a POSIX layer? A TCP/IP stack? Ummn, no. It's simply never going to happen. There are plenty of tiny OS's around on which lean systems could be based, but they were written and designed to be dense and elegant, not to prove some weird point about asm programming.
Actually, it's unlikely that by the time Moore's law slows done, that hand-coded assembly will be very common at all. Even now, most RISC compilers do a better job than hand-coders unless there are specialized instructions to be taken advantage of or some special knowledge that the coder has that the compiler doesn't.
... architectures.
Take a look at the hardware documentation for the 21264 pipeline or the architecture documentation for the IA-64 and think about how much fun it would be to write reams of hand-coded assembly for them.
If Moore's law slows down, I suspect that we'll see even more work on aggresive optimizing compilers and profile-driven optimization, not on trying to hand-optimize general-purpose code for 16-way issue multithreaded speculative out-of-order
Ok, that's an overly inflammatory subject header. However, I once had a chat with a friend and we tried to work out what practical effect it would have on the world if you could solve NP problems reliably in polynomial time. I'm sure a lot of things would become slightly better, but neither of us could think of any revolutionary new applications that would become possible that weren't previously possible.
Remember that a lot of the problems that are in NP can be approximated pretty well. If you were actually routing travelling salesmen around the US, you might not mind a solution that's off by a few percent (particularly when there are going to be a whole pile of other sources of error; your salesmen getting stuck in traffic jams or dallying with housewives, etc.).
Can anyone offer some problem domains where quantum computing would offer more than an incremental improvement (discounting crypto, which seems to be a case of gain a little, lose a little)? I'm not claiming that such domains don't exist, btw. I'd be delighted if someone could point out a few.
I doubt that Microsoft will continue this course of action.
Reading that article about how Microsoft is up to 86% market share (and still rising), I wouldn't be suprised if, in the next couple years, that Microsoft itself starts trying to make life a little easier for Netscape users. Why? Same reason that they've bailed out Apple and other, safe, weak competitors.
99% market share in the browser market isn't that much better than 86% for all practical purposes, and brings the attention of the Federal Government rather sharply.
Huh? Sports tend to involve far more socialization (not all of it good, mind you) than games, and develop gross-body coordination and fitness.
The idea that the adults should blatantly pander to the game consoles by "hanging around" their kids in a desperate attempt at togetherness is ridiculous. It seem to stem from a peculiar, very American idea that if kids want to spend endless amounts of time on the Playstation or in front of the TV, there's nothing that can or should be done to stop them.
That being said, I don't think that the realism of the X-box is going to make any difference. Games have reached the level of sophistication that you can play them for hours - they did that a long time ago for most people. Hell, I wasted long, long hours on Elite on the C-64.
"That 400 megahertz processor operates on about 4 times more CPU instructions per clock cycle than your X86 chip."
In theory, yes, but in practice Sparcs have never been very successful at getting the theoretical issue rate out of real code. The fight between "brainiacs" (like SPARC & PPC which try for high issue rates at the expense of clock speed) and "speed demons" (like the Alpha, which tries for the highest possible clock rates) is constantly won by the speed demon camp. Look at the historical performance of the Sparcs on (say) the SPECmarks (which granted, are not perfect, but they're certainly adequate for evaluating CPU cores if not memory systems).
Well, no need to get snotty. Don't assume that if you like something complicated, anyone who doesn't like that thing obviously isn't intellectually up to that level of complexity. That style of writing is not to everyone's taste; I found it pretentious and tiresome. I don't like Douglas Coupland either, so I guess our tastes diverge. Personally, I'd rather have heard a little more about the actual facility and the science behind it and a little less in the way of wild philosophical speculations, references to the "copulation of fusion" and other such drivel.
This is an interesting idea, and I guess we should praise Mr. Tonneau for a great deal of originality and a lot of hard work. Beyond that...
Quite frankly, I see some serious flaws with the reasoning behind this project. It seems that Mr. Tonneau believes that the reason a lot of Free Software is bad (read big, buggy, etc) is not bad design or hasty implementation, but because it was written in C or C++.
He makes some very strange remarks about the desirability of language flexibility; talking about how limited C/C++ is because you can't build your own "for" operator. This strikes me as completely bogus. Yes, I understand that it is possible to extend a language in this way.
However, your brand new language construct (say "really_cool_for") either remains private to your own code - in which case you increase the cognitive load of someone else reading your code - or it achieves some level of public use, gradually adding to the size of the language in common use. Eventually, you'll have several loosely overlapping versions of the same language - all perfectly legal of course, according to the way he's defined the language - just not necessarily intelligible.
Granted, C and C++ can feel like this at times. The first time you hit someone else's wonderful object hierarchy with loads of overloaded operators and inheritance or macros can feel like you're programming in a new language. However, you're not.
He also points out that while C++ can define new classes (eg. complex), it doesn't really understand how to optimize them. Instead, he offers, we should tell the compiler what optimizations are possible.
His example, as it happens (concerning complex numbers - implementing x*2 as x+x) is simply wrong, suggesting that he doesn't really understand what's happening in modern compilers very well (constant propagation and inlining would take care of this example quite well). However, let's assume that we're talking about more complicated classes, which the optimizer cannot handle so easily.
Even in this case, such a feature is broken. The "how to optimize this code" module becomes part of the code; yet another thing that can be done wrong, or that can get out of step with the actual code. This will wind up as a glorified #pragma feature, with all the same problems.
In fact, I'd say the the language extensibility will make it very hard to do any complicated optimizations on the code at all. About the only thing that makes interprocedural optimizations doable in C/C++ is the fact that procedure semantics are almost always preserved in C/C++; when they aren't, it's obvious (exceptions, setjmp/longjump, a call to exit). If you can magic up your own funky interprocedural control structures (maybe because you _really_ miss continuations), then either the optimizer punts, or you have to build a nice little language so that you can tell the optimizer just what the hell is going on with your gnarly new control structure.
There are serious interactions between new language features and optimizations; garbage collection interacts with practically everything, for example. Dealing with these interactions in conventional compilers (where you know what language features exist) is hard work for smart compiler guys. Doing it automatically is probably going to be just too hard.
Another thing: I find it a little odd that the guy can write several paragraphs on how hard it is to write code across multiple platforms, a few more on some of the more obvious flaws and weaknesses of C/C++, and not mention Java. Yes, I know his language does rather more things than Java (most of them bizarre). It's just a bit disingenuous to go from talking about more obvious C/C++ lossage (that's addressed, however well or poorly, by Java) to "you need to use my language", without mentioning that C/C++ are hardly the only game in town.
And finally, what I thought was the funniest quote by far:
"For big-project managers, Pliant resembles an ideal linker since you can write various
pieces of a project in various programming
styles, with each part using the optimum style
for its functional needs..."
How's your blood pressure, "big-project managers"?
Hmmn. I can't wait to see the People's Own Chip, designed entirely without worry about whether or not existing fab technology can actually build the chip at a price that makes it worth building.
I hate to break this to you. You better sit down. Ok, good. There is no Santa Claus, you're adopted, and Intel is in the business of making money of its chips, not designing the coolest chip possible whatever the expense.
How do you go about it? Screen shots? PAH! Put it on your web site and let them play with the real thing.
Ummn, this seems like a bit of a fantasy.
If the Web phone or the PDA isn't actually built yet, you'll be probably faking up the interface in Flash (pick your tool), for marketing and practical reasons. After all, all of the rest of the system probably isn't built either - the hardware, the server if it uses on-line resources, etc. You probably built a nice demo for the program to sell the program to management and do early user testing - that's probably a lot better suited to this kind of application than is getting the application itself running like a fish out of water on a PC screen.
If the hardware does exist, and the potential customer exists, why not give people a demo version, or let them play with the device in a store? That way they can actually experience the whole interface experience, not a simulation of what it would look like to fetch movie times on a tiny, unreadable screen for more than the cost of a newspaper in connect time (or whatever people are doing with next-gen web phones these days).
I just found Populous II (in my mind, better than I) on www.theunderdog.org. This is a PC port and works under Win 98 if you enable expanded memory. Populous III is a botch. The idiots who designed it essentially threw away the part that made Populous I & II interesting - namely the autonomy of your little worshippers (you can't just give each one orders). They turned it into a Command & Conquer clone, and seemed offensively proud of having done so (I saw a review where they basically said "screw you" to Populous II purists). It's pretty sad when the gaming community understands why a game worked better than the developers of a sequel.
Actually, like you suggested, the chances are pretty low. Most of China's landmass is very sparsely inhabited. Amazes me, but its true. Ah well, if you can fit 300 million into Indonesia, nothing should suprise me.
Whaa? "the real problem in the cache world is miss rate and branch prediction, not capacity" indeed. How do you suppose things get kicked out of cache? Little elves? While conflicts are still a problem (particularly if you don't do I-cache optimizations and/or have low associativity), increasing the capacity of the cache makes a lots of problems go away. There's a reason Intel gets to charge more for the Xeon, you know, and it's not just the groovy name.
Well, actually, given the number of crypto-freaks that I've met that seem to enjoy using encryption whenever possible, I suspect that a lot of what would get revealed would be pretty picayune.
One could even make a case for totally gratuitous use of such features (anonymizers for e-mail and IP, crypto, etc.) as a means of helping to conceal genuine uses of same. You never know when you yourself might need them for real.
I agree with you here. I myself connect to my home server through Port 25 - why should I have to change my e-mail affiliation (or set up some dodgy forwarding thingy) every time I change an ISP? If my cable modem company said otherwise, I'd say bye-bye.
Prevention is better than cure, in this case.
This seems like an area where a faster, tougher response to spam would greatly reduce their problems with it. If UUnet were to have a working group that spanned the legal, sales and abuse departments that pretty much responded to spam within 12 hours (or some similar short time) and expedited dealing with it, sure, that would cost them a lot more money.
On the other hand, they'd ultimately have less work to do; how many spammers would use the service if they knew that they'd get about 24 hours of use.
A further twist would be placing some sort of brake on large amounts of outgoing mail - perhaps every 10 complaints received reduces the # of messages per hour by 10% that UUnet will handle from these people (or further, artificially choking off the bandwidth of outgoing packets that are directed at port 25 - although that might be infeasible technically). If it turns out the complaints are not well-founded, then the brake could be removed.
Of course, if the ISPs are colluding with the spammers, there's not a whole lot one can do.
If I've got a magnetic field, then presumably something is generating the magnetic field, right? If this is working off Earth's magnetic field, then the "reaction" is that it's giving the planet Earth a tiny push - similarly for any other magnetic field that this operates in.
Am I missing some important point here?
This sounds like a competent job, at the least. Am I the only one feeling faintly dizzy at the idea of people taking the source material of science fiction movies somewhat seriously?
I'd be happy to see an end to the days where fantasy and science fiction books were pillaged for a few cool gimmmicks, while having their plots and characters completely eviscerated. Witness the Lynch Dune (I guess I'm one of those "few" who hold some real animosity towards it - see the CNN sidebar) and Bladerunner.
[ Before I get flamed to a crisp, I must point out that I enjoyed Bladerunner, but the "androids are people too" sentimentalism of the plot was pretty much a Scott creation]
It will be interesting to see if non-Herbert readers can make any sense of the mini-series, though. It's a astonishingly complicated book, even for a six-hour mini-series. I guess I'll have to wait for it to make it to Australia...
"I'll grant that Unix may not have been what it was without Ken, but was the institutional culture such at Bell Labs that Bell Labs would have come out with something similar to Unix even if he had never been involved?"
In a word, no. I'm sure you think you're being very clever as you ask this rhetorical question. However, if you had even the faintest idea about the early history of Unix, you'd realize how off-track you are being. Without Ken Thompson and Dennis Ritchie, there would have been no UNIX or UNIX-like operating system at Bell Labs at all.
Bell Labs management didn't ask for a simple OS coded in a HLL at all. KT and DR just went ahead and did it for their own reasons. I do attibute the success of Unix to "institutional culture" (Bell Labs providing a happy little sandbox for people to play in, Bell Labs not being able to commercialize Unix immediately because of the consent decree). However, no such project would have ever happened without the individual contributions of two very unique and talented personalities.
By all means criticize Slashdot heroes, if you like. I don't worship everything that came out of the Unix Room at Bell Labs or idolize KT et al. But get your facts correct, first.
This is simply incorrect. Ken Thompson wrote the first C compiler; the design of C (as far as I remember) was shared between him and Dennis Ritchie. Ken is a fantastically talented person with a talent for seeing in one step what would take a normal person several plodding steps. Wonderfully weird sense of humor, too.
I've heard about these kind of free wireless networks before, and it sounds pretty cool. One thing I don't understand is this: how do these wireless nets connect to the mainstream internet? More specifically, who pays for the link?
I have a always-on connection, but my ISP would crucify me on the telephone pole outside my house if I started routing random wireless user's packets through their network.
For an experimental setup, there are plenty of universities and the like willing to donate bandwidth. I don't see this scaling, though.