I think it's pretty obvious that in one case EVERYTHING running on you system stops where as in the other only your current desktop closes
Which in the case of graphical apps (what 99% of the users do) means you've to stop all your apps - MUA, xchat client, AIM client, desktop environment. You need to stop all the apps and start them again. That's pretty much the same effect than rebooting your system (and this was my point)
Sorry, but would that be really that useful? I mean, how many times per day are you planning to update your drivers?
How many times per day do you update your libc, and does it requires rebooting? Why not? What about servers?
Besides, "server migration" is not useful just for this. It's needed for wireless connections - if you're using a remote X app and you move, suddenly your app breaks, it disconnected from your X server. The Right Thing to do would be to restore the communication when the connection goes back - this does not happen today. Although perhaps it's not called "server migration" but "supporting app disconnection"...dunno, I'm not expert.
Actually, upgrading drivers without rebooting IS difficult. We don't have this in the OSS field - in order to update a driver, be it 2D or 3D or whatever, you need at least to reboot X. That means switching off all your apps, and what current desktops that's pretty much like "rebooting" your computer. Yes, you're not rebooting, but with graphic apps in practice you're pretty much doing it.
What we need is to modify xlib to support "server migration" - we could move all the windows from a xserver to a kind of/dev/null-like fake server, then update x.org drivers, rmmod the old drivers, insmod the new ones, launch xorg, and move all the windows to your new xserver, switch off the fake xserver. Or something like that. (Suggestions?)
This was suggested. The kernel.org people didn't seem to have interest in it. Those light http servers are probably good for lots of small static html files. kernel.org is not like that - it needs to serve + 20 MB files and CD ISOs. Your benchmarks don't measure that. I can bet the kernel.org people knows what they're using and they'd have switched if it'd be really useful.
Can you even get the server to TELL you what the load is when it's that high??
Of course you can. "Normal people" often get high load values when they run out of mem and the box starts swapping and you can't control the box. Try running some thousands of "cp/dev/zero/dev/null" or something like that to get high load value, you still will be able to control the box - specially with a reniced root shell.
"Before we get a bunch of complaints about the fact that most binaries generated by GCC 4.0 are only marginally faster (and some a bit slower) than those compiled with 3.4, let me point out a few things that I've gathered from casually browsing the GCC development lists. I'm neither a GCC contributor nor a compiler expert.
Prior to GCC 4.0, the implementation of optimizations was mostly language-specific; there was little or no integration of optimization techniques across all languages. The main goal of the 4.0 release is to roll out a new, unified optimization framework (Tree-SSA), and to begin converting the old, fragmented optimization strategies to the unified framework.
Major improvements to the quality of the generated code aren't expected to arrive until later versions, when GCC contributors will have had a chance to really begin to leverage the new optimization infrastructure instead of just migrating to it.
So, although GCC 4.0 brings fairly dramatic benefits to compilation speed, the speed of generated binaries isn't expected to be markedly better than 3.4; that latter speedup isn't expected until later installments in the 4.x series."
Guys, just wait until Itanium is ready... This is just a 64bit extention to a 32bit extention to a 16 bit architecture...
Opteron actually IS a 64-bit extension to the x86 hell. Same instruction set - they just extended it to 64 bit, they didn't changed anything. The success of the x86-64 architecture is being just a "extension", making very easy for compilers, software developers etc. to switch to the "new" architecture. They only added 8 registers more to the typical 8 - PPC and almost every 64-bit cpu from the past decade has 32, in a 20 years timeframe (we'll be running software in x86-64 compatible CPUs just for compatibility all that time just like happened with 32-bit x86) and they won't be enought - just like today 8 are too few
Same crap. Itanium may not be great, but at least it has been built from scratch to be a real 64-bit CPU, I'd get a real 64-bit CPU anytime. The shiny x86-64 still runs the 20-years-old 16-bit ms-dos, and it's not by chance. They are damn fast just because of internal changes, not because it really is a "real 64 bit CPU"
Anyone who is willing to switch there entire network over to something only out of beta for a few days is an idiot. It's that simple.
In fact they've running it for months, even before the RTM date. Do you have a better way to debug the OS than putting it in servers which receive 30 millions of visits each day? (They have a farm of those to serve those 30 millions, so if one of them crashes and you lose one connection is not a big deal)
BTW, OSDL did the same by putting linux 2.5 development versions in all their servers (getting uptimes of 200+ days in some cases BTW)
Apparently, the number of servers that run messenger went from 250 32-bit servers to 25 64-bit servers. Apparently it was due to a limit in the number of network connections in the 32-bit edition
What are the "network limits" of linux, BSD, etc BTW?
I'm tired of reviews which spent all the time saying how great spothlight and dashboard is. And sure, they are great, but I'm sure there ARE more things in Tiger. Spothlight isn't exactly a killer app for people like which spend 100% of the time in a command shell (yes, I know there's a spotligh command-line equivalent) and running vim. If spotlight and dashboard is everything Tigert has, it's way more deceiving than Longhorn.
And dashboard is everything but a killer app - I already tried gdesklets and I just do NOT need to have a widget telling me the weather, I've windows in my house, thanks. Neither I need a stock tracker, or a currency converter, and much less a calculator or a calendar or a fligh tracker or a world clock (Why on earth would 99.9% of the global population want to know what time is in other part of the world?)
So why not Apple die-hard-fans stop talking about all that bullshit and start talking about launchd, and all those REAL features which are really interesting?
If you've ever tried comparing KDE or Gnome from Slackware/RedHat/Debian with Gentoo, you will see that the optimizations are very effective. I've used Slack,RH/FC, Deb, LFS and Gentoo. It takes me less than half the time to open Mozilla on gentoo. I like that.
I'm sorry to say this to you, but real performance doesn't come from microoptimizations, but from the algorithms and data structures. I don't understand what on earth people smokes these days to think that a compiler switch is going to make gnome, kde, mozilla and openoffice suddenly less bloated and faster, and convert O(N^N) algorithms in O(1) or something.
Mozilla is slow in gentoo, and is slow in other distros because it is the same damned code. If it's really faster (and give me numbers, not sensations, it's very easy to make people think something is faster by just telling him its faster. Quoting Linus: "If we can't measure it, it doesn't exist") I will be pleased to analyze for you what it's making it faster - prelink, who knows.
Usually, only asm paths hand-coded by programmers in the code really benefit from microoptimizations. Forget about most of the rest.
How is gentoo cool for servers? You want stability of libraries and code plus rigourous testing/QA on servers doing anything approximating real work. How does (can?) gentoo provide that?
Gentoo is about personalization, it allows you to personalize every compilation switch, using different compilers, etc. So it is a bit harder to track down bugs because of the higher number of variables involved
On the other hand, it's personalization what makes that happen, so if you use the defaults I guess there won't be many problems. "Personalization" here also means "easier to shoot your own foot". You can use it for good and for bad
Oh, and you may not like to recompile libc because of a security bug when your server is under load. Or you might want to patch and compile it by hand, because it's faster than moving all the package-based distro mechanism (you can do this also in a package-based distro anyway...) Choice...
Paul has been following the Longhorn evolution for a couple of years. When he says "the makings a train wreck" he means that there has been basically ZERO evolution since the 2004 winhec.
Not a surprise, it's know that 90% or more of the windows division spent its time working on SP2 until SP2 got released.
Actually, Debian could drop most of their mirrors by using apt-torrent, people would download all their packages with torrent (which fits very well in the APT model, just add a deb http://127.0.0.1:6968/ line to your sources.list) and mirrors wouldn't be neccesary.
Only MD5's and PGP signs would have to be stored at debian.org servers - to save yourself from hacked packages.
No, it's not strange - corruption happens everywhere. What is *scaring* is that American politics are so full of shit that they don't need to bother to hide it.
Hey, we're removing delegates from a public comission, despite of the fact that those delegates might be good enought to do their work without being biased by politics, and despite of the fact that America is supposing to be a DEMOCRACY, even if Bush is the president there're more partys in the congress, but bushies don't seem to care a lot about it
It's just like the war: "Hey sorry, we were a bit confused when we said Irak had masive-destruction weapons. But it doesn't matter because too people is so stupid they aren't able to vote against bush even if bush himself recognized the reasons he gave to start the war were wrong". When a president recognizes he has failed and people reelects him anyway the best thing you can do is emigrate.
And actually it has helped the kernel developers to make the kernel better, so in fact it has been a good thing, I don't agree at all that non-free software is always bad, it can help...
Well...despite of being closed software, I must admited Opera knows how to support linux - if you go to their download page they have packages for the main distros, ej: they've separated packages for debain woody, sarge and sid, different versions of fedora, etc...in fact I'm a a debian user (in case you haven't noticed) and I've the following line in my sources.list:
Thanks!, so this page is the pointer I needed *duck*. Nice, it has lots of formats, from/etc/hosts to C:\windows\system32\drivers\etc\hosts, to BIND and MS XML files....
Slashdot Mirror
I think it's pretty obvious that in one case EVERYTHING running on you system stops where as in the other only your current desktop closes
Which in the case of graphical apps (what 99% of the users do) means you've to stop all your apps - MUA, xchat client, AIM client, desktop environment. You need to stop all the apps and start them again. That's pretty much the same effect than rebooting your system (and this was my point)
Sorry, but would that be really that useful? I mean, how many times per day are you planning to update your drivers?
How many times per day do you update your libc, and does it requires rebooting? Why not? What about servers?
Besides, "server migration" is not useful just for this. It's needed for wireless connections - if you're using a remote X app and you move, suddenly your app breaks, it disconnected from your X server. The Right Thing to do would be to restore the communication when the connection goes back - this does not happen today. Although perhaps it's not called "server migration" but "supporting app disconnection"...dunno, I'm not expert.
Actually, upgrading drivers without rebooting IS difficult. We don't have this in the OSS field - in order to update a driver, be it 2D or 3D or whatever, you need at least to reboot X. That means switching off all your apps, and what current desktops that's pretty much like "rebooting" your computer. Yes, you're not rebooting, but with graphic apps in practice you're pretty much doing it.
/dev/null-like fake server, then update x.org drivers, rmmod the old drivers, insmod the new ones, launch xorg, and move all the windows to your new xserver, switch off the fake xserver. Or something like that. (Suggestions?)
What we need is to modify xlib to support "server migration" - we could move all the windows from a xserver to a kind of
DONT USE APACHE.
This was suggested. The kernel.org people didn't seem to have interest in it. Those light http servers are probably good for lots of small static html files. kernel.org is not like that - it needs to serve + 20 MB files and CD ISOs. Your benchmarks don't measure that. I can bet the kernel.org people knows what they're using and they'd have switched if it'd be really useful.
Can you even get the server to TELL you what the load is when it's that high??
/dev/zero /dev/null" or something like that to get high load value, you still will be able to control the box - specially with a reniced root shell.
Of course you can. "Normal people" often get high load values when they run out of mem and the box starts swapping and you can't control the box. Try running some thousands of "cp
"let's use this blog thing for marketing"
I found this in the osnews announcement
"Before we get a bunch of complaints about the fact that most binaries generated by GCC 4.0 are only marginally faster (and some a bit slower) than those compiled with 3.4, let me point out a few things that I've gathered from casually browsing the GCC development lists. I'm neither a GCC contributor nor a compiler expert.
Prior to GCC 4.0, the implementation of optimizations was mostly language-specific; there was little or no integration of optimization techniques across all languages. The main goal of the 4.0 release is to roll out a new, unified optimization framework (Tree-SSA), and to begin converting the old, fragmented optimization strategies to the unified framework.
Major improvements to the quality of the generated code aren't expected to arrive until later versions, when GCC contributors will have had a chance to really begin to leverage the new optimization infrastructure instead of just migrating to it.
So, although GCC 4.0 brings fairly dramatic benefits to compilation speed, the speed of generated binaries isn't expected to be markedly better than 3.4; that latter speedup isn't expected until later installments in the 4.x series."
They know perfectly what they've to do: Open the source as much as you can, open the standards, play nice with your competitors
This is just free advertisement. They know perfectly what to do. They can hire people who can tell them what to do.
Guys, just wait until Itanium is ready... This is just a 64bit extention to a 32bit extention to a 16 bit architecture...
Opteron actually IS a 64-bit extension to the x86 hell. Same instruction set - they just extended it to 64 bit, they didn't changed anything. The success of the x86-64 architecture is being just a "extension", making very easy for compilers, software developers etc. to switch to the "new" architecture. They only added 8 registers more to the typical 8 - PPC and almost every 64-bit cpu from the past decade has 32, in a 20 years timeframe (we'll be running software in x86-64 compatible CPUs just for compatibility all that time just like happened with 32-bit x86) and they won't be enought - just like today 8 are too few
Same crap. Itanium may not be great, but at least it has been built from scratch to be a real 64-bit CPU, I'd get a real 64-bit CPU anytime. The shiny x86-64 still runs the 20-years-old 16-bit ms-dos, and it's not by chance. They are damn fast just because of internal changes, not because it really is a "real 64 bit CPU"
MS crippled the tcp stack on purpose (to stop worms believe it or not), google will reveal more
Only in XP not in servers. Why wouldn't Microsoft create a modified kernel version for those servers if they want, anyway?
Anyone who is willing to switch there entire network over to something only out of beta for a few days is an idiot. It's that simple.
In fact they've running it for months, even before the RTM date. Do you have a better way to debug the OS than putting it in servers which receive 30 millions of visits each day? (They have a farm of those to serve those 30 millions, so if one of them crashes and you lose one connection is not a big deal)
BTW, OSDL did the same by putting linux 2.5 development versions in all their servers (getting uptimes of 200+ days in some cases BTW)
Apparently, the number of servers that run messenger went from 250 32-bit servers to 25 64-bit servers. Apparently it was due to a limit in the number of network connections in the 32-bit edition
What are the "network limits" of linux, BSD, etc BTW?
I'm tired of reviews which spent all the time saying how great spothlight and dashboard is. And sure, they are great, but I'm sure there ARE more things in Tiger. Spothlight isn't exactly a killer app for people like which spend 100% of the time in a command shell (yes, I know there's a spotligh command-line equivalent) and running vim. If spotlight and dashboard is everything Tigert has, it's way more deceiving than Longhorn.
And dashboard is everything but a killer app - I already tried gdesklets and I just do NOT need to have a widget telling me the weather, I've windows in my house, thanks. Neither I need a stock tracker, or a currency converter, and much less a calculator or a calendar or a fligh tracker or a world clock (Why on earth would 99.9% of the global population want to know what time is in other part of the world?)
So why not Apple die-hard-fans stop talking about all that bullshit and start talking about launchd, and all those REAL features which are really interesting?
A decent compiler with a little bit of optimization will do what? convert that to something more like:
Everybody uses those optimizations. CPU-specific optimizations don't do that
If you've ever tried comparing KDE or Gnome from Slackware/RedHat/Debian with Gentoo, you will see that the optimizations are very effective. I've used Slack,RH/FC, Deb, LFS and Gentoo. It takes me less than half the time to open Mozilla on gentoo. I like that.
I'm sorry to say this to you, but real performance doesn't come from microoptimizations, but from the algorithms and data structures. I don't understand what on earth people smokes these days to think that a compiler switch is going to make gnome, kde, mozilla and openoffice suddenly less bloated and faster, and convert O(N^N) algorithms in O(1) or something.
Mozilla is slow in gentoo, and is slow in other distros because it is the same damned code. If it's really faster (and give me numbers, not sensations, it's very easy to make people think something is faster by just telling him its faster. Quoting Linus: "If we can't measure it, it doesn't exist") I will be pleased to analyze for you what it's making it faster - prelink, who knows.
Usually, only asm paths hand-coded by programmers in the code really benefit from microoptimizations. Forget about most of the rest.
People asks. People answer. People asked if gentoo is good or bad for severs and I answered. You seem to have lost that part.
You don't know how to use a web browser, do you?
How is gentoo cool for servers? You want stability of libraries and code plus rigourous testing/QA on servers doing anything approximating real work. How does (can?) gentoo provide that?
Gentoo is about personalization, it allows you to personalize every compilation switch, using different compilers, etc. So it is a bit harder to track down bugs because of the higher number of variables involved
On the other hand, it's personalization what makes that happen, so if you use the defaults I guess there won't be many problems. "Personalization" here also means "easier to shoot your own foot". You can use it for good and for bad
Oh, and you may not like to recompile libc because of a security bug when your server is under load. Or you might want to patch and compile it by hand, because it's faster than moving all the package-based distro mechanism (you can do this also in a package-based distro anyway...) Choice...
Paul has been following the Longhorn evolution for a couple of years. When he says "the makings a train wreck" he means that there has been basically ZERO evolution since the 2004 winhec.
Not a surprise, it's know that 90% or more of the windows division spent its time working on SP2 until SP2 got released.
microsoft has announced it will continue supporting itanium
Actually, Debian could drop most of their mirrors by using apt-torrent, people would download all their packages with torrent (which fits very well in the APT model, just add a deb http://127.0.0.1:6968/ line to your sources.list) and mirrors wouldn't be neccesary.
Only MD5's and PGP signs would have to be stored at debian.org servers - to save yourself from hacked packages.
No, it's not strange - corruption happens everywhere. What is *scaring* is that American politics are so full of shit that they don't need to bother to hide it.
Hey, we're removing delegates from a public comission, despite of the fact that those delegates might be good enought to do their work without being biased by politics, and despite of the fact that America is supposing to be a DEMOCRACY, even if Bush is the president there're more partys in the congress, but bushies don't seem to care a lot about it
It's just like the war: "Hey sorry, we were a bit confused when we said Irak had masive-destruction weapons. But it doesn't matter because too people is so stupid they aren't able to vote against bush even if bush himself recognized the reasons he gave to start the war were wrong". When a president recognizes he has failed and people reelects him anyway the best thing you can do is emigrate.
And actually it has helped the kernel developers to make the kernel better, so in fact it has been a good thing, I don't agree at all that non-free software is always bad, it can help...
So much pr0n can't be good...
Well...despite of being closed software, I must admited Opera knows how to support linux - if you go to their download page they have packages for the main distros, ej: they've separated packages for debain woody, sarge and sid, different versions of fedora, etc...in fact I'm a a debian user (in case you haven't noticed) and I've the following line in my sources.list:
deb http://deb.opera.com/opera/ stable non-free
Thanks!, so this page is the pointer I needed *duck*. Nice, it has lots of formats, from /etc/hosts to C:\windows\system32\drivers\etc\hosts, to BIND and MS XML files....