And yet my Windows 7 will have better gaming and application support than your Loonix desktop for the next decade well after mainstream support is dropped
Windows definitely has much better support for new games, but for older games I've found that WINE often works better. I have a number of games designed for Windows 9x that work fine in WINE on OS X (which is very much a tier 2 platform for WINE), yet won't even run on Windows 2000 or XP, let alone anything newer.
I wish they'd managed to keep the old application remote display mechanism (NXHOST =~ DISPLAY), but the unreasonably-licensed Adobe Display PostScript pretty much put an end to that... alas.
Not really. Replacing PostScript with PDF for the display model was sensible, because PostScript is Turing-complete and there were lots of fun things you could do by being able to run PS code on someone else's monitor (some security holes, but even without them a denial of service was easy). Apple made the decision not to provide a stable protocol for client to window server communication. This has some advantages - they can modify it in every point release if they want to without having to care about backwards compatibility - but it means they sacrifice remote display.
OS X does not use the GPU for font rendering. It renders each character to a texture on the CPU and just composites on the GPU. This was added with Quartz Extreme. X11 also does the same thing via the XRENDER extension, and so does Vista via Direct2D stuff. The MSR paper that I am referring to described how to store the bezier control points on the GPU and then construct the glyph with pixel shaders.
I thought the disk IO was worth thousands of times more costly in operations like these
A lot of people thought that. Then SSDs came along and made a lot of these assumptions incorrect. Developers for most major operating systems are currently optimising everything on the I/O path because assumptions about the relative speed of I/O and computing that have been true for the last 30 years are suddenly wrong (especially on mobile devices).
Most likely it will improve it. In general, running tasks on the GPU uses less power than on the CPU. It's almost always more power-efficient to use dedicated silicon than general purpose, and while a GPU is a general-purpose processor these days it's still heavily optimised for this kind of task, whereas the CPU is not.
It's also worth noting that MS has had a long time to tune this. The original implementation of GPU-accelerated font rendering was done by MSR about a decade ago. In the time it's taken them to transfer the technology from research to a product, academic research projects have spun out companies, had them bought by MS, and had their products integrated into the MS lineup. This is a pretty good case study of what's wrong with Microsoft's interaction with its research division.
Not true. Many embedded systems use microkernels that can't do this. The driver can issue DMA requests, but it must call into the microkernel to request some memory for the target or the IOMMU will raise an exception.
It's increasingly easy to implement operating systems where buggy drivers can't trash the entire system now that most consumer CPUs come with an IOMMU. If you're using an nVidia GPU, almost all of the complex logic is actually in userspace. All that the kernel-space driver does is set up a context on the GPU with a command submission buffer mapped into userspace and allocate memory in VRAM or in main memory accessible from the GPU. The card can only DMA to regions registered in the GART, so there's basically nothing a malicious or buggy userspace program can do except trash its own memory and fill the image buffer that he windowing system will composite for its window with nonsense. High end NICs (e.g. infiniband) have also been designed in this way for a long time, because the overhead of going via the kernel was too high.
Microsof tried to push tablets that ran applications written for a two-button-mouse-and-keyboard interface, with a floating on-screen keyboard and a stylus. It was clunky because it was not a very natural fit for the UI. Apple did two things right. The first was not to attempt to run desktop apps: at the very least, you need to rewrite your UI if you port from OS X to iOS. The second was to wait until multitouch capacitive touchscreens were cheap.
FireWire was already dying by that point. I had a couple of drives that had FireWire 400, FireWire 800 and USB 2. FW800 was clearly superior: I could plug both of them into a single port on my laptop and still not be bottlenecked by the connection. USB2 couldn't quite match FW400 in real-world usage. But there were very few computers with FW400, let alone FW800, so there was little incentive for device manufacturers to use FireWire (and produce more expensive devices, since FireWire was no host-device model).
The EU regulation has an loophole, which means that as long as Apple puts a dock to micro-USB adaptor in the box they comply. As to the dock vs micro-USB issue, most people seem to be missing the fact that the dock connector does a lot more than USB. It also contains audio line in and out, component and s-video, serial, USB, and FireWire (not in recent ones). USB is a step backwards. A modern replacement for the dock should ideally contain display port or HDMI video and a switchable line out / S/PDIF or similar. Personally, I'd love to see Thunderbolt ports on mobile devices.
The real issue was that Intel put USB controllers on south bridge chips. This meant that motherboard makers got USB almost for free - they just needed to put the sockets on the board and connect them to the chip. To support FireWire, they needed to add another chip and connect it to the south bridge via the PCI bus as well as to the socket. The south bridge already had traces going to the PS/2, serial and parallel ports, so adding USB did almost nothing to increase motherboard complexity - they just had to run a few extra traces alongside existing ones. Adding FireWire meant a lot of effort in board layout.
I clicked on a few ads, back when Google was the Internet startup with the novel ideal of unobtrusive text-based ads that were relevant to the content of the page. I can't remember the last time I even saw a relevant ad though, let alone considered clicking on one.
'The city' is usually used in England (if not referring to the nearest city) to mean The City of London, which is about a square mile containing all of the banks and associated surplus population.
What is it with people claiming that everyone who draws attention to Apple's market placement is an Apple fan? For the record, I own an Android phone...
It's possible to do that. Creating a web app that stores the data locally in a form that is useful even if the web app becomes inaccessible, however, is far more difficult. I can still access spreadsheet files created on my Psion Series 3 that I dumped on a hard disk 20 almost years ago in their native format by running the Series 3a emulator in DOSBox and exporting them in some other format (although CSV is the only one that I can open directly, but I can use an old version of another spreadsheet as an intermediate for better preservation of formulae. Good luck doing that with a web app...
Well, except in the whole old-style 'making money' sense. Apple devices still occupy the most profitable 10% of the market. I suspect that Google and Apple are both happy with this: Google wants lots of Android users so it can collect information about them for advertising and make money out of them, Apple wants the high-margin part of the market to make the biggest profit from direct sales.
Isn't it interesting how you just assume that everyone agrees with you.
So, just to clarify, you judge accuracy of a news source in terms of popularity? You're really not doing anything to dispel the stereotype of a Fox News watcher...
That's true now, but I'd invite you to go back and read the original Ethernet papers from PARC. They describe, among other things, a single (coax) wiring model, with support for up to 256 computers on a single broadcast domain sharing a 3Mb/s channel. Numerous parts of the specification are based on limits of the technology at the time, such as the number of RAM chips it was possible to fit on the board and the I/O speed of the Alto.
The evaluation paper on the Alto, published in 1979, points out that it's possible to imagine a network of thousands of personal computers.
Someone has to drive the buses, sweep the streets, flip the burgers and operate the checkout at the supermarket et cetera, et cetera
I've been to several cities where busses have been replaced by automated trams. Street sweeping isn't done by guys with brooms anymore, it's done by guys driving around (slow-moving) vehicles. They're no harder to automate than a roomba. Most supermarkets have self-service checkouts and just one security guard to watch half a dozen or more of them, and even that wouldn't be required with RFID on the product tags. Burger flipping is probably around for a little while longer - it's not hard to design a machine that would cook and assemble fast food burger (it's simpler than many automated factory tasks), but the human is so cheap in comparison to the machine that it would take a good few years to break even and the human is more flexible when you want to change the menu.
If these people had been stakeholders in the businesses introducing automation, then it would have been fine: as they were replaced by robots they'd have just had more free time and less work. Unfortunately, we've concentrated ownership in a small subset of the population and are trying to fudge the gap with welfare payments, paid out of a general fund and not by the people making profits from the trend.
In an ideal world, everyone would have everything that they want. That's not possible at our current level of social and technological development, so some form of rationing is required. We use tokens known as money to implement the rationing system, with the basic idea that people who create things or offer services that are perceived to have value should be able to get these tokens in return. If someone does something that is very valuable, or moderately valuable to a lot of people, then they get more of the tokens and can, in turn, buy more goods or services with them. This serves as a motivation.
There are two problems. The first is the ratio and the second is how the value is determined. It's easy to look around and see that some people are contributing more to society than others by one order of magnitude. Look a bit more, and you can probably find a differential of two orders of magnitude. Someone in the USA on minimum wage makes around $20K. A CEO of a loss-making bank or hedge fund management company in the last few years probably made $20M in a single year. It's hard to argue that the bank CEO made a more positive contribution to society than, say, a toilet cleaner. It's even harder to justify that their contributions were a thousand times more beneficial.
And that's not even going near the super rich. The three richest people in the world, between them, control more wealth than the 48 poorest countries. When one person is effectively able to buy 16 countries then there's a problem. You began with a strawman, so I'll end with one: the extreme case of wealth inequality is one person being able to own another.
Don't forget that the 240ms RTT is an absolute minimum - that's the distance to the first hop, and not including any processing delays. You're looking at 300-350ms as a minimum for a complete path. Still, the bandwidth is better than the ADSL that my mother gets, so if it's not stupidly expensive then I can imagine it being useful. The 4Mb/s downstream is (just!) enough to stream iPlayer HD.
No, WHATWG is basically the everyone-except-Microsoft group. It was originally Apple, Opera, and Mozilla, now Google has joined. There are a lot of things that you can blame Microsoft for, but this is not one of them, unless you count the stagnation of Internet standards post IE6 as a motivation for the formation of WHATWG.
Slashdot Mirror
And yet my Windows 7 will have better gaming and application support than your Loonix desktop for the next decade well after mainstream support is dropped
Windows definitely has much better support for new games, but for older games I've found that WINE often works better. I have a number of games designed for Windows 9x that work fine in WINE on OS X (which is very much a tier 2 platform for WINE), yet won't even run on Windows 2000 or XP, let alone anything newer.
I wish they'd managed to keep the old application remote display mechanism (NXHOST =~ DISPLAY), but the unreasonably-licensed Adobe Display PostScript pretty much put an end to that... alas.
Not really. Replacing PostScript with PDF for the display model was sensible, because PostScript is Turing-complete and there were lots of fun things you could do by being able to run PS code on someone else's monitor (some security holes, but even without them a denial of service was easy). Apple made the decision not to provide a stable protocol for client to window server communication. This has some advantages - they can modify it in every point release if they want to without having to care about backwards compatibility - but it means they sacrifice remote display.
OS X does not use the GPU for font rendering. It renders each character to a texture on the CPU and just composites on the GPU. This was added with Quartz Extreme. X11 also does the same thing via the XRENDER extension, and so does Vista via Direct2D stuff. The MSR paper that I am referring to described how to store the bezier control points on the GPU and then construct the glyph with pixel shaders.
I thought the disk IO was worth thousands of times more costly in operations like these
A lot of people thought that. Then SSDs came along and made a lot of these assumptions incorrect. Developers for most major operating systems are currently optimising everything on the I/O path because assumptions about the relative speed of I/O and computing that have been true for the last 30 years are suddenly wrong (especially on mobile devices).
Most likely it will improve it. In general, running tasks on the GPU uses less power than on the CPU. It's almost always more power-efficient to use dedicated silicon than general purpose, and while a GPU is a general-purpose processor these days it's still heavily optimised for this kind of task, whereas the CPU is not.
It's also worth noting that MS has had a long time to tune this. The original implementation of GPU-accelerated font rendering was done by MSR about a decade ago. In the time it's taken them to transfer the technology from research to a product, academic research projects have spun out companies, had them bought by MS, and had their products integrated into the MS lineup. This is a pretty good case study of what's wrong with Microsoft's interaction with its research division.
Not true. Many embedded systems use microkernels that can't do this. The driver can issue DMA requests, but it must call into the microkernel to request some memory for the target or the IOMMU will raise an exception.
It's increasingly easy to implement operating systems where buggy drivers can't trash the entire system now that most consumer CPUs come with an IOMMU. If you're using an nVidia GPU, almost all of the complex logic is actually in userspace. All that the kernel-space driver does is set up a context on the GPU with a command submission buffer mapped into userspace and allocate memory in VRAM or in main memory accessible from the GPU. The card can only DMA to regions registered in the GART, so there's basically nothing a malicious or buggy userspace program can do except trash its own memory and fill the image buffer that he windowing system will composite for its window with nonsense. High end NICs (e.g. infiniband) have also been designed in this way for a long time, because the overhead of going via the kernel was too high.
There's one bus that takes me within ten blocks of work and if I miss it in the morning, it ends up being faster to walk the ~1.5 miles to work.
Seriously? With a commute that short, why would you bother with a bus or a car? It's a perfect distance for walking or cycling.
Microsof tried to push tablets that ran applications written for a two-button-mouse-and-keyboard interface, with a floating on-screen keyboard and a stylus. It was clunky because it was not a very natural fit for the UI. Apple did two things right. The first was not to attempt to run desktop apps: at the very least, you need to rewrite your UI if you port from OS X to iOS. The second was to wait until multitouch capacitive touchscreens were cheap.
FireWire was already dying by that point. I had a couple of drives that had FireWire 400, FireWire 800 and USB 2. FW800 was clearly superior: I could plug both of them into a single port on my laptop and still not be bottlenecked by the connection. USB2 couldn't quite match FW400 in real-world usage. But there were very few computers with FW400, let alone FW800, so there was little incentive for device manufacturers to use FireWire (and produce more expensive devices, since FireWire was no host-device model).
The EU regulation has an loophole, which means that as long as Apple puts a dock to micro-USB adaptor in the box they comply. As to the dock vs micro-USB issue, most people seem to be missing the fact that the dock connector does a lot more than USB. It also contains audio line in and out, component and s-video, serial, USB, and FireWire (not in recent ones). USB is a step backwards. A modern replacement for the dock should ideally contain display port or HDMI video and a switchable line out / S/PDIF or similar. Personally, I'd love to see Thunderbolt ports on mobile devices.
The real issue was that Intel put USB controllers on south bridge chips. This meant that motherboard makers got USB almost for free - they just needed to put the sockets on the board and connect them to the chip. To support FireWire, they needed to add another chip and connect it to the south bridge via the PCI bus as well as to the socket. The south bridge already had traces going to the PS/2, serial and parallel ports, so adding USB did almost nothing to increase motherboard complexity - they just had to run a few extra traces alongside existing ones. Adding FireWire meant a lot of effort in board layout.
I clicked on a few ads, back when Google was the Internet startup with the novel ideal of unobtrusive text-based ads that were relevant to the content of the page. I can't remember the last time I even saw a relevant ad though, let alone considered clicking on one.
I stopped reading there. What magic do "real pro financial analysts" have which slashdotters do not have?
They can convincingly spout bullshit.
in England, they call London "the city"
'The city' is usually used in England (if not referring to the nearest city) to mean The City of London, which is about a square mile containing all of the banks and associated surplus population.
What is it with people claiming that everyone who draws attention to Apple's market placement is an Apple fan? For the record, I own an Android phone...
It's possible to do that. Creating a web app that stores the data locally in a form that is useful even if the web app becomes inaccessible, however, is far more difficult. I can still access spreadsheet files created on my Psion Series 3 that I dumped on a hard disk 20 almost years ago in their native format by running the Series 3a emulator in DOSBox and exporting them in some other format (although CSV is the only one that I can open directly, but I can use an old version of another spreadsheet as an intermediate for better preservation of formulae. Good luck doing that with a web app...
Well, except in the whole old-style 'making money' sense. Apple devices still occupy the most profitable 10% of the market. I suspect that Google and Apple are both happy with this: Google wants lots of Android users so it can collect information about them for advertising and make money out of them, Apple wants the high-margin part of the market to make the biggest profit from direct sales.
Isn't it interesting how you just assume that everyone agrees with you.
So, just to clarify, you judge accuracy of a news source in terms of popularity? You're really not doing anything to dispel the stereotype of a Fox News watcher...
That's true now, but I'd invite you to go back and read the original Ethernet papers from PARC. They describe, among other things, a single (coax) wiring model, with support for up to 256 computers on a single broadcast domain sharing a 3Mb/s channel. Numerous parts of the specification are based on limits of the technology at the time, such as the number of RAM chips it was possible to fit on the board and the I/O speed of the Alto.
The evaluation paper on the Alto, published in 1979, points out that it's possible to imagine a network of thousands of personal computers.
Someone has to drive the buses, sweep the streets, flip the burgers and operate the checkout at the supermarket et cetera, et cetera
I've been to several cities where busses have been replaced by automated trams. Street sweeping isn't done by guys with brooms anymore, it's done by guys driving around (slow-moving) vehicles. They're no harder to automate than a roomba. Most supermarkets have self-service checkouts and just one security guard to watch half a dozen or more of them, and even that wouldn't be required with RFID on the product tags. Burger flipping is probably around for a little while longer - it's not hard to design a machine that would cook and assemble fast food burger (it's simpler than many automated factory tasks), but the human is so cheap in comparison to the machine that it would take a good few years to break even and the human is more flexible when you want to change the menu.
If these people had been stakeholders in the businesses introducing automation, then it would have been fine: as they were replaced by robots they'd have just had more free time and less work. Unfortunately, we've concentrated ownership in a small subset of the population and are trying to fudge the gap with welfare payments, paid out of a general fund and not by the people making profits from the trend.
So where's Big Pharma? They should be out-competing the MIC for these tax dollars using these statistics!
In an ideal world, everyone would have everything that they want. That's not possible at our current level of social and technological development, so some form of rationing is required. We use tokens known as money to implement the rationing system, with the basic idea that people who create things or offer services that are perceived to have value should be able to get these tokens in return. If someone does something that is very valuable, or moderately valuable to a lot of people, then they get more of the tokens and can, in turn, buy more goods or services with them. This serves as a motivation.
There are two problems. The first is the ratio and the second is how the value is determined. It's easy to look around and see that some people are contributing more to society than others by one order of magnitude. Look a bit more, and you can probably find a differential of two orders of magnitude. Someone in the USA on minimum wage makes around $20K. A CEO of a loss-making bank or hedge fund management company in the last few years probably made $20M in a single year. It's hard to argue that the bank CEO made a more positive contribution to society than, say, a toilet cleaner. It's even harder to justify that their contributions were a thousand times more beneficial.
And that's not even going near the super rich. The three richest people in the world, between them, control more wealth than the 48 poorest countries. When one person is effectively able to buy 16 countries then there's a problem. You began with a strawman, so I'll end with one: the extreme case of wealth inequality is one person being able to own another.
it's like the british monarchy also ran apple, ibm, microsoft, google, ge, and gm
A better historical example: The East India Company.
Don't forget that the 240ms RTT is an absolute minimum - that's the distance to the first hop, and not including any processing delays. You're looking at 300-350ms as a minimum for a complete path. Still, the bandwidth is better than the ADSL that my mother gets, so if it's not stupidly expensive then I can imagine it being useful. The 4Mb/s downstream is (just!) enough to stream iPlayer HD.
No, WHATWG is basically the everyone-except-Microsoft group. It was originally Apple, Opera, and Mozilla, now Google has joined. There are a lot of things that you can blame Microsoft for, but this is not one of them, unless you count the stagnation of Internet standards post IE6 as a motivation for the formation of WHATWG.