Final Fantasy 4 on the DS is great. Being able to play GBA games on a lighted screen without the buttons being scrunched is always a Good Thing (tm).
Of course, FF4 GBA has tons of bugs. Berserk is buggy, slowing down your character to the point of uselessness. The airship flying is choppy (which is dumb because the GBA has better mode 7 than the SNES did). The "darkness" debuff does nothing. Worst of all, do things in the wrong order on the menu screen and your saves will be erased. (Yes that bug is in the American release too.)
There is no technical reason that this can't be done. Even if you have a type with a destructor instead of "char", there is nothing bad about this. The compiler can store a copy of size for later and iteratively destruct them one by one. That's what would happen anyway if you used new[]. (If there is only one variable-sized array being used in the function, it can actually avoid storing the value entirely because it can deduce it by comparing ebp with esp.) GCC implements this feature already and there are no breaking problems with it.
Another thing C++ needs is a standard implementation of big integers. C# and Java have this. Even Perl has this. It's such a pain when doing cryptography to either reinvent the wheel or use some library with annoying DLL/SO dependencies (GMP).
Finally, someone needs to get on Microsoft's case and force them to implement stdint.h like everyone else has.
When backing up, generate a random "tape" key. Encrypt this "tape key" using a block cipher and your official key. Store the encrypted tape key several times at several locations on the tape. The locations of the key must be known without needing to read the tape to find them.
With that set up, encrypt the main contents of the tape with a stream cipher (say, RC4) with the tape key.
This way, damage to a certain area of the tape will not result in a complete loss of data. Using a random key for each tape eliminates the big cryptographic no-no of using a stream cipher key twice.
Actually, Phantasy Star Online had a back door, not a buffer overflow. A packet that Sega called RcvProgramPatch could be sent to the client containing assembly code that the game would then execute. This allowed Sega to patch holes in the game and check for cheats, but it eventually led to the downfall of the Gamecube security system. (Dreamcast PSO had this feature as well, but Dreamcast had other security problems =) )
Melissa
Making "New" fork the current page
on
Firefox Secrets
·
· Score: 1
In Firefox, the "New Window" command (control-N) makes a new window at your default page. Is there any way to make it so that "New Window" makes another window at the current page, effectively forking the current window? This is how Internet Exploder works.
Up until a few years ago, it was illegal to sell used games in Japan. If you have an imported game from this period it's likely it will have a NO RESALE icon on the back. A Japanese court decided it was illegal to do this, but it took an act of Parliament to actually reverse it (their legal system does not have precedent).
The system failed in Japan because it was against consumer rights.
Melissa
Re:Throw your Microsoft boxes into Boston Harbor!
on
The Demise of IP?
·
· Score: 1
Being under government edict is not an excuse under US patent law. Non-analog video recorders are required by copyright law to be susceptible to VHS Macrovision. Since blocking recording requires detecting Macrovision, and detecting Macrovision is patented by Macrovision, all digital recorder manufacturers must pay royalties to Macrovision. This is despite being under a government edict to do so.
Look on the back of a digital recorder for patents that are labeled "Licensed for limited viewing use only".
Visual Studio 8 has so many problems right now that it's not worth trying to use. Unless you use.NET, there is very little you stand to upgrade. So many outstanding problems were not fixed. It is quite obvious that Microsoft spent all their time on the new.NET and web features but almost no time with updates to the core standard C++ language.
Broken compiler - Complicated floating-point code will many times result in bad code generation without warning. Sometimes, the x87 register allocation algorithm fails and produces code that causes an x87 stack overflow exception at runtime. - Complicated templates, such as with 8 parameters or more, can cause internal compiler errors.
Annoyances - The default deprecation of much of the standard C library in favor of the Microsoft-only so-called "secure" versions is annoying. Compiling standards-compliant code now requires #pragmas or compiler switches to disable warnings. - The _M_AMD64 flag macro used in the prototype 7.1 x64 compiler was changed to _M_X64, grr.
Missing features - VC8 still does not have stdint.h and cstdint like every other compiler out there. - VC8 does not have built-in variable-sized stack allocation (IE, "char buffer[x]" with x a variable). GCC has had this for years now, and it is being seriously considered for the C++ standard. (It's not very hard to implement, after all.) - VC8 does not have __asm in x64. This is a terrible mistake. It would have been very easy for Microsoft to have implemented this feature (it's not very different from x86)! This is the most severe problem my company has with Visual Studio 8. We have resorted to using a GCC cross-compiler to compile our x64 C programs with embedded x64 assembly language.
The few minor features that were added - *printf now understand "%llx" and similar "ll" constructions. This was something they forgot when they finally added support for "long long" in VC7.1. - *printf now allow ordered arguments; this is important for localization. - swprintf now has the length parameter as standardized. (I don't like how swprintf takes this parameter - that should be called "snwprintf" - but I accept it because it's standard.) - OpenMP is implemented. I don't like OpenMP but at least it's a standard. -/hotpatch to put small stubs at the beginning of functions to allow hooking by placing a jmp instruction. These are a very bad idea to use in a game company like mine, but are otherwise good. - UTF-8 source files work, as do Unicode identifiers..NET scares me. It seems to indicate that Microsoft considers unmanaged code wholly deprecated. Here is a line from MSDN showing evidence of this:
Future versions of Windows will increasingly require components and applications to be verifiable.
Melissa
A single decision doomed Windows security
on
Ignore Vista Until 2008
·
· Score: 3, Interesting
People don't understand the truth about Windows security. It can be traced to a single fateful decision.
In Win32, every module (EXE/DLL) is contiguous in the process virtual address space. The code and data are next to each other in each module, but not between modules. The stack and heap are allocated as blocks at essentially random addresses. The memory layout looks like this:
empty code data empty stack empty code data empty heap...
The problem is that there is no single address that you could choose that says "only code is allowed below this address, and no code is allowed above this address".
On the x86, before AMD64, it was impossible to tell the processor that certain memory addresses cannot be executed. Anything that was readable was also executable. This means it is possible to execute from the data areas, a fatal flaw.
However, the x86 *does* have a feature that allows you to say "no code is allowed above this address". This is known as the "CS limit". By setting this, any attempt to execute from a data area would crash the program. Crashing the program is a lot better than taking over your computer.
Win32's memory layout prevents this feature from being used, because if you try to set a limit, either you have data in the code area, allowing exploits, or you have code in the data area, preventing legitimate code from executing.
AMD tried to correct this with the NX bit in the AMD64 chips, but it was too late. Too many Win32 programs rely on the ability to execute from a data section. As a result, in XP SP2 and Vista, the feature is only enabled by default in a few programs. You can turn it on for all, but then a lot of copy-protected games won't run.
Linux usually has the same problem. However, because most Linux programs come with source, it is possible to modify every application in the system to work this way.
Copyright is ownership of a specific number. Patents are ownership of a certain class of numbers *.
But trademarks are something else entirely. Their true use is fraud and libel prevention - you can't go around using someone else's name to sell your product or bash your competitors'.
Of course, there have been many recent abuses of trademark law, but at least trademark law has a foundation in reality. I don't see a problem with making a NES clone, but I do have a problem with calling it a NES.
* Or equivalently, a formula used to generate a class of numbers.
I prefer ignoring the fact that there's encryption. I hook the decryption function. When the place I need to patch gets decrypted, I found out and I write over that instruction.
This way, I don't need to bother with most of the protection.
What you just described is called the Xbox 360. The X360 dedicates one of its cores as a "hypervisor" that runs code in ROM inside the CPU. This code verifies that the kernel (the "supervisor") has not been tampered with. As this happens entirely inside the CPU, it will be *extremely* difficult to hack. The hypervisor intentionally reads from the CPU cache of the supervisor so that hypervisor RAM read bus cycles can't be singled out and attacked.
A lot is known about the X360 protection system.
Melissa
I'll wait for RSA-2048
on
RSA-640 Factored
·
· Score: 3, Interesting
I won't be interested until they do RSA-2048. Then we could factor the Xbox private key and do whatever we want.
Windows Vista has a partial solution to this. When you insert a CD, it asks you what you want to do with it. You can run the autorun script, or load it directly in Windows Media Player (which effectively bypasses the protection). Or do nothing and run your favorite ripping program.
It is a felony for someone to tell you how to remove it, or how to avoid installing it. Hence the "Post Anonymously".
What greatly disturbs me is that the Sony DRM is permanent. If you completely format your hard drive to start over, you are still committing a felony. The DMCA makes no distinction.
At some point, you *have* to trust the client. The speed and teleport hacks in WoW work because of a tradeoff. It would be extremely expensive for the server to be doing world geometry collision checking for 3000+ characters and who knows how many objects.
Sure, you could do things like prevent moving more than a certain amount of distance in a game tick. In fact, we know the server *already* does this, because people got banned for teleporting even before Warden (which according to WoW!Sharp was 1.6.1). Even with this, you can still do things like walk through walls, swim in the air, and walk up cliffs, because the server has no concept of the world geometry.
The ultimate limit of implementing things on the server would be for the server to render the game for you and stream it to you over video. But even then you're not safe. You can write bots that read the pixels and react accordingly. Now your only two options are to place Turing tests into your game, or modify the reward structure so that automated processess aren't worth it. (Turing tests could be something blatant, like the captcha on this Slashdot "reply" page, or something less obvious, like randomizing colors and textures so that human object recognition is required for analyzing the position of game objects.)
Can you imagine what kind of resources the server would need to have in order to check the X/Y/Z coordinates of every single client? The WoW client is barely able to do this on its own, with *1* user...
Teleportation and speed hacking can be detected, but not things like swimming in the air or walking up mountains.
Science can not prove that there is no God, science can likewise not prove that there are no blue swans with yellow spots or a Tooth Fairy. You can't prove the non-existance of something.
When you give finite bounds to your system, you can prove things do not exist. For example, if we consider the Universe to be a large thermodynamic system, science has proven that there cannot exist a refrigerator that doesn't consume energy or produce heat.
It's more accurate that science can't disprove the existence of something outside of its framework. Someday, we'll prove that a God cannot exist inside our universe. But science could never determine if a God existed outside the Universe.
The opposite is true as well. Neither science nor religion could ever prove the existence of something outside the system.
This is exactly the problem with Macrovision. The law *specifically requires* devices to refuse to record Macrovision-protected content. In order to make a device do that, you must pay Macrovision patent royalties.
The Government has no right to say "you must license patent X", when X doesn't cover what you want to make.
This would be equivalent to building the only road between two cities through private property then expecting you to pay tolls to the property owner.
Melissa
Re:Hypothetical question....
on
Patents vs. Secrecy
·
· Score: 2, Insightful
I think that they can only classify things if the federal government pays for it.
If I'm wrong, and you manage to factor big numbers quickly, I suggest you do three things:
- Immediately get it on the Internet as fast as you can. Make sure Slashdot sees it, especially. Place a notice that patent rights are reserved.
- Arrange to have it published somewhere, preferably by a lesser-known journal that would be less inclined to listen to the Feds. You have 1 year to do this under American law.
If the Feds never know about it before you reveal it massively, there is little that they can do. If they arrest you, it would look terrible to the administration in power. It's all about the PR.
Melissa
Pedophiles who prey on children they find on the Internet are rare, compared to those that prey on children they already know. The fact that they know the children makes them more able to abuse the sense of trust the child has for them. It also means the children are more accessible.
Slashdot Mirror
Final Fantasy 4 on the DS is great. Being able to play GBA games on a lighted screen without the buttons being scrunched is always a Good Thing (tm).
Of course, FF4 GBA has tons of bugs. Berserk is buggy, slowing down your character to the point of uselessness. The airship flying is choppy (which is dumb because the GBA has better mode 7 than the SNES did). The "darkness" debuff does nothing. Worst of all, do things in the wrong order on the menu screen and your saves will be erased. (Yes that bug is in the American release too.)
Melissa
I want variable-sized stack arrays:
...
void function(size_t size)
{
char array[size];
}
There is no technical reason that this can't be done. Even if you have a type with a destructor instead of "char", there is nothing bad about this. The compiler can store a copy of size for later and iteratively destruct them one by one. That's what would happen anyway if you used new[]. (If there is only one variable-sized array being used in the function, it can actually avoid storing the value entirely because it can deduce it by comparing ebp with esp.) GCC implements this feature already and there are no breaking problems with it.
Another thing C++ needs is a standard implementation of big integers. C# and Java have this. Even Perl has this. It's such a pain when doing cryptography to either reinvent the wheel or use some library with annoying DLL/SO dependencies (GMP).
Finally, someone needs to get on Microsoft's case and force them to implement stdint.h like everyone else has.
Melissa
When backing up, generate a random "tape" key. Encrypt this "tape key" using a block cipher and your official key. Store the encrypted tape key several times at several locations on the tape. The locations of the key must be known without needing to read the tape to find them.
With that set up, encrypt the main contents of the tape with a stream cipher (say, RC4) with the tape key.
This way, damage to a certain area of the tape will not result in a complete loss of data. Using a random key for each tape eliminates the big cryptographic no-no of using a stream cipher key twice.
Melissa
Actually, Phantasy Star Online had a back door, not a buffer overflow. A packet that Sega called RcvProgramPatch could be sent to the client containing assembly code that the game would then execute. This allowed Sega to patch holes in the game and check for cheats, but it eventually led to the downfall of the Gamecube security system. (Dreamcast PSO had this feature as well, but Dreamcast had other security problems =) )
Melissa
In Firefox, the "New Window" command (control-N) makes a new window at your default page. Is there any way to make it so that "New Window" makes another window at the current page, effectively forking the current window? This is how Internet Exploder works.
Melissa
I couldn't ever get a date as a geek guy so I became a geek girl. That's one way to solve the problem =)
Melissa
Up until a few years ago, it was illegal to sell used games in Japan. If you have an imported game from this period it's likely it will have a NO RESALE icon on the back. A Japanese court decided it was illegal to do this, but it took an act of Parliament to actually reverse it (their legal system does not have precedent).
The system failed in Japan because it was against consumer rights.
Melissa
Being under government edict is not an excuse under US patent law. Non-analog video recorders are required by copyright law to be susceptible to VHS Macrovision. Since blocking recording requires detecting Macrovision, and detecting Macrovision is patented by Macrovision, all digital recorder manufacturers must pay royalties to Macrovision. This is despite being under a government edict to do so.
Look on the back of a digital recorder for patents that are labeled "Licensed for limited viewing use only".
Melissa
Visual Studio 8 has so many problems right now that it's not worth trying to use. Unless you use .NET, there is very little you stand to upgrade. So many outstanding problems were not fixed. It is quite obvious that Microsoft spent all their time on the new .NET and web features but almost no time with updates to the core standard C++ language.
/hotpatch to put small stubs at the beginning of functions to allow hooking by placing a jmp instruction. These are a very bad idea to use in a game company like mine, but are otherwise good. .NET scares me. It seems to indicate that Microsoft considers unmanaged code wholly deprecated. Here is a line from MSDN showing evidence of this:
Broken compiler
- Complicated floating-point code will many times result in bad code generation without warning. Sometimes, the x87 register allocation algorithm fails and produces code that causes an x87 stack overflow exception at runtime.
- Complicated templates, such as with 8 parameters or more, can cause internal compiler errors.
Annoyances
- The default deprecation of much of the standard C library in favor of the Microsoft-only so-called "secure" versions is annoying. Compiling standards-compliant code now requires #pragmas or compiler switches to disable warnings.
- The _M_AMD64 flag macro used in the prototype 7.1 x64 compiler was changed to _M_X64, grr.
Missing features
- VC8 still does not have stdint.h and cstdint like every other compiler out there.
- VC8 does not have built-in variable-sized stack allocation (IE, "char buffer[x]" with x a variable). GCC has had this for years now, and it is being seriously considered for the C++ standard. (It's not very hard to implement, after all.)
- VC8 does not have __asm in x64. This is a terrible mistake. It would have been very easy for Microsoft to have implemented this feature (it's not very different from x86)! This is the most severe problem my company has with Visual Studio 8. We have resorted to using a GCC cross-compiler to compile our x64 C programs with embedded x64 assembly language.
The few minor features that were added
- *printf now understand "%llx" and similar "ll" constructions. This was something they forgot when they finally added support for "long long" in VC7.1.
- *printf now allow ordered arguments; this is important for localization.
- swprintf now has the length parameter as standardized. (I don't like how swprintf takes this parameter - that should be called "snwprintf" - but I accept it because it's standard.)
- OpenMP is implemented. I don't like OpenMP but at least it's a standard.
-
- UTF-8 source files work, as do Unicode identifiers.
Future versions of Windows will increasingly require components and applications to be verifiable.
Melissa
People don't understand the truth about Windows security. It can be traced to a single fateful decision.
...
In Win32, every module (EXE/DLL) is contiguous in the process virtual address space. The code and data are next to each other in each module, but not between modules. The stack and heap are allocated as blocks at essentially random addresses. The memory layout looks like this:
empty code data empty stack empty code data empty heap
The problem is that there is no single address that you could choose that says "only code is allowed below this address, and no code is allowed above this address".
On the x86, before AMD64, it was impossible to tell the processor that certain memory addresses cannot be executed. Anything that was readable was also executable. This means it is possible to execute from the data areas, a fatal flaw.
However, the x86 *does* have a feature that allows you to say "no code is allowed above this address". This is known as the "CS limit". By setting this, any attempt to execute from a data area would crash the program. Crashing the program is a lot better than taking over your computer.
Win32's memory layout prevents this feature from being used, because if you try to set a limit, either you have data in the code area, allowing exploits, or you have code in the data area, preventing legitimate code from executing.
AMD tried to correct this with the NX bit in the AMD64 chips, but it was too late. Too many Win32 programs rely on the ability to execute from a data section. As a result, in XP SP2 and Vista, the feature is only enabled by default in a few programs. You can turn it on for all, but then a lot of copy-protected games won't run.
Linux usually has the same problem. However, because most Linux programs come with source, it is possible to modify every application in the system to work this way.
Melissa
Copyright is ownership of a specific number.
Patents are ownership of a certain class of numbers *.
But trademarks are something else entirely. Their true use is fraud and libel prevention - you can't go around using someone else's name to sell your product or bash your competitors'.
Of course, there have been many recent abuses of trademark law, but at least trademark law has a foundation in reality. I don't see a problem with making a NES clone, but I do have a problem with calling it a NES.
* Or equivalently, a formula used to generate a class of numbers.
Melissa
I prefer ignoring the fact that there's encryption. I hook the decryption function. When the place I need to patch gets decrypted, I found out and I write over that instruction.
This way, I don't need to bother with most of the protection.
Melissa
...fourteen years...
Patents last for 20 years, not 14. Exception: drug patents last for 7.
Melissa
The USPTO has stated that their policy on perpetual motion patents can be changed by providing them with a working model.
Melissa
What you just described is called the Xbox 360. The X360 dedicates one of its cores as a "hypervisor" that runs code in ROM inside the CPU. This code verifies that the kernel (the "supervisor") has not been tampered with. As this happens entirely inside the CPU, it will be *extremely* difficult to hack. The hypervisor intentionally reads from the CPU cache of the supervisor so that hypervisor RAM read bus cycles can't be singled out and attacked.
A lot is known about the X360 protection system.
Melissa
I won't be interested until they do RSA-2048. Then we could factor the Xbox private key and do whatever we want.
Melissa
I forgot to check "post anonymously".
Windows Vista has a partial solution to this. When you insert a CD, it asks you what you want to do with it. You can run the autorun script, or load it directly in Windows Media Player (which effectively bypasses the protection). Or do nothing and run your favorite ripping program.
It is a felony for someone to tell you how to remove it, or how to avoid installing it. Hence the "Post Anonymously".
What greatly disturbs me is that the Sony DRM is permanent. If you completely format your hard drive to start over, you are still committing a felony. The DMCA makes no distinction.
At some point, you *have* to trust the client. The speed and teleport hacks in WoW work because of a tradeoff. It would be extremely expensive for the server to be doing world geometry collision checking for 3000+ characters and who knows how many objects.
Sure, you could do things like prevent moving more than a certain amount of distance in a game tick. In fact, we know the server *already* does this, because people got banned for teleporting even before Warden (which according to WoW!Sharp was 1.6.1). Even with this, you can still do things like walk through walls, swim in the air, and walk up cliffs, because the server has no concept of the world geometry.
The ultimate limit of implementing things on the server would be for the server to render the game for you and stream it to you over video. But even then you're not safe. You can write bots that read the pixels and react accordingly. Now your only two options are to place Turing tests into your game, or modify the reward structure so that automated processess aren't worth it. (Turing tests could be something blatant, like the captcha on this Slashdot "reply" page, or something less obvious, like randomizing colors and textures so that human object recognition is required for analyzing the position of game objects.)
Melissa
Can you imagine what kind of resources the server would need to have in order to check the X/Y/Z coordinates of every single client? The WoW client is barely able to do this on its own, with *1* user...
Teleportation and speed hacking can be detected, but not things like swimming in the air or walking up mountains.
Melissa
Windows NT already allows forward slashes in the Win32 API. CreateFileW translates them to backslashes automatically before calling NtCreateFile.
Melissa
Science can not prove that there is no God, science can likewise not prove that there are no blue swans with yellow spots or a Tooth Fairy. You can't prove the non-existance of something.
When you give finite bounds to your system, you can prove things do not exist. For example, if we consider the Universe to be a large thermodynamic system, science has proven that there cannot exist a refrigerator that doesn't consume energy or produce heat.
It's more accurate that science can't disprove the existence of something outside of its framework. Someday, we'll prove that a God cannot exist inside our universe. But science could never determine if a God existed outside the Universe.
The opposite is true as well. Neither science nor religion could ever prove the existence of something outside the system.
Melissa
This is exactly the problem with Macrovision. The law *specifically requires* devices to refuse to record Macrovision-protected content. In order to make a device do that, you must pay Macrovision patent royalties.
The Government has no right to say "you must license patent X", when X doesn't cover what you want to make.
This would be equivalent to building the only road between two cities through private property then expecting you to pay tolls to the property owner.
Melissa
I think that they can only classify things if the federal government pays for it. If I'm wrong, and you manage to factor big numbers quickly, I suggest you do three things: - Immediately get it on the Internet as fast as you can. Make sure Slashdot sees it, especially. Place a notice that patent rights are reserved. - Arrange to have it published somewhere, preferably by a lesser-known journal that would be less inclined to listen to the Feds. You have 1 year to do this under American law. If the Feds never know about it before you reveal it massively, there is little that they can do. If they arrest you, it would look terrible to the administration in power. It's all about the PR. Melissa
Pedophiles who prey on children they find on the Internet are rare, compared to those that prey on children they already know. The fact that they know the children makes them more able to abuse the sense of trust the child has for them. It also means the children are more accessible.
Melissa