You can choose to pay per-CPU on Sun servers, too. [...] Sun provides this pricing option to reduce the price-barrier to high-end servers when a customer truly cannot afford all the CPUs now
But, you see, the fact that Sun _can_ provide this option shows they are something of a monopoly. Imagine a hundred Sun competitors all selling mainframes. If it is profitable for Sun to ship the multiple-CPU box and charge (arbitrarily) for only one CPU, then it would be profitable for Sun or any of its competitors to just ship the hardware uncrippled at that price. One company would do it and the others would have to follow.
Look at the PC market - can you even imagine RAM manufacturers producing 128Mbyte modules and deactivating half of the capacity to sell at a lower price? Of course not, not unless they have some special cheap manufacturing process which their competitors cannot emulate. Intel, on the other hand, might be able to get away with deliberately downclocking its CPUs to maintain price differentials, the full-speed ones costing a lot extra. To a certain extent this already happens (CPUs tested at x MHz being relabelled to y MHz, to address shortages of the cheaper part without disrupting the market segmentation).
I know IBM is charging according to consumer costs, 'what the market will bear'. When a company can do this, it shows that company has market power and there is imperfect competition. In a fully competitive market the consumer cost will be equal to the manufacturing cost, because if it were any higher then competitors would step in and offer the same product at a lower price (while still making nonzero profit on each unit). Of course this situation never arises in practice.
Standard Oil charged what the market would bear. Microsoft charges what the market will bear (which is high for Office, but strangely much lower for things like MS Money which have strong competition). When the price companies can charge is higher than the marginal cost, this shows imperfect competition. In the real world all markets are like this, but to a greater or lesser extent. In the case of IBM mainframes the price is a lot higher than manufacturing cost, so it's reasonable to suspect IBM has a dominant position. This is borne out by other facts.
I deliberately didn't go into bad thing vs good thing in my original post - I was just pointing out that factually, IBM is either a monopolist in mainframes or rather close to it. It certainly has a monopoly in IBM mainframes:-).
The fact that IBM can make it cheaper for the customer by crippling the machine, and still make a profit, shows how high the original prices are relative to the manufacturing cost of the machine.
If there were two IBMs competing with each other, then one of them would surely realize: if we can sell a machine with one CPU activated and make a profit then we could sell it with six CPUs activated, at the same cost to us, and still make a profit. Plus we would take business from the other company. So unless the two companies colluded with each other, mainframe prices would come down to only a small amount above the hardware cost.
The fact that Sun and HP also have these pricing policies shows that their products are not perfect substitutes for IBM mainframes, or at least that they're not that interested in competing with IBM on price. Or maybe they would be interested but they want to keep the cosy price segmentation in their own market.
I'm not saying consumers would be better off if manufacturers had to ship uncrippled hardware. It would probably end up with everyone paying the higher prices, in order for the manufacturers to recoup R&D costs. But it's a fact that differential pricing usually indicates market power and less-intense price competition.
The problem might be if IBM Global Services markets itself as impartial but has a bias towards IBM products. Just like reports produced by Arthur Andersen on development projects tend, totally by coincidence, to recommend proposals put forward by Arthur Andersen.
It's clear that IBM has a monopoly in the mainframe market, or at least something close to it. You can deduce this simply by looking at their pricing policies.
If you buy a mainframe then it often comes with say six CPUs, of which only one is activated. If you pay IBM some extra subscription money they will send an engineer round to activate the second CPU, or up to all six depending on how much you pay. It costs them almost nothing to do this, and it would cost nothing extra to simply have all six enabled when the machine leaves the factory, but IBM charges extra for it.
You can consider this as market segmentation - selling the same product to different parts of the market and charging different prices, so as to squeeze the most out of each consumer. If there were plenty of competition in the market, then IBM would need to sell mainframes with the best price/performance possible and would ship with six CPUs by default, at a price close to the manufacturing cost. The fact that they can get away with this pricing scheme shows they have considerable market power, if not an outright monopoly.
A more positive way of looking at the situation is that the cost of a mainframe reflects less the manufacturing costs (marginal cost), and more the R&D effort that went into desigining it or the expense of building the factory (fixed costs). In this case IBM's charging different prices, despite the marginal cost to them being no different, is just like Novell charging different prices for a 10-client Netware licence and a 100-client licence. So IBM has a monopoly on that particular mainframe design in the same way Novell has a monopoly on Netware. This is still not ideal for the consumer, but it's often considered a necessary evil to provide incentive to invest in new designs.
AFAIK: The kernel doesn't use the glibc C library. It has its own memory management code which presumably the kernel zlib code uses. This memory manager may or may not guard against free()ing the same area twice.
But what is the use of zlib in the kernel anyway? Just to uncompress the vmlinuz image before the kernel starts? If so it's not much of a vulnerability, if you can corrupt the vmlinuz file then you can control the whole system anyway.
Under what circumstances might nuclear weapons be used under the new posture? The NPR says they "could be employed against targets able to withstand nonnuclear attack," or in retaliation for the use of nuclear, biological, or chemical weapons, or "in the event of surprising military developments."
Better keep those donations coming... or you never know what might happen.
Perhaps Google doesn't do such a good job when searching on non-geek topics. I use the Web mostly for computer stuff and random urban legend / Kevin Bacon searches, so I wouldn't know. But maybe if you want to book a holiday a semi-automated index like Yahoo does a better job than just counting links (after all, who links to a competitor's site?).
With the version number jump, you'd think they would have taken the opportunity to rename it from GTK+ back to GTK, which is what everyone calls it anyway.
Funny - in Britain, Boomerang is the only Cartoon Network channel that's available gratis, so many more people get it than their other channels. It doesn't run very much advertising either, so I wonder what their reason is. Perhaps to act as an advertisement for the other CN channels.
The command-line.net runtime and compiler are freely downloadable. I wonder whether they run under Wine. This would be preferable to downloading the 'shared source' version, since there is no chance of becoming contaminated by looking at the source.
The 'web' part of a webcam wouldn't - but isn't the actual camera just an ordinary video camera capable of 60Hz refresh? Hmm, 60Hz isn't fast enough, that would mean 60b/s from a single LED and a measly 60Kb/s from a bank of 1000. And maybe it's even worse than that because the camera might be capable of only ten frames per second (seeing as it was designed for sending jerky pictures across slow links anyway).
Hmm, need to rethink this one. I wonder if there is any cheap part capable of the high sample rates needed. That RONJA system another poster mentioned looks interesting, but a bit complex to set up. I think the ideal has to be eight resistors plugged into a parallel port:-).
It's surprising that you can actually construct a real data signal from the LED flashes - I thought that an LED would be too slow to respond to a rapidly changing signal so it would just be half-on all the time. But on page 2 of the report they show an LED emitting light that allows you to perfectly reconstruct a 9600b/s signal. I guess LEDs are rather different from lights based on resistors getting hot; they don't need time to warm up or cool down.
This sounds like a dirt-cheap way to construct wireless links, with no risk to human health (unlike lasers). An LED taped to one window and a $29 webcam in the building opposite could get speeds approaching those of a modem, if you designed a protocol specifically for this purpose. The authors of this paper managed to reconstruct data even without a specially-designed protocol.
A bank of say 1000 LEDs, with a zoom lens at the other end to make sure each one is distinguishable, could transmit *at least* 9.6Mb/s, ie more than a megabyte per second. You could do this by taping a pair of binoculars to your webcam.
If there's a shared 'cypherpunks' (or even 'cipherpunks') account you couldn't use it to post messages, or set preferences. It would be no better than Anonymous Coward, worse in fact since there'd always be some loser changing all the preferences to weird settings (block all stories _except_ JonKatz). All it would give you is no ads, and you can get that anyway with Junkbuster or Mozilla.
FWIW: Personally I'd pay for the ability to post to Slashdot by email, treating it as a kind of mailing list. If they threw that in as part of a fixed-rate subscription package I'd sign up at once.
We all know that pay-per-view or micropayments don't work. This is because users just don't like the idea of someone standing behind them metering the pages they view. If any payment system is to work it should be 'fit and forget'.
I'd happily sign up for a reasonable *fixed fee* per month, but being charged more the more pages I view seems, well, unpleasant. OTOH if the Slashdot crew can justify it from bandwidth costs I might pay up.
FWIW, I think the karma money exchange is worth considering. As long as users can opt out, that is tick a box saying 'monetary considerations will have no effect on the comment scores I see'. But personally I'd be happy to sell a bit of karma (can always whore it back again), and I'd like to see a comment promoted if the author thought enough of it to put some money behind it.
Perhaps the tech industry should get together and draft an Act that would restrict the output of Hollywood in order to protect computer manufacturers' profits.
For example, a movie would not be allowed to show computers in a negative light, nor to glorify 'hacking'. And there would be a statutory upper limit on how entertaining films could be, because otherwise people might spend too much time at cinemas and too little time at home playing computer games.
We have allowed the irresponsible 'film industry' to damage our profits for too long. If they can't sort it out, government should step in.
I'm not so sure that people will ever need even the full 64 bits, let alone 128 or more. You start getting to the point where every atom in the known universe could have its own video diary and you'd still have used only a fraction of the space.
If you're talking about just a large address space rather than a large memory then it's more reasonable (eg every TCP port on every IPv6 address, that's about 80 bits of space), but there's no pressing reason why the processor itself should have a 128-bit wide memory bus just for that reason. It's just a waste of silicon.
And as for 128-bit *integer* arithmetic: can you give some examples? Certainly there are integer computations going on with more than that accuracy using bignums (often to represent rational numbers as a pair of large integers), but 128 bits won't be enough for those either, you'll still have to use bignums. At best a 128-bit CPU could do those calculations twice as fast as a 64-bit one.
If it gets to a point where processors need to do arithmetic on 128-bit IPv6 addresses for a large part of the time, and these operations have to be speed critical, and for some really odd reason you have to store things at an address pointed to by a 128-bit number (which would be an unfeasibly large address space), then yes, 128-bit CPUs might be handy. Until then, any integer quantity anyone wants to handle fits in 64 bits. Floating point is a different matter...
Slashdot Mirror
But, you see, the fact that Sun _can_ provide this option shows they are something of a monopoly. Imagine a hundred Sun competitors all selling mainframes. If it is profitable for Sun to ship the multiple-CPU box and charge (arbitrarily) for only one CPU, then it would be profitable for Sun or any of its competitors to just ship the hardware uncrippled at that price. One company would do it and the others would have to follow.
Look at the PC market - can you even imagine RAM manufacturers producing 128Mbyte modules and deactivating half of the capacity to sell at a lower price? Of course not, not unless they have some special cheap manufacturing process which their competitors cannot emulate. Intel, on the other hand, might be able to get away with deliberately downclocking its CPUs to maintain price differentials, the full-speed ones costing a lot extra. To a certain extent this already happens (CPUs tested at x MHz being relabelled to y MHz, to address shortages of the cheaper part without disrupting the market segmentation).
I know IBM is charging according to consumer costs, 'what the market will bear'. When a company can do this, it shows that company has market power and there is imperfect competition. In a fully competitive market the consumer cost will be equal to the manufacturing cost, because if it were any higher then competitors would step in and offer the same product at a lower price (while still making nonzero profit on each unit). Of course this situation never arises in practice.
:-).
Standard Oil charged what the market would bear. Microsoft charges what the market will bear (which is high for Office, but strangely much lower for things like MS Money which have strong competition). When the price companies can charge is higher than the marginal cost, this shows imperfect competition. In the real world all markets are like this, but to a greater or lesser extent. In the case of IBM mainframes the price is a lot higher than manufacturing cost, so it's reasonable to suspect IBM has a dominant position. This is borne out by other facts.
I deliberately didn't go into bad thing vs good thing in my original post - I was just pointing out that factually, IBM is either a monopolist in mainframes or rather close to it. It certainly has a monopoly in IBM mainframes
The fact that IBM can make it cheaper for the customer by crippling the machine, and still make a profit, shows how high the original prices are relative to the manufacturing cost of the machine.
If there were two IBMs competing with each other, then one of them would surely realize: if we can sell a machine with one CPU activated and make a profit then we could sell it with six CPUs activated, at the same cost to us, and still make a profit. Plus we would take business from the other company. So unless the two companies colluded with each other, mainframe prices would come down to only a small amount above the hardware cost.
The fact that Sun and HP also have these pricing policies shows that their products are not perfect substitutes for IBM mainframes, or at least that they're not that interested in competing with IBM on price. Or maybe they would be interested but they want to keep the cosy price segmentation in their own market.
I'm not saying consumers would be better off if manufacturers had to ship uncrippled hardware. It would probably end up with everyone paying the higher prices, in order for the manufacturers to recoup R&D costs. But it's a fact that differential pricing usually indicates market power and less-intense price competition.
The problem might be if IBM Global Services markets itself as impartial but has a bias towards IBM products. Just like reports produced by Arthur Andersen on development projects tend, totally by coincidence, to recommend proposals put forward by Arthur Andersen.
It's clear that IBM has a monopoly in the mainframe market, or at least something close to it. You can deduce this simply by looking at their pricing policies.
If you buy a mainframe then it often comes with say six CPUs, of which only one is activated. If you pay IBM some extra subscription money they will send an engineer round to activate the second CPU, or up to all six depending on how much you pay. It costs them almost nothing to do this, and it would cost nothing extra to simply have all six enabled when the machine leaves the factory, but IBM charges extra for it.
You can consider this as market segmentation - selling the same product to different parts of the market and charging different prices, so as to squeeze the most out of each consumer. If there were plenty of competition in the market, then IBM would need to sell mainframes with the best price/performance possible and would ship with six CPUs by default, at a price close to the manufacturing cost. The fact that they can get away with this pricing scheme shows they have considerable market power, if not an outright monopoly.
A more positive way of looking at the situation is that the cost of a mainframe reflects less the manufacturing costs (marginal cost), and more the R&D effort that went into desigining it or the expense of building the factory (fixed costs). In this case IBM's charging different prices, despite the marginal cost to them being no different, is just like Novell charging different prices for a 10-client Netware licence and a 100-client licence. So IBM has a monopoly on that particular mainframe design in the same way Novell has a monopoly on Netware. This is still not ideal for the consumer, but it's often considered a necessary evil to provide incentive to invest in new designs.
Seems strangely appropriate for a microkernel-based system :-).
AFAIK: The kernel doesn't use the glibc C library. It has its own memory management code which presumably the kernel zlib code uses. This memory manager may or may not guard against free()ing the same area twice.
But what is the use of zlib in the kernel anyway? Just to uncompress the vmlinuz image before the kernel starts? If so it's not much of a vulnerability, if you can corrupt the vmlinuz file then you can control the whole system anyway.
But does it run Linux?
ISTR that Linux on Palms is kinda primitive ATM.
...a wireless Beowulf cluster of these!
Perhaps Google doesn't do such a good job when searching on non-geek topics. I use the Web mostly for computer stuff and random urban legend / Kevin Bacon searches, so I wouldn't know. But maybe if you want to book a holiday a semi-automated index like Yahoo does a better job than just counting links (after all, who links to a competitor's site?).
With the version number jump, you'd think they would have taken the opportunity to rename it from GTK+ back to GTK, which is what everyone calls it anyway.
You can also get 23-second kernel compiles in software using Compilercache :-).
'very few people get Boomerang'
Funny - in Britain, Boomerang is the only Cartoon Network channel that's available gratis, so many more people get it than their other channels. It doesn't run very much advertising either, so I wonder what their reason is. Perhaps to act as an advertisement for the other CN channels.
The command-line .net runtime and compiler are freely downloadable. I wonder whether they run under Wine. This would be preferable to downloading the 'shared source' version, since there is no chance of becoming contaminated by looking at the source.
The 'web' part of a webcam wouldn't - but isn't the actual camera just an ordinary video camera capable of 60Hz refresh? Hmm, 60Hz isn't fast enough, that would mean 60b/s from a single LED and a measly 60Kb/s from a bank of 1000. And maybe it's even worse than that because the camera might be capable of only ten frames per second (seeing as it was designed for sending jerky pictures across slow links anyway).
:-).
Hmm, need to rethink this one. I wonder if there is any cheap part capable of the high sample rates needed. That RONJA system another poster mentioned looks interesting, but a bit complex to set up. I think the ideal has to be eight resistors plugged into a parallel port
It's surprising that you can actually construct a real data signal from the LED flashes - I thought that an LED would be too slow to respond to a rapidly changing signal so it would just be half-on all the time. But on page 2 of the report they show an LED emitting light that allows you to perfectly reconstruct a 9600b/s signal. I guess LEDs are rather different from lights based on resistors getting hot; they don't need time to warm up or cool down.
This sounds like a dirt-cheap way to construct wireless links, with no risk to human health (unlike lasers). An LED taped to one window and a $29 webcam in the building opposite could get speeds approaching those of a modem, if you designed a protocol specifically for this purpose. The authors of this paper managed to reconstruct data even without a specially-designed protocol.
A bank of say 1000 LEDs, with a zoom lens at the other end to make sure each one is distinguishable, could transmit *at least* 9.6Mb/s, ie more than a megabyte per second. You could do this by taping a pair of binoculars to your webcam.
If there's a shared 'cypherpunks' (or even 'cipherpunks') account you couldn't use it to post messages, or set preferences. It would be no better than Anonymous Coward, worse in fact since there'd always be some loser changing all the preferences to weird settings (block all stories _except_ JonKatz). All it would give you is no ads, and you can get that anyway with Junkbuster or Mozilla.
FWIW: Personally I'd pay for the ability to post to Slashdot by email, treating it as a kind of mailing list. If they threw that in as part of a fixed-rate subscription package I'd sign up at once.
We all know that pay-per-view or micropayments don't work. This is because users just don't like the idea of someone standing behind them metering the pages they view. If any payment system is to work it should be 'fit and forget'.
I'd happily sign up for a reasonable *fixed fee* per month, but being charged more the more pages I view seems, well, unpleasant. OTOH if the Slashdot crew can justify it from bandwidth costs I might pay up.
FWIW, I think the karma money exchange is worth considering. As long as users can opt out, that is tick a box saying 'monetary considerations will have no effect on the comment scores I see'. But personally I'd be happy to sell a bit of karma (can always whore it back again), and I'd like to see a comment promoted if the author thought enough of it to put some money behind it.
Perhaps the tech industry should get together and draft an Act that would restrict the output of Hollywood in order to protect computer manufacturers' profits.
For example, a movie would not be allowed to show computers in a negative light, nor to glorify 'hacking'. And there would be a statutory upper limit on how entertaining films could be, because otherwise people might spend too much time at cinemas and too little time at home playing computer games.
We have allowed the irresponsible 'film industry' to damage our profits for too long. If they can't sort it out, government should step in.
I'm not so sure that people will ever need even the full 64 bits, let alone 128 or more. You start getting to the point where every atom in the known universe could have its own video diary and you'd still have used only a fraction of the space.
If you're talking about just a large address space rather than a large memory then it's more reasonable (eg every TCP port on every IPv6 address, that's about 80 bits of space), but there's no pressing reason why the processor itself should have a 128-bit wide memory bus just for that reason. It's just a waste of silicon.
And as for 128-bit *integer* arithmetic: can you give some examples? Certainly there are integer computations going on with more than that accuracy using bignums (often to represent rational numbers as a pair of large integers), but 128 bits won't be enough for those either, you'll still have to use bignums. At best a 128-bit CPU could do those calculations twice as fast as a 64-bit one.
If it gets to a point where processors need to do arithmetic on 128-bit IPv6 addresses for a large part of the time, and these operations have to be speed critical, and for some really odd reason you have to store things at an address pointed to by a 128-bit number (which would be an unfeasibly large address space), then yes, 128-bit CPUs might be handy. Until then, any integer quantity anyone wants to handle fits in 64 bits. Floating point is a different matter...
Sorry - confused by the 'No comments' setting which was set in the URL for some reason :-(.
Huh? Why are there no comments to this story?
64 bits should be enough for anyone.
No really, I mean it.