You forgot the most important point... Gentoo can be installed from a livecd, which is what the full installer cd is. You can use the system for various tasks while it's even performing the initial installation phase.
Programs which check for updates like that are incredibly annoying... Having a whole heap of programs looking in different places for updates is horrendously stupid. The OS should provide a centralised place from which you can update the entire OS and all your apps in a centralised and consistent manner.
Ethereal only requires root if you want to actively sniff the interface with it (as opposed to reading logs you captured earlier), there are obvious reasons why non root users can't sniff network traffic especially on a system which was designed to be multi user rather than having multi-user support kludged in as an afterthought.
In many unixes nowadays you can use capabilities, to give a program that normally would require root, whatever access it requires without giving it full root (such as raw socket capability etc)...
Also, you have to be careful *WHICH* programs you give increased privileges to, some are simply not designed with that in mind, or perhaps just poorly programmed.
A huge proportion of computer hardware is manufactured in China and has been for years, not to mention countless other things... What's to stop the Chinese from sending bugged components instead of full machines?
However it raises an interesting point, it's much easier to hide back doors in software, so by this reckoning china should ban the use of american software... If this started happening, i`m sure microsoft would make it's pet government back down.
VMWare was originally designed to run under linux, and there are still some advantages to running it this way:
If a usb device has no driver under linux then it can be passed straight through and driven by an os running under vmware (you have to unload native linux drivers for any device you want passed to vmware), the windows version works differently in that you must have a native driver installed before you can pass a device to vmware. This issue has manifested itself many times when we've been at customer sites and presented with a random usb device (usb to serial adapters mainly) for which windows requires extra drivers (and linux includes drivers in the default kernel).
Performance - networking runs much faster when vmware is running atop linux, this is especially important for me as i`m often doing pentesting which involves lots of network scanning...
Security - you can nat your windows images behind your base linux install, your base linux can have everything turned off to minimise the chances of it being exploited (windows will often not let you turn some services off)
And finally, try vmware server as opposed to workstation, you can run it headless and only attach a gui when you want one..
When the first itanium chips came out, they were still being beaten on floating point by alpha and power4... Now that time has moved on, the alpha has stood still, but power5 is still faster than itanium.
It seems to depend upon the cards too... There are a large number of cards based on atheros chipsets, i have several myself... My experience is that the d-link one (650G or something, i forget the model number, it has 802.11a disabled) is very flakey, while my cisco cards (also based on the same atheros chipset, but with 802.11a still enabled) work perfectly.
If you ever want to do anything "dodgy" with wireless, like sniffing or packet injection, atheros cards are the ones to go for and i would recommend the cisco ones.
That's an apples and oranges comparison... X11 is far more flexible, and was always designed to be network transparent... If you really want something to compare windows against, i suggest MacOS9 or earlier, or AmigaOS. Both of these are simplistic interfaces designed to draw an interface to a locally attached display system.
Try comparing remote use of windows (or amigaos, or macos 9) to remote use of X11 and the performance tables are turned heavily in favour of X, i can watch video over a LAN using X11 and mplayer, windows rdesktop and windows media player is incapable of displaying watchable video over the same network.
As for garbage and flicker, this has to do with the drivers, and wether or not they sync their redraws to the monitor's refresh rate.
As for message passing and context switching, now your back to the monolithic vs microkernel approach.
Your right about inefficient use of the X11 protocol, but that's the fault of the applications themselves, and is a problem which will occur on any platform when people write inefficient apps.
In terms of raw pixel pushing speed, i've found well supported hardware to be faster on linux, an example being nvidia cards running quake3, linux tends to have an edge of around 5%.
I found intel's thermal throttling to work quite well, so while i can overclock a P4 quite considerably, it doesn't run at the full overclocked speed for very long... After a while, it will get too hot and start clocking itself down, probably somewhere down near it's originally rated speed.
Trying to secure msoffice so that users can't use it to execute arbitrary programs is a mammoth task, especially if those users will demand access to macros and such...
Having done many pentests on citrix/rdp based systems, none of them have been secure, and the ones which give you access to msoffice are by far the easiest ones to break.
One thing to note is the way shared memory works on unix/linux compared to windows.
Windows can do shared memory on DLLs only, and it goes based on the name of the dll.
Unix does shared memory on libraries and binaries, and it works based on inode...
If you have apps with a tiny executeable and the rest of the program code is loaded from dlls/libraries then this doesnt make much difference, if your running apps with big executeables then windows will be at a huge disadvantage.
Also, they way dlls are referenced (by their names), if one user on windows loads a dll with a particular name and keeps it resident, subsequent users will share that dll, if the user is able to load a malicious dll with the same name as a legitimate one you've got problems, this won't happen on unix.
Running an opteron in 32bit mode is a mistake... You could have a 64bit kernel and a 32bit userland, ala solaris... But with a 32bit kernel, you significantly lose performance when you go over 1gig of ram since you have to use nasty kludges like highmem.
Many of the so-called "thin" clients nowadays are based on XP embedded... I would avoid these things like the plague!
Of the ones i saw, they all had 256-512mb of ram, and XP embedded is almost a full copy of XP, complete with the normal problems but harder to install patches on. It seems incredibly stupid to use a fully featured OS to run a remote desktop client and nothing else, users could break out of the rdesktop client and access/abuse the local machine.
Considering that the most privileged user of a system will always need the ability to modify core parts of the OS, such as loading or modifying kernel drivers etc...
The first hurdle, is to ensure seperation higher up the chain before meddling with the kernel. No matter how well compartmentalised your microkernel OS is, if your user runs all his apps as a privileged user and one of those apps becomes infected with a virus, it will still be able to work it's way down to any part of the core OS it chooses.
If an unprivileged user gets infected with a virus, it won't be able to trash the OS and the quickest dirtiest fix would be to logout, and log back in as a clean user.
narrator.device was present in Workbench 1.3 aswell, possibly earlier versions too (i've never used earlier versions myself so can't be sure)... It was later dropped from Workbench 3.0, tho could still be installed seperately.
Any application could write speech, by saving text to the speak: device (speak: is treated like a normal drive, you just copy files there same as you would to a normal disk)
Well if these pressure groups have the financial and manpower resources to force microsoft to listen to them, then hiring a few programmers won't be beyond them either.
Rather than complaining, why don't they actually do something about it? Once they've implemented accessibility features for opendocument the situation will greatly benefit them, as these features will be under the control of the people who need them, instead of beholden to a company who's sole concern is making profit.
OpenDocument gives exactly this opportunity. These groups, instead of wasting their time complaining that current opendocument supporting apps don't support these features, should be working on developing applications specifically for the disabled market which cater exactly to their needs in the most efficient way possible, instead of bolting on "accessibility" features to existing applications which are designed with visual use in mind.
While HTML may support ABBR/ACRONYM tags, who uses them? More importantly, when someone is typing text into a word processor how often do they mark acronyms as such?
If it doesn't know how to pronounce a word, it can take a best guess, or simply read out the letters... There's no reason that someone blind wouldn't be able to spell, so they can quite easy extrapolate the meaning based on the letters involved, just like people who read the text have to.
So rather than wasting piles and piles of cash bludgeoning sun into supporting these technologies in star/open office, why not devote that same time and money towards contributing such code themselves, or even writing their own applications with accessibility as the primary goal?
Surely it makes more sense for the people who are asking for these accessibility features to play the largest possible role in creating them? That way they will suit their needs all the better.
The article comments that microsoft had to be heavily prodded and dragged to improve accessibility, they didn't really want to bother at all, and only did so because interested parties put pressure on them.
With OpenDocument, these pressure groups don't need to waste their time and money pressurising others, they can employ programmers to provide this accessibility themselves. They could modify existing applications, or even write their own applications from scratch with accessibility as the primary goal.
Doing it this way, the features they require will be written by people with the goal of making those features work as best they can, instead of an organisation that has to weigh up the pressure from the groups against the cost of implementation.
But you've also not migrated to anything else, and if you find yourself needing licenses for more workstations in the near future you might have a problem.
Image editing is one area where proprietary software doesn't suck... There are standardised image formats, regardless of what software you use. Proprietary image editing software doesn't keep you locked in to it's own formats, so publishers of such software have to compete on product quality rather than relying on you being forced to keep buying their latest versions.
Slashdot Mirror
You forgot the most important point...
Gentoo can be installed from a livecd, which is what the full installer cd is. You can use the system for various tasks while it's even performing the initial installation phase.
Programs which check for updates like that are incredibly annoying...
Having a whole heap of programs looking in different places for updates is horrendously stupid. The OS should provide a centralised place from which you can update the entire OS and all your apps in a centralised and consistent manner.
Ethereal only requires root if you want to actively sniff the interface with it (as opposed to reading logs you captured earlier), there are obvious reasons why non root users can't sniff network traffic especially on a system which was designed to be multi user rather than having multi-user support kludged in as an afterthought.
In many unixes nowadays you can use capabilities, to give a program that normally would require root, whatever access it requires without giving it full root (such as raw socket capability etc)...
Also, you have to be careful *WHICH* programs you give increased privileges to, some are simply not designed with that in mind, or perhaps just poorly programmed.
A huge proportion of computer hardware is manufactured in China and has been for years, not to mention countless other things... What's to stop the Chinese from sending bugged components instead of full machines?
However it raises an interesting point, it's much easier to hide back doors in software, so by this reckoning china should ban the use of american software... If this started happening, i`m sure microsoft would make it's pet government back down.
VMWare was originally designed to run under linux, and there are still some advantages to running it this way:
If a usb device has no driver under linux then it can be passed straight through and driven by an os running under vmware (you have to unload native linux drivers for any device you want passed to vmware), the windows version works differently in that you must have a native driver installed before you can pass a device to vmware. This issue has manifested itself many times when we've been at customer sites and presented with a random usb device (usb to serial adapters mainly) for which windows requires extra drivers (and linux includes drivers in the default kernel).
Performance - networking runs much faster when vmware is running atop linux, this is especially important for me as i`m often doing pentesting which involves lots of network scanning...
Security - you can nat your windows images behind your base linux install, your base linux can have everything turned off to minimise the chances of it being exploited (windows will often not let you turn some services off)
And finally, try vmware server as opposed to workstation, you can run it headless and only attach a gui when you want one..
When the first itanium chips came out, they were still being beaten on floating point by alpha and power4...
Now that time has moved on, the alpha has stood still, but power5 is still faster than itanium.
It seems to depend upon the cards too...
There are a large number of cards based on atheros chipsets, i have several myself...
My experience is that the d-link one (650G or something, i forget the model number, it has 802.11a disabled) is very flakey, while my cisco cards (also based on the same atheros chipset, but with 802.11a still enabled) work perfectly.
If you ever want to do anything "dodgy" with wireless, like sniffing or packet injection, atheros cards are the ones to go for and i would recommend the cisco ones.
And when they made 2000, they had BSDi write them the new stack for 2000, which is based upon BSD code.
That's an apples and oranges comparison...
X11 is far more flexible, and was always designed to be network transparent... If you really want something to compare windows against, i suggest MacOS9 or earlier, or AmigaOS. Both of these are simplistic interfaces designed to draw an interface to a locally attached display system.
Try comparing remote use of windows (or amigaos, or macos 9) to remote use of X11 and the performance tables are turned heavily in favour of X, i can watch video over a LAN using X11 and mplayer, windows rdesktop and windows media player is incapable of displaying watchable video over the same network.
As for garbage and flicker, this has to do with the drivers, and wether or not they sync their redraws to the monitor's refresh rate.
As for message passing and context switching, now your back to the monolithic vs microkernel approach.
Your right about inefficient use of the X11 protocol, but that's the fault of the applications themselves, and is a problem which will occur on any platform when people write inefficient apps.
In terms of raw pixel pushing speed, i've found well supported hardware to be faster on linux, an example being nvidia cards running quake3, linux tends to have an edge of around 5%.
My gateway is only 66mhz, and handles my 2mb line through it's pair of old ISA nics just fine.
I found intel's thermal throttling to work quite well, so while i can overclock a P4 quite considerably, it doesn't run at the full overclocked speed for very long... After a while, it will get too hot and start clocking itself down, probably somewhere down near it's originally rated speed.
Trying to secure msoffice so that users can't use it to execute arbitrary programs is a mammoth task, especially if those users will demand access to macros and such...
Having done many pentests on citrix/rdp based systems, none of them have been secure, and the ones which give you access to msoffice are by far the easiest ones to break.
One thing to note is the way shared memory works on unix/linux compared to windows.
Windows can do shared memory on DLLs only, and it goes based on the name of the dll.
Unix does shared memory on libraries and binaries, and it works based on inode...
If you have apps with a tiny executeable and the rest of the program code is loaded from dlls/libraries then this doesnt make much difference, if your running apps with big executeables then windows will be at a huge disadvantage.
Also, they way dlls are referenced (by their names), if one user on windows loads a dll with a particular name and keeps it resident, subsequent users will share that dll, if the user is able to load a malicious dll with the same name as a legitimate one you've got problems, this won't happen on unix.
Running an opteron in 32bit mode is a mistake...
You could have a 64bit kernel and a 32bit userland, ala solaris... But with a 32bit kernel, you significantly lose performance when you go over 1gig of ram since you have to use nasty kludges like highmem.
Many of the so-called "thin" clients nowadays are based on XP embedded... I would avoid these things like the plague!
Of the ones i saw, they all had 256-512mb of ram, and XP embedded is almost a full copy of XP, complete with the normal problems but harder to install patches on.
It seems incredibly stupid to use a fully featured OS to run a remote desktop client and nothing else, users could break out of the rdesktop client and access/abuse the local machine.
Considering that the most privileged user of a system will always need the ability to modify core parts of the OS, such as loading or modifying kernel drivers etc...
The first hurdle, is to ensure seperation higher up the chain before meddling with the kernel.
No matter how well compartmentalised your microkernel OS is, if your user runs all his apps as a privileged user and one of those apps becomes infected with a virus, it will still be able to work it's way down to any part of the core OS it chooses.
If an unprivileged user gets infected with a virus, it won't be able to trash the OS and the quickest dirtiest fix would be to logout, and log back in as a clean user.
narrator.device was present in Workbench 1.3 aswell, possibly earlier versions too (i've never used earlier versions myself so can't be sure)...
It was later dropped from Workbench 3.0, tho could still be installed seperately.
Any application could write speech, by saving text to the speak: device (speak: is treated like a normal drive, you just copy files there same as you would to a normal disk)
Well if these pressure groups have the financial and manpower resources to force microsoft to listen to them, then hiring a few programmers won't be beyond them either.
Rather than complaining, why don't they actually do something about it? Once they've implemented accessibility features for opendocument the situation will greatly benefit them, as these features will be under the control of the people who need them, instead of beholden to a company who's sole concern is making profit.
OpenDocument gives exactly this opportunity.
These groups, instead of wasting their time complaining that current opendocument supporting apps don't support these features, should be working on developing applications specifically for the disabled market which cater exactly to their needs in the most efficient way possible, instead of bolting on "accessibility" features to existing applications which are designed with visual use in mind.
While HTML may support ABBR/ACRONYM tags, who uses them?
More importantly, when someone is typing text into a word processor how often do they mark acronyms as such?
If it doesn't know how to pronounce a word, it can take a best guess, or simply read out the letters...
There's no reason that someone blind wouldn't be able to spell, so they can quite easy extrapolate the meaning based on the letters involved, just like people who read the text have to.
So rather than wasting piles and piles of cash bludgeoning sun into supporting these technologies in star/open office, why not devote that same time and money towards contributing such code themselves, or even writing their own applications with accessibility as the primary goal?
Surely it makes more sense for the people who are asking for these accessibility features to play the largest possible role in creating them? That way they will suit their needs all the better.
The article comments that microsoft had to be heavily prodded and dragged to improve accessibility, they didn't really want to bother at all, and only did so because interested parties put pressure on them.
With OpenDocument, these pressure groups don't need to waste their time and money pressurising others, they can employ programmers to provide this accessibility themselves. They could modify existing applications, or even write their own applications from scratch with accessibility as the primary goal.
Doing it this way, the features they require will be written by people with the goal of making those features work as best they can, instead of an organisation that has to weigh up the pressure from the groups against the cost of implementation.
But you've also not migrated to anything else, and if you find yourself needing licenses for more workstations in the near future you might have a problem.
Image editing is one area where proprietary software doesn't suck...
There are standardised image formats, regardless of what software you use. Proprietary image editing software doesn't keep you locked in to it's own formats, so publishers of such software have to compete on product quality rather than relying on you being forced to keep buying their latest versions.