And the main reason that people run IP is that it is the only real standard for accessing network resources under Unix. >>>>>>>>>>>>>>>>> True. IP is a terrible protocol. It has a lot of overhead, plus it has a lot of connection startup time. Who decided to use TCP/IP for internet anyway?
And the main reason people use HTTP for web browsing is that it is the only real standard for transporting hypertext. >>>>>>>> HTTP is also pretty limited. Read the/. article on what they're trying to replace it with.
And the main reason people use SMTP, POP, and/or IMAP is -- you guessed it -- they are the standards for accessing email. >>>>>>>> Let's see... never mind, I couldn't care less what protocol my email system uses.
You say "standard" like it is a bad thing. Believe it or not, there are those of us who prefer interoperability and stability over flashy looks. >>>>>>>>>> X11 doesn't have more stability than Aqua. However, it is a lot faster and more powerful. Interoperability seems to be the only thing left, and 99% of Mac users couldn't care less about that.
Aqua may be cool and all, but will it run on any computer from an IBM mainframe to a Palm Pilot? No, I didn't think so. But X11 will. This doesn't make Aqua inferior or X11 better, but it is a distinction to be aware of. >>>>>>>>> I don't own an IBM mainframe or a Palm Pilot. Neither do most MacOS users. And even if they did, the PP doesn't have an X server.
You have to take this in contex. He is talking about X11 within the context of a desktop OS. In that context, X11 is worse than the Win9x GDI.
Re:Misguided anti-Aqua sentiment
on
MacOSX and X11
·
· Score: 2
Apple has made a break from X on UNIX, not on its own machines. IE> Instead of going to normal way and using X with UNIX, they used something else. Also, this isn't the closest they've gotten to X. MacOS already has a few X servers.
I'm curious what kind of hardware you're using. I tend to be on the bleeding edge (I'm stupid, right now I'm running Mandrake 7.1 with kernel 2.4, ReiserFS, the beta NVIDIA drivers, and KDE2 Beta 2. It's surprising the bloody thing boots.) Windows tends to handle modern hardware a LOT better. Also the quality of the HW company matters. I usually stick with Matrox, NVIDIA, Creative, Diamond, etc. If you're using anything from ATI, then good luck to you.
Consumer hardware has gotten REALLY fast.
on
ATI Radeon Released
·
· Score: 4
For those of you who haven't been on the scene lately, I'll clue you in. With the arrival of the GeForce, consumer 3D hardware has gotten REALLY fast. Right now, the GeForce 2 GTS is nearly the fastest 3D card availabe on PCs for workstation tasks. If you head over to Intergraphs's website, you'll see their comparisons between the Elsa's NVIDIA Quadro-based card and Intergraphs Wildcat 4210, which is currently the fastest workstation card availabe (more than twice as fast as the SGI Visual Workstation series in awedvs tests.) However, the Quadro-based card is nearly 50-70% the speed of the intergraph machine. Considering that the Quadro is only 135MHz compared to the GeForce2 GTS's 200MHz, plus the fact that the GeForce2 has twice as many pipes, it means that a GeForce2 is probably close to the performance of a Wildcat 4210. Thus, you can get nearly all of the $2000+Wildcat's performance in Hercules's $400 64MB GeForce2 card (which can be run at 235MHz core and 200+ MHz RAM). That sound you just heard was a collective orgasm from all the 3D Studio users who just realized that a $5000 PC can take the place of their $10,000 intergraph.
My point was that this was a game related article, and had nothing to do with general networking. From a gamer's point of view, Linux has POS networking because it won't work on his modem and/or his ISP.
Maybe, maybe not. Some people have the money to burn on a new vidcard every 6 months, so what's wrong with that? Plus they're a big help to people like me (only a softcore gamer) because they bring in the high-end hardware so I can run 3D apps without buying an SGI. Also, the sheer fact that they're willing to pony up the cash to buy so much stuff proves my point that they're very influentail on the PC market.
Its never that easy:) In all my years of using Linux, I have yet to be able to install a driver off an RPM. ALSA doesn't offer RPMS, the RedHat NVIDIA RPMS don't like me, etc. If it worked, I'd think, OK, Linux has a decent driver interface. Actually, treating everything as a package (from software to drivers) makes even more sense. But even though the capability is there, nobody ever uses it.
SYes it's flexible, but to tell the truth, why make people who have better things to do than compile their kernel live with a sub-standard kernel? RedHat is particularly bad for this kind of thing, the stock kernels are shit. In other systems (NT or BeOS) if you don't use something, like SCSI or ISDN or all the networking stuff, it's not loaded. However, in most stock linux kernels, everything from firewalling, to NAT, to SCSI and ISDN are loaded. Yes, most of the drivers are compiled as modules but the subsystem is still there. In NT, the entire subsystem is just not loaded if it isn't used.
NT has quite a bit better OpenGL performance than Win98 does. In my testing its at least 10 or 15% faster and much smoother. For example, Quake III's best platform is probably NT 4.0. Also, I don't see the NVIDIA drivers getting appreciably faster. Even though the drivers are beta, they are from FINAL code. NVIDIA uses the same exact driver code for all platforms, with a platform specific glue layer for each platform. So continued tweeking may make the driver more stable, but I doubt they're going to get appreciably faster. (After all, how much can you tweek a glue layer?) Win2K is a little weird on the performance front. It seems like it should be really fast, but there are a lot of problems with the OS right now. Before you condem it, wait for Service Pack 1, and see if the glitches are fixed.
Gamers are very influential consumers with a lot of clout in the consumer sector. They've got money, and are willing to spend it. What, you think Intel pumps out 1GHz CPUs for accountants? That they added SSE to make Excel run faster? I DO understand it, it seems to me that you haven't studied the demographics of the situation.
Drivers that come with the kernel are not what he's talking about. Drivers in the Windows world get updated often (not because they suck, NVIDIA for example is constantly improving their drivers, (adding features, tweeking speed, etc) even they they are higher quality than 90% of Linux drivers.) There has to be an easy standardized way to install updated drivers. People will not put up with the current driver mess, where often you have to compile the driver, and all the different types of drivers have a different (though usually CLI) installation method.
1) Most windows drivers are also loadable kernel modules. 2) People shouldn't have to run a shell script, plus, its never that easy is it? To upgrade ALSA I have to do all kinds of scary things (what the hell does lsmod do?) I don't even know where I can go to configure my USB devices. The problem is that Linux treats drivers on a case by case basis. Graphics drivers have special requirements, sound drivers have a different set of installation requirements, etc. People can get the hang of installing drivers in Windows, just like I can get the hang of compiling the kernel. However, I by the time people get the hang of all the different procedures for all the different driver types, they've given up and rebooted.
Nothing in Linux is out of box. I'm running kernel 2.4, and it still boinks out on me once in awhile. Also, you have to custom compile your kernel, which isn't an appealing option. Anti-aliasing isn't needed, but it makes those menu's look nicer doncha think?
These command line vs. GUI things make me sick. There is a use for both, and YES it is possible to use both transparently. Linux doesn't do it, because a lot of things, like configuring anything or installing RPMS (package managers are yucky! I should just be able to right click and install. Also, stuff like nodeps or replace files gets confusing with the current interfaces.) Windows doesn't do it, because the DOS CLI sucks. So far, of the three OSs that I've used, only BeOS does it. (Not a shameless plug!) Switching between GUI and CLI is wonderfully natural. In things that do work better from a GUI standpoint (configuring stuff) it has GUI tools. However, if I want to copy a bunch of files with a certain attribute or install software via my custom scripts I can just pop up a CLI (bash) and have at it.
In terms of real end user support. This is the real world, not the hacking-induced hallucination that was Linux up until recently. Until a year ago, (for gamers anyway) Linux didn't exist. Also, the movement to make Linux competent as a consumer and game OS didn't really start until recently either.
But he's not talking about general networking is he? This isn't theoretical. Right here, right now. If you're a gamer, Linux networking sucks compared to Windows. For games the better TCP/IP stack isn't really noticable, and it has much less support for modems and services.
True, Windows is and for the forseeable future will stay faster than Linux. Why? Mainly because of DirectX. DirectX is essentially a tool to allow developers to shove aside Windows. While running a DirectX application, Windows is effectivally shut out for a great deal of the time. Running in exclusive mode DirectDraw, for example, gives around 90-something% of the processor time to the application. It allows direct access to hardware, and essentially behaves like DOS with standardized hardware acceleration. This is the exact kind of behaviour Linux (and UNIX in general) prevents against. While multiple users, and the general abstraction and generality of UNIX's design may be great for some things, it is not great for games. Case in point: DRI. DRI is an attempt to to put direct hardware access into Linux. It does it in the UNIX-way (client-server, abstracted, general and portable) and it seems hurts performance in the process. Otherwise NVIDIA probably would have chosen THAT for its drivers instead of creating its own kernel driver (which I think probably behaves a lot more rudely to the OS than does DRI). Before you flame me, think of the reasons behind their desicion. Making their own system certainly wasn't to protect source code, they could have made binary DRI drivers. It certainly hurt them, because they have to spend much more effort to maintain drivers in sync with kernel versions and XFree86 updates, plus they lose any other *NIX users. The only reason they could have chosen to do their own is because DRI wasn't fast enough. And if it doesn't work for NVIDIA, that's a big warning sign. They have the fastest hardware available, and will for the near future, and if it doesn't serve their needs, then something is wrong with the system. Also, it seems to me, that DRI just won't cut it for the broad range of cards at the consumer level. They all depend on varying different models of rendering, and by not allowing apps direct access to the drivers, DRI inherently preaches a particular way of rendering. For a marginal OS, that is very dangerous, because the hardware vendors will NOT give a thought to the needs of DRI while they're designing hardware. SGI pulled this off, because THEY design the hardware AND software, but Linux doesn't have this luxury. I use this point (DRI) to try to explain that the "UNIX-way" just won't work with games. While its flexibility, stability, and tweekability make it a great OS for many tasks, it just doesn't work for games. In fact many of these traits (its flexibility and the abstraction that leads to stability and portability) actually work AGAINST it being a good gaming OS.
Not that kind of standards. He is saying that Linux is very customizable. You can have different skins for different things etc. However, he is also saying that Linux doesn't really adhere to any standards in terms of UI or directory structure. The two are not mutually exclusive. Ideally, you'd have the customizability of Linux, while retaining the standard approach. For example, picking you own directory structure is useless. Thus, have a standardized directory structure. Having different buttons like Apply, OK, Try are dumb because they mean the same thing. So, standardize the widgets, and allow people to move stuff around and skin them all they want. Make skins GLOBAL. Thus, a user can change the UI of their desktop and only have to learn the interface once, rather than relearning it every time for different apps. As for standards he's saying this. Okay, make one desktop environment, but make it very customizable. Standardize on one set of libraries and backend programs (stuff people don't are about like the libc version or wither the distro uses mawk or gawk.) If 1% of people end up hating it, well, then that will have to do. You can't please everyone all the time. However, if you try to, you get the current mess Linux is in with 2 major competing widgets with two major (incompatible) versions each and three different sets of two major libraries (libc and libstdc++) etc.
By the 99% thing I was pointing out that 99% of people will use this on a PC. What are they trying to target? Are they making a back-end renderer (Gecko) a web browser for PCs, or a standardized interface for web pads? As for you opinion of the speed, that is very subjective. For example, I feel that even KDE2 runs pretty slow on my 300MHz 128MB computer. Of course, that's after using BeOS for years. Other feel that KDE runs perfectly even on a 100MHz machine.
I was responding to the person above who said it barely runs on a 700MHz Athlon. Also, running great is totally subjective. Some people think that GNOME and KDE (where you can actually SEE redraw) run great. However, after using BeOS, I tend to think that anything that doesn't offer instant response runs lousy. Even some BeOS apps (early version of TaskManager for example) don't get my apporval because they tend to slow the resizing process. In the end, what matters is that does the load make sense for the type of app. Web browsers are relativly simple creatures. I can understand my 3D render using up 128MB of RAM, but my web browser or windowing system doing the same is just ridiculous. (No, I'm not saying Mozilla uses 128MB, it uses more than it should.) Why put up with bloatware anyway? Why run something "acceptably." That's the mantra behind MS software. (It runs okay on the midrange 500MHz CPUs so lets ship it.) Plus, if Mozilla has any hope in the embeeded market, it has to run great on a Pentium classic class CPU. Mozilla definately does not.
The widget set is the foundation for the UI. The UI is rendered through XUL files. Thus the widgets are intimately tied with the XUL layout engine. Also, the problem is that people DO care about the platform. Most people spend nearly all their time in one platform. Thus it is more important to make all apps on that platform similar than it is to make that one app similar on all platforms. What happens more often? A person bouncing from *nix to Mac to Windows, or someone using the app only in one OS? The problem with using the same widget set is that it is not the NATIVE widget set. Thus it is out of place and presents a learning curve to those using it. There is a serious case for making all apps look the same. One of the prime reasons that GNOME and KDE are so popular is because A) Most apps look the same, and B) They look like Windows. Once a person learns one UI mindset (no matter how confusing the UI, it will happen) then it is very easy to learn a new application with uses that mindset. By breaking the mindset, Mozilla is hurting its users.
Oh come on. Sure internet devices may be the wave of the future, but there is no point in cluttering up the interface for the time being where there ARE no internet devices. This is the 99% thing again. 99% of people will use this thing on their desktop. Why totally ruin the experience (or severely hamper it) for those 99% so you can access that 1%? Also, the Mozilla UI barely runs on a 700MHz Athlon, so you really think it will run on Internet stations with 5X less power? I seriously doubt they can tweek it THAT much!
Yes the choice was easy. Given the fact that they'd have to spend the same amount of time debugging multiple widget sets, they should have done that. A good design would minimize the amount of widget specific code, while using native widgets. In the end, they chose to create a custom widget because it made them feel cool. Thus it was better from THEIR point of view. However, it detracts from the USER'S experience. Thus the choice should have been to use and debug native widgets.
Some of the Truespace UI makes it genuinely more useful. I find it more convinient to have menubar along with the tools, but you CAN switch it. The interface has a lot of touches, especially in the way of allowing you to manipulate the objects very freely with the mouse instead of having to click rotate, rotate the object, then move, and move the object, etc.
Is it just me, or almost all the high-profile OSS projects feature-laden, MS-type, bloatware? The only major project that isn't overly bloated seems to be the kernel itself. Even X doesn't suffer from feature-craze, it is bloated for other reasons. I continually beat my head against a wall thinking about how my RAM is disappearing down the toilet. Not only does my average desktop load dozens of libraries when one or two would suffice (compat libc and libc++ for Netscape, KDE 1 and 2 libraries, Qt 1 and 2, GTK, GNOME, etc) but they keep adding more stuff to it. FOr example, Mozilla has a great rendering engine (Gecko) with a cruddy piece of bloatware wrapped around it (everything else in Mozilla.) Who cares about total customizability through XUL, who cares that everything is tied to a java script. Aren't scripts slow anyway? People bash MS for making active desktop, but this is even worse. Do you really want your programs interpreting XML to do your user interface? This is carried to GNOME (and KDE to a slightly lesser extent) too. One one hand, MS is using the selvte COM, realizing that if people really need network-transparency they can use DCOM or use a custom CORBA extension. However, GNOME is using CORBA for local components! Are they crazy? MS is famous for introducing bloatware. They put in features 99% of people never use, which impacts perfomance very heavily. However, they have to keep selling more versions. OSS stuff doesn't have to do that. They have the freedom to make software that is nice and light with just enough features to please 99% of the people. Why don't they?
Slashdot Mirror
And the main reason that people run IP is that it is the only real standard for accessing network resources under Unix.
/. article on what they're trying to replace it with.
>>>>>>>>>>>>>>>>>
True. IP is a terrible protocol. It has a lot of overhead, plus it has a lot of connection startup time. Who decided to use TCP/IP for internet anyway?
And the main reason people use HTTP for web browsing is that it is the only real standard for transporting hypertext.
>>>>>>>>
HTTP is also pretty limited. Read the
And the main reason people use SMTP, POP, and/or IMAP is -- you guessed it -- they are the standards for accessing email.
>>>>>>>>
Let's see... never mind, I couldn't care less what protocol my email system uses.
You say "standard" like it is a bad thing. Believe it or not, there are those of us who prefer interoperability and stability over flashy looks.
>>>>>>>>>>
X11 doesn't have more stability than Aqua. However, it is a lot faster and more powerful. Interoperability seems to be the only thing left, and 99% of Mac users couldn't care less about that.
Aqua may be cool and all, but will it run on any computer from an IBM mainframe to a Palm Pilot? No, I didn't think so. But X11 will. This doesn't make
Aqua inferior or X11 better, but it is a distinction to be aware of.
>>>>>>>>>
I don't own an IBM mainframe or a Palm Pilot. Neither do most MacOS users. And even if they did, the PP doesn't have an X server.
You have to take this in contex. He is talking about X11 within the context of a desktop OS. In that context, X11 is worse than the Win9x GDI.
Apple has made a break from X on UNIX, not on its own machines. IE> Instead of going to normal way and using X with UNIX, they used something else. Also, this isn't the closest they've gotten to X. MacOS already has a few X servers.
I'm curious what kind of hardware you're using. I tend to be on the bleeding edge (I'm stupid, right now I'm running Mandrake 7.1 with kernel 2.4, ReiserFS, the beta NVIDIA drivers, and KDE2 Beta 2. It's surprising the bloody thing boots.) Windows tends to handle modern hardware a LOT better. Also the quality of the HW company matters. I usually stick with Matrox, NVIDIA, Creative, Diamond, etc. If you're using anything from ATI, then good luck to you.
For those of you who haven't been on the scene lately, I'll clue you in. With the arrival of the GeForce, consumer 3D hardware has gotten REALLY fast. Right now, the GeForce 2 GTS is nearly the fastest 3D card availabe on PCs for workstation tasks. If you head over to Intergraphs's website, you'll see their comparisons between the Elsa's NVIDIA Quadro-based card and Intergraphs Wildcat 4210, which is currently the fastest workstation card availabe (more than twice as fast as the SGI Visual Workstation series in awedvs tests.) However, the Quadro-based card is nearly 50-70% the speed of the intergraph machine. Considering that the Quadro is only 135MHz compared to the GeForce2 GTS's 200MHz, plus the fact that the GeForce2 has twice as many pipes, it means that a GeForce2 is probably close to the performance of a Wildcat 4210. Thus, you can get nearly all of the $2000+Wildcat's performance in Hercules's $400 64MB GeForce2 card (which can be run at 235MHz core and 200+ MHz RAM). That sound you just heard was a collective orgasm from all the 3D Studio users who just realized that a $5000 PC can take the place of their $10,000 intergraph.
My point was that this was a game related article, and had nothing to do with general networking. From a gamer's point of view, Linux has POS networking because it won't work on his modem and/or his ISP.
Maybe, maybe not. Some people have the money to burn on a new vidcard every 6 months, so what's wrong with that? Plus they're a big help to people like me (only a softcore gamer) because they bring in the high-end hardware so I can run 3D apps without buying an SGI. Also, the sheer fact that they're willing to pony up the cash to buy so much stuff proves my point that they're very influentail on the PC market.
Its never that easy:) In all my years of using Linux, I have yet to be able to install a driver off an RPM. ALSA doesn't offer RPMS, the RedHat NVIDIA RPMS don't like me, etc. If it worked, I'd think, OK, Linux has a decent driver interface. Actually, treating everything as a package (from software to drivers) makes even more sense. But even though the capability is there, nobody ever uses it.
SYes it's flexible, but to tell the truth, why make people who have better things to do than compile their kernel live with a sub-standard kernel? RedHat is particularly bad for this kind of thing, the stock kernels are shit. In other systems (NT or BeOS) if you don't use something, like SCSI or ISDN or all the networking stuff, it's not loaded. However, in most stock linux kernels, everything from firewalling, to NAT, to SCSI and ISDN are loaded. Yes, most of the drivers are compiled as modules but the subsystem is still there. In NT, the entire subsystem is just not loaded if it isn't used.
NT has quite a bit better OpenGL performance than Win98 does. In my testing its at least 10 or 15% faster and much smoother. For example, Quake III's best platform is probably NT 4.0. Also, I don't see the NVIDIA drivers getting appreciably faster. Even though the drivers are beta, they are from FINAL code. NVIDIA uses the same exact driver code for all platforms, with a platform specific glue layer for each platform. So continued tweeking may make the driver more stable, but I doubt they're going to get appreciably faster. (After all, how much can you tweek a glue layer?) Win2K is a little weird on the performance front. It seems like it should be really fast, but there are a lot of problems with the OS right now. Before you condem it, wait for Service Pack 1, and see if the glitches are fixed.
Gamers are very influential consumers with a lot of clout in the consumer sector. They've got money, and are willing to spend it. What, you think Intel pumps out 1GHz CPUs for accountants? That they added SSE to make Excel run faster? I DO understand it, it seems to me that you haven't studied the demographics of the situation.
Drivers that come with the kernel are not what he's talking about. Drivers in the Windows world get updated often (not because they suck, NVIDIA for example is constantly improving their drivers, (adding features, tweeking speed, etc) even they they are higher quality than 90% of Linux drivers.) There has to be an easy standardized way to install updated drivers. People will not put up with the current driver mess, where often you have to compile the driver, and all the different types of drivers have a different (though usually CLI) installation method.
1) Most windows drivers are also loadable kernel modules.
2) People shouldn't have to run a shell script, plus, its never that easy is it? To upgrade ALSA I have to do all kinds of scary things (what the hell does lsmod do?) I don't even know where I can go to configure my USB devices. The problem is that Linux treats drivers on a case by case basis. Graphics drivers have special requirements, sound drivers have a different set of installation requirements, etc. People can get the hang of installing drivers in Windows, just like I can get the hang of compiling the kernel. However, I by the time people get the hang of all the different procedures for all the different driver types, they've given up and rebooted.
Nothing in Linux is out of box. I'm running kernel 2.4, and it still boinks out on me once in awhile. Also, you have to custom compile your kernel, which isn't an appealing option. Anti-aliasing isn't needed, but it makes those menu's look nicer doncha think?
These command line vs. GUI things make me sick. There is a use for both, and YES it is possible to use both transparently. Linux doesn't do it, because a lot of things, like configuring anything or installing RPMS (package managers are yucky! I should just be able to right click and install. Also, stuff like nodeps or replace files gets confusing with the current interfaces.) Windows doesn't do it, because the DOS CLI sucks. So far, of the three OSs that I've used, only BeOS does it. (Not a shameless plug!) Switching between GUI and CLI is wonderfully natural. In things that do work better from a GUI standpoint (configuring stuff) it has GUI tools. However, if I want to copy a bunch of files with a certain attribute or install software via my custom scripts I can just pop up a CLI (bash) and have at it.
In terms of real end user support. This is the real world, not the hacking-induced hallucination that was Linux up until recently. Until a year ago, (for gamers anyway) Linux didn't exist. Also, the movement to make Linux competent as a consumer and game OS didn't really start until recently either.
But he's not talking about general networking is he? This isn't theoretical. Right here, right now. If you're a gamer, Linux networking sucks compared to Windows. For games the better TCP/IP stack isn't really noticable, and it has much less support for modems and services.
True, Windows is and for the forseeable future will stay faster than Linux. Why? Mainly because of DirectX. DirectX is essentially a tool to allow developers to shove aside Windows. While running a DirectX application, Windows is effectivally shut out for a great deal of the time. Running in exclusive mode DirectDraw, for example, gives around 90-something% of the processor time to the application. It allows direct access to hardware, and essentially behaves like DOS with standardized hardware acceleration. This is the exact kind of behaviour Linux (and UNIX in general) prevents against. While multiple users, and the general abstraction and generality of UNIX's design may be great for some things, it is not great for games. Case in point: DRI. DRI is an attempt to to put direct hardware access into Linux. It does it in the UNIX-way (client-server, abstracted, general and portable) and it seems hurts performance in the process. Otherwise NVIDIA probably would have chosen THAT for its drivers instead of creating its own kernel driver (which I think probably behaves a lot more rudely to the OS than does DRI). Before you flame me, think of the reasons behind their desicion. Making their own system certainly wasn't to protect source code, they could have made binary DRI drivers. It certainly hurt them, because they have to spend much more effort to maintain drivers in sync with kernel versions and XFree86 updates, plus they lose any other *NIX users. The only reason they could have chosen to do their own is because DRI wasn't fast enough. And if it doesn't work for NVIDIA, that's a big warning sign. They have the fastest hardware available, and will for the near future, and if it doesn't serve their needs, then something is wrong with the system. Also, it seems to me, that DRI just won't cut it for the broad range of cards at the consumer level. They all depend on varying different models of rendering, and by not allowing apps direct access to the drivers, DRI inherently preaches a particular way of rendering. For a marginal OS, that is very dangerous, because the hardware vendors will NOT give a thought to the needs of DRI while they're designing hardware. SGI pulled this off, because THEY design the hardware AND software, but Linux doesn't have this luxury. I use this point (DRI) to try to explain that the "UNIX-way" just won't work with games. While its flexibility, stability, and tweekability make it a great OS for many tasks, it just doesn't work for games. In fact many of these traits (its flexibility and the abstraction that leads to stability and portability) actually work AGAINST it being a good gaming OS.
Not that kind of standards. He is saying that Linux is very customizable. You can have different skins for different things etc. However, he is also saying that Linux doesn't really adhere to any standards in terms of UI or directory structure. The two are not mutually exclusive. Ideally, you'd have the customizability of Linux, while retaining the standard approach. For example, picking you own directory structure is useless. Thus, have a standardized directory structure. Having different buttons like Apply, OK, Try are dumb because they mean the same thing. So, standardize the widgets, and allow people to move stuff around and skin them all they want. Make skins GLOBAL. Thus, a user can change the UI of their desktop and only have to learn the interface once, rather than relearning it every time for different apps. As for standards he's saying this. Okay, make one desktop environment, but make it very customizable. Standardize on one set of libraries and backend programs (stuff people don't are about like the libc version or wither the distro uses mawk or gawk.) If 1% of people end up hating it, well, then that will have to do. You can't please everyone all the time. However, if you try to, you get the current mess Linux is in with 2 major competing widgets with two major (incompatible) versions each and three different sets of two major libraries (libc and libstdc++) etc.
By the 99% thing I was pointing out that 99% of people will use this on a PC. What are they trying to target? Are they making a back-end renderer (Gecko) a web browser for PCs, or a standardized interface for web pads? As for you opinion of the speed, that is very subjective. For example, I feel that even KDE2 runs pretty slow on my 300MHz 128MB computer. Of course, that's after using BeOS for years. Other feel that KDE runs perfectly even on a 100MHz machine.
I was responding to the person above who said it barely runs on a 700MHz Athlon. Also, running great is totally subjective. Some people think that GNOME and KDE (where you can actually SEE redraw) run great. However, after using BeOS, I tend to think that anything that doesn't offer instant response runs lousy. Even some BeOS apps (early version of TaskManager for example) don't get my apporval because they tend to slow the resizing process. In the end, what matters is that does the load make sense for the type of app. Web browsers are relativly simple creatures. I can understand my 3D render using up 128MB of RAM, but my web browser or windowing system doing the same is just ridiculous. (No, I'm not saying Mozilla uses 128MB, it uses more than it should.) Why put up with bloatware anyway? Why run something "acceptably." That's the mantra behind MS software. (It runs okay on the midrange 500MHz CPUs so lets ship it.) Plus, if Mozilla has any hope in the embeeded market, it has to run great on a Pentium classic class CPU. Mozilla definately does not.
The widget set is the foundation for the UI. The UI is rendered through XUL files. Thus the widgets are intimately tied with the XUL layout engine. Also, the problem is that people DO care about the platform. Most people spend nearly all their time in one platform. Thus it is more important to make all apps on that platform similar than it is to make that one app similar on all platforms. What happens more often? A person bouncing from *nix to Mac to Windows, or someone using the app only in one OS? The problem with using the same widget set is that it is not the NATIVE widget set. Thus it is out of place and presents a learning curve to those using it. There is a serious case for making all apps look the same. One of the prime reasons that GNOME and KDE are so popular is because A) Most apps look the same, and B) They look like Windows. Once a person learns one UI mindset (no matter how confusing the UI, it will happen) then it is very easy to learn a new application with uses that mindset. By breaking the mindset, Mozilla is hurting its users.
Oh come on. Sure internet devices may be the wave of the future, but there is no point in cluttering up the interface for the time being where there ARE no internet devices. This is the 99% thing again. 99% of people will use this thing on their desktop. Why totally ruin the experience (or severely hamper it) for those 99% so you can access that 1%? Also, the Mozilla UI barely runs on a 700MHz Athlon, so you really think it will run on Internet stations with 5X less power? I seriously doubt they can tweek it THAT much!
Yes the choice was easy. Given the fact that they'd have to spend the same amount of time debugging multiple widget sets, they should have done that. A good design would minimize the amount of widget specific code, while using native widgets. In the end, they chose to create a custom widget because it made them feel cool. Thus it was better from THEIR point of view. However, it detracts from the USER'S experience. Thus the choice should have been to use and debug native widgets.
Some of the Truespace UI makes it genuinely more useful. I find it more convinient to have menubar along with the tools, but you CAN switch it. The interface has a lot of touches, especially in the way of allowing you to manipulate the objects very freely with the mouse instead of having to click rotate, rotate the object, then move, and move the object, etc.
Is it just me, or almost all the high-profile OSS projects feature-laden, MS-type, bloatware? The only major project that isn't overly bloated seems to be the kernel itself. Even X doesn't suffer from feature-craze, it is bloated for other reasons. I continually beat my head against a wall thinking about how my RAM is disappearing down the toilet. Not only does my average desktop load dozens of libraries when one or two would suffice (compat libc and libc++ for Netscape, KDE 1 and 2 libraries, Qt 1 and 2, GTK, GNOME, etc) but they keep adding more stuff to it. FOr example, Mozilla has a great rendering engine (Gecko) with a cruddy piece of bloatware wrapped around it (everything else in Mozilla.) Who cares about total customizability through XUL, who cares that everything is tied to a java script. Aren't scripts slow anyway? People bash MS for making active desktop, but this is even worse. Do you really want your programs interpreting XML to do your user interface? This is carried to GNOME (and KDE to a slightly lesser extent) too. One one hand, MS is using the selvte COM, realizing that if people really need network-transparency they can use DCOM or use a custom CORBA extension. However, GNOME is using CORBA for local components! Are they crazy? MS is famous for introducing bloatware. They put in features 99% of people never use, which impacts perfomance very heavily. However, they have to keep selling more versions. OSS stuff doesn't have to do that. They have the freedom to make software that is nice and light with just enough features to please 99% of the people. Why don't they?