What I would *love* is a way to colapse or skip entire threads.
Yes, it's high time Slashdot got on the AJAX train and spruced up its comment pages. There's a lot of potential for a really sweet AJAX comment system and yet I haven't seen one implemented anywhere; Slashdot could really lead the way with some innovations here. Server load might even be reduced by reducing page loads and performing more work on the client (if less page loads is a problem for advertisers, ads could occasionally be inserted near freshly loaded comments).
Video memory is not directly addressable by the processor; it is read/written to using the PCI bus. Even if some of it is mapped into the CPU's virtual address space, it is likely that not all 2 GB will be mapped at the same time (especially considering that the contents of each card's 512 MB will be mostly duplicated except for the frame buffers; what a waste of RAM!).
I hardly think derisive laughter is warranted by a single counterexample. Here in LA there are a couple of toll highways which work well. Several other comments on the original post describe working toll roads. Congestion and poor maintenance on a toll road simply indicates that the toll isn't high enough; although I'm not familiar with Chicago toll roads I'll just bet that the toll is capped by misguided legislation lobbied for by commuters.
It isn't exactly laziness on Firefox's part, more like inertia. Remember, the Firefox codebase is *old*. The image resizing code was likely written long before CPUs had enough power to consider filtering images on the fly, and before 3D hardware came built in to everybody's motherboard. Today it is feasible even on CPUs (though slow; it would still be a noticable performance hit), and trivial for graphics hardware. It's just that nobody has bothered to update the drawing code yet; there is a chicken-and-egg problem; nobody uses browser image scaling because it sucks. The Firefox-Cairo project is an effort to completely modernize Firefox's graphics drawing code, and as a side-effect it will fix this problem. In fact Firefox may use a GDI backend for Cairo instead of the OpenGL one. Either way, image filtering will be fixed.
Note to the clueless: the above post is sarcastic; stop replying with comments about how they do this where you live. The AC was trying to make the point that a 2-tier system *does* work for highways; private toll roads are always well-maintained and congestion-free.
I don't agree that this situation is analagous to a 2-tiered Internet. In the case of highways, private toll roads are always competing with a free baseline service provided by the government (a service that is actually excellent in most respects). The private toll highways must be much better than the free baseline just to survive. In the case of the Internet, private ISP dupopolies will control all tiers of service; there is no free baseline to keep things reasonable. ISPs are free to limit their services to be as crappy as they want, and the insurmountable barriers to entry in their market (both physical and legislative) will keep competition from providing what consumers really want at a price they are willing to pay.
Now, if the government provided a free and decent baseline Internet service, much like the free and decent baseline road service, the situation would become much different. In fact, I would argue that they should do exactly this. The only problem with government Internet service is that it would be much more susceptible to regulation. Copyright cartels could lobby the government to implement draconian monitoring and DRM schemes on the public Internet, and the FBI could have a field day with wiretaps. Then again, on the plus side, free speech over the government's Internet would be truly protected by law. And if you decided you didn't like like what the government was doing with its Internet service, you could still opt to pay for a private ISP, which would by necessity be much better than the government's service.
there's no reason not to expect a browser to be able to scale images
You mean other than the fact that the two most widely used browsers in the world still perform crappy no-interpolation nearest-neighbor image scaling, and yet somehow still manage to suck up gobs of CPU time doing it? (I'm talking here about IE and Firefox, of course.) If you expect a user's browser to have nice image scaling, you will be wrong 99% of the time. It's a ridiculous situation, in this age of dedicated 3D rendering hardware in nearly every PC. A 3D card's *primary function* is drawing scaled and warped images with nice interpolation!
I believe Opera and Safari do perform decent image scaling. I haven't heard if IE7 will, but I doubt it. Firefox 3 is moving to Cairo for graphics, which should allow it to leverage OpenGL-capable graphics hardware for scaling images with nice interpolation. But Firefox 2 will likely still be using the same old terrible image scaling we've been living with for years.
Instead of pressing the "reboot later" button on those windows, you grab the very top left portion of the title bar and drag them down past the clock in the lower right so they are entirely off the screen except for a few pixels. No more will pop up as long as one is still open.
And I meant they don't *become* more difficult to debug because of garbage collection. What garbage collection does is *fix* all the memory leaks which *aren't* hard to debug, which is a completely different thing. And furthermore I dispute your claim that those types of leaks acually are harder to debug; in fact I think it's just the opposite.
There are plenty of ways you can make it easy to tell what's in your leaked memory.
... by doing more work manually. Work you wouldn't have to do if you were using garbage collection.
Right, using limits the lifetime of your allocated resources, not the object itself; but that allows it to fill the same role as a stack-allocated variable in C++ (only without the possibility of passing a dangling pointer to the stack out of your function). And structs are usually allocated on the stack but can be part of objects and even boxed. I'm aware of the distinctions but didn't take the time to lengthen my post with all the details.
I fail to see how following a chain of references to a memory hog is harder than finding a memory leak which has nothing pointing to it at all. In a garbage-collected application, with a proper debugger and profiler you should not have any trouble figuring out exactly what's taking up every byte of your memory, and once you've done that you can easily figure out who has the references to it. I recommend you take a look at Microsoft's awesome CLR profiler; I'm sure a similar tool exists for Java but it may not be free.
It is just as easy to keep references to unneeded objects in C++, so C++ can have the same types of so-called "hard to debug" leaks you blame on garbage collection. But on top of that, if you have a true memory leak, C++ doesn't even tell you what's stored in all that leaked memory. You'll just have to trace back to find the last guy anywhere in the code who threw away a pointer without deleting it, and it could easily be very tough to figure out. And C++ doesn't have a magical solution to leaks in third-party libraries either.
In C# structs are allocated on the stack. But it's almost entirely a performance decision to do so, and doesn't affect object lifetimes as seen by the programmer. For limited-lifetime objects, C# actually has a built-in keyword. It's called using and it works along with the IDisposable interface, regardless of whether your object was allocated on the stack or not. But in a garbage-collected world most objects don't need to have their lifetimes managed, and that's a good thing, not a bad one.
You can't ignore the complexity of manual memory management. You must free all your allocations, and you must police dangling pointers. C++ exposes that complexity all of the time, even though you only need it occasionally, if ever. You can use a smart pointer class, but the more sophisticated of those are simply slow unsafe reference-counting garbage collectors...
On NVIDIA Quadro cards if you drill down through the control panel screens you can set fine details of your resolution and refresh rate down to the timing of various parts of the video signal. I've successfully output 1080i from a Quadro using this panel. I believe GeForce cards can do this too, but perhaps with not quite as much fine control or high-refresh performance.
You're not thinking big enough. Initially we might become a world of perpetual 70-year-olds. But medical research would continue to advance and we would learn to *reverse* the aging process. Eventually this could become a world of perpetual 18-year-olds. Furthermore, we could address not only the physical consequences of aging, but the mental ones too.
Of course, it's a long road to the day that happens, and personally I don't believe anyone alive today will be around to see it. But if we manage not to nuke ourselves into the next ice age between now and then, I see no other real obstacle in the path. Indeed, it seems inevitable.
Actually I think the overpopulation argument against life extension is much more worrying. I just don't buy that people will stop having babies soon enough to prevent a major problem.
Forget the GPL; does the rootkit actually *use* this code? If so, then I think Aple has a pretty clear DMCA case against Sony, since they certainly didn't license FairPlay DRM for Sony to use. And Apple is much more likely than DVD Jon to have both the inclination *and* the means to start a big legal fight about it. Sony breaking Apple's DRM with their rootkit designed to protect their own DRM would be irony too delicious for words.
If I'm not feeling like I want to play game, I can easily plug my iPod into the front of the system and listen to my music.
Of course it's true that you can't do this with music purchased from the iTunes music store. But I also heard that you can't do this with music that is playable on the iPod. That is, in order to get music to play on your XBox, you must not use iTunes to add it to your playable library; instead you have to put it on with Explorer (or Finder) like any other non-music file, and the iPod won't play it when you do that (unless you also put it on with iTunes, using twice the space). Is this true?
Bad accounts are easy to detect. This system could easily be made quite fraud-proof.
Tie accounts to something valuable, like Amazon accounts with a verified purchase history, to prevent mass signps. Easy.
Randomly check answers by randomly posing the same question to two different accounts. This step is infinitely tuneable; more checks means more fraud resistance at a higher overhead cost. New accounts and accounts with suspicious activity could be checked much more often. Reliable accounts could be checked less and used more for checking others.
The system is self-correcting and self-sustaining; money goes in and quality results come out.
This is really a brilliant idea; I predict that the first company to use this technology to weed out search result spam and tune their algorithms is the first company to finally provide better search results than Google. The first company to use this technology to improve targeted ad relevance is the first company to provide more relevant ads than AdSense. I believe Google has too much pride in their algorithms to use this "Mechanical Turk" method unless forced to, so it's wide open for Yahoo or MSN or some upstart to come in and take their crown. If so it would be Google's first big misstep.
Only the clueless would use the middlemen for less when they could be going straight to the source and getting the full price.
The secondary services would have to verify the answers of their users to prevent click fraud; this would mean that they would have to send many questions out to two different users for verification. This would cut into their profits. (The main service must do the same thing, but the secondary services basically have to duplicate that effort, making things even less efficient.)
That expression is usually used with a singular possessive noun. For example, Kelly-Moore Paint advertises itself as "The painter's paint store." Using a plural there looks strange, though I'm not enough of an english weenie to know if it's technically incorrect. At least with an apostrophe there your intent is clear and people will be able to parse the sentence more easily.
Contract law. When you buy the commercial version from TrollTech the license you agree to forbids using it with any code that was developed using the free version.
No, you're right. I imagine if you came to them and offered to pay more than what it would have cost to buy the licenses to begin with, they would be reasonable. I bet they also would make exceptions for commercial applications using the LGPL'd KDE libraries, which would otherwise be forbidden by this policy.
Yes, you can cheat TrollTech out of their money by lying and breaking the law. News flash: you can do the same with any other software company. You could buy one copy of Visual Studio.NET, then download a crack and install it on a million computers and Microsoft would never know. The situation with TrollTech is pretty much the same; the only difference is you don't need to download a crack first, but that doesn't make it any more legal.
Slashdot Mirror
Yes, it's high time Slashdot got on the AJAX train and spruced up its comment pages. There's a lot of potential for a really sweet AJAX comment system and yet I haven't seen one implemented anywhere; Slashdot could really lead the way with some innovations here. Server load might even be reduced by reducing page loads and performing more work on the client (if less page loads is a problem for advertisers, ads could occasionally be inserted near freshly loaded comments).
Video memory is not directly addressable by the processor; it is read/written to using the PCI bus. Even if some of it is mapped into the CPU's virtual address space, it is likely that not all 2 GB will be mapped at the same time (especially considering that the contents of each card's 512 MB will be mostly duplicated except for the frame buffers; what a waste of RAM!).
I hardly think derisive laughter is warranted by a single counterexample. Here in LA there are a couple of toll highways which work well. Several other comments on the original post describe working toll roads. Congestion and poor maintenance on a toll road simply indicates that the toll isn't high enough; although I'm not familiar with Chicago toll roads I'll just bet that the toll is capped by misguided legislation lobbied for by commuters.
It isn't exactly laziness on Firefox's part, more like inertia. Remember, the Firefox codebase is *old*. The image resizing code was likely written long before CPUs had enough power to consider filtering images on the fly, and before 3D hardware came built in to everybody's motherboard. Today it is feasible even on CPUs (though slow; it would still be a noticable performance hit), and trivial for graphics hardware. It's just that nobody has bothered to update the drawing code yet; there is a chicken-and-egg problem; nobody uses browser image scaling because it sucks. The Firefox-Cairo project is an effort to completely modernize Firefox's graphics drawing code, and as a side-effect it will fix this problem. In fact Firefox may use a GDI backend for Cairo instead of the OpenGL one. Either way, image filtering will be fixed.
I don't agree that this situation is analagous to a 2-tiered Internet. In the case of highways, private toll roads are always competing with a free baseline service provided by the government (a service that is actually excellent in most respects). The private toll highways must be much better than the free baseline just to survive. In the case of the Internet, private ISP dupopolies will control all tiers of service; there is no free baseline to keep things reasonable. ISPs are free to limit their services to be as crappy as they want, and the insurmountable barriers to entry in their market (both physical and legislative) will keep competition from providing what consumers really want at a price they are willing to pay.
Now, if the government provided a free and decent baseline Internet service, much like the free and decent baseline road service, the situation would become much different. In fact, I would argue that they should do exactly this. The only problem with government Internet service is that it would be much more susceptible to regulation. Copyright cartels could lobby the government to implement draconian monitoring and DRM schemes on the public Internet, and the FBI could have a field day with wiretaps. Then again, on the plus side, free speech over the government's Internet would be truly protected by law. And if you decided you didn't like like what the government was doing with its Internet service, you could still opt to pay for a private ISP, which would by necessity be much better than the government's service.
You mean other than the fact that the two most widely used browsers in the world still perform crappy no-interpolation nearest-neighbor image scaling, and yet somehow still manage to suck up gobs of CPU time doing it? (I'm talking here about IE and Firefox, of course.) If you expect a user's browser to have nice image scaling, you will be wrong 99% of the time. It's a ridiculous situation, in this age of dedicated 3D rendering hardware in nearly every PC. A 3D card's *primary function* is drawing scaled and warped images with nice interpolation!
I believe Opera and Safari do perform decent image scaling. I haven't heard if IE7 will, but I doubt it. Firefox 3 is moving to Cairo for graphics, which should allow it to leverage OpenGL-capable graphics hardware for scaling images with nice interpolation. But Firefox 2 will likely still be using the same old terrible image scaling we've been living with for years.
Instead of pressing the "reboot later" button on those windows, you grab the very top left portion of the title bar and drag them down past the clock in the lower right so they are entirely off the screen except for a few pixels. No more will pop up as long as one is still open.
There are plenty of ways you can make it easy to tell what's in your leaked memory.
Right, using limits the lifetime of your allocated resources, not the object itself; but that allows it to fill the same role as a stack-allocated variable in C++ (only without the possibility of passing a dangling pointer to the stack out of your function). And structs are usually allocated on the stack but can be part of objects and even boxed. I'm aware of the distinctions but didn't take the time to lengthen my post with all the details.
You're not the only one. In a garbage collected language, instead of talking about memory leaks, refer to "memory hogs". It makes more sense.
It is just as easy to keep references to unneeded objects in C++, so C++ can have the same types of so-called "hard to debug" leaks you blame on garbage collection. But on top of that, if you have a true memory leak, C++ doesn't even tell you what's stored in all that leaked memory. You'll just have to trace back to find the last guy anywhere in the code who threw away a pointer without deleting it, and it could easily be very tough to figure out. And C++ doesn't have a magical solution to leaks in third-party libraries either.
In C# structs are allocated on the stack. But it's almost entirely a performance decision to do so, and doesn't affect object lifetimes as seen by the programmer. For limited-lifetime objects, C# actually has a built-in keyword. It's called using and it works along with the IDisposable interface, regardless of whether your object was allocated on the stack or not. But in a garbage-collected world most objects don't need to have their lifetimes managed, and that's a good thing, not a bad one.
You can't ignore the complexity of manual memory management. You must free all your allocations, and you must police dangling pointers. C++ exposes that complexity all of the time, even though you only need it occasionally, if ever. You can use a smart pointer class, but the more sophisticated of those are simply slow unsafe reference-counting garbage collectors...
On NVIDIA Quadro cards if you drill down through the control panel screens you can set fine details of your resolution and refresh rate down to the timing of various parts of the video signal. I've successfully output 1080i from a Quadro using this panel. I believe GeForce cards can do this too, but perhaps with not quite as much fine control or high-refresh performance.
Why were you forced to use the vendor's compiler instead of GCC?
Of course, it's a long road to the day that happens, and personally I don't believe anyone alive today will be around to see it. But if we manage not to nuke ourselves into the next ice age between now and then, I see no other real obstacle in the path. Indeed, it seems inevitable.
Actually I think the overpopulation argument against life extension is much more worrying. I just don't buy that people will stop having babies soon enough to prevent a major problem.
Forget the GPL; does the rootkit actually *use* this code? If so, then I think Aple has a pretty clear DMCA case against Sony, since they certainly didn't license FairPlay DRM for Sony to use. And Apple is much more likely than DVD Jon to have both the inclination *and* the means to start a big legal fight about it. Sony breaking Apple's DRM with their rootkit designed to protect their own DRM would be irony too delicious for words.
Of course it's true that you can't do this with music purchased from the iTunes music store. But I also heard that you can't do this with music that is playable on the iPod. That is, in order to get music to play on your XBox, you must not use iTunes to add it to your playable library; instead you have to put it on with Explorer (or Finder) like any other non-music file, and the iPod won't play it when you do that (unless you also put it on with iTunes, using twice the space). Is this true?
- Tie accounts to something valuable, like Amazon accounts with a verified purchase history, to prevent mass signps. Easy.
- Randomly check answers by randomly posing the same question to two different accounts. This step is infinitely tuneable; more checks means more fraud resistance at a higher overhead cost. New accounts and accounts with suspicious activity could be checked much more often. Reliable accounts could be checked less and used more for checking others.
- The system is self-correcting and self-sustaining; money goes in and quality results come out.
This is really a brilliant idea; I predict that the first company to use this technology to weed out search result spam and tune their algorithms is the first company to finally provide better search results than Google. The first company to use this technology to improve targeted ad relevance is the first company to provide more relevant ads than AdSense. I believe Google has too much pride in their algorithms to use this "Mechanical Turk" method unless forced to, so it's wide open for Yahoo or MSN or some upstart to come in and take their crown. If so it would be Google's first big misstep.Sorry, doesn't add up.
That expression is usually used with a singular possessive noun. For example, Kelly-Moore Paint advertises itself as "The painter's paint store." Using a plural there looks strange, though I'm not enough of an english weenie to know if it's technically incorrect. At least with an apostrophe there your intent is clear and people will be able to parse the sentence more easily.
Contract law. When you buy the commercial version from TrollTech the license you agree to forbids using it with any code that was developed using the free version.
No, you're right. I imagine if you came to them and offered to pay more than what it would have cost to buy the licenses to begin with, they would be reasonable. I bet they also would make exceptions for commercial applications using the LGPL'd KDE libraries, which would otherwise be forbidden by this policy.
Yes, you can cheat TrollTech out of their money by lying and breaking the law. News flash: you can do the same with any other software company. You could buy one copy of Visual Studio .NET, then download a crack and install it on a million computers and Microsoft would never know. The situation with TrollTech is pretty much the same; the only difference is you don't need to download a crack first, but that doesn't make it any more legal.
Your sig is missing an apostrophe. Without the possessive form ("programmer's") it doesn't make any sense.