That's my point. You said 2.96 had "performance issues" and I pointed out that it would most likely make Apache *faster*. Thus, had an official compiler been used, like you suggested, performance would be worse.
It's easy to see, but its a pain to manage all the symlinks. Config files like this are probably where GUIs make the most sense, since they are both easy to manage *and* can give all the info at a glance of the screen. Still, with grep and all, BSD init is nearly as easy to see and definately more easy to manage.
I'm not done yet. Why aren't you suggesting getting the latest CVS build of FreeBSD? Probably because you're not up-to-date on any fixes in that tree! Second, GCC 2.96 does have performance issues. It makes things faster (since its based on the 3.0 code)
Good god, this happens every time someone does a benchmark. If he had used an AC kernel you would have complained about that. If he had used RedHat, you would have suggested Debian. There is always another web-server/patch level/config tweek that can be done, isn't there? I think the only thing this test proves is that, out of box, Linux blows goats! Either use the latest software and live with the instability, use older software and live with the slowness, or spend 3 years tweeking to make everything work both stable and fast!
I'd be happy to give Linux a try, but I'm only going to do it when it's distributed by Microsoft. Yes folks, I'm afraid I'm another one of those ActiveUpdate bigots. I've no particular desire to go clambering up the Linux learning curve for its own sake, any more than I'm interested in doing the same with Windows NT. MIcrosoft Linux sounds like a fine idea to me, though.
Actually, one could argue, that *because* Windows does it (hence most people are used to it) it *is* the right way. Using a side-stick controller would probably be more efficient for cars, but everyone knows how to use a steering wheel, so why change?
Sorry I got carried away. While Linux does have some innovative projects (like EVAS), I would still argue that much of Linux is about freedom (OSS) and incremental quality increases, rather revolutionary ideas.
They do compete somewhat on API, but at some point you just have to standardize the thing. Take OpenGL for example. One reason it has been so damn popular is because it is a standard API which vendors can write their own implementations for. Some people (me) don't like their lack of object orientation. Still, I appreciate the consistancy and universality of the API. You can achive this standardization two ways. First, you can decree that there will be one API with one implementation. This is essentially what Win32 and Direct3D are. Or, you can say there will be one API, that anybody can write an implementation for. (POSIX and OpenGL). The third way is just as bad as the first, where you setup a free-for-all of competing APIs. The second way is the One True Way (TM).
PS> I much prefer Win32 and D3D to POSIX and OpenGL, but guess which one I program in?
Because the practice of using different toolkits on the same machine should be stamped out with extreme prejudice! It boggles the mind that Linux-lusers, all caught up in their self-rightous jihad to have as many different toolkits as possible, under the excuse of fostering competition, never realize what the best way to foster competition is. Open STANDARDS. There needs to be ONE (uno, un, ack, 1 25^0) standard API. Toolkits should serve as a back-end to that API. Then toolkits can compete based on quality/speed/stability alone, not the artificial bounds of software base! It takes power away from the evil software developers and puts it back into the hands of the USER! THAT is real freedom!
Wait... I'm not done yet. Are you implying that BeOS is somehow *not* innovative? That somehow the speed, the technological tricks, the pervasive use of messaging and its attendant benifets (apps work seamlessly together, common systems such as contact databases are shared through standard APIs, alls apps are automagically scriptable, drag and drop actually WORKS, etc), the ease of use, etc, is less innovative than anything GNOME or KDE have managed to spew out? Aside from Rasterman's EVAS and some other fringe (relativly) projects, I have yet to see anything as innovative come out from any Linux projects. Refined? Sure. Tweeked? Why not. Evolutionary? Maybe. Innovative? Hell no.
Will you please come down from your ivory tower and talk some specifics? The Windows UI is absolutely kick-ass from the clicky point of view. People are used to (in real life) of pointing (or touching) things and being able to do different things to them. The left-click signifies grabing onto the object, and the double-click signifies tapping the object (kinda like hitting a button). The right click is probably the best part of the Windows UI (and something that ALL other desktop environments don't do enough of) because it gives a context menu. Humans like context. Our language is more simple because of it, communication is faster, etc. The context menu essentially allows the user to choose from a list of actions larger than the number of buttons on the mouse. Your left-click/right-click thing is not only unnatural, but innefficient (you have to FIND the other executable). However, right clicking and opening an "open-with" context menu is the most efficient thing that anyone's thought of.
I have used bootman (the BeOS bootloader) to boot several versions of Linux from both ext2 and ReiserFS filesystems, I've used it to boot QNX RtP from QNX-FS, I've used to to boot NT4 from NTFS, Win98 from FAT32 and FAT16, Win2K from NTFS and Fat32, FreeBSD on FFS, hell, it'll even boot Plan9! And yes, bootman does support adding and removing entries, all you have to do is type "bootman" at the CLI, it detects all your partitions, asks from names for each, and then lets you chose the default one. Much simpler, I must say, than Lilo OR GRUB.
Do you not know how to read? Damn, we need to fix these inner city DC schools! Their students can't even recognize the difference between the character patterns "BeOS" and "NT4" I mean, even if they had some basic counting skills they could see that one had four letters and the other three.
PS> Of course, this message is wasted anyway. Its not like you can read it...
Instead you should learn to think for yourself, and hack for yourself!
>>>>>>>>>>>>>
Well, that's a stupid idea. Because the minute the person learns how to do that, they come up with yet another stupid GUI toolkit for *NIX.
Umm, it would be stupid to implement each toolkit into a graphics card because the toolkit does a lot more than drawing. GNOME, for example, as an entire application API. This means that the graphics card would go from being an ASIC (application specific integrated circut) to a general purpose CPU, since most of the toolkit code is just like regular application code. Also, the end result would be slower since most graphics cards run at a measly 200MHz and have only limited on-board resources, and aren't designed to run regular code, but do a very specific drawing task. Then, the fact that the graphics card is really a data-processor, rather than a command executer, would hurt performance. Lastly, apps constantly call toolkit code, and if that were on the graphics card, you'd have to access all app data over the AGP bus, which is a lot slower than the CPU accessing it over the system bus. What would make more sense, and this is what I think you were getting at, would be a graphics card that implemented something like Quartz into hardware. That way you could get all the nifty eye-candy, without the CPU cost.
PS> You can get he benifets of having another processor doing toolkit code by getting an SMP machine;)
There are several reasons why crashing X is as bad as crashing the OS. Not everyone uses Linux for servers, and not many servers run X (all the time.) Thus if X crashes, the person most likely loses the work in the X program that he/she was working on. On a server, its good that it drops back into the command line, because most services are run from there. On a desktop however, most stuff is run from the GUI (even CLI apps are usually run from Xterms) and when X goes down, they all dissapear.
Monkey. That source was released only recently and you know it. There is still a difference between GNOME and CDE, despite the corperations. If anything screws up GNOME, it will be the fact that the developers have no perspective on what priorities are important in a desktop environment, and the temptation for the developers to do stuff just because it is cool.
Another problem. Why should the bootloader even be an issue? I have no clue what version of bootman I'm using, and I won't ever need to. It just works, it configs easily, and it never breaks. If you have to get tech support for your *bootloader* than your OS has some serious usability problems.
Slashdot Mirror
That's my point. You said 2.96 had "performance issues" and I pointed out that it would most likely make Apache *faster*. Thus, had an official compiler been used, like you suggested, performance would be worse.
It's easy to see, but its a pain to manage all the symlinks. Config files like this are probably where GUIs make the most sense, since they are both easy to manage *and* can give all the info at a glance of the screen. Still, with grep and all, BSD init is nearly as easy to see and definately more easy to manage.
You, my friend, are insance. How could anyone *like* those god-awful SysV startup scripts? BSD-initscripts all the way!
I'm not done yet. Why aren't you suggesting getting the latest CVS build of FreeBSD? Probably because you're not up-to-date on any fixes in that tree! Second, GCC 2.96 does have performance issues. It makes things faster (since its based on the 3.0 code)
Good god, this happens every time someone does a benchmark. If he had used an AC kernel you would have complained about that. If he had used RedHat, you would have suggested Debian. There is always another web-server/patch level/config tweek that can be done, isn't there? I think the only thing this test proves is that, out of box, Linux blows goats! Either use the latest software and live with the instability, use older software and live with the slowness, or spend 3 years tweeking to make everything work both stable and fast!
For "fastest, most cleanly written OS" I'd have to give the award to BeOS or QNX. Certianly not any *NIX.
I'd be happy to give Linux a try, but I'm only going to do it when it's distributed by Microsoft. Yes folks, I'm afraid I'm another one of those ActiveUpdate bigots. I've no particular desire to go clambering up the Linux learning curve for its own sake, any more than I'm interested in doing the same with Windows NT. MIcrosoft Linux sounds like a fine idea to me, though.
If you do everything just because everyone else does, you might as well run Windows ;)
Actually, one could argue, that *because* Windows does it (hence most people are used to it) it *is* the right way. Using a side-stick controller would probably be more efficient for cars, but everyone knows how to use a steering wheel, so why change?
Sorry I got carried away. While Linux does have some innovative projects (like EVAS), I would still argue that much of Linux is about freedom (OSS) and incremental quality increases, rather revolutionary ideas.
No, he's right. Virtual memory is a combination of both physical and swap memory.
They do compete somewhat on API, but at some point you just have to standardize the thing. Take OpenGL for example. One reason it has been so damn popular is because it is a standard API which vendors can write their own implementations for. Some people (me) don't like their lack of object orientation. Still, I appreciate the consistancy and universality of the API. You can achive this standardization two ways. First, you can decree that there will be one API with one implementation. This is essentially what Win32 and Direct3D are. Or, you can say there will be one API, that anybody can write an implementation for. (POSIX and OpenGL). The third way is just as bad as the first, where you setup a free-for-all of competing APIs. The second way is the One True Way (TM).
PS> I much prefer Win32 and D3D to POSIX and OpenGL, but guess which one I program in?
Because the practice of using different toolkits on the same machine should be stamped out with extreme prejudice! It boggles the mind that Linux-lusers, all caught up in their self-rightous jihad to have as many different toolkits as possible, under the excuse of fostering competition, never realize what the best way to foster competition is. Open STANDARDS. There needs to be ONE (uno, un, ack, 1 25^0) standard API. Toolkits should serve as a back-end to that API. Then toolkits can compete based on quality/speed/stability alone, not the artificial bounds of software base! It takes power away from the evil software developers and puts it back into the hands of the USER! THAT is real freedom!
Wait... I'm not done yet. Are you implying that BeOS is somehow *not* innovative? That somehow the speed, the technological tricks, the pervasive use of messaging and its attendant benifets (apps work seamlessly together, common systems such as contact databases are shared through standard APIs, alls apps are automagically scriptable, drag and drop actually WORKS, etc), the ease of use, etc, is less innovative than anything GNOME or KDE have managed to spew out? Aside from Rasterman's EVAS and some other fringe (relativly) projects, I have yet to see anything as innovative come out from any Linux projects. Refined? Sure. Tweeked? Why not. Evolutionary? Maybe. Innovative? Hell no.
Exactly *what* is that supposed to mean? BeOS is nothing like UNIX. About as much as NT is like UNIX.
Will you please come down from your ivory tower and talk some specifics? The Windows UI is absolutely kick-ass from the clicky point of view. People are used to (in real life) of pointing (or touching) things and being able to do different things to them. The left-click signifies grabing onto the object, and the double-click signifies tapping the object (kinda like hitting a button). The right click is probably the best part of the Windows UI (and something that ALL other desktop environments don't do enough of) because it gives a context menu. Humans like context. Our language is more simple because of it, communication is faster, etc. The context menu essentially allows the user to choose from a list of actions larger than the number of buttons on the mouse. Your left-click/right-click thing is not only unnatural, but innefficient (you have to FIND the other executable). However, right clicking and opening an "open-with" context menu is the most efficient thing that anyone's thought of.
I have used bootman (the BeOS bootloader) to boot several versions of Linux from both ext2 and ReiserFS filesystems, I've used it to boot QNX RtP from QNX-FS, I've used to to boot NT4 from NTFS, Win98 from FAT32 and FAT16, Win2K from NTFS and Fat32, FreeBSD on FFS, hell, it'll even boot Plan9! And yes, bootman does support adding and removing entries, all you have to do is type "bootman" at the CLI, it detects all your partitions, asks from names for each, and then lets you chose the default one. Much simpler, I must say, than Lilo OR GRUB.
Do you not know how to read? Damn, we need to fix these inner city DC schools! Their students can't even recognize the difference between the character patterns "BeOS" and "NT4" I mean, even if they had some basic counting skills they could see that one had four letters and the other three.
PS> Of course, this message is wasted anyway. Its not like you can read it...
Instead you should learn to think for yourself, and hack for yourself!
>>>>>>>>>>>>>
Well, that's a stupid idea. Because the minute the person learns how to do that, they come up with yet another stupid GUI toolkit for *NIX.
Umm, it would be stupid to implement each toolkit into a graphics card because the toolkit does a lot more than drawing. GNOME, for example, as an entire application API. This means that the graphics card would go from being an ASIC (application specific integrated circut) to a general purpose CPU, since most of the toolkit code is just like regular application code. Also, the end result would be slower since most graphics cards run at a measly 200MHz and have only limited on-board resources, and aren't designed to run regular code, but do a very specific drawing task. Then, the fact that the graphics card is really a data-processor, rather than a command executer, would hurt performance. Lastly, apps constantly call toolkit code, and if that were on the graphics card, you'd have to access all app data over the AGP bus, which is a lot slower than the CPU accessing it over the system bus. What would make more sense, and this is what I think you were getting at, would be a graphics card that implemented something like Quartz into hardware. That way you could get all the nifty eye-candy, without the CPU cost.
;)
PS> You can get he benifets of having another processor doing toolkit code by getting an SMP machine
There are several reasons why crashing X is as bad as crashing the OS. Not everyone uses Linux for servers, and not many servers run X (all the time.) Thus if X crashes, the person most likely loses the work in the X program that he/she was working on. On a server, its good that it drops back into the command line, because most services are run from there. On a desktop however, most stuff is run from the GUI (even CLI apps are usually run from Xterms) and when X goes down, they all dissapear.
Monkey. That source was released only recently and you know it. There is still a difference between GNOME and CDE, despite the corperations. If anything screws up GNOME, it will be the fact that the developers have no perspective on what priorities are important in a desktop environment, and the temptation for the developers to do stuff just because it is cool.
Another problem. Why should the bootloader even be an issue? I have no clue what version of bootman I'm using, and I won't ever need to. It just works, it configs easily, and it never breaks. If you have to get tech support for your *bootloader* than your OS has some serious usability problems.
Exactly why I have just replaced NT4 for all my 3D modeling, 2D graphics needs. You really need to get something called "perspective."
Then it would be Ximian=&Helix.
;)
Yea, just showing of my god-like coding skills