Don't think the GPL has only positive benefits. There have been a lot of architectural decisions by projects like linux and gcc to limit architecture for what amounts to 'political' reasoning.
For example, there was a gcc compiler to java bytecode which was rejected because RMS considered the bytecode to be enough of an intermediate language that it could be used to decouple the front and back-end parts of GCC, allowing people to reuse the language frontends for new platforms and add new languages to the existing platforms without needing to contribute the code under the GPL.
Great, I can have 'unlimited' songs now for $5-10/month, but in two years the labels will start insisting that it go up, up, up. And unlike the Apple system, your only choices are to pay the new price or lose all of your music from the last few years.
If it is their computer, why shouldn't they be able to install any software, anytime, anyplace? Why must I give an installer root access to my system in order to get an application on? With a macintosh, for example, most installers consist of 'drag this application wherever you want to run it from'.
The macintosh experience isn't even the best it could be, it would be very possible to one-up them and provide for native application sandboxing (basically, applications should never inherit permissions based on the user or parent process that executes them). The issue is that the developer-centric audience of open-source development prevents this from happening - the people who are writing code are not scratching these sorts of itches.
Re:High-level VFS does *not* belong in the kernel!
on
GNOME 2.12 Released
·
· Score: 1
Let's clarify what the "underlying OS" is... Are you considering the Finder to be part of it? Last time I checked, Mac OS didn't have the ability to treat URL's as filenames. Its VFS layer is a combination of an automounter and the Finder displaying network shares and volumes. What do you want, a kernel-level HTTP client?
You cannot open arbitrary URLs like you can on windows (how do you save?), but you can deal with URLs backed by WebDav. This is done, of course, with a userspace filesystem driver, and not some odd library that only provides some applications in the system writing to the right APIs the ability to see and manipulate files.
And yes, I use ubuntu for my linux box, although it has been a while since I've even bothered to let gdm start. It is like debian with a release schedule and authorative leadership.
Why would you want to tie any desktop code into the kernel? Should the kernel then include code for *all* desktop environments? KDE? XFCE? Blackbox? Enlightenment? WindowMaker?... Talk aout bloat!
I'm not talking about desktop code to the kernel, I'm talking about using leveraging the kernel's vfs rather than creating your own.
1) You don't need to compile anything if you're using a distro that supports apt, yum, up2date, YaST, etc...
2) Not only is install-as-admin a security feature, but also
3) It's rather trivial in most distros to install "from" a non-admin account -- ie. in SuSE just click the RPM, enter the admin password, and it installs. If all fails, there's sudo.
apt/yum/up2date don't neccessarily support me installing from CD media
install-as-sudo-admin is not a security feature, it is a security checkpoint. Indeed, you have just given the installer scripts root access to your system
you just said use an admin password or sudo. How is this defeating my argument again?
I think it's better to leave this stuff in userspace - it shouldn't have to matter whether the files are sitting in ext3, iso9660, udf, fat32, or coming in via nfs or smb/cifs. Also gnome is cross-platform now so you don't want to extend it into a paticular OS. Despite the name it isn't a file system anyway.
My particular gripe is that gnome draws a line at being a desktop environment, then goes over it by faking userspace filesystems. Users of the system suffer because files which are accessible through gnome may not be accessible through kde/gtk/xlib/regular shell apps.
I like the way you imply that gnome didn't have a clipboard until now. What they added was a very advanced way of handling many clipboard entries. Something which no other desktop does as well yet.
New clipboard management, based on the Freedesktop.org specification and tightly integrated with GNOME, allows for objects to persist in the clipboard longer than the lifetime of an application. This means that if you cut or copy an object and then exit that application, the item you put on the clipboard will remain until you replace it.
This sounds to me like up until now, the clipboard has never been centrally managed, but has relied on the original application being around to provide clipboard data. The concept of a clipboard that lives beyond the application data was cut/copied from has been around a long time in other desktops
Have GnomeVFS built into the underlying OS and not as a IO library wrapper/hack
Well this shows me I shouldn't have bothered writing the above because you clearly don't know what you're talking about.
GnomeVFS is a nightmare for user experience because only some applications understand it. If you use regular X applications, KDE applications, gtk applications, or a shell, there is no compatibility - you just sometimes can see the files, and sometimes are unable to access them at all.
But thats part of my point. Gnome as a project stops far short of what someone like Apple tries to achieve - a complete user experience.
Creating a user-centric and not administrator-centric environment is not something that I've seen distributions tackle, as they generally either do integration work of the various components they want to ship, or work to attract enterprise customers (who want things to be even more administrator-centric). Distributions also have a tendancy to have NIH syndrome when it comes to bundling projects created by other distribution-makers.
To put it bluntly - if creating a user-centric environment is left up to the distributions, it will probably never catch on.
That isn't what RT means. RT has to do with starting processes at precise times, and is used in critical systems. When you bring in real-time requirements, you make sacrifices in other areas of efficiency.
So by having a real-time kernel, you wind up sacrificing the raw CPU power needed to make multimedia happen. What you want for multimedia is priority processing, which has nothing to do with realtime.
The lost efficiency for real-time kernel scheduling is why there isn't real-time scheduling built into mainline linux.
You are killing right? Its been how many years, better part of a decade and they just added freakin' clipboard services.
Call me back when they:
Have GnomeVFS built into the underlying OS and not as a IO library wrapper/hack
Applications stop steaking focus a minute after I told them to start, gave up, and went on to do something else
Icons and windowing theme are standardized. By standardized, i mean "ship with clearlooks or don't call it Gnome"
are possible for users to use, including installing and removing commercial applications, without learning such concepts as "compilers" or "administrative users"
They obviously modified it; it supports the google user cookie format as an authentication mechanism. Which makes me wonder if they will use a modified connection component and AJAX for a web-based client.
The problem is that the companies are outsourcing _everything_ - they take a hardware reference implementation from someone like Broadcom and have software written for it by people in India. The people managing all this are typical middle management - they know just enough to scrape along. The usual technical leads that care about the company and escalate issues like 'license compliance' don't exist anymore.
Darcs works off a non-centralized model (see arch, codeville, monotone, bitkeeper, cogito) instead of a centralized model (cvs, subversion, perforce, clearcase). Rather than tracking revisions, it tracks changes. This means that rather than merging all changes into a new revision, changes are pulled (or pushed) to create a tree.
Of the non-centralized tools out there, darcs is probably simplest to learn to use. However, the use of Haskell has always made me apprehensive - this and performance/scalability problems have limited my use.
The patch model is innovative, but the flip side is that it is unique, and has trade-offs in usage. While other systems generate patches around changes which can be exchanged and even signed to prevent tampering, darcs patches are altered when applied.
If the disk did not have the cache, it would have to write directly to system memory. Only one device (CPU, hard disk, video card, etc.) can write to main memory at a time, so this would significantly increase the amount of time you have to wait for memory acccesses from the cpu - everything would be waiting on the disk to stop filling cache.
Now, as to why such a large cache - the lack of power back-up makes it less useful. It would be better if there was less memory on the disk, and a lot more on a controller.
The identity system does not act on your behalf, but rather exposes the user as a gatekeeper between people who are willing to assert your identity (ISP, banks, employers) and people who are willing to trust those asserting parties and consume those identities.
The actual information is in a very extensible format (they are using SAML right now), so the actual passed information can be just about anything. It could be authorization information such as user roles or specific permissions, or could be a reference to a machine where authorization questions can be asked.
The key to ANY identity system is the central identify repository. That's the box that holds the criteria to prove that you are you.
Incorrect. That would be the key to a global identity system. Trying to market passport as a global identity system is the reason it failed - the larger you get, the less people trust you.
Who will trust Microsoft to maintain that?
Read the seven laws. They describe requirements for a successful identity infrastructure, and do exclude the concept of a central authority on many levels.
You have too narrow a view of what is meant by identity - there is no requirement for a unique identifier. Your identity could very well be "An authorized, card-carrying member of the tinfoil hat brigade" without anything else - no contact information, no membership #
The tax is on the computers Australians buy - my bet is about $80 AUS winds up going to a certain Washington-state company for almost every computer sold. The added charge on every computer sold is commonly referred to as the "Windows Tax"
Depends if Barnes and Nobles started selling two weeks before smaller shops even managed to get it into stock, and they wound up filing a class action against Scholastic for favoring certain chains. That could probably cost them a significant portion of the book profits.
They are probably also worried about someone taking a copy and that extra time and setting up shop selling scanned copies before release, or posting translations into other languages (both of which happened for the last book)
Seriously folks, its almost like you've never heard of release dates before. You should hear some of the financial penalties they make shops agree to for other media (music, movie) releases.
When is the last time you heard of a movie theatre having a public showing of a movie a few days before the studio asked, because they got the reels in a little earlier than normal, had them already spliced up and reviewed and said 'eh, what the hell...'
Slashdot Mirror
Don't think the GPL has only positive benefits. There have been a lot of architectural decisions by projects like linux and gcc to limit architecture for what amounts to 'political' reasoning.
For example, there was a gcc compiler to java bytecode which was rejected because RMS considered the bytecode to be enough of an intermediate language that it could be used to decouple the front and back-end parts of GCC, allowing people to reuse the language frontends for new platforms and add new languages to the existing platforms without needing to contribute the code under the GPL.
... Like the fucking plague.
Great, I can have 'unlimited' songs now for $5-10/month, but in two years the labels will start insisting that it go up, up, up. And unlike the Apple system, your only choices are to pay the new price or lose all of your music from the last few years.
It uses two factor auth, i.e. it doesn't just broadcast out a static credit card # to anyone who can induct a current through it.
You cannot support 10.2.x, because they didn't have universal binary support from what I remember.
But you can have different library dependancies per platform in teh fat mach-o binary, meaning x86 could require 10.4 while ppc requires 10.3
If it is their computer, why shouldn't they be able to install any software, anytime, anyplace? Why must I give an installer root access to my system in order to get an application on? With a macintosh, for example, most installers consist of 'drag this application wherever you want to run it from'.
The macintosh experience isn't even the best it could be, it would be very possible to one-up them and provide for native application sandboxing (basically, applications should never inherit permissions based on the user or parent process that executes them). The issue is that the developer-centric audience of open-source development prevents this from happening - the people who are writing code are not scratching these sorts of itches.
Let's clarify what the "underlying OS" is... Are you considering the Finder to be part of it? Last time I checked, Mac OS didn't have the ability to treat URL's as filenames. Its VFS layer is a combination of an automounter and the Finder displaying network shares and volumes. What do you want, a kernel-level HTTP client?
You cannot open arbitrary URLs like you can on windows (how do you save?), but you can deal with URLs backed by WebDav. This is done, of course, with a userspace filesystem driver, and not some odd library that only provides some applications in the system writing to the right APIs the ability to see and manipulate files.
And yes, I use ubuntu for my linux box, although it has been a while since I've even bothered to let gdm start. It is like debian with a release schedule and authorative leadership.
I'm not talking about desktop code to the kernel, I'm talking about using leveraging the kernel's vfs rather than creating your own.
1) You don't need to compile anything if you're using a distro that supports apt, yum, up2date, YaST, etc... 2) Not only is install-as-admin a security feature, but also 3) It's rather trivial in most distros to install "from" a non-admin account -- ie. in SuSE just click the RPM, enter the admin password, and it installs. If all fails, there's sudo.My particular gripe is that gnome draws a line at being a desktop environment, then goes over it by faking userspace filesystems. Users of the system suffer because files which are accessible through gnome may not be accessible through kde/gtk/xlib/regular shell apps.
This sounds to me like up until now, the clipboard has never been centrally managed, but has relied on the original application being around to provide clipboard data. The concept of a clipboard that lives beyond the application data was cut/copied from has been around a long time in other desktops
Have GnomeVFS built into the underlying OS and not as a IO library wrapper/hack Well this shows me I shouldn't have bothered writing the above because you clearly don't know what you're talking about.
GnomeVFS is a nightmare for user experience because only some applications understand it. If you use regular X applications, KDE applications, gtk applications, or a shell, there is no compatibility - you just sometimes can see the files, and sometimes are unable to access them at all.
But thats part of my point. Gnome as a project stops far short of what someone like Apple tries to achieve - a complete user experience.
Creating a user-centric and not administrator-centric environment is not something that I've seen distributions tackle, as they generally either do integration work of the various components they want to ship, or work to attract enterprise customers (who want things to be even more administrator-centric). Distributions also have a tendancy to have NIH syndrome when it comes to bundling projects created by other distribution-makers.
To put it bluntly - if creating a user-centric environment is left up to the distributions, it will probably never catch on.
That isn't what RT means. RT has to do with starting processes at precise times, and is used in critical systems. When you bring in real-time requirements, you make sacrifices in other areas of efficiency.
So by having a real-time kernel, you wind up sacrificing the raw CPU power needed to make multimedia happen. What you want for multimedia is priority processing, which has nothing to do with realtime.
The lost efficiency for real-time kernel scheduling is why there isn't real-time scheduling built into mainline linux.
Gnome is the OSX killer.
You are killing right? Its been how many years, better part of a decade and they just added freakin' clipboard services.
Call me back when they:
They obviously modified it; it supports the google user cookie format as an authentication mechanism. Which makes me wonder if they will use a modified connection component and AJAX for a web-based client.
The problem is that the companies are outsourcing _everything_ - they take a hardware reference implementation from someone like Broadcom and have software written for it by people in India. The people managing all this are typical middle management - they know just enough to scrape along. The usual technical leads that care about the company and escalate issues like 'license compliance' don't exist anymore.
How about just returning an object tag which references the firefox activex control ;-)
Darcs works off a non-centralized model (see arch, codeville, monotone, bitkeeper, cogito) instead of a centralized model (cvs, subversion, perforce, clearcase). Rather than tracking revisions, it tracks changes. This means that rather than merging all changes into a new revision, changes are pulled (or pushed) to create a tree.
Of the non-centralized tools out there, darcs is probably simplest to learn to use. However, the use of Haskell has always made me apprehensive - this and performance/scalability problems have limited my use.
The patch model is innovative, but the flip side is that it is unique, and has trade-offs in usage. While other systems generate patches around changes which can be exchanged and even signed to prevent tampering, darcs patches are altered when applied.
If the disk did not have the cache, it would have to write directly to system memory. Only one device (CPU, hard disk, video card, etc.) can write to main memory at a time, so this would significantly increase the amount of time you have to wait for memory acccesses from the cpu - everything would be waiting on the disk to stop filling cache.
Now, as to why such a large cache - the lack of power back-up makes it less useful. It would be better if there was less memory on the disk, and a lot more on a controller.
The identity system does not act on your behalf, but rather exposes the user as a gatekeeper between people who are willing to assert your identity (ISP, banks, employers) and people who are willing to trust those asserting parties and consume those identities.
The actual information is in a very extensible format (they are using SAML right now), so the actual passed information can be just about anything. It could be authorization information such as user roles or specific permissions, or could be a reference to a machine where authorization questions can be asked.
Incorrect. That would be the key to a global identity system. Trying to market passport as a global identity system is the reason it failed - the larger you get, the less people trust you.
Who will trust Microsoft to maintain that?
Read the seven laws. They describe requirements for a successful identity infrastructure, and do exclude the concept of a central authority on many levels.
You have too narrow a view of what is meant by identity - there is no requirement for a unique identifier. Your identity could very well be "An authorized, card-carrying member of the tinfoil hat brigade" without anything else - no contact information, no membership #
I sense T-Shirts.. sooon
http://www.ietf.org/rfc/rfc1738.txt?number=1738
the IETF thinks differently re: URLs
yes, although the 'in Xcode' part is still shaky.
The tax is on the computers Australians buy - my bet is about $80 AUS winds up going to a certain Washington-state company for almost every computer sold. The added charge on every computer sold is commonly referred to as the "Windows Tax"
Depends if Barnes and Nobles started selling two weeks before smaller shops even managed to get it into stock, and they wound up filing a class action against Scholastic for favoring certain chains. That could probably cost them a significant portion of the book profits.
They are probably also worried about someone taking a copy and that extra time and setting up shop selling scanned copies before release, or posting translations into other languages (both of which happened for the last book)
Seriously folks, its almost like you've never heard of release dates before. You should hear some of the financial penalties they make shops agree to for other media (music, movie) releases.
When is the last time you heard of a movie theatre having a public showing of a movie a few days before the studio asked, because they got the reels in a little earlier than normal, had them already spliced up and reviewed and said 'eh, what the hell...'