"How do you handle the steady steam of security patches needed? It's a lot quickler to slap on RPMs than to compile form source, but for that you need a distro with a longer support horizon, which costs more, at least in RedHat land."
I am not directly involved with the maintenance of these computers, but first and foremost, these machines are not publicly accessible. They're only accessible to the userbase of the WAN. Certain protocols, like SSH, won't route to the box for local users at the site that it sits at, and will only accept connections from the IS network. We also have our network vlanned all to hades and back, so one site won't necessarily be able to access another.
One thing that I can tell you is that the computers are only running what they need. This pretty much comes down to samba and appletalk for the fileservers. If what I think is correct actually is correct, they have placed samba in a chroot jail. They do update it, but also bear in mind that some protocol exploits require a fairly specific set of circumstances in order to work, for the vulnerable function might be some slightly obscure thing. If the features in use don't include this there is not critical need to update it.
I'm not entirely certain how they roll out changes. I suspect that they have one of the fileserver-style machines in the office, and they simply make a note of the state of the filesystem, make the change, diff that, and make a tarball to deploy, but I'd have to ask. That's how I'd do it if I were the one maintaining the fileservers.
Remember, Slackware was designed around many of the older UNIX paradigms, which relied more heavily on root to know what (s)he was doing. Building scripts to deploy stuff, or to poll to see if something new needs to be retrieved shouldn't be out of the bounds of what a SysOp should be able to do. The hardware that we run these machines on is extremely cheap, at this point I'd guess that you could buy one of our fileservers for $150 in parts from the used surplus dealer in your hometown. Most of the machines are PII's if memory serves. There has been no need to replace them. Because of the lack of expense, it is easy to justify having a few laying around the office to use for guinea-pigs.
"The price gap is slowly converging, on the hand because Linux is simply costing more then it did 2 years ago..."
Really? Where? I run Debian and Slackware at home and have absolutely no problems with costs whatsoever. At work, we have about 120 linux servers, all tweaked-out Slackware machines. We didn't pay a cent for the OS. We use them for print servers for a massive WAN, for site-based fileservers, and for routing.
In fact, it was cheaper for us to use a Linux box with a bunch of fiber ethernet cards to handle our main network switching than it would have been for us to have purchased a router.
There is no need to pay for Linux at all, if you have people competent enough to implement it on their own, or to find a free implementation and tweak it to make it even more suitable.
Scroll lock actually does something at the console in Linux, as does a pageup and down function. This is more reason why I stick with Linux, this stuff was fairly obvious, and didn't require me to play with another application to find out.
The trick on custom stuff, like MS-Access applications and databases, is timing. If your Access solution is getting near no longer working due to age coupled with Microsoft's abandonment of older binaries, you have incentive to rewrite something. Now, are you going to spend $500 for OS and applications per year per computer, and $2000 porting it to a newer version of Access, to keep spending $500 per per year per computer, or are you going to spend $20,000, as an example, rewriting it entirely for a new platform that you'll spend -$0- for per computer per year?
Depending on the number of computers, in your case, 30, if you are a good little Microsoft customer and spending your $500/year, you are paying $15,000 per year, and still paying a developer to update your Access database. So, conversion, after a couple of years, could pay for itself if it's properly done. I'd guess that with 30 computers, you could probably outsource maintenance for when you need support, or on a regular schedule for updating, and not spend as much as you would with Windows. Viruses alone wouldn't be nearly so big a problem.
" You don't lose data when you upgrade a Mac from System 7.x to OS 9.x... what are you smoking?"
When the hard disk drive is 1.2GB and there's not enough free space to do an upgrade, so you reinstall from scratch, you do. Also, we use image deployment rather than hand-installs most of the time, so we overwrite everything on the machine. With 35,000 computers, doing a standard hand-load of the OS would take far too long.
Well, being the 'new darling' that it is, is gets a lot of attention. It also gets a lot of support and development, however, so it's not vapourware like a lot of other "new hotness" IT developments are. Comparing it to other POSIX OSes, it's development might be younger, but its license ensures that a developer's contribution remains free to everyone until it's replaced by something better, so from a developer perspective I can understand developing for Linux over BSD.
Linux has also, from my perspective, just felt more 'right' than BSD has. Take the default text console on FreeBSD. It just feels clunky, like it isn't handling all of the display formatting right. I've never felt that way about Linux's console. It's a small thing, and a thing that most people don't even deal with anymore, but for those of us who have worked with both locally without a GUI, it stands out. Also, for GUI itself, Linux with X has felt, again to me, to be more responsive than FreeBSD with X. Maybe I didn't do something right in FreeBSD, but with my knowledge at the time, Linux's working properly with less hassle made me a believer. Even the kernel stuff for Linux feels more geared to the person working with it, since there are multiple ways to go about defining what one wants in and not, and it feels intuitive. Granted, most advanced users only replace a kernel when it's actually necessary, it still feels better.
I don't say that BSD is bad, but I'm just more accustomed to Linux, as a lot of people that I know are.
Regarding your reference to Borland's suite, you'd think that people have already forgotten how they used to write applications for MS-DOS, Windows, and OS/2 back before all of the free licenses became mainstream...
I won't go quite that far, but the availability of source code in such quantity means that a program won't suddenly be orphaned because the new version of the OS doesn't support that binary anymore, like Microsoft is doing with programs that were originally released for Windows 3.1 and Windows NT 3.5. Functionality can be maintained through a little amount of work.
If you're developing on it. If you're using it for regular users who need email and web and word processing, it doesn't matter what the licensing is. Your memo written in ABIWord doesn't have to contain the GPL.
And if you're developing, there are commercial libraries available to you. There are BSD-licensed libraries too. You don't have to use Stallman's libraries, you can get them elsewhere. Hell, IBM even builds compilers, as does Intel. The entire point of GPLed stuff is for it to remain for everyone. If you don't like that, build it yourself, buy it, or find another non-GPL one.
It's not impossible to do this. It just takes brains and research. I'd rather sink my money into that than into a mindless purchase of a product that goes "BOOM!" far too frequently and forces one into paid upgrades.
Thing that I've noticed is that if a large organization gets into Linux, even if they buy it, it's theirs for the duration and all of the upgrades that they can work into it, instead of requiring either yearly site license fees or massive payouts every so many years for new versions of software to do essentially the same thing. Even paying a consulting company or services company to deploy Debian would make sense in a way, as long as the apt server were the organization's, versus a public server, so that as long as someone is maintaining the package database on the local apt server, they can keep updating the workstations.
Large organizations usually have some form of IS department, so instead of paying them to run around and fix Windows Millennium or XP problems, pay them to keep the network deployed OS current, and fix the bulk of the problems from their desks.
Not quite...
on
MIT Roofnet
·
· Score: 2, Interesting
Apparently you've never worked for an institution.
Insititutions routinely cut something off and wait for the users to complain before finding another solution, if any at all, for them. Where I work at, we've been changing our IP address scheme from an older public IP scheme to a ten-net, and once we felt that we had sufficiently changed enough systems, we turned off the ability to route the old public ones through our WAN. We then waited for the users to call to complain about not getting internet access, fileserver access, or email, and then we would send someone out to fix it. Some of our older systems, Macintoshes running 7.5 or 7.6, required us to reinstall them with 9 in order to make stuff work right, and the users often lost data because they couldn't reach their network share to back up. It wasn't considered a big deal by administration.
If the Department of Homeland Security were to be highly concerned about security, they wouldn't even have workstations with off-the-shelf distributions on them. They'd download the source code themselves, inspect it, and compile the distribution as an internal thing. And even according to the GPL, if it remains internal, i.e. no distribution to other parties, then they don't even have to say what their changes are.
In fact, they would be able to use a framework for distribution through their computer network modelled after Debian's or Slackware's or RedHat's, but with only their own versions software in the update tree. This way, they can hire staff with existing administrative knowledge of the flavour of distribution that they choose, and the person will not really have much of a learning curve. Or, if they're really paranoid, they can write it themselves.
I'd personally recommend against having any personal computer on the user's desk. Give them an X Term that uses some kind if high-encryption tunnelling scheme to deliver the applications to the X Server, and have departmental-sized or building-sized computers for the users to work on. This ensures much better physical security for the equipment, with a fraction of the physical assets to watch, better data integrity since it would be stored on some fault-tolerant medium like RAID5. With a properly implemented security scheme for user login, either with some kind of biometric ID or an actually decent password scheme, it would be relatively difficult to break in compared to normaly corporate environments.
As for local security on the application servers, it would require a fairly decent file security model, but big computers have been done before. The implementers would have to work to ensure no local root exploits, but that would be good for the community as a whole.
"Linux, currently a software system mostly used to power big servers and personal computers, is also now emerging as a small set of computing code to drive devices like mobile phones, remote controls and smart vibrators."
In a large corporate environment, the IS department should put together a basic environment, WM, and theme to be default. This promotes standardization across the board. Let users play, though.
One of the biggest complaints that I've had with Microsoft shells until I stopped using Windows (around the time that 2000 came out) is that while color and fonts and the like can be changed, the results don't seem to be that pretty. Style remains very sterile with just color shift. Granted, XP has some theming, but it's behind the ball in that regard, and still has that Redmond feel.
"Then don't buy a stupid marketing gimmick. All it is, is an Intell processor and a wireless card. Buy a regular Intel/AMD and a PCMCIA wireless card and your set. And you don't have to pay extra for the marketing crap or be locked into proprietary junk."
Wholeheartedly agreed. Remembering when the first winmodems came out, or the first all-in-one 486 and Pentium era boards came out, that stuff was crap. Decent hardware integrators either use real components or leave expansion available. A favorite of mine is Mini-PCI, since it's not locked into one vendor, is a standard, and allows me to customize my laptop slightly for long-duration use devices, which is Ethernet right now.
I have never seen a legitimate computer technology whose name ended in "o". They've all been marketting ploys.
"Good algorithm design and analysis transcends linux vs windows vs mac osx."
That may be, but remember, Sun donated or sold very inexpensively LOTS of computers to university-level education. They had a very significant market share when commercial Internet ventures began, before the whole dotcom fiasco.
Had the stock market not taken quite such a feeding frenzy on dotcoms, having been a little more sane, I wouldn't be surprised if Solaris had stayed top commercial server OS for TCP/IP. This isn't to say that Linux is bad, but commerical products tend to make executives feel more secure, for whatever reason.
that as it may be on a purchasing level...
on
Big Company on Campus
·
· Score: 2, Interesting
...We're all aware that computers that can run Microsoft Windows are also capable of running Linux, many versions of BSD, and Solaris/x86. So, we end up with several free OSes, and a few commercial OSes (counting some of the commercial BSDs) that will run on the hardware. My favourite computer science professor had a computer at his desk that had a windows license sticker, but he never even booted into Redmond's OS before he wiped it and installed Linux.
Many large colleges have UNIX clusters of some form. ASU has the "general" cluster, on Solaris machines. U of A has the "U" cluster. I don't think that UNIX is going anywhere, these systems have thousands of simultaneous users and seem to be fairly stable considering all of the local accounts.
It could also be that maybe colleges are trying to keep their licensing in full compliance instead of getting sloppy about it, for fear of the retribution that could come later.
Yeah, I was trying to be sarcastic. Unfortunately I didn't find outlandish enough examples to make it work that way. There is an element of truth to it, though, in that we do have our problems, and they are very difficult to find a treatment for.
"...but I live in the US, not one of you more violent countries..."
since when?! We've had civil war, forced annexation of lands belonging to other countries, wars against several major European and Asian powers, one of the largest arsenals of nukes ever assembled, and that's all on a government level. On a personal level, we have gang violence like drive-by shootings, muggings, armed robbery, thrill-killing, killing of people because they are in a neighbourhood with the wrong skin pigmentation, domestic and foreign terrorism, and many other problems...
"I am anti-MS because I am tired of rebooting, and know that I could design their apps much better than they ever will. If they have some of the best programmers in the world, why are their applications so bad?"
Hell, I'd be happy if their OS didn't crash, even if the applications did from time to time.
I've been using Linux at home for many years, and I've noticed that applications do crash. Mozilla crashes, ABIWord crashes, StarOffice crashes, but there are two important points to this. First, the applications that I've described are either free or inexpensive. So, I haven't shelled out $500 for a suite of applications that is faulty. Second, it's only the one application that goes down in flames. It isn't the OS, it usually isn't the GUI interface (though X is a hair weak for what I'd like to see), and the other programs remain running without issue.
I don't think that an application should have the ability to crash an OS. That is absolutely ridiculous.
See, I would think that your average car equivalent to an A+ computer tech would be qualified to work at Pep Boys. I don't have a lot of respect for A+ certified technicians by paper alone. I was playing with the hardware inside of my PC when I was twelve.
Ah. I received my a+ certification back when one tested on Windows 3.1/MS-DOS 6.2, In 1996. I had to know stuff like how the MCI control panel worked, and how to fix the boot-related files when they broke and whatnot. I guess that my perception of things is a bit skewed by that.
Slashdot Mirror
ew... webmin...
But what If I don't like administering my server by mouse?
*grin*
"How do you handle the steady steam of security patches needed? It's a lot quickler to slap on RPMs than to compile form source, but for that you need a distro with a longer support horizon, which costs more, at least in RedHat land."
I am not directly involved with the maintenance of these computers, but first and foremost, these machines are not publicly accessible. They're only accessible to the userbase of the WAN. Certain protocols, like SSH, won't route to the box for local users at the site that it sits at, and will only accept connections from the IS network. We also have our network vlanned all to hades and back, so one site won't necessarily be able to access another.
One thing that I can tell you is that the computers are only running what they need. This pretty much comes down to samba and appletalk for the fileservers. If what I think is correct actually is correct, they have placed samba in a chroot jail. They do update it, but also bear in mind that some protocol exploits require a fairly specific set of circumstances in order to work, for the vulnerable function might be some slightly obscure thing. If the features in use don't include this there is not critical need to update it.
I'm not entirely certain how they roll out changes. I suspect that they have one of the fileserver-style machines in the office, and they simply make a note of the state of the filesystem, make the change, diff that, and make a tarball to deploy, but I'd have to ask. That's how I'd do it if I were the one maintaining the fileservers.
Remember, Slackware was designed around many of the older UNIX paradigms, which relied more heavily on root to know what (s)he was doing. Building scripts to deploy stuff, or to poll to see if something new needs to be retrieved shouldn't be out of the bounds of what a SysOp should be able to do. The hardware that we run these machines on is extremely cheap, at this point I'd guess that you could buy one of our fileservers for $150 in parts from the used surplus dealer in your hometown. Most of the machines are PII's if memory serves. There has been no need to replace them. Because of the lack of expense, it is easy to justify having a few laying around the office to use for guinea-pigs.
"I thought the AppleMac only displayed the bomb as an icon. Does the new version actually make it explode?"
Well, there was that situation with some of the PowerPC laptops and their batteries catching fire...
"The price gap is slowly converging, on the hand because Linux is simply costing more then it did 2 years ago..."
Really? Where? I run Debian and Slackware at home and have absolutely no problems with costs whatsoever. At work, we have about 120 linux servers, all tweaked-out Slackware machines. We didn't pay a cent for the OS. We use them for print servers for a massive WAN, for site-based fileservers, and for routing.
In fact, it was cheaper for us to use a Linux box with a bunch of fiber ethernet cards to handle our main network switching than it would have been for us to have purchased a router.
There is no need to pay for Linux at all, if you have people competent enough to implement it on their own, or to find a free implementation and tweak it to make it even more suitable.
Scroll lock actually does something at the console in Linux, as does a pageup and down function. This is more reason why I stick with Linux, this stuff was fairly obvious, and didn't require me to play with another application to find out.
The trick on custom stuff, like MS-Access applications and databases, is timing. If your Access solution is getting near no longer working due to age coupled with Microsoft's abandonment of older binaries, you have incentive to rewrite something. Now, are you going to spend $500 for OS and applications per year per computer, and $2000 porting it to a newer version of Access, to keep spending $500 per per year per computer, or are you going to spend $20,000, as an example, rewriting it entirely for a new platform that you'll spend -$0- for per computer per year?
Depending on the number of computers, in your case, 30, if you are a good little Microsoft customer and spending your $500/year, you are paying $15,000 per year, and still paying a developer to update your Access database. So, conversion, after a couple of years, could pay for itself if it's properly done. I'd guess that with 30 computers, you could probably outsource maintenance for when you need support, or on a regular schedule for updating, and not spend as much as you would with Windows. Viruses alone wouldn't be nearly so big a problem.
" You don't lose data when you upgrade a Mac from System 7.x to OS 9.x... what are you smoking?"
When the hard disk drive is 1.2GB and there's not enough free space to do an upgrade, so you reinstall from scratch, you do. Also, we use image deployment rather than hand-installs most of the time, so we overwrite everything on the machine. With 35,000 computers, doing a standard hand-load of the OS would take far too long.
Well, being the 'new darling' that it is, is gets a lot of attention. It also gets a lot of support and development, however, so it's not vapourware like a lot of other "new hotness" IT developments are. Comparing it to other POSIX OSes, it's development might be younger, but its license ensures that a developer's contribution remains free to everyone until it's replaced by something better, so from a developer perspective I can understand developing for Linux over BSD.
Linux has also, from my perspective, just felt more 'right' than BSD has. Take the default text console on FreeBSD. It just feels clunky, like it isn't handling all of the display formatting right. I've never felt that way about Linux's console. It's a small thing, and a thing that most people don't even deal with anymore, but for those of us who have worked with both locally without a GUI, it stands out. Also, for GUI itself, Linux with X has felt, again to me, to be more responsive than FreeBSD with X. Maybe I didn't do something right in FreeBSD, but with my knowledge at the time, Linux's working properly with less hassle made me a believer. Even the kernel stuff for Linux feels more geared to the person working with it, since there are multiple ways to go about defining what one wants in and not, and it feels intuitive. Granted, most advanced users only replace a kernel when it's actually necessary, it still feels better.
I don't say that BSD is bad, but I'm just more accustomed to Linux, as a lot of people that I know are.
Regarding your reference to Borland's suite, you'd think that people have already forgotten how they used to write applications for MS-DOS, Windows, and OS/2 back before all of the free licenses became mainstream...
"Overall, the system is still the same."
I won't go quite that far, but the availability of source code in such quantity means that a program won't suddenly be orphaned because the new version of the OS doesn't support that binary anymore, like Microsoft is doing with programs that were originally released for Windows 3.1 and Windows NT 3.5. Functionality can be maintained through a little amount of work.
If you're developing on it. If you're using it for regular users who need email and web and word processing, it doesn't matter what the licensing is. Your memo written in ABIWord doesn't have to contain the GPL.
And if you're developing, there are commercial libraries available to you. There are BSD-licensed libraries too. You don't have to use Stallman's libraries, you can get them elsewhere. Hell, IBM even builds compilers, as does Intel. The entire point of GPLed stuff is for it to remain for everyone. If you don't like that, build it yourself, buy it, or find another non-GPL one.
It's not impossible to do this. It just takes brains and research. I'd rather sink my money into that than into a mindless purchase of a product that goes "BOOM!" far too frequently and forces one into paid upgrades.
Thing that I've noticed is that if a large organization gets into Linux, even if they buy it, it's theirs for the duration and all of the upgrades that they can work into it, instead of requiring either yearly site license fees or massive payouts every so many years for new versions of software to do essentially the same thing. Even paying a consulting company or services company to deploy Debian would make sense in a way, as long as the apt server were the organization's, versus a public server, so that as long as someone is maintaining the package database on the local apt server, they can keep updating the workstations.
Large organizations usually have some form of IS department, so instead of paying them to run around and fix Windows Millennium or XP problems, pay them to keep the network deployed OS current, and fix the bulk of the problems from their desks.
Apparently you've never worked for an institution.
Insititutions routinely cut something off and wait for the users to complain before finding another solution, if any at all, for them. Where I work at, we've been changing our IP address scheme from an older public IP scheme to a ten-net, and once we felt that we had sufficiently changed enough systems, we turned off the ability to route the old public ones through our WAN. We then waited for the users to call to complain about not getting internet access, fileserver access, or email, and then we would send someone out to fix it. Some of our older systems, Macintoshes running 7.5 or 7.6, required us to reinstall them with 9 in order to make stuff work right, and the users often lost data because they couldn't reach their network share to back up. It wasn't considered a big deal by administration.
If the Department of Homeland Security were to be highly concerned about security, they wouldn't even have workstations with off-the-shelf distributions on them. They'd download the source code themselves, inspect it, and compile the distribution as an internal thing. And even according to the GPL, if it remains internal, i.e. no distribution to other parties, then they don't even have to say what their changes are.
In fact, they would be able to use a framework for distribution through their computer network modelled after Debian's or Slackware's or RedHat's, but with only their own versions software in the update tree. This way, they can hire staff with existing administrative knowledge of the flavour of distribution that they choose, and the person will not really have much of a learning curve. Or, if they're really paranoid, they can write it themselves.
I'd personally recommend against having any personal computer on the user's desk. Give them an X Term that uses some kind if high-encryption tunnelling scheme to deliver the applications to the X Server, and have departmental-sized or building-sized computers for the users to work on. This ensures much better physical security for the equipment, with a fraction of the physical assets to watch, better data integrity since it would be stored on some fault-tolerant medium like RAID5. With a properly implemented security scheme for user login, either with some kind of biometric ID or an actually decent password scheme, it would be relatively difficult to break in compared to normaly corporate environments.
As for local security on the application servers, it would require a fairly decent file security model, but big computers have been done before. The implementers would have to work to ensure no local root exploits, but that would be good for the community as a whole.
"Linux, currently a software system mostly used to power big servers and personal computers, is also now emerging as a small set of computing code to drive devices like mobile phones, remote controls and smart vibrators."
And who said that Women didn't like Eunuchs?
A modification to your sig:
Eric S. Raymond: I decry as Norm
Cheers...
In a large corporate environment, the IS department should put together a basic environment, WM, and theme to be default. This promotes standardization across the board. Let users play, though.
One of the biggest complaints that I've had with Microsoft shells until I stopped using Windows (around the time that 2000 came out) is that while color and fonts and the like can be changed, the results don't seem to be that pretty. Style remains very sterile with just color shift. Granted, XP has some theming, but it's behind the ball in that regard, and still has that Redmond feel.
"Then don't buy a stupid marketing gimmick. All it is, is an Intell processor and a wireless card. Buy a regular Intel/AMD and a PCMCIA wireless card and your set. And you don't have to pay extra for the marketing crap or be locked into proprietary junk."
Wholeheartedly agreed. Remembering when the first winmodems came out, or the first all-in-one 486 and Pentium era boards came out, that stuff was crap. Decent hardware integrators either use real components or leave expansion available. A favorite of mine is Mini-PCI, since it's not locked into one vendor, is a standard, and allows me to customize my laptop slightly for long-duration use devices, which is Ethernet right now.
I have never seen a legitimate computer technology whose name ended in "o". They've all been marketting ploys.
"Good algorithm design and analysis transcends linux vs windows vs mac osx."
That may be, but remember, Sun donated or sold very inexpensively LOTS of computers to university-level education. They had a very significant market share when commercial Internet ventures began, before the whole dotcom fiasco.
Had the stock market not taken quite such a feeding frenzy on dotcoms, having been a little more sane, I wouldn't be surprised if Solaris had stayed top commercial server OS for TCP/IP. This isn't to say that Linux is bad, but commerical products tend to make executives feel more secure, for whatever reason.
...We're all aware that computers that can run Microsoft Windows are also capable of running Linux, many versions of BSD, and Solaris/x86. So, we end up with several free OSes, and a few commercial OSes (counting some of the commercial BSDs) that will run on the hardware. My favourite computer science professor had a computer at his desk that had a windows license sticker, but he never even booted into Redmond's OS before he wiped it and installed Linux.
Many large colleges have UNIX clusters of some form. ASU has the "general" cluster, on Solaris machines. U of A has the "U" cluster. I don't think that UNIX is going anywhere, these systems have thousands of simultaneous users and seem to be fairly stable considering all of the local accounts.
It could also be that maybe colleges are trying to keep their licensing in full compliance instead of getting sloppy about it, for fear of the retribution that could come later.
Yeah, I was trying to be sarcastic. Unfortunately I didn't find outlandish enough examples to make it work that way. There is an element of truth to it, though, in that we do have our problems, and they are very difficult to find a treatment for.
"...but I live in the US, not one of you more violent countries..."
since when?! We've had civil war, forced annexation of lands belonging to other countries, wars against several major European and Asian powers, one of the largest arsenals of nukes ever assembled, and that's all on a government level. On a personal level, we have gang violence like drive-by shootings, muggings, armed robbery, thrill-killing, killing of people because they are in a neighbourhood with the wrong skin pigmentation, domestic and foreign terrorism, and many other problems...
"I am anti-MS because I am tired of rebooting, and know that I could design their apps much better than they ever will. If they have some of the best programmers in the world, why are their applications so bad?"
Hell, I'd be happy if their OS didn't crash, even if the applications did from time to time.
I've been using Linux at home for many years, and I've noticed that applications do crash. Mozilla crashes, ABIWord crashes, StarOffice crashes, but there are two important points to this. First, the applications that I've described are either free or inexpensive. So, I haven't shelled out $500 for a suite of applications that is faulty. Second, it's only the one application that goes down in flames. It isn't the OS, it usually isn't the GUI interface (though X is a hair weak for what I'd like to see), and the other programs remain running without issue.
I don't think that an application should have the ability to crash an OS. That is absolutely ridiculous.
See, I would think that your average car equivalent to an A+ computer tech would be qualified to work at Pep Boys. I don't have a lot of respect for A+ certified technicians by paper alone. I was playing with the hardware inside of my PC when I was twelve.
Ah. I received my a+ certification back when one tested on Windows 3.1/MS-DOS 6.2, In 1996. I had to know stuff like how the MCI control panel worked, and how to fix the boot-related files when they broke and whatnot. I guess that my perception of things is a bit skewed by that.