The fair use provisions of the Copyright Act are falling by the wayside due to the "Digital Millenium Copyright Act." Take a look at the DeCSS case - read the transcript thats available at http://www.2600.com/news/2000/0121-tra ns.txt.
The DMCA allows corporations to make an end-run around fair use rights. This is a point that's missed by a lot of people commenting on this issue. The DeCSS case is not about copying. Even the MPAA's lawyers admit this.
The suit is filed under a section of the law (1201(a)(2)) which prohibits anyone from offering or distributing any technology "that is primarily designed or produced for the purpose of circumventing a technological measure that effectively controls access to a work protected." This lawsuit is *not* filed under the slightly less objectionable section 1201(b)(1)(A) which offers a similar prohibition against technology whose primary purpose is "circumventing protection that... protects a right of a copyright holder."
Now, if this suit were about 1201(b), it is likely that the defendants could successfully defend themselves, as it is not a "right of a copyright holder" to restrict fair use. As a result, a piece of software that enables fair use would likely not run afoul of this section.
Unfortunately, section 1201(a) is much more insidious. Though it has an exemption for fair use, the defendants are unable to avail themselves of that exemption, as -they- are not exercising fair use by distributing the software. Rather, they are enabling -others- to exercise fair use. There is no protection for this activity (other than potentially speech under the 1st Amendment...)
So, in the absence of specific exemptions in the law, let's examine whether the defendants comitted an offence that falls under s. 1201(a) of the DMCA. Do the defendants offer or distribute the technology in question? Yes (at least those defendants who are distributing rather than linking). Is the technology in question primarily designed to circumvent a technological measure? Yes - clearly, DeCSS circumvents a technological measure known as encryption. Does that technical measure "control access to a work protected" under the act. Again, yes - Without dealing with the technical measure, one cannot view the data stored on the DVD.
This brings us to the issue of the DMCA itself, and particularly s. 1201(a). This section is simply bad law. The phrase "controls access to a work protected..." is far too sweeping, broad and vague. This law provides a new effective right to copyright holders to "control access" to copyrighted material in any manner they choose. It prevents the average consumer, who has little knowledge about the mechanics of taking advantage of their fair use rights, from obtaining assistance in taking advantage of these fair use rights. What is attacked here is not the right to *use* or to *posess* this tool, what is attacked is the right to distribute this tool.
This issue comes down to a prohibition of speech. We likely agree, and some courts do as well, that source code is speech. What needs to be decided is whether this is *protected* speech.
I've been running Linux on my laptop since '96 (Hitachi M120D - P120/16M/1.4G/Int CD/Ext Floppy). I haven't seen anything quite so stable. In terms of stability, it rocks... recently, mind you, it's been my web server, tethered to the cable at home (a job which it performs admirably - built in UPS with the battery and all...) I understand there are problems with some laptops. I've seen systems next to impossible to get X up on. Eventually, however, everyone will start to have a similar experience to what I've had...
Now, in comparison to NT... Let's talk about hot-swapping PCMCIA cards (oh. NT doesn't support that? ok.)
Let's talk about a situation many users with docking stations have. Consider that I have a docking station with a built in ethernet adapter. This is in my office/cube/whatever. I want to be able to drop my laptop into the station, not hook any cables up or whatever. Now consider that I have a PCMCIA ethernet card. I want to roam somewhere else on the *same network segment* to do a presentation, work with a coworker, etc. without bringing my docking station. I would under Linux, drop the docking station interface, undock, carry it where I need, and bring up the PCMCIA interface. Problem is, under NT, there's no easy way to drop an interface, and even if the device doesn't exist, the routes to it do, so there's no easy way to deal with the situation of two interfaces configured for the same network segment, one of which is up, and one of which is down at any given time...
"I'm hoping that Microsoft's inclusion of limited web server software in W2K will sway high-speed ISPs to rethink their stance against servers."
Well I'm hoping that Microsoft's inclusion of limited web server software in W2K will sway users to rethink their stance on Microsoft "security".
If this is at all like PWS or FPE in quality, there'll be a lot of MS users pissed of at Microsoft for exposing them to crackers. It looks like the average @Home user gets port scanned a couple times a day (and not by @Home admins). How much do you think that's going to accelerate when just about everybody has an insecure (MS) web server running?
I can think of yet another niche-market distro. We have the source, thus we have the power. A Linux distro tailored to the needs of the education community - one that concentrates on interactive teaching and the like, while maintaining Linux's openness and security.
This is why I see Linux (and open source in general) as being so important and useful for education. You have access to all levels of abstraction, and yet you have the ability to also use those abstractions. Part of the growth in computer education is being able to gradually peel back the layers of abstraction that separate us from the machine. As the student becomes capable of dealing with more depth and breadth to his (her) education, he (she) can peel back another layer of abstraction.
K, so what do we need? techies, educators and techie-educators. If you can express it, we can build it.
"Every once in a while, I'll get the bomb on the iMac. Usually its because I'm playing with beta software tho..."
This is one of the biggest problem I have with MacOS and Windows. I don't care how badly coded the app is, user mode apps should not bring down the system. The perceived reliability and stability of an operating system should never be judged by how it treats well-behaved apps. The perceived stability and reliability of an operating system must be judged by how it deals with poorly behaved apps and unforeseen circumstances. Think about it - any teacher can teach well-behaved, respectful kids. It takes a special kind to teach poorly behaved kids.
Take a look at the other books (Speaker for the Dead, Xenocide and Children of the Mind). Ender's Game was rather juvenile, but the other three were definitely not. I can't speak for Ender's Shadow (I just started reading it), but I certainly enjoyed the first four. Card definitely takes a look at some interesting subjects in Speaker, Xenocide, and Children - further, they're not shoot-em-up action books like EG.
The foreword to Ender's Shadow describes how the project was initially undertaken, and how Mr. Card "selfishly" (his words, not mine) swiped the offer to write ES from another author...
So use another bootloader... When we've got very friendly, nice bootloaders like GRUB (my personal favourite), GAG (nice and graphical), etc... there's no reason to be bitching about LILO.
I see... The attempt at humour was made painfully obvious by the statement - "The hackers wrote of their own "rooting" exploits (that is, hacking the root directory of a server)"
This is _not_ only about piracy. A big part of this is being able to _view_ DVDs on a system of _our_ choice. This is about _not_ being forced to use proprietary software for any purpose - entertainment or otherwise.
This is a side effect of using just about any file system. Most file systems do not zero out a file's blocks when it gets deleted, but rather add them to a free list. The difference here is in the free block allocation policy supported by the particular fs. If you're relying on rm or del or whatever to eliminate all traces of a file, you're being very foolish.
"calming the nerves of worried parents" is absolutely NOT "a perfectly valid reason" to invade the privacy of kids with unproven technology. It is not a valid reason to spend time and money categorizing, separating and splitting kids into profile groups, rather than actually providing them with the support they need to be unique individuals.
Now, lets talk about the science (or lack thereof), here. This software is developed from interviews with violent kids. This means that its use as a predictor is problematic. The statement "most violent kids exhibit characteristics a, b, and c - therefore most kids that exhibit characteristics a, b, and c are violent" is as much a logical fallacy as saying that "most rapists are male - therefore most males are rapists."
It is important to note that the vast majority of Microsofr's donations are software. While most other software companies assess value of donations as either cost or wholesale value, Microsoft assesses all of their donations at _retail_value_. Any comparisons between Microsoft and other software companies in this field are irrelevant.
On a quick scan of both Microsoft's and Red Hat's site, it becomes apparent that there are about as many NT4-PostSP5 hotfixes as there are software errata listed at Red Hat's site. When you add in the errata for Exchange, IIS, Word, and other MS apps, Microsoft's burying themselves (and their customers) in bug fixes much faster than Red Hat is. Moreover, Red Hat has an easy method of obtaining info about errata *before*they*become*an*issue*. Microsoft's method basically consists of - if you've got a problem, search Knowledge Base - there might be a fix.
Come on. Everybody has bugs. What matters is how you deal with the bugs. Do you sweep them under the carpet like Microsoft? Or should you stand up point them out, and point to the fixes?
By the time most Linux vulnerabilities could hit the press, they've been fixed. This is an example of that. The crond hole has been fixed. By the time a Windows exploit hits the press, they're determining how much to charge us for the fix. Put simply, bugs get fixed faster with open source. You don't have to wait on a corporate timetable to get fixes that you _need_.
Mail order fraud is more challenging than you describe. Most credit card companies require that products purchased by mail order must be shipped to the billing address for the card. Admittedly, this does not apply to pr0n sites;-) and other situations where the product need not be shipped. As a result, if I get your card number and expiry, I can buy something and have it shipped to you, I can subscribe to some pay website, or I can try to change your billing information so that I can actually get product shipped to me.
I would imagine that changing someone's billing address is quite easy... Whenever I've moved, the only authentication they've tried is phone numbers, birthdate, card number, name, old address. Would I do this? No - it leaves a trail pointing at me.
Since Microsoft relies exclusively on prevntative security measures to protect the private key(s) from being compromised, any measure which would reduce this security is unacceptable. More copies of the key makes it more difficult to maintain security and ensure that the private key is not compromised (as opposed to lost).
Microsoft is right about one of the issues here. A backup key is useful for key loss. If the private primary key were lost in fire, flood, stolen, etc. and there were no backup, CryptoAPI would very quickly cease to be useful, as no newer encryption methods would be able to be implemented under CryptoAPI. This means taht a backup key is useful.
Microsoft has, apparently done some things wrong here, though. As I understand it, the reason the primary key can't be changed is that the default CSP is signed with the primary key, and changing this key would render the system unusable. Had a necessary component been signed with both the backup key and the primary key, and both signatures tested for that component, it would become more difficult to change the keys and retain a useful system.
The other thing that Microsoft has messed up here is the issue of key compromise. Microsoft has focussed entirely on prevention on this issue, excluding contingencies where the keys are compromised. They speak of hardware, software and physical security to prevent key compromise, but there does not appear to be a mechanism for key revocation if and when these security mechanisms fail, and the key _is_ compromised.
But the government _does_not_ sell your SSN and related information to corporations. Some company can't just buy a copy of your tax return. Companies _can_ buy copies of your driver's license and related information.
They are selling the pictures and information to _private_companies_. Telecheck is a private company which provides check verification and insurance to merchants. I'm sorry, but I have a problem with the government selling my information and my picture to some private business. (Much as I have a problem with them providing my mailing address to private parking companies...)
There is much more to this issue than just the government. This is about strengthening the stranglehold that big business has over the United States.
Claiming that Linux is a waste of money while MacOS is not is ludicrous. The fact is, like it or not, Linux is a presence in more businesses than MacOS is, whether we're talking about desktop or server use. To marginalize this OS on _business_software_, particularly for small to medium sized businesses, is a dangerous move. Like it or not, for that segment of the market, MacOS is a _much_ riskier proposition with a _much_ bleaker outlook than Linux.
That said, I've never been a big fan of Intuit's products anyway. I'd much rather see Computer Associates port their low-mid end packages (like Simply Accounting) over. I'd also love to see Canadian tax-prep software on Linux as well... Oooh. I think I smell an open source project coming on...
Okay, for those of you who've read this already, this is my semi-regular rant on Windows, installing apps, and shared libraries. If you've read it already, skip it.
>Yes, you need to reboot much more when >installing or changing software than Linux, but >you can always say no and stop and restart the >service. (Same basic prinicple on Linux)
No. These are very different issues. When it comes to shared libraries, Windows has dropped the ball. Shared libraries mean that some other application (perhaps even the OS) may use them and have them locked. When you are asked to reboot, it is because some shared libraries couldnot be copied because they were locked. These shared libraries will not get copied until you reboot. More concerning, however, is the uninstall. If the uninstall can't delete something (say X.DLL), it marks it for deletion at next reboot, and asks if you want to reboot. If you say no, and go install something else in the meantime, which happens to try to install X.DLL, things will be happy until you reboot, when X.DLL is deleted (as the uninstaller told it to do).
Linux deals with updating shared libraries, partially as a result of naming conventions, and partially as a result of filesystem capabilities.
In Windows, if a program is using a particular shared library, you cannot remove that library file from the filesystem. It is locked. Under Linux, what is locked are the inodes. It is irrelevant whether or not the file has a filename linked to it. When I use the unlink() system call (there's a reason there's no delete_file() sys call), we unlink the file name from the inodes storing the file, preventing more apps from opening it. Only when the file's usage count drops to 0 is it removed from the fs. This means that I can create a new file, with the same name as the one I just unlinked, without affecting the performance of any applications using the original file. I can replace shared libraries without having to stop the application using the shared library.
Further, most Linux applications don't install new shared libraries when they install themselves. It would be seen as very poor practice for some little app to install a new version of glibc as part of its own installation. Yet, we accept this as normal in the NT world.
Even better, Linux allows multiple versions of a library to coexist on a system, using version numbers. Under Windows, you can't even get Microsoft's own ODBC 3.51 to coexist with ODBC 4.0.
Slashdot Mirror
The DMCA allows corporations to make an end-run around fair use rights. This is a point that's missed by a lot of people commenting on this issue. The DeCSS case is not about copying. Even the MPAA's lawyers admit this.
The suit is filed under a section of the law (1201(a)(2)) which prohibits anyone from offering or distributing any technology "that is primarily designed or produced for the purpose of circumventing a technological measure that effectively controls access to a work protected." This lawsuit is *not* filed under the slightly less objectionable section 1201(b)(1)(A) which offers a similar prohibition against technology whose primary purpose is "circumventing protection that ... protects a right of a copyright holder."
Now, if this suit were about 1201(b), it is likely that the defendants could successfully defend themselves, as it is not a "right of a copyright holder" to restrict fair use. As a result, a piece of software that enables fair use would likely not run afoul of this section.
Unfortunately, section 1201(a) is much more insidious. Though it has an exemption for fair use, the defendants are unable to avail themselves of that exemption, as -they- are not exercising fair use by distributing the software. Rather, they are enabling -others- to exercise fair use. There is no protection for this activity (other than potentially speech under the 1st Amendment...)
So, in the absence of specific exemptions in the law, let's examine whether the defendants comitted an offence that falls under s. 1201(a) of the DMCA. Do the defendants offer or distribute the technology in question? Yes (at least those defendants who are distributing rather than linking). Is the technology in question primarily designed to circumvent a technological measure? Yes - clearly, DeCSS circumvents a technological measure known as encryption. Does that technical measure "control access to a work protected" under the act. Again, yes - Without dealing with the technical measure, one cannot view the data stored on the DVD.
This brings us to the issue of the DMCA itself, and particularly s. 1201(a). This section is simply bad law. The phrase "controls access to a work protected..." is far too sweeping, broad and vague. This law provides a new effective right to copyright holders to "control access" to copyrighted material in any manner they choose. It prevents the average consumer, who has little knowledge about the mechanics of taking advantage of their fair use rights, from obtaining assistance in taking advantage of these fair use rights. What is attacked here is not the right to *use* or to *posess* this tool, what is attacked is the right to distribute this tool.
This issue comes down to a prohibition of speech. We likely agree, and some courts do as well, that source code is speech. What needs to be decided is whether this is *protected* speech.
On the Hemos question! ;-)
I've been running Linux on my laptop since '96 (Hitachi M120D - P120/16M/1.4G/Int CD/Ext Floppy). I haven't seen anything quite so stable. In terms of stability, it rocks... recently, mind you, it's been my web server, tethered to the cable at home (a job which it performs admirably - built in UPS with the battery and all...) I understand there are problems with some laptops. I've seen systems next to impossible to get X up on. Eventually, however, everyone will start to have a similar experience to what I've had...
Now, in comparison to NT... Let's talk about hot-swapping PCMCIA cards (oh. NT doesn't support that? ok.)
Let's talk about a situation many users with docking stations have. Consider that I have a docking station with a built in ethernet adapter. This is in my office/cube/whatever. I want to be able to drop my laptop into the station, not hook any cables up or whatever. Now consider that I have a PCMCIA ethernet card. I want to roam somewhere else on the *same network segment* to do a presentation, work with a coworker, etc. without bringing my docking station. I would under Linux, drop the docking station interface, undock, carry it where I need, and bring up the PCMCIA interface. Problem is, under NT, there's no easy way to drop an interface, and even if the device doesn't exist, the routes to it do, so there's no easy way to deal with the situation of two interfaces configured for the same network segment, one of which is up, and one of which is down at any given time...
"I'm hoping that Microsoft's inclusion of limited web server software in W2K will sway high-speed ISPs to rethink their stance against servers."
Well I'm hoping that Microsoft's inclusion of limited web server software in W2K will sway users to rethink their stance on Microsoft "security".
If this is at all like PWS or FPE in quality, there'll be a lot of MS users pissed of at Microsoft for exposing them to crackers. It looks like the average @Home user gets port scanned a couple times a day (and not by @Home admins). How much do you think that's going to accelerate when just about everybody has an insecure (MS) web server running?
I can think of yet another niche-market distro. We have the source, thus we have the power. A Linux distro tailored to the needs of the education community - one that concentrates on interactive teaching and the like, while maintaining Linux's openness and security.
This is why I see Linux (and open source in general) as being so important and useful for education. You have access to all levels of abstraction, and yet you have the ability to also use those abstractions. Part of the growth in computer education is being able to gradually peel back the layers of abstraction that separate us from the machine. As the student becomes capable of dealing with more depth and breadth to his (her) education, he (she) can peel back another layer of abstraction.
K, so what do we need? techies, educators and techie-educators. If you can express it, we can build it.
Hope I get this one on meta-mod tomorrow... (NB. It would be fair to mod my comment down, but this one was clearly on-topic.)
This is one of the biggest problem I have with MacOS and Windows. I don't care how badly coded the app is, user mode apps should not bring down the system. The perceived reliability and stability of an operating system should never be judged by how it treats well-behaved apps. The perceived stability and reliability of an operating system must be judged by how it deals with poorly behaved apps and unforeseen circumstances. Think about it - any teacher can teach well-behaved, respectful kids. It takes a special kind to teach poorly behaved kids.
Take a look at the other books (Speaker for the Dead, Xenocide and Children of the Mind). Ender's Game was rather juvenile, but the other three were definitely not. I can't speak for Ender's Shadow (I just started reading it), but I certainly enjoyed the first four. Card definitely takes a look at some interesting subjects in Speaker, Xenocide, and Children - further, they're not shoot-em-up action books like EG.
The foreword to Ender's Shadow describes how the project was initially undertaken, and how Mr. Card "selfishly" (his words, not mine) swiped the offer to write ES from another author...
So use another bootloader... When we've got very friendly, nice bootloaders like GRUB (my personal favourite), GAG (nice and graphical), etc... there's no reason to be bitching about LILO.
Allow slashdot to become useful - let us moderate down the features themselves - not just the repsonses!!!
I see... The attempt at humour was made painfully obvious by the statement - "The hackers wrote of their own "rooting" exploits (that is, hacking the root directory of a server)"
This is _not_ only about piracy. A big part of this is being able to _view_ DVDs on a system of _our_ choice. This is about _not_ being forced to use proprietary software for any purpose - entertainment or otherwise.
This is a side effect of using just about any file system. Most file systems do not zero out a file's blocks when it gets deleted, but rather add them to a free list. The difference here is in the free block allocation policy supported by the particular fs. If you're relying on rm or del or whatever to eliminate all traces of a file, you're being very foolish.
"calming the nerves of worried parents" is absolutely NOT "a perfectly valid reason" to invade the privacy of kids with unproven technology. It is not a valid reason to spend time and money categorizing, separating and splitting kids into profile groups, rather than actually providing them with the support they need to be unique individuals.
Now, lets talk about the science (or lack thereof), here. This software is developed from interviews with violent kids. This means that its use as a predictor is problematic. The statement "most violent kids exhibit characteristics a, b, and c - therefore most kids that exhibit characteristics a, b, and c are violent" is as much a logical fallacy as saying that "most rapists are male - therefore most males are rapists."
It is important to note that the vast majority of Microsofr's donations are software. While most other software companies assess value of donations as either cost or wholesale value, Microsoft assesses all of their donations at _retail_value_. Any comparisons between Microsoft and other software companies in this field are irrelevant.
On a quick scan of both Microsoft's and Red Hat's site, it becomes apparent that there are about as many NT4-PostSP5 hotfixes as there are software errata listed at Red Hat's site. When you add in the errata for Exchange, IIS, Word, and other MS apps, Microsoft's burying themselves (and their customers) in bug fixes much faster than Red Hat is. Moreover, Red Hat has an easy method of obtaining info about errata *before*they*become*an*issue*. Microsoft's method basically consists of - if you've got a problem, search Knowledge Base - there might be a fix.
Come on. Everybody has bugs. What matters is how you deal with the bugs. Do you sweep them under the carpet like Microsoft? Or should you stand up point them out, and point to the fixes?
By the time most Linux vulnerabilities could hit the press, they've been fixed. This is an example of that. The crond hole has been fixed. By the time a Windows exploit hits the press, they're determining how much to charge us for the fix. Put simply, bugs get fixed faster with open source. You don't have to wait on a corporate timetable to get fixes that you _need_.
Mail order fraud is more challenging than you describe. Most credit card companies require that products purchased by mail order must be shipped to the billing address for the card. Admittedly, this does not apply to pr0n sites ;-) and other situations where the product need not be shipped. As a result, if I get your card number and expiry, I can buy something and have it shipped to you, I can subscribe to some pay website, or I can try to change your billing information so that I can actually get product shipped to me.
I would imagine that changing someone's billing address is quite easy... Whenever I've moved, the only authentication they've tried is phone numbers, birthdate, card number, name, old address. Would I do this? No - it leaves a trail pointing at me.
Since Microsoft relies exclusively on prevntative security measures to protect the private key(s) from being compromised, any measure which would reduce this security is unacceptable. More copies of the key makes it more difficult to maintain security and ensure that the private key is not compromised (as opposed to lost).
Microsoft is right about one of the issues here. A backup key is useful for key loss. If the private primary key were lost in fire, flood, stolen, etc. and there were no backup, CryptoAPI would very quickly cease to be useful, as no newer encryption methods would be able to be implemented under CryptoAPI. This means taht a backup key is useful.
Microsoft has, apparently done some things wrong here, though. As I understand it, the reason the primary key can't be changed is that the default CSP is signed with the primary key, and changing this key would render the system unusable. Had a necessary component been signed with both the backup key and the primary key, and both signatures tested for that component, it would become more difficult to change the keys and retain a useful system.
The other thing that Microsoft has messed up here is the issue of key compromise. Microsoft has focussed entirely on prevention on this issue, excluding contingencies where the keys are compromised. They speak of hardware, software and physical security to prevent key compromise, but there does not appear to be a mechanism for key revocation if and when these security mechanisms fail, and the key _is_ compromised.
But the government _does_not_ sell your SSN and related information to corporations. Some company can't just buy a copy of your tax return. Companies _can_ buy copies of your driver's license and related information.
They are selling the pictures and information to _private_companies_. Telecheck is a private company which provides check verification and insurance to merchants. I'm sorry, but I have a problem with the government selling my information and my picture to some private business. (Much as I have a problem with them providing my mailing address to private parking companies...)
There is much more to this issue than just the government. This is about strengthening the stranglehold that big business has over the United States.
Claiming that Linux is a waste of money while MacOS is not is ludicrous. The fact is, like it or not, Linux is a presence in more businesses than MacOS is, whether we're talking about desktop or server use. To marginalize this OS on _business_software_, particularly for small to medium sized businesses, is a dangerous move. Like it or not, for that segment of the market, MacOS is a _much_ riskier proposition with a _much_ bleaker outlook than Linux.
That said, I've never been a big fan of Intuit's products anyway. I'd much rather see Computer Associates port their low-mid end packages (like Simply Accounting) over. I'd also love to see Canadian tax-prep software on Linux as well... Oooh. I think I smell an open source project coming on...
Okay, for those of you who've read this already, this is my semi-regular rant on Windows, installing apps, and shared libraries. If you've read it already, skip it.
>Yes, you need to reboot much more when
>installing or changing software than Linux, but
>you can always say no and stop and restart the
>service. (Same basic prinicple on Linux)
No. These are very different issues. When it comes to shared libraries, Windows has dropped the ball. Shared libraries mean that some other application (perhaps even the OS) may use them and have them locked. When you are asked to reboot, it is because some shared libraries couldnot be copied because they were locked. These shared libraries will not get copied until you reboot. More concerning, however, is the uninstall. If the uninstall can't delete something (say X.DLL), it marks it for deletion at next reboot, and asks if you want to reboot. If you say no, and go install something else in the meantime, which happens to try to install X.DLL, things will be happy until you reboot, when X.DLL is deleted (as the uninstaller told it to do).
Linux deals with updating shared libraries, partially as a result of naming conventions, and partially as a result of filesystem capabilities.
In Windows, if a program is using a particular shared library, you cannot remove that library file from the filesystem. It is locked. Under Linux, what is locked are the inodes. It is irrelevant whether or not the file has a filename linked to it. When I use the unlink() system call (there's a reason there's no delete_file() sys call), we unlink the file name from the inodes storing the file, preventing more apps from opening it. Only when the file's usage count drops to 0 is it removed from the fs. This means that I can create a new file, with the same name as the one I just unlinked, without affecting the performance of any applications using the original file. I can replace shared libraries without having to stop the application using the shared library.
Further, most Linux applications don't install new shared libraries when they install themselves. It would be seen as very poor practice for some little app to install a new version of glibc as part of its own installation. Yet, we accept this as normal in the NT world.
Even better, Linux allows multiple versions of a library to coexist on a system, using version numbers. Under Windows, you can't even get Microsoft's own ODBC 3.51 to coexist with ODBC 4.0.