That is the exact kind of snobish attitude that pisses me off. Windows devlopers are GOOD. A lot of them could kick the ass of half of the Linux kernel hackers, including Linus himself. They choose to program for windows because thats what everyone uses. Face it Linux is not mainstream yet. Everyone has heard of it, but 5 people outside the nerd community use it on their desktops. Just because they program for the OS that makes the most business sense doesn't mean their bad programmers. I'm sure the guys at Adobe could whoop the guys at GIMP in a coding contest. Same thing for the guys at a dozen other companies that put out good products despite their OS limitations. Hell even the guys at Microsoft (esp. the guys who programmed DirectDraw and DirectSound and DirectInput) could teach the Linux developers a few lessons. Don't be a Linux bigot. Its just as bad as a window bigot.
Thats untrue. Very few REAL developers use MFC because it is A) A bitch to code. B) A bitch to maintain and C) A bitch with runtime performance. All MFC developers should be forced to run their own applications.
Whatever you do don't use MFC. Performance sucks and it is a bitch to maintian MFC code. Someone ought to drag the MFC developers over to BeOS's website and make them learn a GOOD C++ API.
Yes its an old protocol. Yes its a bitch to implement on a large network, and yes no-one uses it anymore. But it is probably one of the fastest network protocols available. Between two machines over a 10 megabit ethernet connection, I get 400 kbps over TCP in Linux, 350Kbps in windows (when it works) and on NetBUI I get 575 kbps in windows. There is something to be said for fast and light if you're just connecting a few computers and don't need very many features.
When MS does it people laugh. When Linux does it its cool? I don't know about you, but Linux has one strong point. Thats small servers and large workstations. Of course its possible to get Linux to run on small to huge platforms, but is it the optimum thing? Say I'm a company putting Linux on a 256 CPU box to do some mission critical task. Will anyone from the Linux community tell me I've lost it?
My point is that by saying, "32 meg is plenty" you implied that you though that Linux would get the entire 32 meg to run. But shrink it down to 6 or 8 meg, with linux taking up 4 yet qnx only taking up 1, you get quite a different picture.
I'm not one to nitpick, but yes technically Linux is becoming more like a microkernel. I never said that one day linux will become a microkernel, I said it is moving in that direction. Thus if Linux ends up being a mix of the two, then it IS becoming more like a microkernel and less like a monolithic kernel. Perhaps a diagram will help Monolithic--------- Inbetween ---------microkernel Direction A -----> Thus if Linux starts out at Monolithic, and continues along direction A with the eventual goal of inbetween, then it is moving towards the microkernel, hence it is becoming more like a microkernel.
No, QNX does not start taking up space. Have you ever seen their demos? In a floppy disk they fit the GUI, a web browser and server and a bunch of other stuff. Photon is ridiculously tiny. Its is a hell of a lot smaller than Linux because it was designed for embedded markets. And does developer support really matter? Most PDA users won't run command line apps on their PDA, and you're not telling me they tried to shoehorn X into there are you? If anything, samsung has a custom GUI, and developers will have to learn that. Not much different from QNX, which is a POSIX system and the developers having to learn Photon. QNX is also extremely modular, for most purposes, just as modular as Linux.
You forget that most of the RAM is used for application storage. Most WinCE machines I've seen only have a few meg of free RAM cuz every piece of data you put in there take up some of you're RAM. You put 6 MP3s, suddenly you're down to 4 meg of RAM.
From all that I can find on ARM's and Samsung's website, it appears that the CPU in the machine is some member of the ARM9 family, probably the ARM9E-S, because it includes DSP instructions. The spec sheet for the PDA says it has a 3D audio codec (probably dolby because the ARM9 dsp can handle that) plays MP3s and MPEG video, all of which the ARM9E-S is perfectly suited to do. You can go too this website http://www.arm.com/Pro+Peripherals/Cores/ARM9ES/ to find out some more about the CPU. Its a really nice CPU, puts out about 200 something MIPS and uses about 1.3 or 2.5 mW/MHz depending on the what voltage its running at.
When you hear about something like this, you really have to ask yourself, is it practical? Linux is a UNIX. No matter how slimmed down, it still carries a lot of UNIX baggage. It still has a lot of complexity that a PDA really doesn't need. MS has fallen into the same trap by shoe-horning windows into PDAs and ended up with the monstrosity that is WinCE. Even in text mode, Linux needs about 4 meg to run comfortably. Thats great when you have a 32 meg RAM PC, but think about it, that 32 meg is mostly going to storage for apps. So in the end, you really end up with about 8 meg or so, and if half that is taken up by the OS, that leaves precious little for the OS. PDAs are in a strange position. They are too big for a PalmOS type OS, but too small for a desktop or laptop OS. Instead of shrinking the desktop OS, doesn't it make sense to enlarge the small OS? I would think that a better starting point for a device like this would be QNX. The kernel is 32K, includeds networking, and Photon is really small and fast. Of course you still have the problem that make WinCE handhelds a pain to use. The desktop metaphor reeks on a 4" screen! Ideally, you'd have a taskbar with a start menu type thing, and would switch apps just by clicking on the taskbar. Throw out the rest of this desktop metaphor. There is a reason that palms are so popular; the interface fits on a PDA. It seems to me that Palm, GeOS, and Netwon are the only ones who ever "got it."
Its increadible, I just managed to read 15 trolls in a row! Anyway, onto the matter at hand. People who are using the dev kernels, would you like to enlighten us about how they work so far? Is 2.4 going to be nice and stable when it comes out? Finally, its coming out soon, right? I remember hearing something about XFree 4.0 needing a kernel driver thats only in 2.4. On a slightly unrelated note, has anyone noticed that Linux is becoming more and more like a microkernel everyday? Stuff is being moved out into user space, and the whole XFree server in user space with small kernel driver is exactly how BeOS and Chorus graphics drivers are implemented.
I never said porting would be easy. I said that a few thousand lines of critical ASM is easy enough just to rewrite. And check out some code written for MASM at www.gamedev.net it look like C with short function names. Yes optimizing compilers work great these days (even gcc works decently) but some things don't get optimized correctly and is just easier to write in ASM. I agree that writing an entire base OS in assembly is stupid, I never said otherwise. The type of ASM I'm talking about is small, critical sections of code written in ASM to really enhance performance. Back in the day before OpenGL on the PC, people would write their games in C, but write something like a triangle rasterizer (a few hundred lines of code) in ASM for maximum performance. The guys at Be are doing this with their OpenGL implementation, writing some critical, but small, portions in ASM for speed.
I'm not sure about the statistics of how many companies are in the position of having extensive support contracts, so I may be wrong as to the size of the support market, but my post really wasn't about companies like Sun anyway. Thats not the market the Linux in its present form is targeting. Instead Caldera and Redhat are mainly targeting the smaller server and workstation business, ie. everyone who uses NT. I can bet you that most shops that use NT have their own IT staff handle things, not MS staff. In the future, when Linux makes inroads in the high end, selling support might make sense, but in the market that Linux is in now, selling support is not a very good business proposition.
Sure, take pot shots at ASM programmers. The ASM is better, depending on what you're doing. For example if you're writing some small low-level code, A) Its short so porting isn't really an issue. A few thousand lines of ASM code is just as easy to re-write for another arch rather than port. B) ASM is just as readable as C these days. take a look at MASM (Microsofts ASM compiler) C)Its fast as hell, and for small critical sections of programs, its really usefull.
I really don't know if any of the Linux companies are worth their massive stock valuations. Companies like amazon.com also aren't worth it, but have to potential to be. Linux however, probably doesn't have the potential to make anyone any money. How does a Linux company make money? By selling support, right? Yet, how many companies actually use the support from their OS provider? Most have support contracts for their hardware, but software issues are usually handled by internal IT staff. Companies like SUN and SGI can make money by giving their OS away because they have hardware to sell. Caldera on the other hand, only make money from the distro and from any support it sells. That means a company that previously bought 100 copies of NT from MS, can now just buy 1 copy of Linux from cheapbytes.com. Its not like they used MS support anyway, they just had their IT guy fix it. Thus the market for a corperate linux distro shrinks from everyone who uses the OS, to those who use the OS, but can't install it themselves and can't afford IT staff. Thats a very small piece of the market. Another way for the company to make money would be to write propriatory extensions for Linux. Caldera actually tried this before with Caldera Linux 1.3. Of course it failed. In the wide open world of Linux, proriatory Linux extensions won't fly. This can easily be seen in Redhat. You'd think that a company that entirely owns the corperate linux market would be making money? Redhat is losing something like $35 million a year. As for VA Linux, what the hell are they thinking? Why would I buy from some upstart Linux company when I can get systems from Dell, a proven, reliable company that actually has good service and support! Believe it or not, corperate types don't buy into the labor of love that is Linux. They are business, and they will do whatever makes sense, not what makes them feel good by helping out some upstart linux company.
Its interesting how many people complain how a standard GUI would lead to lack of flexibility. In Linux's current state, this is true, but doesn't have to be. The reason that writing a standard GUI Linux style wouldn't work has to do with the fact that Linux GUIs are terribly non-modular. They define everything from a drag and drop protocol to widget sets, to window handling, to system level APIs and other unrelated services. There are a set of standard things that NEVER change no matter what app or GUI one uses. Things like a drag and drop protocol should be standard on the system. I seriously doubt that the drag and drop protocol defined by GNOME is terribly different from the one defined by KDE. A lot of the services that KDE and GNOME provide should really be at a lower level in the system than the window manager. KDE and GNOME bill themselves as "desktop environments" because they provide many services to an application. These services are all essentially the same between GNOME and KDE. Its that part of the desktop that should be standardized. All the stuff above that should be in a window mangager which can be configurable. Programmatically it would look like this... You call the GUI functions defined by the API, but how they are implemented would be handled by the window manager. You might ask for a color picker widget, but depending on what window manager you had it would look and act different, long as it had a standard interface.
Why was this moderated up? You do realize that not every computer is in a minitower. These things are in massive cases. I know I've seen pictures of SGI origins with 64 CPUs in one box. I also remember that some massively parallel computers (IBM POWER series I think) don't have casings on the CPU. They do this weird thing where they just take a big piece of silicon, burn the chips into it with connection inbetween, then encase the whole thing. Thats leads to like 64 CPUs in a case the size of a pizza.
Its silly to say that most non-technical users let somebody configure their computer for them. Not everyone has their own personal IT staff. In my experiance, the majority of homes users are at the point that they can understand right clicking on the desktop, hitting display configuration, and changing their resolution, wallpaper, etc from there. They can also install hardware, because they can plug it in, turn on the computer, and follow the instructions from the plug and play installer. Some more advanced users (ie. teenagers) can even download drivers and stuff. Of course, even at the skill level (considered an intermediate Windows user) they still can't figure out how to change their mouse cursor in X. You know why? Because it is in some stupid CLI app called xsetroot. What was the person who wrote that thinking? Why the hell is all the X configuration stuff in CLI apps? Do you want to be able to change your mouse cursor from the CLI? And don't tell me its for flexibility, in windows, I can have an app give me a random cursor everytime I boot up! KDE and GNOME have put some of the basic configuration stuff into their control panels, but its still ridiculous what one has to do to get stuff configured. I have a pretty vanilla system, TNT graphics, AWE64 soundcard. Yet, to get the TNT working, I had to run the command line app, xconfigurator, which asked my what resolutions I wanted to enable? Hello! Why can every other OS on the planet figure it out themselves, yet Linux can't! To get my soundcard working was even more of a chore. I had to use sndconfig, (which isn't run automatically) and even though its plug and play, I had to give it IRQs and DMAs. And I don't see it getting any better, unless some infrastructure is put in to make some CONSISTANT (I know Linux users hate that word) hardware installation schemes. BeOS is at the point that where it can dynamicly detect anything plugged into a PCI, ISA, or USB slot/port, and install the driver. If the driver is on a floppy, then all you have to do it copy into the driver directory. Voila, consistant, easy, and about as flexible as you can get.
No, I don't think he is excluding the others from being GUIs, but I think he is saying that to get the same GUI experiance as in Windows and MacOS (standard drag and drop, standard widget sets, a large set of libraries for graphics programming) you have to use one of the desktop environments. Also, remember, from a non-linux user point of view, is there any GUIs besides KDE and GNOME? What comes installed standard on RedHat and every distro that is a copy of Redhat? Gnome or KDE! Second, yes a keyboard is there for typing stuff. NOT TYPING COMMANDS! A GUI should not require you to ever drop out into CLI, or use an xterm to configure stuff. Thats why its a GUI. You can argue, GUI vs. CLI all you want, but if a GUI requires some CLI use to configure, then it is a badly designed GUI! Do the Linux CLI require you to boot up X to configure its font? Of course not, that would be ridiculous. In BeOS and Windows, I have yet to have to drop into the CLI to get something done. (Aside from installing Xwindows in BeOS) In Linux I find myself having to do it all the time.
Whoa, MII is way ahead of what I'm talking about. The MII is technically the 6x86MX, the version after the Cyrix 6x86. Back in the 6x86 days, Pentium MMXs didn't even exist, and the 6x86 FPU and integer unit whopped anything available. Think 1995-1996 and you have the idea.
I just want to clear up some of the confusion some people are having about Cyrix and their FPUs. In the past, Cyrix chips were the fastest you could buy, in integer OR fpu. It wasn't until the Pentium MMX and its pipelined FPU that the Cyrix name became synonomous with crappy FPU performance.
I don't understand all the people saying how Transmeta is competition to Intel, even at the low end. If you hadn't read the specs, even the 700MHz version, offers pretty bad FP performance compared to a P6 architecture. It has good integer performance, but in low end consumer space, thats irrelevant. Most apps that really need power are FP based and a chip with an FPU that can't match a 500MHz Celeron definately won't compete. Transmeta is for lightweight portables, nothing more.
Slashdot Mirror
That is the exact kind of snobish attitude that pisses me off. Windows devlopers are GOOD. A lot of them could kick the ass of half of the Linux kernel hackers, including Linus himself. They choose to program for windows because thats what everyone uses. Face it Linux is not mainstream yet. Everyone has heard of it, but 5 people outside the nerd community use it on their desktops. Just because they program for the OS that makes the most business sense doesn't mean their bad programmers. I'm sure the guys at Adobe could whoop the guys at GIMP in a coding contest. Same thing for the guys at a dozen other companies that put out good products despite their OS limitations. Hell even the guys at Microsoft (esp. the guys who programmed DirectDraw and DirectSound and DirectInput) could teach the Linux developers a few lessons. Don't be a Linux bigot. Its just as bad as a window bigot.
Thats untrue. Very few REAL developers use MFC because it is A) A bitch to code. B) A bitch to maintain and C) A bitch with runtime performance. All MFC developers should be forced to run their own applications.
Whatever you do don't use MFC. Performance sucks and it is a bitch to maintian MFC code. Someone ought to drag the MFC developers over to BeOS's website and make them learn a GOOD C++ API.
USB is ridiculously expensive. Retail a 4 port USB card goes about $40!
Yes its an old protocol. Yes its a bitch to implement on a large network, and yes no-one uses it anymore. But it is probably one of the fastest network protocols available. Between two machines over a 10 megabit ethernet connection, I get 400 kbps over TCP in Linux, 350Kbps in windows (when it works) and on NetBUI I get 575 kbps in windows.
There is something to be said for fast and light if you're just connecting a few computers and don't need very many features.
When MS does it people laugh. When Linux does it its cool? I don't know about you, but Linux has one strong point. Thats small servers and large workstations. Of course its possible to get Linux to run on small to huge platforms, but is it the optimum thing? Say I'm a company putting Linux on a 256 CPU box to do some mission critical task. Will anyone from the Linux community tell me I've lost it?
My point is that by saying, "32 meg is plenty" you implied that you though that Linux would get the entire 32 meg to run. But shrink it down to 6 or 8 meg, with linux taking up 4 yet qnx only taking up 1, you get quite a different picture.
I'm not one to nitpick, but yes technically Linux is becoming more like a microkernel. I never said that one day linux will become a microkernel, I said it is moving in that direction. Thus if Linux ends up being a mix of the two, then it IS becoming more like a microkernel and less like a monolithic kernel. Perhaps a diagram will help
Monolithic--------- Inbetween ---------microkernel
Direction A ----->
Thus if Linux starts out at Monolithic, and continues along direction A with the eventual goal of inbetween, then it is moving towards the microkernel, hence it is becoming more like a microkernel.
No, QNX does not start taking up space. Have you ever seen their demos? In a floppy disk they fit the GUI, a web browser and server and a bunch of other stuff. Photon is ridiculously tiny. Its is a hell of a lot smaller than Linux because it was designed for embedded markets. And does developer support really matter? Most PDA users won't run command line apps on their PDA, and you're not telling me they tried to shoehorn X into there are you? If anything, samsung has a custom GUI, and developers will have to learn that. Not much different from QNX, which is a POSIX system and the developers having to learn Photon. QNX is also extremely modular, for most purposes, just as modular as Linux.
You forget that most of the RAM is used for application storage. Most WinCE machines I've seen only have a few meg of free RAM cuz every piece of data you put in there take up some of you're RAM. You put 6 MP3s, suddenly you're down to 4 meg of RAM.
From all that I can find on ARM's and Samsung's website, it appears that the CPU in the machine is some member of the ARM9 family, probably the
ARM9E-S, because it includes DSP instructions. The spec sheet for the PDA says it has a 3D audio codec (probably dolby because the ARM9 dsp can handle that) plays MP3s and MPEG video, all of which the ARM9E-S is perfectly suited to do. You can go too this website http://www.arm.com/Pro+Peripherals/Cores/ARM9ES/
to find out some more about the CPU. Its a really nice CPU, puts out about 200 something MIPS and uses about 1.3 or 2.5 mW/MHz depending on the what voltage its running at.
When you hear about something like this, you really have to ask yourself, is it practical? Linux is a UNIX. No matter how slimmed down, it still carries a lot of UNIX baggage. It still has a lot of complexity that a PDA really doesn't need. MS has fallen into the same trap by shoe-horning windows into PDAs and ended up with the monstrosity that is WinCE. Even in text mode, Linux needs about 4 meg to run comfortably. Thats great when you have a 32 meg RAM PC, but think about it, that 32 meg is mostly going to storage for apps. So in the end, you really end up with about 8 meg or so, and if half that is taken up by the OS, that leaves precious little for the OS. PDAs are in a strange position. They are too big for a PalmOS type OS, but too small for a desktop or laptop OS. Instead of shrinking the desktop OS, doesn't it make sense to enlarge the small OS? I would think that a better starting point for a device like this would be QNX. The kernel is 32K, includeds networking, and Photon is really small and fast. Of course you still have the problem that make WinCE handhelds a pain to use. The desktop metaphor reeks on a 4" screen! Ideally, you'd have a taskbar with a start menu type thing, and would switch apps just by clicking on the taskbar. Throw out the rest of this desktop metaphor. There is a reason that palms are so popular; the interface fits on a PDA. It seems to me that Palm, GeOS, and Netwon are the only ones who ever "got it."
Xfree doesn't require Linux, but DRI requires a small kernel driver that is supposedly easily portable to different Unicies.
Its increadible, I just managed to read 15 trolls in a row! Anyway, onto the matter at hand. People who are using the dev kernels, would you like to enlighten us about how they work so far? Is 2.4 going to be nice and stable when it comes out? Finally, its coming out soon, right? I remember hearing something about XFree 4.0 needing a kernel driver thats only in 2.4. On a slightly unrelated note, has anyone noticed that Linux is becoming more and more like a microkernel everyday? Stuff is being moved out into user space, and the whole XFree server in user space with small kernel driver is exactly how BeOS and Chorus graphics drivers are implemented.
I never said porting would be easy. I said that a few thousand lines of critical ASM is easy enough just to rewrite. And check out some code written for MASM at www.gamedev.net it look like C with short function names. Yes optimizing compilers work great these days (even gcc works decently) but some things don't get optimized correctly and is just easier to write in ASM. I agree that writing an entire base OS in assembly is stupid, I never said otherwise. The type of ASM I'm talking about is small, critical sections of code written in ASM to really enhance performance. Back in the day before OpenGL on the PC, people would write their games in C, but write something like a triangle rasterizer (a few hundred lines of code) in ASM for maximum performance. The guys at Be are doing this with their OpenGL implementation, writing some critical, but small, portions in ASM for speed.
I'm not sure about the statistics of how many companies are in the position of having extensive support contracts, so I may be wrong as to the size of the support market, but my post really wasn't about companies like Sun anyway. Thats not the market the Linux in its present form is targeting. Instead Caldera and Redhat are mainly targeting the smaller server and workstation business, ie. everyone who uses NT. I can bet you that most shops that use NT have their own IT staff handle things, not MS staff. In the future, when Linux makes inroads in the high end, selling support might make sense, but in the market that Linux is in now, selling support is not a very good business proposition.
Sure, take pot shots at ASM programmers.
The ASM is better, depending on what you're doing.
For example if you're writing some small low-level code,
A) Its short so porting isn't really an issue. A few thousand lines of ASM code is just as easy to re-write for another arch rather than port.
B) ASM is just as readable as C these days. take a look at MASM (Microsofts ASM compiler)
C)Its fast as hell, and for small critical sections of programs, its really usefull.
I really don't know if any of the Linux companies are worth their massive stock valuations. Companies like amazon.com also aren't worth it, but have to potential to be. Linux however, probably doesn't have the potential to make anyone any money. How does a Linux company make money? By selling support, right? Yet, how many companies actually use the support from their OS provider? Most have support contracts for their hardware, but software issues are usually handled by internal IT staff. Companies like SUN and SGI can make money by giving their OS away because they have hardware to sell. Caldera on the other hand, only make money from the distro and from any support it sells. That means a company that previously bought 100 copies of NT from MS, can now just buy 1 copy of Linux from cheapbytes.com. Its not like they used MS support anyway, they just had their IT guy fix it. Thus the market for a corperate linux distro shrinks from everyone who uses the OS, to those who use the OS, but can't install it themselves and can't afford IT staff. Thats a very small piece of the market. Another way for the company to make money would be to write propriatory extensions for Linux. Caldera actually tried this before with Caldera Linux 1.3. Of course it failed. In the wide open world of Linux, proriatory Linux extensions won't fly. This can easily be seen in Redhat. You'd think that a company that entirely owns the corperate linux market would be making money? Redhat is losing something like $35 million a year. As for VA Linux, what the hell are they thinking? Why would I buy from some upstart Linux company when I can get systems from Dell, a proven, reliable company that actually has good service and support! Believe it or not, corperate types don't buy into the labor of love that is Linux. They are business, and they will do whatever makes sense, not what makes them feel good by helping out some upstart linux company.
Its interesting how many people complain how a standard GUI would lead to lack of flexibility. In Linux's current state, this is true, but doesn't have to be. The reason that writing a standard GUI Linux style wouldn't work has to do with the fact that Linux GUIs are terribly non-modular. They define everything from a drag and drop protocol to widget sets, to window handling, to system level APIs and other unrelated services. There are a set of standard things that NEVER change no matter what app or GUI one uses. Things like a drag and drop protocol should be standard on the system. I seriously doubt that the drag and drop protocol defined by GNOME is terribly different from the one defined by KDE. A lot of the services that KDE and GNOME provide should really be at a lower level in the system than the window manager. KDE and GNOME bill themselves as "desktop environments" because they provide many services to an application. These services are all essentially the same between GNOME and KDE. Its that part of the desktop that should be standardized. All the stuff above that should be in a window mangager which can be configurable. Programmatically it would look like this... You call the GUI functions defined by the API, but how they are implemented would be handled by the window manager. You might ask for a color picker widget, but depending on what window manager you had it would look and act different, long as it had a standard interface.
Why was this moderated up? You do realize that not every computer is in a minitower. These things are in massive cases. I know I've seen pictures of SGI origins with 64 CPUs in one box. I also remember that some massively parallel computers (IBM POWER series I think) don't have casings on the CPU. They do this weird thing where they just take a big piece of silicon, burn the chips into it with connection inbetween, then encase the whole thing. Thats leads to like 64 CPUs in a case the size of a pizza.
Its silly to say that most non-technical users let somebody configure their computer for them. Not everyone has their own personal IT staff. In my experiance, the majority of homes users are at the point that they can understand right clicking on the desktop, hitting display configuration, and changing their resolution, wallpaper, etc from there. They can also install hardware, because they can plug it in, turn on the computer, and follow the instructions from the plug and play installer. Some more advanced users (ie. teenagers) can even download drivers and stuff. Of course, even at the skill level (considered an intermediate Windows user) they still can't figure out how to change their mouse cursor in X. You know why? Because it is in some stupid CLI app called xsetroot. What was the person who wrote that thinking? Why the hell is all the X configuration stuff in CLI apps? Do you want to be able to change your mouse cursor from the CLI? And don't tell me its for flexibility, in windows, I can have an app give me a random cursor everytime I boot up! KDE and GNOME have put some of the basic configuration stuff into their control panels, but its still ridiculous what one has to do to get stuff configured. I have a pretty vanilla system, TNT graphics, AWE64 soundcard. Yet, to get the TNT working, I had to run the command line app, xconfigurator, which asked my what resolutions I wanted to enable? Hello! Why can every other OS on the planet figure it out themselves, yet Linux can't! To get my soundcard working was even more of a chore. I had to use sndconfig, (which isn't run automatically) and even though its plug and play, I had to give it IRQs and DMAs. And I don't see it getting any better, unless some infrastructure is put in to make some CONSISTANT (I know Linux users hate that word) hardware installation schemes. BeOS is at the point that where it can dynamicly detect anything plugged into a PCI, ISA, or USB slot/port, and install the driver. If the driver is on a floppy, then all you have to do it copy into the driver directory. Voila, consistant, easy, and about as flexible as you can get.
No, I don't think he is excluding the others from being GUIs, but I think he is saying that to get the same GUI experiance as in Windows and MacOS (standard drag and drop, standard widget sets, a large set of libraries for graphics programming) you have to use one of the desktop environments. Also, remember, from a non-linux user point of view, is there any GUIs besides KDE and GNOME? What comes installed standard on RedHat and every distro that is a copy of Redhat? Gnome or KDE! Second, yes a keyboard is there for typing stuff. NOT TYPING COMMANDS! A GUI should not require you to ever drop out into CLI, or use an xterm to configure stuff. Thats why its a GUI. You can argue, GUI vs. CLI all you want, but if a GUI requires some CLI use to configure, then it is a badly designed GUI! Do the Linux CLI require you to boot up X to configure its font? Of course not, that would be ridiculous. In BeOS and Windows, I have yet to have to drop into the CLI to get something done. (Aside from installing Xwindows in BeOS) In Linux I find myself having to do it all the time.
Whoa, MII is way ahead of what I'm talking about. The MII is technically the 6x86MX, the version after the Cyrix 6x86. Back in the 6x86 days, Pentium MMXs didn't even exist, and the 6x86 FPU and integer unit whopped anything available. Think 1995-1996 and you have the idea.
I just want to clear up some of the confusion some people are having about Cyrix and their FPUs. In the past, Cyrix chips were the fastest you could buy, in integer OR fpu. It wasn't until the Pentium MMX and its pipelined FPU that the Cyrix name became synonomous with crappy FPU performance.
I don't understand all the people saying how Transmeta is competition to Intel, even at the low end. If you hadn't read the specs, even the 700MHz version, offers pretty bad FP performance compared to a P6 architecture. It has good integer performance, but in low end consumer space, thats irrelevant. Most apps that really need power are FP based and a chip with an FPU that can't match a 500MHz Celeron definately won't compete. Transmeta is for lightweight portables, nothing more.