My UNIX history is pretty good... it's not "history" for me: I provided some corrections for Peter Salus' UNIX timeline.
Mono applications on Linux only use the ECMA APIs, which are not covered by any patents and which are explicitly open.
And if Microsoft changes CLI they'll have to track it anyway, to retain compatibility with Silverlight, so it's really no more open than OOXML... standardizing a proprietary system doesn't make it less proprietary if the proprietor retains control.
Yeh, I have a stake. I have a stake in open systems. If a system is under the thumb of a single organization, it's not open even if you have the source code... whether it's Mono, Java, or GCC.
Any that have been viable and widely used products in the past decade? Legacy software in maintainance mode like the Mentex RSX-11, the DECsystem-20 emulators, or Unisys' emulators for Burroughs and Sperry mainframes don't count.
IBM has a boatload of operating systems under VM, and whatever they call the AS400 this week. You can run Linux in parallel with anything else under VM, and last I looked the AS400 hardware and AIX hardware were basically the same.
The original POSIX subsystem was deliberately crippled by Microsoft, and the crippled version is the only one shipping with any non-server version of Windows. The updated version, Interix, was written by Softway Systems using a licensed copy of the original POSIX source code. They were bought by Microsoft (this was announced during a Usenix convention where Steve Walli showed himself remarkably capable of dissimulation, because I'd asked him about the possibility at the beginning of the conference and he gave me NO indication that it was about to happen) and the prodcut was apparently shelved then finally came out as a free-but-unadvertised download. Yes, I've used it myself, it's great, but it doesn't count as something shipping with Windows even now.
Obviously when I wrote "developed after UNIX" I mean "the development process started after UNIX".
I'm exquisitely familiar[*] with The Univac/Sperry/Unisys 1100/2200 series. I'm familiar with MCP inly through the literature, admittedly. And I know why SX1100 was called "SUX 1100". By 1997 the writing was on the wall even for the Swift, and for some years now Exec and MCP have only been sold under emulation.
I suppose I could qualify that with "by 1997 there was no operating system that wasn't clearly a lame duck that..." but really, you're stretching here.
They've said explicitly that OOXML is easier for them to work with then OpenOffice's format. Period. It basically comes down to laziness, one of the primary virtues in hackerdom. So, lay off 'em, unless you're already in your editor writing code.
I'm already in my editor writing code, so I don't have to lay off them. Laziness in the defense of proprietary systems is no virtue.
It's free as in beer, yes. That's the "free" that matters to the the masses.
Free as in speech? Halfway there. It's open source, but if it moves away from open systems that's not good enough. CLI, like OOXML, is a proprietary system, under the direct control of one company... and one that has proven actively hostile to Open Source. I was already unhappy about the spread of Java and Sun is nowhere near the threat that Microsoft is.
And what makes you think that gnome is even remotely compatible with windows outside of mono?
It's not... yet. But that's the way it's trending, and Mono is part of it.
And for the last two decades, people were pushing AT&T stuff, some of it patented.
AT&T donated the key patent for UNIX (the setuid patent) into the public domain. The UNIX APIs were designed to be independent of the underlying hardware and implementation, and they never made any attempt to enforce any potential copyrights on the UNIX programmer's manuals. The only product I know of that felt it necessary to avoid using the precise APIs described in the manual, ever, was Idris... presumably because it was by a former Bell Labs employee. There are, so far as I can tell, only two significant operating systems started after the publication of the 1976 Bell System Technical Journal that were not based primarily on the UNIX "software tools" environment: Mac OS, and Windows... and both of those were instead based on the Xerox environment. I'm not counting MS-DOS, because it was an 8086 port of CP/M by Tim Patterson of Seattle Computer Systems, and starting with MS-DOS 2.x it was increasingly adopting UNIX APIs.
By 1987 (two decades ago) the UNIX environment had been re-implemented dozens of times, both standalone and hosted on top of other operating systems. By 1997 (one decade ago) there was no operating system in the world that wasn't either UNIX-based, transitioning to UNIX, or shipping with a functional hosted UNIX environment... other than Windows.
And by that time AT&T had sold all their rights in UNIX to Novell, who had publicly disclaimed any intellectual property in the APIs.
If there was any remaining danger in these APIs, the results of the Caldera (the new SCO) suit have completely defanged it.
So far, there is no indication that there is any more risk to Mono from Microsoft than there was to Linux from AT&T.
On the one hand we have a set of APIs that were already in the public domain both because of explicit donation and due to being published without copyright notice before the US joined the Berne convention, and have since been been proven safe to use, and on the other hand we have a set of APIs that are actively controlled by a company that has a history of using submarine patents, and who is currently attempting to monetize them... with some success.
If you can't see there is a difference there you're deliberately not looking at it.
The handful of people I know who use newer versions of office are pissed off everytime they forget to "save as" into an older format and that is because of where I work. All work computers are setup with an older version of office (can't remember the version number off the top of my head). These people usually end up switching to an older version of office and I have been able to con some of them into using openoffice.org.
Every time there's a new version of Office Microsoft has made it impossible to maintain a heterogenous environment. You either have to stick with the older apps, or you have to do a mass upgrade... because even if you don't use any of the newer features and even when they've used allegedly backwards compatible formats Word uses them in the saved files.
OOXML is just the latest one-way format change. They're not doing anything to standardise on an open and portable format, they're just taking advantage of the standards process to push through another forced mass upgrade for Office.
You don't have to clone Windows to produce a viable alternative platform... and in fact if you do end up cloning Windows you'll have eliminated everything that makes the result "alternative".
Why should I care about a Linux-based system where your applications are written in.NET and your formats and user interface and APIs are driven by compatibility with Windows? If I want to run an OS that's compatible with Windows, I've already got that option.
What makes Linux an alternative is that it's an Open Systems environment that happens to be Open Source as well. That applications written for it aren't locked in to Linux, they'll run on any Open Systems platform. If the interfaces and protocols it uses are Microsoft's, then why should anyone care whether it's got a Linux or an NT kernel under the hood?
A quick hack would require a valid CD key before you download PS (the CD key would come from a website purchase). The installer would prompt you for the CD key and check it against the server. Then the CD key would be used to encrypt the download and the deb would be sent to you.
That's basically how it currently works for a lot of commercial software, except that the package you get installs directly and doesn't need you to have a separate installer already sitting on the system, and they take care of the CD key stuff when you install the software the first time, so their upgrader works invisibly in the background. The thing is that you'll need to do that for Adobe's server to update Photoshop, and for Autodesk's server to download Maya, and so on... unless you go with my idea of having someone like Amazon doing it...
Of course this would be done against a "commercial software" repo server
Right.
So basically you just described the same thing that I suggested... you just described "copy protection, DRM, and all that crap" in more detail than I did, as if that somehow made it go away.
There is no technical or business problem with making this work when it becomes common to distribute commercial software to Ubuntu.
The question is, will it become common to distribute commercial software to Ubuntu without "all this" being in place?
Adobe could easily replace their current distribution with a repository with no change in their copy protection, payment, or other models. And, as a bonus, they'd get much easier software upgrades and maintenance.
Only if they could get the 90% of the market running Windows and the 7% running OSX to go along with the 3% running Linux.
The reason they don't is simply that they haven't ported any of their for-pay software to Linux.
And I'm sure the idea that they should change their business model for 3% of the market has nothing to do with that.
Centralized package management for open source products was a great idea when the FreeBSD project started the Ports system, back when Linux was still catch-as-catch-can. But it depends on a common trust model... even interfering with that model slightly, like RHEL does, orlike Sun used to with Java, makes it a pain. Depending on dozens of companies maintaining their own repositories with their own access controls just isn't going to fly... the operation will end up as manual as it is now.
If you commit patent infringement, of course, you have to pay damages; willful infringement just triples them.
Oops, my bad. The point is that willful infringement makes things worse.
If it's patented, distros won't incorporate it.
I think you're repeating my point there. If you use this documentation that Microsoft is reluctantly shipping, you're reducing the value of your software to the open source community. Thus what Microsoft's actually released has no value at all to open source. Which is my point.
1. That one is a *Microsoft* problem, acknowledged as such by them. 2. Was Microsoft being unreasonably slammed "around here" (in/.) about it? 3. The people worried about what APE might be doing to applications obviously haven't looked at what antivirus software does.
Seriously... antivirus software modifies every application and installs drivers and does everything APE does, and much much more. It's the most dangerous software you can run on your computer, and I personally won't run it even on Windows. I understand there are some people who actually run it on OS X, and I sincerely hope they're keeping good backups.
You do realise that Adobe has software packaged in the Ubuntu repository?
Adobe has freeware there. Acrobat Reader is a teaser product that they're happy to bundle with anything you let them... to give you a reason to buy their full-on Acrobat software... which they don't include. They don't have Photoshop or their other real commercial software there either.
Adobe is not going to ship Photoshop through a distro repository until someone like Amazon is doing their own distro with support for credit card payments in AZpkg, copy protection, maybe even DRM and all the other crap that (justifiably) makes most Linux users wince.
The functionality makes APE work has been REMOVED on Leopard which means it will simply IGNORE APE, won't load it.
Input Managers have not been removed in Leopard. It simply asks the user if they want to run them, once. And rosyna has already said that APE is not dependent on the Input Manager mechanism. Input Managers are a red herring.
I won't argue with you about whether Apple was right to try and dissuade people from using Input Managers, because I happen to agree with you, but it's beside the point.
I won't argue with you about whether Apple is responsible for people needing to use APE, because I agree with you: I use APE to remove some of Apple's annoying bling myself. But, again, it's beside the point.
Unsanity warns you that they are making deep modifications to the user interface and patching applications dynamically when they are loaded. They are not "just apps", they are not even like Parallels which does come with a kernel extension... because it runs as a regular application otherwise. Not to mention that Apple considers Parallels a key player, but has a definite problem with Unsanity.
I'm not saying that people who are having this problem are morons, but I don't think they should be so quick to blame Apple for this... unless you're suggesting that Apple actively broke APE or something?
You have to be carefull with terms like OS and application, they mean different things to different people.
Indeed, and I have already seen messages claiming that Unsanity is patching the kernel, which why as a side issue I am making sure that I'm correcting that misapprehension when it looks like it may be made.
The real issue here is that Apple has been forcing bling down our throats, and slowing down the GUI, and the only way to get rid of it is through programs like Unsanity's.
Every time MS releases a significant upgrade there are articles on Slashdot talking about problems with the latest games, lists of third-party applications a service pack "has broken", drivers that don't work with the new OS, etc.
And there will be more apps that are broken on Leopard, and there will be stories about them. And when Apple is responsible, then I'll be there throwing bricks too. But this one isn't Apple's fault, you know that by now if you're actually following it, so don't try and claim that Apple should be slammed for it out of some misplaced desire for "balance".
If you want to install commercial software, you subscribe to it through the package manager, just like anything else.
I can see Blizzard or Adobe going for that. Not.
That attitude towards commercial software is why Linux is not going to develop a commercial ecosystem.
Maybe that's true for a small number of people who need specialty apps.
That would be just about everyone. People buy computers to do the "specialty" stuff they need to do win their life.
Almost all the commercial apps I have ever bought for OS X turned out to be either useless or were there to fill in some functionality that is included for free in recent Linux distributions.
I've used Gimp, and I've used Photoshop, and I know which one is "useless". And it's not the one included for free in recent Linux distributions.
In fact just last week I was looking for an excuse not to pay the bucks to upgrade to CS3, and looked at the available open source, shareware, and commercial painting applications.
There were three open source apps that ran under X11. Gimp and two derivatives of Gimp. None of them were worth pouring out of a boot. Even the cheapest shareware OS X application was better than them.
The areas where the free UNIX environment wins are really important, yes, but they're really "specialty apps" for most people. THe free UNIX ecosystem produces the best "Tools for Nerds" (to borrow Slashdot's slogan), but when it comes to other specialties for the majority of the population who aren't 100% hardcore computer geeks it really doesn't work as well.
There's some really good open source apps of OSX, though. But they don't run on Linux and they're built by people who don't think that "Tools for Nerds" are all you need.
Windows XP SP2 failing miserably as people tried to upgrade [...] third party patched ones.
I'd call that pilot error too.
[...] as people tried to upgrade malware infected systems [...]
I'd almost call that pilot error as well, except that Microsoft ships the biggest malware target (the Windows HTML control) and trains people to get infected by malware by popping up warning dialogs all the time so that when they finally get that "Blaster wants to infect your system. (Infect me) (Cancel)" dialog they click on "Infect me" by reflex.
Of course, I must admit Apple trains people to patch their system by refusing to let people turn off bling. But at least people know what extensions they've installed and can remove them before a major upgrade or do what I do and not upgrade for the "dotzero" release.
dpkg doesn't upgrade anything that you've installed that didn't come from the dpkg repositories. If you installed anything yourself, you're out of luck.
The only reason this isn't an issue on Linux is that there isn't any ecosystem of commercial software... there's a handful of commercial applications you can buy for Linux, and pretty much all of them are server apps. And it's that commercial ecosystem that really makes OSX worthwhile... if it was just a nice-looking free-UNIX distribution with alll the software downloaded via Fink or DarwinPorts nobody would use it.
Yeah, and they use documented, official, supported methods of extending the OS, specifically so that crazy unpredictable behavior is avoided even when the OS is updated.
Well, I don't know about you, but I wasn't planning to upgrade to Leopard without removing my enhanced drivers and kernel extensions either.
But my core point is that if Apple didn't insist on forcing bling on us, people wouldn't go to this kind of extreme just to remove the bling.
You have a bunch of people who don't have anything to do with the author of an application writing code that mass-modifies dozens of applications, libraries, etc., essentially doing binary patching on the running OS.
I've done that as well. I mean, binary patching the actual OS, not just applications running in userland. Sometimes you gotta do it.
If you knew what GNU libc does to try and avoid having to make people rebuild applications when upgrading libraries, you'd run screaming. They have code in there to look for libraries at runtime and dynamically load different variants of other libraries depending on what you're using and what you have installed. The glibc team has people who do nothing but look for cases where they have to adapt for different libraries and different kernel versions.
The reason that you don't have more of a problem on Linux is that there's no central Steve Jobs for Linux who dictates the way the GUI works, so if you don't like the way Enlightenment or fvwm or Windowmaker behaves, you can change it. The downside of this is that there's no single framework you can modify or replace to make global changes. There used to be, back when everyone used Athena Widgets, and you could replace libXaw with libXaw95 to get a Windows 95 look, or with libXawSTeP to get a NeXTSTeP look. Now, instead, you get Battluin GUIs between the Gnome and KDE yobbos.
And there's all kinds of Windows hacks that do similar stuff to APE, from development tools to simple user interface enhancements. And, yes, they can cause problems and break in new versions of the OS.
What Unsanity has done is to create a framework that makes this kind of thing relatively safe compared to having everyone build their own. Unfortunately since they're not at Apple or someone that Apple is willing to support (because they are undoing the things that The Steve has decreed) there's an unhealthy passive-aggressive relationship between Apple and Unsanity that doesn't exist between (say) Debian and the glibc team.
And, yes, they should be disabling themselves on upgrade. And Apple should look at the things that people are using Haxies for and make the things they are trying to get rid of optional.
The other thing is, on Windows people simply don't put their trust in having an upgrade work. They do clean installs. And they wait on upgrading Windows until this kind of thing gets shaken out.
Slashdot Mirror
My UNIX history is pretty good... it's not "history" for me: I provided some corrections for Peter Salus' UNIX timeline.
Mono applications on Linux only use the ECMA APIs, which are not covered by any patents and which are explicitly open.
And if Microsoft changes CLI they'll have to track it anyway, to retain compatibility with Silverlight, so it's really no more open than OOXML... standardizing a proprietary system doesn't make it less proprietary if the proprietor retains control.
Yeh, I have a stake. I have a stake in open systems. If a system is under the thumb of a single organization, it's not open even if you have the source code... whether it's Mono, Java, or GCC.
Any that have been viable and widely used products in the past decade? Legacy software in maintainance mode like the Mentex RSX-11, the DECsystem-20 emulators, or Unisys' emulators for Burroughs and Sperry mainframes don't count.
IBM has a boatload of operating systems under VM, and whatever they call the AS400 this week. You can run Linux in parallel with anything else under VM, and last I looked the AS400 hardware and AIX hardware were basically the same.
What have I missed?
The original POSIX subsystem was deliberately crippled by Microsoft, and the crippled version is the only one shipping with any non-server version of Windows. The updated version, Interix, was written by Softway Systems using a licensed copy of the original POSIX source code. They were bought by Microsoft (this was announced during a Usenix convention where Steve Walli showed himself remarkably capable of dissimulation, because I'd asked him about the possibility at the beginning of the conference and he gave me NO indication that it was about to happen) and the prodcut was apparently shelved then finally came out as a free-but-unadvertised download. Yes, I've used it myself, it's great, but it doesn't count as something shipping with Windows even now.
Obviously when I wrote "developed after UNIX" I mean "the development process started after UNIX".
I'm exquisitely familiar[*] with The Univac/Sperry/Unisys 1100/2200 series. I'm familiar with MCP inly through the literature, admittedly. And I know why SX1100 was called "SUX 1100". By 1997 the writing was on the wall even for the Swift, and for some years now Exec and MCP have only been sold under emulation.
I suppose I could qualify that with "by 1997 there was no operating system that wasn't clearly a lame duck that..." but really, you're stretching here.
[*] Read as "exquisitely painfully familiar".
They've said explicitly that OOXML is easier for them to work with then OpenOffice's format. Period. It basically comes down to laziness, one of the primary virtues in hackerdom. So, lay off 'em, unless you're already in your editor writing code.
I'm already in my editor writing code, so I don't have to lay off them. Laziness in the defense of proprietary systems is no virtue.
The driving difference is that it's free.
It's free as in beer, yes. That's the "free" that matters to the the masses.
Free as in speech? Halfway there. It's open source, but if it moves away from open systems that's not good enough. CLI, like OOXML, is a proprietary system, under the direct control of one company... and one that has proven actively hostile to Open Source. I was already unhappy about the spread of Java and Sun is nowhere near the threat that Microsoft is.
And what makes you think that gnome is even remotely compatible with windows outside of mono?
It's not... yet. But that's the way it's trending, and Mono is part of it.
And for the last two decades, people were pushing AT&T stuff, some of it patented.
AT&T donated the key patent for UNIX (the setuid patent) into the public domain. The UNIX APIs were designed to be independent of the underlying hardware and implementation, and they never made any attempt to enforce any potential copyrights on the UNIX programmer's manuals. The only product I know of that felt it necessary to avoid using the precise APIs described in the manual, ever, was Idris... presumably because it was by a former Bell Labs employee. There are, so far as I can tell, only two significant operating systems started after the publication of the 1976 Bell System Technical Journal that were not based primarily on the UNIX "software tools" environment: Mac OS, and Windows... and both of those were instead based on the Xerox environment. I'm not counting MS-DOS, because it was an 8086 port of CP/M by Tim Patterson of Seattle Computer Systems, and starting with MS-DOS 2.x it was increasingly adopting UNIX APIs.
By 1987 (two decades ago) the UNIX environment had been re-implemented dozens of times, both standalone and hosted on top of other operating systems. By 1997 (one decade ago) there was no operating system in the world that wasn't either UNIX-based, transitioning to UNIX, or shipping with a functional hosted UNIX environment... other than Windows.
And by that time AT&T had sold all their rights in UNIX to Novell, who had publicly disclaimed any intellectual property in the APIs.
If there was any remaining danger in these APIs, the results of the Caldera (the new SCO) suit have completely defanged it.
So far, there is no indication that there is any more risk to Mono from Microsoft than there was to Linux from AT&T.
On the one hand we have a set of APIs that were already in the public domain both because of explicit donation and due to being published without copyright notice before the US joined the Berne convention, and have since been been proven safe to use, and on the other hand we have a set of APIs that are actively controlled by a company that has a history of using submarine patents, and who is currently attempting to monetize them... with some success.
If you can't see there is a difference there you're deliberately not looking at it.
The handful of people I know who use newer versions of office are pissed off everytime they forget to "save as" into an older format and that is because of where I work. All work computers are setup with an older version of office (can't remember the version number off the top of my head). These people usually end up switching to an older version of office and I have been able to con some of them into using openoffice.org.
Every time there's a new version of Office Microsoft has made it impossible to maintain a heterogenous environment. You either have to stick with the older apps, or you have to do a mass upgrade... because even if you don't use any of the newer features and even when they've used allegedly backwards compatible formats Word uses them in the saved files.
OOXML is just the latest one-way format change. They're not doing anything to standardise on an open and portable format, they're just taking advantage of the standards process to push through another forced mass upgrade for Office.
You don't have to clone Windows to produce a viable alternative platform... and in fact if you do end up cloning Windows you'll have eliminated everything that makes the result "alternative".
.NET and your formats and user interface and APIs are driven by compatibility with Windows? If I want to run an OS that's compatible with Windows, I've already got that option.
Why should I care about a Linux-based system where your applications are written in
What makes Linux an alternative is that it's an Open Systems environment that happens to be Open Source as well. That applications written for it aren't locked in to Linux, they'll run on any Open Systems platform. If the interfaces and protocols it uses are Microsoft's, then why should anyone care whether it's got a Linux or an NT kernel under the hood?
A quick hack would require a valid CD key before you download PS (the CD key would come from a website purchase). The installer would prompt you for the CD key and check it against the server. Then the CD key would be used to encrypt the download and the deb would be sent to you.
That's basically how it currently works for a lot of commercial software, except that the package you get installs directly and doesn't need you to have a separate installer already sitting on the system, and they take care of the CD key stuff when you install the software the first time, so their upgrader works invisibly in the background. The thing is that you'll need to do that for Adobe's server to update Photoshop, and for Autodesk's server to download Maya, and so on... unless you go with my idea of having someone like Amazon doing it...
Of course this would be done against a "commercial software" repo server
Right.
So basically you just described the same thing that I suggested... you just described "copy protection, DRM, and all that crap" in more detail than I did, as if that somehow made it go away.
There is no technical or business problem with making this work when it becomes common to distribute commercial software to Ubuntu.
The question is, will it become common to distribute commercial software to Ubuntu without "all this" being in place?
Adobe could easily replace their current distribution with a repository with no change in their copy protection, payment, or other models. And, as a bonus, they'd get much easier software upgrades and maintenance.
Only if they could get the 90% of the market running Windows and the 7% running OSX to go along with the 3% running Linux.
The reason they don't is simply that they haven't ported any of their for-pay software to Linux.
And I'm sure the idea that they should change their business model for 3% of the market has nothing to do with that.
Centralized package management for open source products was a great idea when the FreeBSD project started the Ports system, back when Linux was still catch-as-catch-can. But it depends on a common trust model... even interfering with that model slightly, like RHEL does, orlike Sun used to with Java, makes it a pain. Depending on dozens of companies maintaining their own repositories with their own access controls just isn't going to fly... the operation will end up as manual as it is now.
If you commit patent infringement, of course, you have to pay damages; willful infringement just triples them.
Oops, my bad. The point is that willful infringement makes things worse.
If it's patented, distros won't incorporate it.
I think you're repeating my point there. If you use this documentation that Microsoft is reluctantly shipping, you're reducing the value of your software to the open source community. Thus what Microsoft's actually released has no value at all to open source. Which is my point.
1. That one is a *Microsoft* problem, acknowledged as such by them. /.) about it?
2. Was Microsoft being unreasonably slammed "around here" (in
3. The people worried about what APE might be doing to applications obviously haven't looked at what antivirus software does.
Seriously... antivirus software modifies every application and installs drivers and does everything APE does, and much much more. It's the most dangerous software you can run on your computer, and I personally won't run it even on Windows. I understand there are some people who actually run it on OS X, and I sincerely hope they're keeping good backups.
You do realise that Adobe has software packaged in the Ubuntu repository?
Adobe has freeware there. Acrobat Reader is a teaser product that they're happy to bundle with anything you let them... to give you a reason to buy their full-on Acrobat software... which they don't include. They don't have Photoshop or their other real commercial software there either.
Adobe is not going to ship Photoshop through a distro repository until someone like Amazon is doing their own distro with support for credit card payments in AZpkg, copy protection, maybe even DRM and all the other crap that (justifiably) makes most Linux users wince.
The functionality makes APE work has been REMOVED on Leopard which means it will simply IGNORE APE, won't load it.
Input Managers have not been removed in Leopard. It simply asks the user if they want to run them, once. And rosyna has already said that APE is not dependent on the Input Manager mechanism. Input Managers are a red herring.
I won't argue with you about whether Apple was right to try and dissuade people from using Input Managers, because I happen to agree with you, but it's beside the point.
I won't argue with you about whether Apple is responsible for people needing to use APE, because I agree with you: I use APE to remove some of Apple's annoying bling myself. But, again, it's beside the point.
No, really, the Windows kernel and userland are just as distinct as any other OS.
Making changes in the GUI, which is what APE does, rarely if ever involves mods to the kernel in any OS.
Unsanity warns you that they are making deep modifications to the user interface and patching applications dynamically when they are loaded. They are not "just apps", they are not even like Parallels which does come with a kernel extension... because it runs as a regular application otherwise. Not to mention that Apple considers Parallels a key player, but has a definite problem with Unsanity.
I'm not saying that people who are having this problem are morons, but I don't think they should be so quick to blame Apple for this... unless you're suggesting that Apple actively broke APE or something?
You have to be carefull with terms like OS and application, they mean different things to different people.
Indeed, and I have already seen messages claiming that Unsanity is patching the kernel, which why as a side issue I am making sure that I'm correcting that misapprehension when it looks like it may be made.
The real issue here is that Apple has been forcing bling down our throats, and slowing down the GUI, and the only way to get rid of it is through programs like Unsanity's.
Every time MS releases a significant upgrade there are articles on Slashdot talking about problems with the latest games, lists of third-party applications a service pack "has broken", drivers that don't work with the new OS, etc.
And there will be more apps that are broken on Leopard, and there will be stories about them. And when Apple is responsible, then I'll be there throwing bricks too. But this one isn't Apple's fault, you know that by now if you're actually following it, so don't try and claim that Apple should be slammed for it out of some misplaced desire for "balance".
If you want to install commercial software, you subscribe to it through the package manager, just like anything else.
I can see Blizzard or Adobe going for that. Not.
That attitude towards commercial software is why Linux is not going to develop a commercial ecosystem.
Maybe that's true for a small number of people who need specialty apps.
That would be just about everyone. People buy computers to do the "specialty" stuff they need to do win their life.
Almost all the commercial apps I have ever bought for OS X turned out to be either useless or were there to fill in some functionality that is included for free in recent Linux distributions.
I've used Gimp, and I've used Photoshop, and I know which one is "useless". And it's not the one included for free in recent Linux distributions.
In fact just last week I was looking for an excuse not to pay the bucks to upgrade to CS3, and looked at the available open source, shareware, and commercial painting applications.
There were three open source apps that ran under X11. Gimp and two derivatives of Gimp. None of them were worth pouring out of a boot. Even the cheapest shareware OS X application was better than them.
The areas where the free UNIX environment wins are really important, yes, but they're really "specialty apps" for most people. THe free UNIX ecosystem produces the best "Tools for Nerds" (to borrow Slashdot's slogan), but when it comes to other specialties for the majority of the population who aren't 100% hardcore computer geeks it really doesn't work as well.
There's some really good open source apps of OSX, though. But they don't run on Linux and they're built by people who don't think that "Tools for Nerds" are all you need.
Windows XP SP2 failing miserably as people tried to upgrade [...] third party patched ones.
I'd call that pilot error too.
[...] as people tried to upgrade malware infected systems [...]
I'd almost call that pilot error as well, except that Microsoft ships the biggest malware target (the Windows HTML control) and trains people to get infected by malware by popping up warning dialogs all the time so that when they finally get that "Blaster wants to infect your system. (Infect me) (Cancel)" dialog they click on "Infect me" by reflex.
Of course, I must admit Apple trains people to patch their system by refusing to let people turn off bling. But at least people know what extensions they've installed and can remove them before a major upgrade or do what I do and not upgrade for the "dotzero" release.
dpkg doesn't upgrade anything that you've installed that didn't come from the dpkg repositories. If you installed anything yourself, you're out of luck.
The only reason this isn't an issue on Linux is that there isn't any ecosystem of commercial software... there's a handful of commercial applications you can buy for Linux, and pretty much all of them are server apps. And it's that commercial ecosystem that really makes OSX worthwhile... if it was just a nice-looking free-UNIX distribution with alll the software downloaded via Fink or DarwinPorts nobody would use it.
Yeah, and they use documented, official, supported methods of extending the OS, specifically so that crazy unpredictable behavior is avoided even when the OS is updated.
Well, I don't know about you, but I wasn't planning to upgrade to Leopard without removing my enhanced drivers and kernel extensions either.
But my core point is that if Apple didn't insist on forcing bling on us, people wouldn't go to this kind of extreme just to remove the bling.
You have a bunch of people who don't have anything to do with the author of an application writing code that mass-modifies dozens of applications, libraries, etc., essentially doing binary patching on the running OS.
I've done that as well. I mean, binary patching the actual OS, not just applications running in userland. Sometimes you gotta do it.
If you knew what GNU libc does to try and avoid having to make people rebuild applications when upgrading libraries, you'd run screaming. They have code in there to look for libraries at runtime and dynamically load different variants of other libraries depending on what you're using and what you have installed. The glibc team has people who do nothing but look for cases where they have to adapt for different libraries and different kernel versions.
The reason that you don't have more of a problem on Linux is that there's no central Steve Jobs for Linux who dictates the way the GUI works, so if you don't like the way Enlightenment or fvwm or Windowmaker behaves, you can change it. The downside of this is that there's no single framework you can modify or replace to make global changes. There used to be, back when everyone used Athena Widgets, and you could replace libXaw with libXaw95 to get a Windows 95 look, or with libXawSTeP to get a NeXTSTeP look. Now, instead, you get Battluin GUIs between the Gnome and KDE yobbos.
And there's all kinds of Windows hacks that do similar stuff to APE, from development tools to simple user interface enhancements. And, yes, they can cause problems and break in new versions of the OS.
What Unsanity has done is to create a framework that makes this kind of thing relatively safe compared to having everyone build their own. Unfortunately since they're not at Apple or someone that Apple is willing to support (because they are undoing the things that The Steve has decreed) there's an unhealthy passive-aggressive relationship between Apple and Unsanity that doesn't exist between (say) Debian and the glibc team.
And, yes, they should be disabling themselves on upgrade. And Apple should look at the things that people are using Haxies for and make the things they are trying to get rid of optional.
The other thing is, on Windows people simply don't put their trust in having an upgrade work. They do clean installs. And they wait on upgrading Windows until this kind of thing gets shaken out.