That doesn't make any sense. The Mesa guys *implemented* an open specification. And it took about a year and a half for Mesa 1.0 to come out. Reverse-engineering an existing driver would be a whole lot more complicated.
Besides, only a fool buys a $300 video card. Yes, I may get flamed for that comment, but that doesn't detract from the fact that (with a few exceptions) it's true. First, $300 isn't unreasonable at all, especially if you're a heavy gamer. $300 is about the point where you get a top-of-the-line card, without spending extra change for features that don't add any performance. Spend less than about $200, and you give up significant amounts of performance or next-generation features. To those who spend a lot of time gaming, an extra $150 for the ability to turn the detail to maximum in all theri games is worth it. Especially when you consider that the amortized cost is only about $12.50 per month if you upgrade once a year. Second, not everybody uses their cards just for gaming. $300 cards aren't fast enough when you've got a deadline coming up and SolidEdge feels slower and slower by the minute...
No, the point is that we can't[1] code something better because they won't release the necessary programming information. Check out this driver comparison on the Radeon 9000. The DRI drivers perform very poorly. Yet, the documentation for chips up to the Radeon 9200 *was* supplied to the DRI project. Indeed, the DRI project doesn't support any cards that don't have documentation. So "no documentation" is no excuse. Its just the simple fact that ATI knows the hardware because they designed it, and they are in a much better position to write these drivers than the OSS community.
I'd rather have a slightly slower driver than a fast clsoed source one
At least frame it correctly. Its not "slightly slower." Its much slower, sometimes by several times. The trade-off is not a little bit of speed for a lot of freedom. Its turning your $300 graphics card into a $75 card, for a very little bit of extra freedom. Unlike an OS, or major software app, you're not tied into drivers. If the constraints of a closed-source driver become too much, I can switch out my card and its drivers in a matter of hours.
As far as I'm concerned, begging manufacturers to open source their drivers is whining. I don't think the OSS spirit is about forcing people to open their code. If they choose to, great. If they don't, well, we can just code something better. That's precisely what GNU did to UNIX. Now the issue at hand is that the OSS community has *not* been able to produce something that can replace this proprietory code. Until that happens, they really don't have a solid bargaining position.
That's because VIA hardware *SUCKS* Same thing with Matrox. Their 3D cards sucked, so they released the specs for it. When they got a card that didn't suck (Parhelia) they stopped releasing specs.
Look, I care about freedom too. I use Open Source software whenever possible, and have sent in patches and stuff for it too. However, I'm willing to draw the line somewhere. For me, that line is: I'll use OSS software, unless it doesn't do what I need it to do. This is a case where OSS software doesn't do what I need it to do.
There are NO good OSS 3D drivers. By good I mean as fast and as full-featured as the Windows drivers for the same card. The DRI ATI drivers don't even support shaders! When OSS manages to make a good set of drivers for the R100 or R200 (whose specs *are* available) then they can say "OSS drivers would be better!" Until then, its just whining.
1) Because most other drivers are an order of magnitude simpler. Your modem driver, for example, does not need to do complex reordering of display lists to optimize performance. Also, OpenGL drivers are an entire OpenGL implementation. So its like open-sourcing not just your modem driver, but your whole IP stack.
2) No slight to the XFree86 developers, but:
a) Many of the XFree86 drivers (eg: nv) are significantly slower than their proprietory counterparts and in any case 2D is much easier to do drivers for than 3D,
b) The DRI project has not released any drivers that take full-advantage of the hardware. NVIDIA's binary driver will get you exact same performance and the exact same features the Windows driver has. The DRI drivers for Radeons are not only *much* slower, but don't even support basic features like vertex/pixel shaders!
If their binary is anything like NVIDIA's driver, that'd be a dozen megabytes of code to reverse engineer! Remember, a graphics driver is *not* a simple register-banger. Its an entire implementation of OpenGL. Much of that code is more or less hardware-independent, and has to do with optimizing display lists and whatnot. Reverse engineering the driver would be extremely difficult.
Listen to yourself! You payed $1000 for an 800MHz G4??? And you claim that that parts aren't overpriced? The G4 is a terrible CPU. Its completely bottlenecked by an ancient SDR memory bus. At today's prices, that G4 + motherboard is the equivilent of $70-$80 PC hardware.
And where do you get G5 motherboards? I can't find a place that sells them. Even if they do, they are most likely Apple-refurb parts, and carry the same insane prices the G4 refurb parts did.
Not FUD. I'm in the same position. I don't want to pay for "Apple Quality Hardware(TM)" (whatever that means, since Apple fans continually claim that Apple uses the exact same hardware). I wanna pay pricewatch prices for pieces of a computer I can stick into pieces I already have. Apple won't let me do that. Whereas I could put together a dual 2.8 GHz Xeon for about $1000 (I already have a harddrive, etc) a comparable (to be generous) dual 1.8 GHz G5 would set me back $2500. Sure, I could get a G4 iMac for dirt cheap, but who wants to use that POS?
Well, I don't put a lot of stock in GPA either. There is way too much variation between schools that try to go easy on their students GPA, and schools that relentlessy trash their students GPA. I think the best thing to do is skill-drills during interviews, to see how much they really know. I've found that people who look great on paper absolutely break down in such a situation.
If you graduated several years ago, I wouldn't be surprised that Java wasn't taught. It wasn't highly-hyped language at the time. Schools that teach mainly Java are much more prevalent these days where its all the rage. There are holdouts, still (kudos to MIT for doing their freshmen classes in Scheme according to the SICP) but not enough.
And the I wasn't really attacking Java. I think its a silly language, but it can be useful because of its vendor and library support. My point was more that its currently a fad language, and in response many places are graduating Java programmers just to feed the market.
(Toolkit designers: please please PLEASE give us a way to render a widget into a pixmap. That alone would solve a lot of problems. Ask and ye shall receive.
You would be really surprised how low the bar as slipped these days. Where did you go to school, anyway? There are schools turning out good computer scientists, but there are a whole lot more churning out Java programmers.
That's because we just came out of several straight quarters of the worst quarterly earnings, revenue, and GDP growth in many years. The question is, where are the numbers now relative to how they were doing before the recession?
Eugenia did a review of KDE 3.2-beta2 a while back. It was actually quite spot-on. As a KDE user, disagree with her ideologically on some points (eg. the configure menu comment) but on the whole I think it was a balanced evaluation of KDE, from a traditional UI designer's point of view.
What the fuck is wrong with you? You've posted the same thing dozens of times already. We get the picture!
And I've got 20 total entries in the root of my KMenu. I'd like to see how you get 20 = 30,000. Hell, a well-outfitted GNOME setup has at least a dozen!
How many items do you have in your home folder??? Mine has about a dozen (not counting hidden ones) and loads in less than a second (KDE 3.2, Debian, 2.0GHz P4)
* The 5-10 second load-time for KDE apps This is a legitimate problem, though getting much better. With prelinking, load times are closer to 2-3 seconds rather than 5-10.
* The 4 second load time just to open a folder Depends on the size of the folder. Normal-sized folders open pretty much instantly on my machine. Performance on large folders (/usr/bin with thousands of items) could be improved, though.
* The cramped and embarrassing K menu, with a 100 different groups and completely illogical redundancies like "Preferences," "System Settings," and "Control Center" Um, my stock Debian KMenu has a dozen folders. "Preferences" is gone in 3.2. There is a "settings" folder and a "system" folder, which is logical --- one is for preferences, the other is for system utilities. Kinda analagous to Windows's "Control Panel" and "Accessories/System Tools" And "Control Center" is an entry in "Settings"! Are you sure you're not mistaking the quick- access area at the top for default menu items? And the categories seem perfectly logical to me. "Graphics," "internet," "multimedia," etc. Sounds perfectly logical to me. Certainly, more logical than the Windows method of giving each company a top-level entry in the menu.
* The poor naming scheme that--despite close to five years of bitching--hasn't been changed in favor of something sane KDE will drop the 'k' when GNOME drops the 'g', Apple drops the 'i', and Microsoft drops the 'MS'. Remember, its "MS Word" not just "Word."
* The convoluted Control Center that is an example of poor interface design with 3,000 items and subitems, grouped together under a cursor that for some reason won't stop changing to the hand icon Actually, having all preferences in one program seems a lot more easy to navigate for me than a folder full of applets, one for each task. Work is on-going to make KControl more logical (there was an OSNews entry recently) but the "centralized control" aspect will remain. And Microsoft does it too --- consider the Windows NT administration console.
* The fact that the cursor changes to a hand icon when it moves over taskbar buttons (cursor changes are confusing, disorienting, and annoying to newbies and power users alike) The cursor changes to know when you can click on something. And you pulled that "cursor changes are confusing" thing out of your ass. People manage to use Internet Explorer on a daily basis with the cursor changing over links. You just don't like it because its different from what you're used to.
* The fact that when you tell KDE to put application menus at the top like MacOS...which is faster than slowing the cursor and pinpointing a menu in a floating window like in WindowsWorks fine on my machine. I'm guessing that you're not running 3.2...
* The seeming need for every new version of KDE to add five more sidebars, buttons, and features to KPanel/Konquerer/anything else beginning with K, instead of cleaning the interface and making things faster Every release of KDE since 2.0 has gotten faster. Every release of KDE since 3.0 has become more streamlined.
I could go on and on. I don't get why it is so slow. KDE isn't slow. Its got slow load-times for applications, but everything else is fast. In terms of user responsiveness, I find KDE to be faster than XP. In terms of redraw, KDE is even faster than XP. For example, try opening up two IE windows to Slashdot, one on top of the other. Resizing the top window will cause the bottom window blank --- for several seconds if you do it at the right speed. On my KDE, Konqueror doesn't do that.
Try to stay with the argument. I'm not arguing against Java. Java does precisely what it was designed to do. However, part of that design (constraints on what the programmer can do) is to commoditize programmers. The more pluggable code is, the more easily code can be passed to someone else and still be maintainable, the less useful the original programmer becomes. Further, the restrictive design of Java allows mediocre programmers to work on projects without hurting themselves. I don't see why anybody is surprised that companies are now taking advantage of this to hire mediocre, cheaper, programmers!
Slashdot Mirror
That doesn't make any sense. The Mesa guys *implemented* an open specification. And it took about a year and a half for Mesa 1.0 to come out. Reverse-engineering an existing driver would be a whole lot more complicated.
Besides, only a fool buys a $300 video card. Yes, I may get flamed for that comment, but that doesn't detract from the fact that (with a few exceptions) it's true.
First, $300 isn't unreasonable at all, especially if you're a heavy gamer. $300 is about the point where you get a top-of-the-line card, without spending extra change for features that don't add any performance. Spend less than about $200, and you give up significant amounts of performance or next-generation features. To those who spend a lot of time gaming, an extra $150 for the ability to turn the detail to maximum in all theri games is worth it. Especially when you consider that the amortized cost is only about $12.50 per month if you upgrade once a year. Second, not everybody uses their cards just for gaming. $300 cards aren't fast enough when you've got a deadline coming up and SolidEdge feels slower and slower by the minute...
No, the point is that we can't[1] code something better because they won't release the necessary programming information.
Check out this driver comparison on the Radeon 9000. The DRI drivers perform very poorly. Yet, the documentation for chips up to the Radeon 9200 *was* supplied to the DRI project. Indeed, the DRI project doesn't support any cards that don't have documentation. So "no documentation" is no excuse. Its just the simple fact that ATI knows the hardware because they designed it, and they are in a much better position to write these drivers than the OSS community.
Gotta love Slashdot posts stolen right out of Johnny Mnemonic...
I'd rather have a slightly slower driver than a fast clsoed source one
At least frame it correctly. Its not "slightly slower." Its much slower, sometimes by several times. The trade-off is not a little bit of speed for a lot of freedom. Its turning your $300 graphics card into a $75 card, for a very little bit of extra freedom. Unlike an OS, or major software app, you're not tied into drivers. If the constraints of a closed-source driver become too much, I can switch out my card and its drivers in a matter of hours.
As far as I'm concerned, begging manufacturers to open source their drivers is whining. I don't think the OSS spirit is about forcing people to open their code. If they choose to, great. If they don't, well, we can just code something better. That's precisely what GNU did to UNIX. Now the issue at hand is that the OSS community has *not* been able to produce something that can replace this proprietory code. Until that happens, they really don't have a solid bargaining position.
That's because VIA hardware *SUCKS* Same thing with Matrox. Their 3D cards sucked, so they released the specs for it. When they got a card that didn't suck (Parhelia) they stopped releasing specs.
Look, I care about freedom too. I use Open Source software whenever possible, and have sent in patches and stuff for it too. However, I'm willing to draw the line somewhere. For me, that line is: I'll use OSS software, unless it doesn't do what I need it to do. This is a case where OSS software doesn't do what I need it to do.
There are NO good OSS 3D drivers. By good I mean as fast and as full-featured as the Windows drivers for the same card. The DRI ATI drivers don't even support shaders! When OSS manages to make a good set of drivers for the R100 or R200 (whose specs *are* available) then they can say "OSS drivers would be better!" Until then, its just whining.
1) Because most other drivers are an order of magnitude simpler. Your modem driver, for example, does not need to do complex reordering of display lists to optimize performance. Also, OpenGL drivers are an entire OpenGL implementation. So its like open-sourcing not just your modem driver, but your whole IP stack.
2) No slight to the XFree86 developers, but:
a) Many of the XFree86 drivers (eg: nv) are significantly slower than their proprietory counterparts and in any case 2D is much easier to do drivers for than 3D,
b) The DRI project has not released any drivers that take full-advantage of the hardware. NVIDIA's binary driver will get you exact same performance and the exact same features the Windows driver has. The DRI drivers for Radeons are not only *much* slower, but don't even support basic features like vertex/pixel shaders!
If their binary is anything like NVIDIA's driver, that'd be a dozen megabytes of code to reverse engineer! Remember, a graphics driver is *not* a simple register-banger. Its an entire implementation of OpenGL. Much of that code is more or less hardware-independent, and has to do with optimizing display lists and whatnot. Reverse engineering the driver would be extremely difficult.
Listen to yourself! You payed $1000 for an 800MHz G4??? And you claim that that parts aren't overpriced? The G4 is a terrible CPU. Its completely bottlenecked by an ancient SDR memory bus. At today's prices, that G4 + motherboard is the equivilent of $70-$80 PC hardware.
And where do you get G5 motherboards? I can't find a place that sells them. Even if they do, they are most likely Apple-refurb parts, and carry the same insane prices the G4 refurb parts did.
Not FUD. I'm in the same position. I don't want to pay for "Apple Quality Hardware(TM)" (whatever that means, since Apple fans continually claim that Apple uses the exact same hardware). I wanna pay pricewatch prices for pieces of a computer I can stick into pieces I already have. Apple won't let me do that. Whereas I could put together a dual 2.8 GHz Xeon for about $1000 (I already have a harddrive, etc) a comparable (to be generous) dual 1.8 GHz G5 would set me back $2500. Sure, I could get a G4 iMac for dirt cheap, but who wants to use that POS?
Well, I don't put a lot of stock in GPA either. There is way too much variation between schools that try to go easy on their students GPA, and schools that relentlessy trash their students GPA. I think the best thing to do is skill-drills during interviews, to see how much they really know. I've found that people who look great on paper absolutely break down in such a situation.
If you graduated several years ago, I wouldn't be surprised that Java wasn't taught. It wasn't highly-hyped language at the time. Schools that teach mainly Java are much more prevalent these days where its all the rage. There are holdouts, still (kudos to MIT for doing their freshmen classes in Scheme according to the SICP) but not enough.
And the I wasn't really attacking Java. I think its a silly language, but it can be useful because of its vendor and library support. My point was more that its currently a fad language, and in response many places are graduating Java programmers just to feed the market.
(Toolkit designers: please please PLEASE give us a way to render a widget into a pixmap. That alone would solve a lot of problems.
Ask and ye shall receive.
You would be really surprised how low the bar as slipped these days. Where did you go to school, anyway? There are schools turning out good computer scientists, but there are a whole lot more churning out Java programmers.
Its as much a science as theoretical physics is. It is, to a large degree, applied mathematics.
That's because we just came out of several straight quarters of the worst quarterly earnings, revenue, and GDP growth in many years. The question is, where are the numbers now relative to how they were doing before the recession?
Are you using Luna or the classic engine? Cuz I've got 3 2.26GHz P4s running XP here, and they *all* blank.
Eugenia did a review of KDE 3.2-beta2 a while back. It was actually quite spot-on. As a KDE user, disagree with her ideologically on some points (eg. the configure menu comment) but on the whole I think it was a balanced evaluation of KDE, from a traditional UI designer's point of view.
Actually, yes they do, at least as of 3.2 :)
WTF?
What the fuck is wrong with you? You've posted the same thing dozens of times already. We get the picture!
And I've got 20 total entries in the root of my KMenu. I'd like to see how you get 20 = 30,000. Hell, a well-outfitted GNOME setup has at least a dozen!
MS Office,
c .
MS Outlook,
MS Excel,
MS Access, etc.
WinDVD,
WinAMP
WinRAR
WinZIP,
WinACE
iDVD,
iTunes,
iMovie
iPhoto,
iLife.
OpenOffice,
OpenWriter,
OpenImpress,
OpenCal
Jesus Christ. Its called pushing a brand.
And how the *FUCK* is a giant foot cooler than a giant, shiny, metal K?
How many items do you have in your home folder??? Mine has about a dozen (not counting hidden ones) and loads in less than a second (KDE 3.2, Debian, 2.0GHz P4)
* The 5-10 second load-time for KDE apps
This is a legitimate problem, though getting much better. With prelinking, load times are closer to 2-3 seconds rather than 5-10.
* The 4 second load time just to open a folder
Depends on the size of the folder. Normal-sized folders open pretty much instantly on my machine. Performance on large folders (/usr/bin with thousands of items) could be improved, though.
* The cramped and embarrassing K menu, with a 100 different groups and completely illogical redundancies like "Preferences," "System Settings," and "Control Center"
Um, my stock Debian KMenu has a dozen folders. "Preferences" is gone in 3.2. There is a "settings" folder and a "system" folder, which is logical --- one is for preferences, the other is for system utilities. Kinda analagous to Windows's "Control Panel" and "Accessories/System Tools" And "Control Center" is an entry in "Settings"! Are you sure you're not mistaking the quick- access area at the top for default menu items? And the categories seem perfectly logical to me. "Graphics," "internet," "multimedia," etc. Sounds perfectly logical to me. Certainly, more logical than the Windows method of giving each company a top-level entry in the menu.
* The poor naming scheme that--despite close to five years of bitching--hasn't been changed in favor of something sane
KDE will drop the 'k' when GNOME drops the 'g', Apple drops the 'i', and Microsoft drops the 'MS'. Remember, its "MS Word" not just "Word."
* The convoluted Control Center that is an example of poor interface design with 3,000 items and subitems, grouped together under a cursor that for some reason won't stop changing to the hand icon
Actually, having all preferences in one program seems a lot more easy to navigate for me than a folder full of applets, one for each task. Work is on-going to make KControl more logical (there was an OSNews entry recently) but the "centralized control" aspect will remain. And Microsoft does it too --- consider the Windows NT administration console.
* The fact that the cursor changes to a hand icon when it moves over taskbar buttons (cursor changes are confusing, disorienting, and annoying to newbies and power users alike)
The cursor changes to know when you can click on something. And you pulled that "cursor changes are confusing" thing out of your ass. People manage to use Internet Explorer on a daily basis with the cursor changing over links. You just don't like it because its different from what you're used to.
* The fact that when you tell KDE to put application menus at the top like MacOS...which is faster than slowing the cursor and pinpointing a menu in a floating window like in WindowsWorks fine on my machine. I'm guessing that you're not running 3.2...
* The seeming need for every new version of KDE to add five more sidebars, buttons, and features to KPanel/Konquerer/anything else beginning with K, instead of cleaning the interface and making things faster
Every release of KDE since 2.0 has gotten faster. Every release of KDE since 3.0 has become more streamlined.
I could go on and on. I don't get why it is so slow.
KDE isn't slow. Its got slow load-times for applications, but everything else is fast. In terms of user responsiveness, I find KDE to be faster than XP. In terms of redraw, KDE is even faster than XP. For example, try opening up two IE windows to Slashdot, one on top of the other. Resizing the top window will cause the bottom window blank --- for several seconds if you do it at the right speed. On my KDE, Konqueror doesn't do that.
Try to stay with the argument. I'm not arguing against Java. Java does precisely what it was designed to do. However, part of that design (constraints on what the programmer can do) is to commoditize programmers. The more pluggable code is, the more easily code can be passed to someone else and still be maintainable, the less useful the original programmer becomes. Further, the restrictive design of Java allows mediocre programmers to work on projects without hurting themselves. I don't see why anybody is surprised that companies are now taking advantage of this to hire mediocre, cheaper, programmers!
Its also got a much better, and more flexible, text and graphics layout engine. And unlike the MS Word UI, it's actually logical.
:)
My dad is one of those Windows users who laments to this day that the US government made him switch to MS Word