You ask "Why the fuck is X11 so damn slow?" I answer "Producers of modern X11 toolkits don't design with low-bandwidth connections in mind." Try running something like xpdf over your 600kb/s connection.:)
Anyway, on to the X11 vs. AJAX question: My understanding of the deep internals of X11 is a little fuzzy, so what I'm going to say here is not completely correct. Please correct me when I fuck something up.
With AJAX, you have a remote server that pushes all of the UI and the instructions to render it to the client. The client handles the rendering. (This is why some AJAX apps are completely fucked up on some browsers and not others.)
With X11, you have a remote server that pushes the state of a remote app to the client. The server handles all of the rendering. (This is why -assuming that your app doesn't require some X11 extension that you don't have-, remote X11 apps look the same, no matter what local X "server" you're using.)
Now, we could do a *much* better job of offloading a lot of work to the client. This would allow us to write X11 apps that would survive being connected to a new "server". Ask google about this:
1) Yes, of course I have, and No, it's not the holy grail you were looking for. KDE isn't the X-Server, it's a layer _on_top_ of the Xserver...
Oh, I know this. It is, however, a desktop environment that -in many ways- rivals anything that's coming out of Redmond or Cupertino.
2) Remove the network layer from the X-Server, combine and compact the GDI to focus on speed and simplicity. The GDI should worry ONLY on how to actually talk to the graphics hardware
a) What is the GDI? The I've only ever seen this acronym officially used when describing an old graphics API for MS Windows. b) You do know that work that's being done with "unprivileged X" and kernel modesetting is making it so's X doesn't have to manage the hardware anymore? c) Do you have benchmarks that demonstrate that X's network transparency harms performance on local clients that utilize the MIT Shared Memory extension? [0] Do you have any benchmarks which compare basic operations under WIN32, Cocoa, and X11?
This is a continuation of point 2. Do you have benchmarks or design insights to support your claims?
Of course you can do all of what you can in MacOS and Windows on a Linux box - at least as far as the eye-candy goes, but at the cost of speed and fuctionality
What hardware and driver revisions did you use KDE 4.2 on? The only *performance* bug that i've found with 3D accell is bouncing between redirected and unredirected mode when using Firefox in fullscreen. I'm running on an AGP R420 (Radeon X800) with xf86-video-ati-6.12.2.
[0] This extension has been available since X11R5, which was released in 1991.
The X-Server has lived its lifetime. It's time to kill it off, and bring something that actually brings the Linux graphical front end into the present time.
0) Have you ever used KDE 4.2? 1) What do you suggest as a replacement for X11R7? 2) What aspects of X11R7 are holding the Linux GUI and/or desktop back?
If you can imagine a flavour of *nix that actually offered TCP/IP as an option. Today it just seems absurd.
Heh. The spirit of this still lives on in Win32/64. One's app has to initialize the system IP DLL (AKA: Winsock) before one can do *anything* with the network... even something as simple as gethostname.
This is why we need to pass a law requiring all citizens to keep a lawyer on retainer at all times. It's the only way to make certain that our rights are respected.
Police aren't supposed to keep a community crime-free. That's not even in their job description. Their job is to clean up the mess afterward...
This is spot on. SCOTUS recognises that a police force has a limited amount of resources at its disposal. They have ruled that the police forces of this nation are under *no* obligation to either protect citizens from or prevent crime.
I would check out the Red, Blue and Orange books for a comprehensive reference library and the OpenGL Superbible for a big tutorial.:)
Be aware that your old OGL knowledge is still useful and perfectly valid. Any hardware/driver combo that supports OGL 2 features is going to support the 1.x features as well.
OTOH, any hardware that supports OGL 3.x will most likely have drivers that support the older bits of OGL.:)
You know, if legitimate software could ever learn how to make software as resilient as malware the world would be a better place.
I'm not sure exactly what you mean... but every single bit of code that I've written in the past couple of years has been bulletproof. If my software dies, you can bet that there's something so *very* wrong with the OS that a reboot is in order. [Assuming that the OS hasn't already hard-locked.]:)
The problem still is that you have to give full access to your machine to an untrusted installer for a friggin game.
Yeah... it really burns my biscuits that a video game's installer has full permissions to my games directory and full permissions to the registry, but *still* requires me to run it as Admin.
Most installers that use Windows Installer are dumb. Throw rocks at them.
GPL can "infect" a company's IP. And that's not a bug, it's a feature. RMS has said so himself and others are also quite clear on this.
The GPL is no more infectious than the "you can't sell the software that you build with this tool" clause of the license that accompanied "student" versions of MSFT's Visual C++ 6 and 7.
From the first link:
GPL and NDA
* To: gcc at gcc dot gnu dot org
* Subject: GPL and NDA
* From: Richard Stallman
* Date: Thu, 19 Jul 2001 05:07:10 -0600 (MDT)
* Reply-to: rms at gnu dot org
GPL-covered code may not be distributed under an NDA. To do so is a violation of the GPL.
If someone asks you to sign an NDA for receiving GPL-covered code that is copyright FSF, please inform the FSF immediately. If it involves GPL-covered code that has some other copyright holder, please inform that copyright holder, just as you would for any other kind of violation of the GPL.
It is possible for a person or company to develop changes to a GPL-covered program and sign an NDA promising not to release these changes *to anyone*. This is a different case. As long as these changes are not distributed at all, a fortiori they are not distributed in a way that violates the GPL.
However, if and when the changes are distributed to another person or outside the company, they must be distributed under the terms of the GPL, not under an NDA.
This doesn't have anything to do with the licensing of works that derive from GPL-licensed code. Your inclusion of this information makes you look either illiterate or careless.
From the second link:
From: Phil Nelson Subject: Inaccurate information in `GNU' section Date: 23 Apr 2004, 09:33:30
Hash: SHA1
On Thursday 22 April 2004 11:45 pm, Peter N Lewis wrote: > > > I'm confused by this. If it "does not contain code from any > > > >> GPL-covered software", then surely you can release it under any > >> license you feel like, whether it can run stand-alone or not. If I > > > >A program can be free of code from a GPL-covered program, but it links > >with a library licensed under the GPL, then it has to be under the GPL > >as well. It is this catch, if you will, that led to the LGPL. > > This could only be true if you ship the binary with it linked.
First, the disclaimer, IANAL.
Second, I wrote gdbm for the FSF and have had many many conversations with Mr. Stallman about the GPL and how it applies to libraries. The key to the GPL is the understanding of "derivative work". I'll use gdbm as an example.
If code directly calls any of the gdbm functions, it *is* a derivative work. Therefore, any work that directly calls gdbm functions and is distributed must be distributed under the terms of the GPL. Of course, the GPL is transitive, thus any program that has a call path that ends up in a gdbm function is required to be under the GPL. (If distributed.)
If your code is written for the original dbm interface it is not a derivative work of gdbm, even if you happen to use the dbm interface to gdbm. It would be considered a derivative work of dbm.
I don't know if this would stand up in a court of law, but I'm very sure this is what Mr. Stallman intended for the GPL. I have requested that gdbm be put under the LGPL so that programs that use gdbm don't have to be under the GPL, but he has chosen to keep the GPL on gdbm. Therefore, whenever people ask about gdbm, the GPL and their programs, I must tell them that if they plan to distri
Not all machines hibernate correctly.:( Moreover, not all of us have the good fortune of being able to work in an environment like emacs that supports true session saving.
Our group was recently informed "the simple act of shutting down PCs at night can save a company with 10,000 PCs over $260,000 a year". We kicked around the idea.
That's an alleged savings of $26/year/computer, or about $0.09/day.
52*2*0.09=$9.36 per PC per year. I wonder how the reduction in hard drive life due to extra power cycling affects the overall cost savings and environmental impact of shutting down the machines on the weekend.
Don't try to screw with me, okay? I remember with SYN floods were new. I was already working with machines (in "IT" as you call it now) when the Morris Worm went around...
I work with an old programmer. He has a lot of knowledge. None of it is relevant. He's a horrible coder and an even worse CM.
Name one mainstream linux distro that has NX and ASLR on...
Oooh. There's that tricky word again.
... (not the weak forms) and is widely used please. Then we'll check its market penetration and add it to Vista's and see if we reach 1/3rd of all machines out there.... Just because ASLR was mentioned in 1999 doesn't mean it's old hat today.
Less than 1/3rd of the machines out there on the internet are using ASLR. And of those, most are running Vista.
ASLR has been known about for ten years and had a Linux implementation for seven. Something that's widely known within a community is old hat. It doesn't matter if the proles aren't using it.
The people hacking it had more trouble than they expected because the machine was running SP1 which was new and the NX and ASLR made it a lot tougher, they had to rework their hacks they had prepared on Vista RTM before the contest.
Holy shit! Hackers who prepared a hack on a soft system had to modify their techniques when attacking a harder system? STOP THE PRESSES! THIS MUST GO ON THE FRONT PAGE!
Also, in 2009 Vista SP1 was hacked. In the latter case, it was hacked through a vulnerability in Java. They never were able to inject code onto the platform, but they didn't need to to hack in.
A system cannot be considered secure if insecure code running as an unprivileged user on the system causes it to be compromised.
If you want to say how the ASLR and NX in Vista SP1 means nothing, you should use research that is about Vista SP1 or later.
many other OSes that present a compromise, like various forms of Linux or Vista.
If you're not a software developer, my Hardened Gentoo systems require no compromise. If, however, you want a usable backtrace, or to be able to set breakpoints in GDB, you're SOL.:(
Slashdot Mirror
You ask "Why the fuck is X11 so damn slow?" I answer "Producers of modern X11 toolkits don't design with low-bandwidth connections in mind." Try running something like xpdf over your 600kb/s connection. :)
Anyway, on to the X11 vs. AJAX question:
My understanding of the deep internals of X11 is a little fuzzy, so what I'm going to say here is not completely correct. Please correct me when I fuck something up.
With AJAX, you have a remote server that pushes all of the UI and the instructions to render it to the client. The client handles the rendering. (This is why some AJAX apps are completely fucked up on some browsers and not others.)
With X11, you have a remote server that pushes the state of a remote app to the client. The server handles all of the rendering. (This is why -assuming that your app doesn't require some X11 extension that you don't have-, remote X11 apps look the same, no matter what local X "server" you're using.)
Now, we could do a *much* better job of offloading a lot of work to the client. This would allow us to write X11 apps that would survive being connected to a new "server". Ask google about this:
aKademy06-KDE_and_Consumer_Electronics_-_The_Lost_Momentum_-_Holger_Freyther
for an example of (IIRC) a GTK app which was written to survive connection to a new X "server".
1) Yes, of course I have, and No, it's not the holy grail you were looking for. KDE isn't the X-Server, it's a layer _on_top_ of the Xserver...
Oh, I know this. It is, however, a desktop environment that -in many ways- rivals anything that's coming out of Redmond or Cupertino.
2) Remove the network layer from the X-Server, combine and compact the GDI to focus on speed and simplicity. The GDI should worry ONLY on how to actually talk to the graphics hardware
a) What is the GDI? The I've only ever seen this acronym officially used when describing an old graphics API for MS Windows.
b) You do know that work that's being done with "unprivileged X" and kernel modesetting is making it so's X doesn't have to manage the hardware anymore?
c) Do you have benchmarks that demonstrate that X's network transparency harms performance on local clients that utilize the MIT Shared Memory extension? [0] Do you have any benchmarks which compare basic operations under WIN32, Cocoa, and X11?
3) Big, sluggish, unneccessarily complex X-Server...
This is a continuation of point 2. Do you have benchmarks or design insights to support your claims?
Of course you can do all of what you can in MacOS and Windows on a Linux box - at least as far as the eye-candy goes, but at the cost of speed and fuctionality
What hardware and driver revisions did you use KDE 4.2 on? The only *performance* bug that i've found with 3D accell is bouncing between redirected and unredirected mode when using Firefox in fullscreen. I'm running on an AGP R420 (Radeon X800) with xf86-video-ati-6.12.2.
[0] This extension has been available since X11R5, which was released in 1991.
The X-Server has lived its lifetime. It's time to kill it off, and bring something that actually brings the Linux graphical front end into the present time.
0) Have you ever used KDE 4.2?
1) What do you suggest as a replacement for X11R7?
2) What aspects of X11R7 are holding the Linux GUI and/or desktop back?
that's what the Linux experiene is like for me, and for many other people I bet.
'Tis a pity. Have you filed a bug report with the distro maintainer?
If you can imagine a flavour of *nix that actually offered TCP/IP as an option. Today it just seems absurd.
Heh. The spirit of this still lives on in Win32/64. One's app has to initialize the system IP DLL (AKA: Winsock) before one can do *anything* with the network... even something as simple as gethostname.
Was that 20MB drive a "Hard Card"?
Oh, the memories....
This is why we need to pass a law requiring all citizens to keep a lawyer on retainer at all times. It's the only way to make certain that our rights are respected.
Police aren't supposed to keep a community crime-free. That's not even in their job description. Their job is to clean up the mess afterward...
This is spot on. SCOTUS recognises that a police force has a limited amount of resources at its disposal. They have ruled that the police forces of this nation are under *no* obligation to either protect citizens from or prevent crime.
...because "stop snitching" only guarantees your community is going to become a criminally infested hell, complete with brutal and corrupt police
Unless, of course, that brutal and corrupt police force employs equally corrupt "snitches" who fabricate evidence to be used against innocent men.
Stopping this behaviour is the true objective of most of those who say "stop snitching". :)
I would check out the Red, Blue and Orange books for a comprehensive reference library and the OpenGL Superbible for a big tutorial. :)
Be aware that your old OGL knowledge is still useful and perfectly valid. Any hardware/driver combo that supports OGL 2 features is going to support the 1.x features as well.
OTOH, any hardware that supports OGL 3.x will most likely have drivers that support the older bits of OGL. :)
Balls! I hate the mile-wide "Post Anonymously" checkbox. This was me with my HD video statistics.
Now, why oh why would I want a possibly bad plant or a nuclear plant...
It's interesting that you make this distinction...
OT comment WRT your new signature:
Heh. Your rural life in Iowa isn't quite so idyllic? That sucks. :(
You know, if legitimate software could ever learn how to make software as resilient as malware the world would be a better place.
I'm not sure exactly what you mean... but every single bit of code that I've written in the past couple of years has been bulletproof. If my software dies, you can bet that there's something so *very* wrong with the OS that a reboot is in order. [Assuming that the OS hasn't already hard-locked.] :)
The question "Do you want the government raising your children?" has already been put to the public, and the answer is obvious: a resounding "No!".
Heh. You don't live where I live. Middle and high-school is considered by most parents to be a gratis day care center.
So you're recommending either extensive regulation, of communal ownership of the "last mile".
Please carefully re-read my post.
If a careful re-reading isn't illuminating, read this comment:
http://slashdot.org/comments.pl?sid=1196517&cid=27557021
At this point it is clear that any increase in capacity will be matched by an increase in use...
Do you have the numbers to back up this claim?
The problem still is that you have to give full access to your machine to an untrusted installer for a friggin game.
Yeah... it really burns my biscuits that a video game's installer has full permissions to my games directory and full permissions to the registry, but *still* requires me to run it as Admin.
Most installers that use Windows Installer are dumb. Throw rocks at them.
It surprised me to see a Mac botnet...
Are you serious?
You do know that trojans can hit any soft system?
You said:
GPL can "infect" a company's IP. And that's not a bug, it's a feature. RMS has said so himself and others are also quite clear on this.
The GPL is no more infectious than the "you can't sell the software that you build with this tool" clause of the license that accompanied "student" versions of MSFT's Visual C++ 6 and 7.
From the first link:
GPL and NDA
* To: gcc at gcc dot gnu dot org
* Subject: GPL and NDA
* From: Richard Stallman
* Date: Thu, 19 Jul 2001 05:07:10 -0600 (MDT)
* Reply-to: rms at gnu dot org
GPL-covered code may not be distributed under an NDA.
To do so is a violation of the GPL.
If someone asks you to sign an NDA for receiving GPL-covered code that
is copyright FSF, please inform the FSF immediately. If it involves
GPL-covered code that has some other copyright holder, please inform
that copyright holder, just as you would for any other kind of
violation of the GPL.
It is possible for a person or company to develop changes to a
GPL-covered program and sign an NDA promising not to release these
changes *to anyone*. This is a different case. As long as these
changes are not distributed at all, a fortiori they are not
distributed in a way that violates the GPL.
However, if and when the changes are distributed to another person or
outside the company, they must be distributed under the terms of the
GPL, not under an NDA.
This doesn't have anything to do with the licensing of works that derive from GPL-licensed code. Your inclusion of this information makes you look either illiterate or careless.
From the second link:
From: Phil Nelson
Subject: Inaccurate information in `GNU' section
Date: 23 Apr 2004, 09:33:30
Hash: SHA1
On Thursday 22 April 2004 11:45 pm, Peter N Lewis wrote:
> > > I'm confused by this. If it "does not contain code from any
> >
> >> GPL-covered software", then surely you can release it under any
> >> license you feel like, whether it can run stand-alone or not. If I
> >
> >A program can be free of code from a GPL-covered program, but it links
> >with a library licensed under the GPL, then it has to be under the GPL
> >as well. It is this catch, if you will, that led to the LGPL.
>
> This could only be true if you ship the binary with it linked.
First, the disclaimer, IANAL.
Second, I wrote gdbm for the FSF and have had many many conversations with Mr.
Stallman about the GPL and how it applies to libraries. The key to the GPL
is the understanding of "derivative work". I'll use gdbm as an example.
If code directly calls any of the gdbm functions, it *is* a derivative work.
Therefore, any work that directly calls gdbm functions and is distributed
must be distributed under the terms of the GPL. Of course, the GPL is
transitive, thus any program that has a call path that ends up in a gdbm
function is required to be under the GPL. (If distributed.)
If your code is written for the original dbm interface it is not a derivative
work of gdbm, even if you happen to use the dbm interface to gdbm.
It would be considered a derivative work of dbm.
I don't know if this would stand up in a court of law, but I'm very sure this
is what Mr. Stallman intended for the GPL. I have requested that gdbm be
put under the LGPL so that programs that use gdbm don't have to be under the
GPL, but he has chosen to keep the GPL on gdbm. Therefore, whenever people
ask about gdbm, the GPL and their programs, I must tell them that if they
plan to distri
Not all machines hibernate correctly. :(
Moreover, not all of us have the good fortune of being able to work in an environment like emacs that supports true session saving.
From the PP:
Our group was recently informed "the simple act of shutting down PCs at night can save a company with 10,000 PCs over $260,000 a year". We kicked around the idea.
That's an alleged savings of $26/year/computer, or about $0.09/day.
52*2*0.09=$9.36 per PC per year.
I wonder how the reduction in hard drive life due to extra power cycling affects the overall cost savings and environmental impact of shutting down the machines on the weekend.
Don't try to screw with me, okay? I remember with SYN floods were new. I was already working with machines (in "IT" as you call it now) when the Morris Worm went around...
I work with an old programmer. He has a lot of knowledge. None of it is relevant. He's a horrible coder and an even worse CM.
Name one mainstream linux distro that has NX and ASLR on...
Oooh. There's that tricky word again.
... (not the weak forms) and is widely used please. Then we'll check its market penetration and add it to Vista's and see if we reach 1/3rd of all machines out there. ... Just because ASLR was mentioned in 1999 doesn't mean it's old hat today.
Less than 1/3rd of the machines out there on the internet are using ASLR. And of those, most are running Vista.
ASLR has been known about for ten years and had a Linux implementation for seven. Something that's widely known within a community is old hat. It doesn't matter if the proles aren't using it.
The people hacking it had more trouble than they expected because the machine was running SP1 which was new and the NX and ASLR made it a lot tougher, they had to rework their hacks they had prepared on Vista RTM before the contest.
Holy shit! Hackers who prepared a hack on a soft system had to modify their techniques when attacking a harder system? STOP THE PRESSES! THIS MUST GO ON THE FRONT PAGE!
Also, in 2009 Vista SP1 was hacked. In the latter case, it was hacked through a vulnerability in Java. They never were able to inject code onto the platform, but they didn't need to to hack in.
A system cannot be considered secure if insecure code running as an unprivileged user on the system causes it to be compromised.
If you want to say how the ASLR and NX in Vista SP1 means nothing, you should use research that is about Vista SP1 or later.
Read this. I've been to this guy's presentations. He knows his stuff:
http://taossa.com/archive/bh08sotirovdowdslides.pdf
many other OSes that present a compromise, like various forms of Linux or Vista.
If you're not a software developer, my Hardened Gentoo systems require no compromise. If, however, you want a usable backtrace, or to be able to set breakpoints in GDB, you're SOL. :(
"will implement"?
IIRC, PaX provides strong ASLR out of the box, and has for quite some time.