Well, the documentation isn't there because the API is still in development. The GNOME roadmap has GTK+ moving to Cairo in mid-late 2005, so it's likely that others will follow suit.
The DRI exists only in the X server and the kernel. Thus, the only two variables are still X and the kernel, just like NVIDIA's drivers.
Face it, ATI's driver team has problems. NVIDIA can keep 90% of their drivers the same between platforms (and thus have the same performance in both platforms). Yet, ATI's Linux drivers are still far inferior to their Windows drivers.
Enums, autoboxing of primitive types, and metadata...I suppose this is "inovation", right? There was no shameless copying of C# in any of these new features. It was all about customer demand.
Well, they weren't innovative when Microsoft did it, so they sure as hell aren't innovtative now. More generally, it'll be at least another decade before Sun or MS put a feature into C# or Java that wasn't already in Lisp or ML a decade ago.
1) Most kernel compatibility can be fixed in the open-source wrapper. I remember using the NVIDIA drivers on 2.5.x when they only supported 2.4 kernels. The only thing I can remember that required changes to the driver itself is the 4k-stack vs 8k-stack thing.
2) You've got to make a judgement call. What are the odds that NVIDIA will go out of business before I buy a new card? Since I only use cards for a couple of years, those odds are slim, so I can deal with them.
I'm not sure the problem is with ATI. Yes, the problem is with ATI. I've owned a Riva TNT, GeForce2 MX, and a GeForce4Go, and all three have always worked flawlessly with NVIDIA's Linux drivers. If NVIDIA can do it, then there is no reason ATI shouldn't be able to.
I see no reason why the drivers cannot be binary just like on Windows.
There are multiple reasons:
1) Windows supports a far smaller range of hardware than Linux. It'd be nice if people on Linux/PPC got to use video drivers too.
2) Windows changes far more slowly. Linux evolves at a faster pace, and can't afford to be tied down with driver compatibility.
3) Microsoft can afford the QA staff to deal with maintaining binary compatibility with dozens of weird binary drivers. Linux cannot.
Remember, Microsoft points to drivers are the #1 source of instability in Windows. The kernel developers don't want to be in the same boat --- if something goes wrong in an open source driver, at least it's their own fault and they can fix it.
Well, there is more in play than just installation. If the NVIDIA drivers work for you (they've always worked for me), you get an OpenGL driver that is every bit as good as their Windows OpenGL driver. That means it's good enough to run not just Quake, but any pro-level 3D app you'd care to run. ATI's Linux drivers, on the other hand, aren't as good as their Windows ones. They are slower and less stable. There is a reason ILM uses Quadros instead of FireGLs.
If your distro doesn't install the NVIDIA drivers automatically for you (SuSE and Gentoo do, among others), it's fairly trivial to install it manually. It's just a matter of using their installer program (making sure your kernel source tree is set up properly --- if you've ever built it, it should be), and then editing one line in XF86Config.
Look, Linux is not homogeneous, it is not straightforward to deliver drivers that will plug into XFree86 and just work on *every* distro with *every* little tweak your average Linux enthusiast has decided to play with. Actually, it's not all that hard either. XFree86 and X.org have a standard binary ABI, and pretty much everyone keeps X11 in the same place (/usr/X11R6). Delivering well-packaged binary drivers should be as easy as NVIDIA makes it out to be.
ATI driver guys are plenty competent at Linux/Unix drivers No they aren't, at least not in relation to the NVIDIA driver guys. ATI's Linux drivers are much slower than their Windows drivers, and much less stable. Driver development has always been a weak point of ATI --- there is a reason they are planning to completely rewrite their OpenGL driver soon.
Superior in what way? The X drivers for NVIDIA don't support OpenGL at all, don't support RENDER acceleration, and are significantly slower on x11perf. They are more stable on certain configurations, but on all the platforms I've used NVIDIA's drivers on, I don't think I've ever had them hard-lock.
Not only do I think that's a very bizarre statement to make you apparently go on to contradict I don't think we're arguing the same point here. Your original statement was: "Most people are casual users of computers at best". This is what I am taking exception to. I don't think this is really true anymore, and you don't really have any evidence that it is.
And you know what's even more productive? Software that's been designed by someone who understands the things it's users will want to do and makes them so easy to do it's users don't need training. Ideally, yes, software would be both highly efficient and have a shallow learning curve. Unfortunately, the real world doesn't work like that. There is often an engineering trade-off between learning curve and efficiency. Take, for example, proposal writing. Many people who work for government contractors spend most of their time writing proposals. A proper typesetting tool like LyX scales much better to the 100+ page, heavily cross-referenced documents than does a WYSIWYG tool like Word. This becomes especially true when there are multiple authors --- LyX removes the need to coordinate or post-process each person's submission to maintain consistant formatting. As efficient as it is, LyX definitely has a higher learning curve than Microsoft Word. Now, the efficiency gain is worth some additional training if you spend several hours every day for several years writing proposals, but few businesses seem to realize this.
Having an intuitive, easy/simple to use interface does not mean the software is only able to be simple in it's functionality. Powerful, efficient interfaces are quite often not intuitive. Ask anybody who uses Softimage XSI. The tool is regarded for it's phenomenal efficiency of workflow, but it is not an "intuitive, easy/simple to use" tool at all.
Most people have a wide variety of demanding things going on in their life (children, work) and only barely grasp how to use their computer. I would say this is a baseless assertion. These days, pretty much every professional worker uses a computer quite often. Most school-age and college-age students use a computer quite often. For both these groups of people, computers are integral to their work and study. Thus, for them, learning how to use a computer efficiently can pay off greatly in increased productivity.
If something is too difficult to do, they don't do it. They have what they consider to be "real lives" to get on with and "time at the computer" does not factor in to that. I would definitely consider "my boss wants this report on his desk in 2 hours" to be part of "real life." If not knowing how to use your computer wastes you an hour of time even a couple of times, that's going to have a very real impact on your "real life."
They precieve computers explicitly as tools for things like email, buying things on ebay and 'helping with homework'. Yes, for the majority of people, a computer is a tool. However, tools are designed to be efficient and productive --- intuitiveness comes second. People would never hire a carpenter that didn't know the ins-and-outs of his saw, but we seem to have no problem with office workers who don't know how to use their computers.
I don't disagree entirely --- there will always be a place for easy, simple, intuitive tools that can cater to the occasional computer user. However, an enormous percentage of the workforce use their computers as a primary tool. For them, it's more productive for to have efficient tools in whose use they are properly trained.
I think you're point II is stupid. Aside from grandma, most computer users I know use computers all the time. People use it hours a day for work (pretty much everyone with a desk job has a computer on it). Students use computers everyday to contact friends and do homework. Computer ownership is pretty much required at universities these days. So who is left? Not to many, I'd think. I'm convinced that in the near future, any usability recommendations that are predicated on peoples' unfamiliarity with computers will become obsolete and burdensome.
Well, you're right in that GNOME copies Apple much more than MS. In fact, if they'd just bite the bullet and support putting the menubar at the top, they'd have MacOS Classic for x86.
On the other hand, it's a bit silly to say that KDE is copying MS. The Konqueror example is a poor one. Konqueror is a KPart host and nothing more. It's neither a web-browser or a file manager (or a PDF viewer or a.doc viewer or whatever), but is a generic document viewer that can use whatever document plug-ins it finds.
NX is a compression and caching layer on top of the X11 protocol. It takes the basic X protocol and performs compression of protocol requests, caching of server responses, compression of images, etc.
1) Console cartridges are just expensive. EPROM chips can cost $25 to make, compared to $0.50 for a CD.
2) Initially, the production cost of CD-ROM games was quite high, because of the expense of tools (eg: multimedia authoring tools) that were required for "multimedia" programs.
3) The cost of games has gone up much faster than inflation between then and now.
It's quite easy to explain the trends without resorting to your train of thought.
I don't think it's a matter of blaming or excusing. I think what the original poster was getting at was that you can condemn the actions of extremist muslims, but you shouldn't have the morally superior outlook that many Christians have. If modern Christians were as poor as most Arabs, as uneducated as most Arabs, and lived in as poorly developed areas as most Arabs, they'd be just as intolerant and barbaric. The history of Christianity proves that. The fact that modern Christian extremists don't (usually) bomb buildings and kill people is mere luck of circumstance, not moral superiority as many believe. Once one realizes that fact, it becomes much harder to be self-rightous about it.
I'll buy that (it depends on your machine), but they have nothing to do with X. Resizing and moving is limited not by the raw display performance of X (which is all X really does), but interactions between X, the window manager, and the toolkit. It does no good blaming X for it when the blame lies elsewhere. In particular, there is no synchronization between the window manager and the toolkit, which means the window manager resizes as quick as it can (100s of times per second), creating an ugly visual effect and starving the app. There is a feature in the current draft of the netwm spec that allowed for synchronzation. On my desktop (2.0GHz P4), a KDE and Qt version with support for that feature resizes and moves windows as fast as Windows.
No you can't. You have no idea what you're talking about. Mozilla using it's own toolkit, and thus looking different from the rest of GNOME (which uses GTK+) is *exactly* the same thing as Office using it's own toolkit, and thus looking different from the rest of the Luna desktop. In both cases, attempts are made to make things look similar, but there are slight and even egregious differences (Office doesn't use the Luna scrollbars! Arrows look slightly different. Widgets are slightly flatter).
1) Sorry, you attempted proof-by-assertion. That is an illegal move in any civilized discussion. You lose, but thank you for playing, try again anytime! 2) They aren't at all similar in concept. If you think so, describe each system and show how they are similar. DRI is much more like modern versions of DirectX than DirectX 3. 3) You said X is not as responsive as a desktop windowing system should be. I pointed out that it is actually quite responsive (as evidenced by the fact that there are fast X apps, like most Qt ones). No assumptions on my part. I think you meant to say that "this or that application is slow or this or that toolkit is slow" which is something entirely different than saying X is slow. Also, suggesting that X should somehow do things to fix broken toolkits just doesn't make any sense. What do you suggest it does? The common answers of "get rid of the network protocol" or "put X in the kernel" will do jack-shit to help apps that redraw the entire scene when a few pixels change.
If you actually benchmark the damn thing, you'd see that it is pretty damn efficient. With a good set of drivers (NVIDIA), 3D is as fast as on Windows, and 2D is within 10% (go to www.rocklyte.com and look at their benchmarks comparing Athene, X, and Windows). My own benchmarks show that with reasonably-sized primitives, you can drive the graphics card at about half it's memory bandwidth, which is very efficient for a general-purpose 2D protocol. Also, you're obviously not a programmer if you think people use single pixels to draw movies or text. Movies use the XVideo extension or OpenGL, which stream pixel data in batches to the graphics card. Text drawing uses server-side pixmaps of each glyph, and the normal case involves simply sending some commands to string together the glyphs that are already in video memory.
Re:Time for X11R7 or even X12
on
The Power of X
·
· Score: 2, Insightful
That's where you are wrong. Windows apps *do* draw their own graphics. That's why Office doesn't have the same scrollbars as Luna apps, and why Visio XP has those god-awful Keramik-style blue-gradient toolbars.
Re:unified desktop
on
The Power of X
·
· Score: 5, Insightful
If Windows has such a "tight" and "well-designed" GUI, then please, tell me:
1) Why does Luna look like a pre-schooler threw up after eating several crayons? 2) Why do MS Office, MS Visio, and MS Visual Studio all look different (hint: they use different toolkits!) 3) Why does every other Windows apps (Winamp, Windows Media Player, Ephpod, etc, etc) use their own weird-looking skin? 4) Why do the buttons on every single installer (Wise, InstallShield, MSI) all look different?
Slashdot Mirror
Except for startup-speed, KDE is faster than GNOME. Poor resize and redraw plagues GNOME apps.
Well, the documentation isn't there because the API is still in development. The GNOME roadmap has GTK+ moving to Cairo in mid-late 2005, so it's likely that others will follow suit.
The DRI exists only in the X server and the kernel. Thus, the only two variables are still X and the kernel, just like NVIDIA's drivers.
Face it, ATI's driver team has problems. NVIDIA can keep 90% of their drivers the same between platforms (and thus have the same performance in both platforms). Yet, ATI's Linux drivers are still far inferior to their Windows drivers.
Enums, autoboxing of primitive types, and metadata...I suppose this is "inovation", right? There was no shameless copying of C# in any of these new features. It was all about customer demand.
Well, they weren't innovative when Microsoft did it, so they sure as hell aren't innovtative now. More generally, it'll be at least another decade before Sun or MS put a feature into C# or Java that wasn't already in Lisp or ML a decade ago.
Well, there are two points to consider:
1) Most kernel compatibility can be fixed in the open-source wrapper. I remember using the NVIDIA drivers on 2.5.x when they only supported 2.4 kernels. The only thing I can remember that required changes to the driver itself is the 4k-stack vs 8k-stack thing.
2) You've got to make a judgement call. What are the odds that NVIDIA will go out of business before I buy a new card? Since I only use cards for a couple of years, those odds are slim, so I can deal with them.
The DRI itself doesn't suck. It's the DRI drivers that generally aren't as good as the closed-source ones.
I'm not sure the problem is with ATI.
Yes, the problem is with ATI. I've owned a Riva TNT, GeForce2 MX, and a GeForce4Go, and all three have always worked flawlessly with NVIDIA's Linux drivers. If NVIDIA can do it, then there is no reason ATI shouldn't be able to.
I see no reason why the drivers cannot be binary just like on Windows.
There are multiple reasons:
1) Windows supports a far smaller range of hardware than Linux. It'd be nice if people on Linux/PPC got to use video drivers too.
2) Windows changes far more slowly. Linux evolves at a faster pace, and can't afford to be tied down with driver compatibility.
3) Microsoft can afford the QA staff to deal with maintaining binary compatibility with dozens of weird binary drivers. Linux cannot.
Remember, Microsoft points to drivers are the #1 source of instability in Windows. The kernel developers don't want to be in the same boat --- if something goes wrong in an open source driver, at least it's their own fault and they can fix it.
Well, there is more in play than just installation. If the NVIDIA drivers work for you (they've always worked for me), you get an OpenGL driver that is every bit as good as their Windows OpenGL driver. That means it's good enough to run not just Quake, but any pro-level 3D app you'd care to run. ATI's Linux drivers, on the other hand, aren't as good as their Windows ones. They are slower and less stable. There is a reason ILM uses Quadros instead of FireGLs.
If your distro doesn't install the NVIDIA drivers automatically for you (SuSE and Gentoo do, among others), it's fairly trivial to install it manually. It's just a matter of using their installer program (making sure your kernel source tree is set up properly --- if you've ever built it, it should be), and then editing one line in XF86Config.
Look, Linux is not homogeneous, it is not straightforward to deliver drivers that will plug into XFree86 and just work on *every* distro with *every* little tweak your average Linux enthusiast has decided to play with.
Actually, it's not all that hard either. XFree86 and X.org have a standard binary ABI, and pretty much everyone keeps X11 in the same place (/usr/X11R6). Delivering well-packaged binary drivers should be as easy as NVIDIA makes it out to be.
ATI driver guys are plenty competent at Linux/Unix drivers
No they aren't, at least not in relation to the NVIDIA driver guys. ATI's Linux drivers are much slower than their Windows drivers, and much less stable. Driver development has always been a weak point of ATI --- there is a reason they are planning to completely rewrite their OpenGL driver soon.
Superior in what way? The X drivers for NVIDIA don't support OpenGL at all, don't support RENDER acceleration, and are significantly slower on x11perf. They are more stable on certain configurations, but on all the platforms I've used NVIDIA's drivers on, I don't think I've ever had them hard-lock.
Not only do I think that's a very bizarre statement to make you apparently go on to contradict
I don't think we're arguing the same point here. Your original statement was: "Most people are casual users of computers at best". This is what I am taking exception to. I don't think this is really true anymore, and you don't really have any evidence that it is.
And you know what's even more productive? Software that's been designed by someone who understands the things it's users will want to do and makes them so easy to do it's users don't need training.
Ideally, yes, software would be both highly efficient and have a shallow learning curve. Unfortunately, the real world doesn't work like that. There is often an engineering trade-off between learning curve and efficiency. Take, for example, proposal writing. Many people who work for government contractors spend most of their time writing proposals. A proper typesetting tool like LyX scales much better to the 100+ page, heavily cross-referenced documents than does a WYSIWYG tool like Word. This becomes especially true when there are multiple authors --- LyX removes the need to coordinate or post-process each person's submission to maintain consistant formatting. As efficient as it is, LyX definitely has a higher learning curve than Microsoft Word. Now, the efficiency gain is worth some additional training if you spend several hours every day for several years writing proposals, but few businesses seem to realize this.
Having an intuitive, easy/simple to use interface does not mean the software is only able to be simple in it's functionality.
Powerful, efficient interfaces are quite often not intuitive. Ask anybody who uses Softimage XSI. The tool is regarded for it's phenomenal efficiency of workflow, but it is not an "intuitive, easy/simple to use" tool at all.
Most people have a wide variety of demanding things going on in their life (children, work) and only barely grasp how to use their computer.
I would say this is a baseless assertion. These days, pretty much every professional worker uses a computer quite often. Most school-age and college-age students use a computer quite often. For both these groups of people, computers are integral to their work and study. Thus, for them, learning how to use a computer efficiently can pay off greatly in increased productivity.
If something is too difficult to do, they don't do it. They have what they consider to be "real lives" to get on with and "time at the computer" does not factor in to that.
I would definitely consider "my boss wants this report on his desk in 2 hours" to be part of "real life." If not knowing how to use your computer wastes you an hour of time even a couple of times, that's going to have a very real impact on your "real life."
They precieve computers explicitly as tools for things like email, buying things on ebay and 'helping with homework'.
Yes, for the majority of people, a computer is a tool. However, tools are designed to be efficient and productive --- intuitiveness comes second. People would never hire a carpenter that didn't know the ins-and-outs of his saw, but we seem to have no problem with office workers who don't know how to use their computers.
I don't disagree entirely --- there will always be a place for easy, simple, intuitive tools that can cater to the occasional computer user. However, an enormous percentage of the workforce use their computers as a primary tool. For them, it's more productive for to have efficient tools in whose use they are properly trained.
I think you're point II is stupid. Aside from grandma, most computer users I know use computers all the time. People use it hours a day for work (pretty much everyone with a desk job has a computer on it). Students use computers everyday to contact friends and do homework. Computer ownership is pretty much required at universities these days. So who is left? Not to many, I'd think. I'm convinced that in the near future, any usability recommendations that are predicated on peoples' unfamiliarity with computers will become obsolete and burdensome.
Well, you're right in that GNOME copies Apple much more than MS. In fact, if they'd just bite the bullet and support putting the menubar at the top, they'd have MacOS Classic for x86.
.doc viewer or whatever), but is a generic document viewer that can use whatever document plug-ins it finds.
On the other hand, it's a bit silly to say that KDE is copying MS. The Konqueror example is a poor one. Konqueror is a KPart host and nothing more. It's neither a web-browser or a file manager (or a PDF viewer or a
NX is a compression and caching layer on top of the X11 protocol. It takes the basic X protocol and performs compression of protocol requests, caching of server responses, compression of images, etc.
Um, there is a simpler explanation:
1) Console cartridges are just expensive. EPROM chips can cost $25 to make, compared to $0.50 for a CD.
2) Initially, the production cost of CD-ROM games was quite high, because of the expense of tools (eg: multimedia authoring tools) that were required for "multimedia" programs.
3) The cost of games has gone up much faster than inflation between then and now.
It's quite easy to explain the trends without resorting to your train of thought.
I don't think it's a matter of blaming or excusing. I think what the original poster was getting at was that you can condemn the actions of extremist muslims, but you shouldn't have the morally superior outlook that many Christians have. If modern Christians were as poor as most Arabs, as uneducated as most Arabs, and lived in as poorly developed areas as most Arabs, they'd be just as intolerant and barbaric. The history of Christianity proves that. The fact that modern Christian extremists don't (usually) bomb buildings and kill people is mere luck of circumstance, not moral superiority as many believe. Once one realizes that fact, it becomes much harder to be self-rightous about it.
I'll buy that (it depends on your machine), but they have nothing to do with X. Resizing and moving is limited not by the raw display performance of X (which is all X really does), but interactions between X, the window manager, and the toolkit. It does no good blaming X for it when the blame lies elsewhere. In particular, there is no synchronization between the window manager and the toolkit, which means the window manager resizes as quick as it can (100s of times per second), creating an ugly visual effect and starving the app. There is a feature in the current draft of the netwm spec that allowed for synchronzation. On my desktop (2.0GHz P4), a KDE and Qt version with support for that feature resizes and moves windows as fast as Windows.
No you can't. You have no idea what you're talking about. Mozilla using it's own toolkit, and thus looking different from the rest of GNOME (which uses GTK+) is *exactly* the same thing as Office using it's own toolkit, and thus looking different from the rest of the Luna desktop. In both cases, attempts are made to make things look similar, but there are slight and even egregious differences (Office doesn't use the Luna scrollbars! Arrows look slightly different. Widgets are slightly flatter).
1) Sorry, you attempted proof-by-assertion. That is an illegal move in any civilized discussion. You lose, but thank you for playing, try again anytime!
2) They aren't at all similar in concept. If you think so, describe each system and show how they are similar. DRI is much more like modern versions of DirectX than DirectX 3.
3) You said X is not as responsive as a desktop windowing system should be. I pointed out that it is actually quite responsive (as evidenced by the fact that there are fast X apps, like most Qt ones). No assumptions on my part. I think you meant to say that "this or that application is slow or this or that toolkit is slow" which is something entirely different than saying X is slow. Also, suggesting that X should somehow do things to fix broken toolkits just doesn't make any sense. What do you suggest it does? The common answers of "get rid of the network protocol" or "put X in the kernel" will do jack-shit to help apps that redraw the entire scene when a few pixels change.
If you actually benchmark the damn thing, you'd see that it is pretty damn efficient. With a good set of drivers (NVIDIA), 3D is as fast as on Windows, and 2D is within 10% (go to www.rocklyte.com and look at their benchmarks comparing Athene, X, and Windows). My own benchmarks show that with reasonably-sized primitives, you can drive the graphics card at about half it's memory bandwidth, which is very efficient for a general-purpose 2D protocol.
Also, you're obviously not a programmer if you think people use single pixels to draw movies or text. Movies use the XVideo extension or OpenGL, which stream pixel data in batches to the graphics card. Text drawing uses server-side pixmaps of each glyph, and the normal case involves simply sending some commands to string together the glyphs that are already in video memory.
All four of those companies are members of X.org.
That's where you are wrong. Windows apps *do* draw their own graphics. That's why Office doesn't have the same scrollbars as Luna apps, and why Visio XP has those god-awful Keramik-style blue-gradient toolbars.
If Windows has such a "tight" and "well-designed" GUI, then please, tell me:
1) Why does Luna look like a pre-schooler threw up after eating several crayons?
2) Why do MS Office, MS Visio, and MS Visual Studio all look different (hint: they use different toolkits!)
3) Why does every other Windows apps (Winamp, Windows Media Player, Ephpod, etc, etc) use their own weird-looking skin?
4) Why do the buttons on every single installer (Wise, InstallShield, MSI) all look different?