One of the barriers is hardware compatibility. I've invested a lot of money on x86 hardware that works with Windows. I'll instantly give Mac OS X a try if it'll ever run on my hardware.
yeah its getting their but I want my installation wizards for programs damnit:)
I've always been intrigued by installer-less application systems. I've never used Mac OS X before, but from what I hear applications can be installed by simply dragging and dropping.
A similar system can be designed for Linux and Windows: applications can be self-contained so that simple copying or extracting will suffice for installation. Features can be advertised via a standardized application description file, so even things like shortcuts or daemons can be detected and used by the operating system without copying files all over the place. Of course, there's the problem of letting the operating system knowing what directories are applications in the first place, but this can be fixed by an extremely simple "register application" command. Applications are automatically unregistered when advertised features detect missing files.
Does anyone know if this is feasible on Windows and Linux systems?
I just block everything that isn't a document of some sort. Haven't had any problems at my company since.
The unfortunate reality is that some viruses may affect you even if you aren't infected. Massive virus outbreaks are like spam: both generate large amounts of junk traffic that slow everyone's connection.
AFAIK, you can't start two copies of Windows Media Player, one for music and one for movies. WinAmp fills in a nice gap by functioning as my one and only music player. WMP also has performance problems managing a play list; it feels slow and clunky.
If there are no benefits, then don't use it. I use and like it because of the benefits.
Which is crap. I don't need that. I'll remove that is to be removed thank you. I don't need megabytes of bloat on the harddrive with copies of everything that went before.
Funny, I still don't know what megabytes of bloat you're referring to. If it kept everything that was previously installed then I'd have an extra couple of gigabytes in use because both Office and Visual Studio.NET use Windows Installer for installation.
Which always amounts to reinstall everything and reboot.
The repair checks for missing or modified files and replaces only the ones that are needed. The same happens with registry keys and shortcuts. I've also NEVER had to reboot because of a repair.
If you don't know this simple fact you are not really competent to comment at all. I track installs, and i noticed it everytime they use the bloatware installers such as Installshield.
My original comment was regarding bloatware with Windows Installer, which is either nonexistent or small enough to not be noticeable by me. I agree that InstallShield feels bloated, but again, even that installer doesn't introduce relevant amounts of bloat.
Ghost a standard image and install from there, no need to uninstall anything
I am unfamiliar with ghosting (as I said in my original post, I set up a server as a learning experience), but why ghost when I can distribute an MSI package to my entire network in under a minute? Like I said, the bloat, if it exists, is negligible. The ease of installing packages far outweighs any bloat you refer to, for me at least.
I do realize that different people may use or prefer different tools, but for my situation and the type of network I deal with, Windows Installer solves many major issues. I realize that you may be more familiar with network administration than I am, and that there may be more advanced software installation mechanisms such as ghosting, but given the small amount of computers I have and the amount of time I have to manage them, Windows Installer satisfies my needs.
When Windows Installer is ported to Linux. Looking at the Windows Installer documentation, it seems very plausible that a clone could be written. That way features such as desktop environment shortcuts can be standardized. Since MSI packages are actually databases, the format can be extended to support Linux features, such as automatic source code compilation.
Why is this still listed as a beta? I thought it was the source of a finalized product that had already been used in multiple installations... Did I miss something?
The blog mentions that WiX has been used internally by several Microsoft teams. It most likely is in beta in a similar way to OSS projects such as Firefox, which are quite usable.
Windows Installer has many benefits over installation software such as NSIS, old InstallShield (the non-Windows Installer versions), or other tools such as Inno Setup.
Windows Installer supports installation rollbacks, so failures restore the system exactly back to the state before installation began. I'm not sure if NSIS or other installers support rollbacks, but I have encountered installers in the past that left files everywhere when they quit after a failure.
I just set up a Windows network as a learning experience, and Windows Installer packages greatly simplify and automate software installation on the network. For example, I can install Office by doing an administrative install, which basically copies all the installation files to a network location, and then assigning Office to network users via a couple clicks. Office is automatically installed when the computer restarts. This works with all Windows Installer packages that support an administrative install; places such ActiveState even provide MSI packages for Python and Perl.
In addition to administrative installs, Windows Installer also supports application advertisement, which basically does things such as add only shortcuts or file associations. The program is automatically installed when the user clicks the shortcut or attempts to open a file.
There is also built-in package repair because Windows Installer keeps track of installed components. You can find more information about these and other features here.
Mind you, Windows Installer technology itself is free, and the database system it uses is documented at MSDN. There are freeware MSI authoring tools such as Advanced Installer, and I recall seeing a web-based tool on SourceForge for modifying MSI packages. Your statement that Windows Installer adds a ton of crap your C drive which are not related to the actual program being installed is also unproven.
If you want small and fast installers with tools such as Inno Setup, by all means go ahead and use them, but they do make things more difficult for network administrators.
Where are we going to be at in 2006? KDE 3.5 and GNOME 2.8, with the same old XFree86 technology running beneath (oh, gee, it might be XServer instead which will, gasp, add transparency). Same old, same old.
What really annoys me is the large amount of coverage silly features such as alpha transparency get in XFree86. I don't care about alpha transparency. I want an easier to set up XFree86, and I've wanted it for five years now.
I also want some standardization with regards to configuration. Sure, configuration files are text files, eliminating the single point of failure with the Windows registry, but there are a million of different Linux config file formats. The best part about the registry is that one tool is used to edit it. There needs to be a single tool, or at least a single file format for Linux config files.
A minor point regarding Longhorn. Win32 isn't being stripped out. A new.NET interface to the OS is being added with varying degrees of integration; some parts of it will lay on top of Win32, while other parts will directly touch the OS. Regardless, Win32 is staying for at least Longhorn.
Do not tempt the gods that way. You're just asking for something.
I didn't specifically ask for it, but for a long time I always felt good at not being infected with spyware. Incidentally, I was setting up a Windows network and decided to do a Windows reinstall. Plain IE 6 on my network was out of the question because of the lack of popup blocking and tabbed browsing, but I came across Avant Browser. I thought it was a godsend until one day, after minimizing everything I saw a Lycos Sidesearch icon on the desktop.
WTF? Lycos Sidesearch? How the hell did it get installed??!? To make a long story short, I looked around in the IE cache but never found the precise URL that installed that crap. I did find other spyware-installing pages, however, and it turned out my copy of IE was lacking a patch that allowed the pages to exploit an IE vulnerability to install themselves. I felt so stupid afterwards.
To be fair, the patch exists but it was about the only patch I didn't install out of all the other available patches. Still, I'm going to play it safe and go back to Firefox. Better to use a browser with non-major market share to avoid the crosshairs of spyware.
Cry at 16 color desktop. Get nvidia drivers. Wonder why sound is messed up. Get new sound drivers. Not all the agp features are working..what! Get via 4-in-1's
There's something wrong with your comparison: first you assume the Linux installation will progress fine as long as the user doesn't have odd hardware and that all hardware will work correctly, but then you go ahead and describe a problematic Windows install with 16-bit color desktop and nonfunctional sound.
At least talk about what happens when hardware doesn't work in Linux, and whether or not Joe will be able to read man pages and edit configuration files.
The NT-based Windows OSs have always had memory protection. Programs cannot overwrite the memory of other programs or system memory, which is precisely what most, if not all, Unix operating systems support.
The problem with Windows is that this memory protection means nothing when a vulnerability is discovered in the core OS.
I went to a presentation of the robots for a robotics course (not the RoboCup team, however) last quarter, and it was extremely interesting seeing how the robots work and what they could do. They appear to run a custom Unix-like or Unix OS by Sony, and one of the students remarked how they could telnet into the dogs for debugging or development (?) purposes.
The robots themselves move quite slowly, and they frequently get themselves into comical positions. Putting two robots side by side and placing a ball in front of them is not good; they'll fight with each other for space. I also witnessed several occasions where the robots got stuck around the goal area.
This was offset, however, by some truly spectacular goals scored by some of the robots, as well as the incredibly interesting ball-searching behavior exhibited by some of the robots.
Overall I had a sense of awe seeing these tiny mechanical things moving around under their own power and making computations themselves. A century ago people couldn't even imagine the state of electronics today, and I think a century from now these robots will likewise do things unimaginable to us.
I would imagine it to be similar to the current state of computer languages, where some languages are well-suited for certain types of components. Python, for instance, isn't written in Python but written in C. In much the same way, the visual designer would probably be designed via traditional processes.
Ah, that sounds like a very interesting idea. Perhaps a standardized XML content language can be defined, similar to HTML, that is "rendered" by an engine for a specific format. Output can be rendered as plain text for consoles, for example, and more sophisticated GUIs created when using a GUI.
This would probably replace GUI toolkits with a HTML-like system; a hybrid approach such as one provided by Longhorn and XAML would probably be best. That way applications can use an HTML-like language and/or a traditional library for UI.
Gentoo also provides command-line tools that make use of colors. While this may seem like simple eye candy and nothing more, it is extremely helpful for Linux n00bs like me. Colors help point out important information such as error messages, and also separates sections of information. It also removes the intimidation of massive amounts of white text.
However, this can be improved even further. What I would like to see is perhaps a standard library for Linux command-line programs that provides an interface for displaying common messages in a specific format. For example, error messages would have a format defined by the user, and instead of programs providing the error message formatting, the formatting library would take care of such details in addition to providing customization abilities.
Judging from other comments in this story, it appears that FAT file system tables are stored in the first few sectors, so overwriting them essentially means you lose information regarding files themselves. A partition damaged in such a manner would be unmountable, so running Knoppix won't help any unless you're dumping raw hard drive data.
Both C# and Java are compiled to bytecode, which is compiled to machine code at runtime via JIT'er. I haven't seen numbers for myself, but apparently the Java JIT'er can possibly produce faster machine code than plain C/C++ since the compilation process is continually optimized based on program behavior, which, if it isn't true, is still something that seems entirely implementable with both.NET and Java.
Additionally, my primary reason for using C# is because it simplifies various little things in C/C++. I can handle memory management in C/C++, and it isn't conceptually challenging, but it's tedious and hard to do perfectly. Higher level languages eliminate a lot of the tedium.
Mono is particularly interesting for people like me, who primarily do Windows coding using Microsoft technologies but are still interested in Linux. Sure, Java and C developers probably don't care about Mono, but I would love to see a Linux implementation of.NET simply because of the possibility of running my current Windows.NET applications on Linux with minimal changes.
I'm not very familiar with Parrot and the type of languages it supports (although I've done quite a lot of Python coding), but it would be really interesting to see Parrot developed into a dynamic language framework, sort of how.NET is a framework that eventually compiles to machine code.
I'm sure I probably worded that wrong, but I can imagine Parrot being used as a scripting system for.NET applications. Currently there are annoying limitations in.NET (cannot unload assemblies) that present severe problems with using.NET languages for scripting.
It would probably be silly for Microsoft to significantly modify the existing.NET libraries and break compatibility with existing.NET programs. Remember that modifications also affect programs and packages beyond Mono and those that use Mono.
Longhorn does include new.NET libraries for Longhorn technologies such as Avalon, as well as general System namespace classes for things such as the web browser (System.Windows.Forms.WebBrowser). Additionally, there is support for emedding Windows.Forms objects inside the new Avalon UI and vice versa, so Windows.Forms isn't immediately obsolete.
Slashdot Mirror
One of the barriers is hardware compatibility. I've invested a lot of money on x86 hardware that works with Windows. I'll instantly give Mac OS X a try if it'll ever run on my hardware.
yeah its getting their but I want my installation wizards for programs damnit :)
I've always been intrigued by installer-less application systems. I've never used Mac OS X before, but from what I hear applications can be installed by simply dragging and dropping.
A similar system can be designed for Linux and Windows: applications can be self-contained so that simple copying or extracting will suffice for installation. Features can be advertised via a standardized application description file, so even things like shortcuts or daemons can be detected and used by the operating system without copying files all over the place. Of course, there's the problem of letting the operating system knowing what directories are applications in the first place, but this can be fixed by an extremely simple "register application" command. Applications are automatically unregistered when advertised features detect missing files.
Does anyone know if this is feasible on Windows and Linux systems?
I just block everything that isn't a document of some sort. Haven't had any problems at my company since.
The unfortunate reality is that some viruses may affect you even if you aren't infected. Massive virus outbreaks are like spam: both generate large amounts of junk traffic that slow everyone's connection.
AFAIK, you can't start two copies of Windows Media Player, one for music and one for movies. WinAmp fills in a nice gap by functioning as my one and only music player. WMP also has performance problems managing a play list; it feels slow and clunky.
None for me.
If there are no benefits, then don't use it. I use and like it because of the benefits.
Which is crap. I don't need that. I'll remove that is to be removed thank you. I don't need megabytes of bloat on the harddrive with copies of everything that went before.
Funny, I still don't know what megabytes of bloat you're referring to. If it kept everything that was previously installed then I'd have an extra couple of gigabytes in use because both Office and Visual Studio.NET use Windows Installer for installation.
Which always amounts to reinstall everything and reboot.
The repair checks for missing or modified files and replaces only the ones that are needed. The same happens with registry keys and shortcuts. I've also NEVER had to reboot because of a repair.
If you don't know this simple fact you are not really competent to comment at all. I track installs, and i noticed it everytime they use the bloatware installers such as Installshield.
My original comment was regarding bloatware with Windows Installer, which is either nonexistent or small enough to not be noticeable by me. I agree that InstallShield feels bloated, but again, even that installer doesn't introduce relevant amounts of bloat.
Ghost a standard image and install from there, no need to uninstall anything
I am unfamiliar with ghosting (as I said in my original post, I set up a server as a learning experience), but why ghost when I can distribute an MSI package to my entire network in under a minute? Like I said, the bloat, if it exists, is negligible. The ease of installing packages far outweighs any bloat you refer to, for me at least.
I do realize that different people may use or prefer different tools, but for my situation and the type of network I deal with, Windows Installer solves many major issues. I realize that you may be more familiar with network administration than I am, and that there may be more advanced software installation mechanisms such as ghosting, but given the small amount of computers I have and the amount of time I have to manage them, Windows Installer satisfies my needs.
when does it run on linux?
When Windows Installer is ported to Linux. Looking at the Windows Installer documentation, it seems very plausible that a clone could be written. That way features such as desktop environment shortcuts can be standardized. Since MSI packages are actually databases, the format can be extended to support Linux features, such as automatic source code compilation.
Why is this still listed as a beta? I thought it was the source of a finalized product that had already been used in multiple installations... Did I miss something?
The blog mentions that WiX has been used internally by several Microsoft teams. It most likely is in beta in a similar way to OSS projects such as Firefox, which are quite usable.
Windows Installer has many benefits over installation software such as NSIS, old InstallShield (the non-Windows Installer versions), or other tools such as Inno Setup.
Windows Installer supports installation rollbacks, so failures restore the system exactly back to the state before installation began. I'm not sure if NSIS or other installers support rollbacks, but I have encountered installers in the past that left files everywhere when they quit after a failure.
I just set up a Windows network as a learning experience, and Windows Installer packages greatly simplify and automate software installation on the network. For example, I can install Office by doing an administrative install, which basically copies all the installation files to a network location, and then assigning Office to network users via a couple clicks. Office is automatically installed when the computer restarts. This works with all Windows Installer packages that support an administrative install; places such ActiveState even provide MSI packages for Python and Perl.
In addition to administrative installs, Windows Installer also supports application advertisement, which basically does things such as add only shortcuts or file associations. The program is automatically installed when the user clicks the shortcut or attempts to open a file.
There is also built-in package repair because Windows Installer keeps track of installed components. You can find more information about these and other features here.
Mind you, Windows Installer technology itself is free, and the database system it uses is documented at MSDN. There are freeware MSI authoring tools such as Advanced Installer, and I recall seeing a web-based tool on SourceForge for modifying MSI packages. Your statement that Windows Installer adds a ton of crap your C drive which are not related to the actual program being installed is also unproven.
If you want small and fast installers with tools such as Inno Setup, by all means go ahead and use them, but they do make things more difficult for network administrators.
Where are we going to be at in 2006? KDE 3.5 and GNOME 2.8, with the same old XFree86 technology running beneath (oh, gee, it might be XServer instead which will, gasp, add transparency). Same old, same old.
What really annoys me is the large amount of coverage silly features such as alpha transparency get in XFree86. I don't care about alpha transparency. I want an easier to set up XFree86, and I've wanted it for five years now.
I also want some standardization with regards to configuration. Sure, configuration files are text files, eliminating the single point of failure with the Windows registry, but there are a million of different Linux config file formats. The best part about the registry is that one tool is used to edit it. There needs to be a single tool, or at least a single file format for Linux config files.
A minor point regarding Longhorn. Win32 isn't being stripped out. A new .NET interface to the OS is being added with varying degrees of integration; some parts of it will lay on top of Win32, while other parts will directly touch the OS. Regardless, Win32 is staying for at least Longhorn.
Do not tempt the gods that way. You're just asking for something.
I didn't specifically ask for it, but for a long time I always felt good at not being infected with spyware. Incidentally, I was setting up a Windows network and decided to do a Windows reinstall. Plain IE 6 on my network was out of the question because of the lack of popup blocking and tabbed browsing, but I came across Avant Browser. I thought it was a godsend until one day, after minimizing everything I saw a Lycos Sidesearch icon on the desktop.
WTF? Lycos Sidesearch? How the hell did it get installed??!? To make a long story short, I looked around in the IE cache but never found the precise URL that installed that crap. I did find other spyware-installing pages, however, and it turned out my copy of IE was lacking a patch that allowed the pages to exploit an IE vulnerability to install themselves. I felt so stupid afterwards.
To be fair, the patch exists but it was about the only patch I didn't install out of all the other available patches. Still, I'm going to play it safe and go back to Firefox. Better to use a browser with non-major market share to avoid the crosshairs of spyware.
Cry at 16 color desktop. Get nvidia drivers. Wonder why sound is messed up. Get new sound drivers. Not all the agp features are working..what! Get via 4-in-1's
There's something wrong with your comparison: first you assume the Linux installation will progress fine as long as the user doesn't have odd hardware and that all hardware will work correctly, but then you go ahead and describe a problematic Windows install with 16-bit color desktop and nonfunctional sound.
At least talk about what happens when hardware doesn't work in Linux, and whether or not Joe will be able to read man pages and edit configuration files.
The NT-based Windows OSs have always had memory protection. Programs cannot overwrite the memory of other programs or system memory, which is precisely what most, if not all, Unix operating systems support.
The problem with Windows is that this memory protection means nothing when a vulnerability is discovered in the core OS.
The CSE department at the University of Washington also has Aibo robots and participates in RoboCup. More information here.
I went to a presentation of the robots for a robotics course (not the RoboCup team, however) last quarter, and it was extremely interesting seeing how the robots work and what they could do. They appear to run a custom Unix-like or Unix OS by Sony, and one of the students remarked how they could telnet into the dogs for debugging or development (?) purposes.
The robots themselves move quite slowly, and they frequently get themselves into comical positions. Putting two robots side by side and placing a ball in front of them is not good; they'll fight with each other for space. I also witnessed several occasions where the robots got stuck around the goal area.
This was offset, however, by some truly spectacular goals scored by some of the robots, as well as the incredibly interesting ball-searching behavior exhibited by some of the robots.
Overall I had a sense of awe seeing these tiny mechanical things moving around under their own power and making computations themselves. A century ago people couldn't even imagine the state of electronics today, and I think a century from now these robots will likewise do things unimaginable to us.
but who will visually debug the visual designer?
I would imagine it to be similar to the current state of computer languages, where some languages are well-suited for certain types of components. Python, for instance, isn't written in Python but written in C. In much the same way, the visual designer would probably be designed via traditional processes.
Ah, that sounds like a very interesting idea. Perhaps a standardized XML content language can be defined, similar to HTML, that is "rendered" by an engine for a specific format. Output can be rendered as plain text for consoles, for example, and more sophisticated GUIs created when using a GUI.
This would probably replace GUI toolkits with a HTML-like system; a hybrid approach such as one provided by Longhorn and XAML would probably be best. That way applications can use an HTML-like language and/or a traditional library for UI.
Gentoo also provides command-line tools that make use of colors. While this may seem like simple eye candy and nothing more, it is extremely helpful for Linux n00bs like me. Colors help point out important information such as error messages, and also separates sections of information. It also removes the intimidation of massive amounts of white text.
However, this can be improved even further. What I would like to see is perhaps a standard library for Linux command-line programs that provides an interface for displaying common messages in a specific format. For example, error messages would have a format defined by the user, and instead of programs providing the error message formatting, the formatting library would take care of such details in addition to providing customization abilities.
Judging from other comments in this story, it appears that FAT file system tables are stored in the first few sectors, so overwriting them essentially means you lose information regarding files themselves. A partition damaged in such a manner would be unmountable, so running Knoppix won't help any unless you're dumping raw hard drive data.
Both C# and Java are compiled to bytecode, which is compiled to machine code at runtime via JIT'er. I haven't seen numbers for myself, but apparently the Java JIT'er can possibly produce faster machine code than plain C/C++ since the compilation process is continually optimized based on program behavior, which, if it isn't true, is still something that seems entirely implementable with both .NET and Java.
Additionally, my primary reason for using C# is because it simplifies various little things in C/C++. I can handle memory management in C/C++, and it isn't conceptually challenging, but it's tedious and hard to do perfectly. Higher level languages eliminate a lot of the tedium.
Mono is particularly interesting for people like me, who primarily do Windows coding using Microsoft technologies but are still interested in Linux. Sure, Java and C developers probably don't care about Mono, but I would love to see a Linux implementation of .NET simply because of the possibility of running my current Windows .NET applications on Linux with minimal changes.
I'm not very familiar with Parrot and the type of languages it supports (although I've done quite a lot of Python coding), but it would be really interesting to see Parrot developed into a dynamic language framework, sort of how .NET is a framework that eventually compiles to machine code.
I'm sure I probably worded that wrong, but I can imagine Parrot being used as a scripting system for .NET applications. Currently there are annoying limitations in .NET (cannot unload assemblies) that present severe problems with using .NET languages for scripting.
It would probably be silly for Microsoft to significantly modify the existing .NET libraries and break compatibility with existing .NET programs. Remember that modifications also affect programs and packages beyond Mono and those that use Mono.
.NET libraries for Longhorn technologies such as Avalon, as well as general System namespace classes for things such as the web browser (System.Windows.Forms.WebBrowser). Additionally, there is support for emedding Windows.Forms objects inside the new Avalon UI and vice versa, so Windows.Forms isn't immediately obsolete.
Longhorn does include new
Try RSS Bandit. It's based on an MSDN article, Building a Desktop News Aggregator, that discusses how to build an RSS aggregator with C# and .NET.
You can get the free ActivePerl distribution for Windows. They also have a Python distribution, which works perfectly and is incredibly useful.