Hacker News - It's not for hackers and it's hardly news.

[AnOminous]

Put simply, the big Silicon Valley companies need to join together to beat down a smaller company that's mad about the blatant copyright violation.

It isn't remotely a copyright violation. These texts are widely publicly available not just in electronic form but in every library in the country as well as anywhere that still sells news on paper. AI can use it just as Google can use it to generate its search engine database. The AI model includes nothing but synopses of content, and generates content indepedently.

 

[UERISIMILITUDO]

It isn't remotely a copyright violation.

I disagree. Now, I hate copyright in all of its forms, but I believe the idea of a world where only large corporations get to have copyright, and everyone else gets to have their work effectively stolen by neural network nonsense, to be even worse.

The AI model includes nothing but synopses of content, and generates content indepedently.

Here's an old example of the machination blatantly copying someone's code:
https://twitter.com/DocSparse/status/1581461734665367554 (nitter) (archive)

Even in the comments there, I see fuckwits arguing against the man enforcing his copyright, because it can get in the way of human progress, supposedly. This is no different than people who excuse self-driving cars running over people in ways humans never would, because it's ultimately for the best, and it just so happens rich people are the main recipients of whatever benefits come from this progress. MicroSoft, Google, and Apple Computer have successfully installed themselves as gods to idiots who pretend to worship science, and gods should get to do whatever they please to us mere mortals.

I wonder if I could get a leaked copy of MicroSoft's various source codes, put them into neural network nonsense, and pretend that whatever comes out be legal. According to MicroSoft, training is fair use.

The only good thing that can possibly come out of this is the total destruction of copyright as a concept.

 

[AnOminous]

The only good thing that can possibly come out of this is the total destruction of copyright as a concept.

Yeah, I'm not really seeing the minus on that one. Computer code is an odd beast, though, because fairly often, simple but specific things can only be done by identical or nearly identical code. When there are only a very few ways to express an idea, or rarely only one, such content is uncopyrightable. There's even a name for this. It's called merger.

 

[UERISIMILITUDO]

Computer code is an odd beast, though, because fairly often, simple but specific things can only be done by identical or nearly identical code.

Trust me, programs which do identical things are very different in languages that are very different. The Common Lisp solution to a problem will very easily be very different from an Ada solution, and those will be very different from a solution in APL or some machine code.

When there are only a very few ways to express an idea, or rarely only one, such content is uncopyrightable.

Yes. The ReactOS project was worried about this, because it turned out reverse engineering some part of Windows resulted in identical machine code, but it was determined to be the only way to do that, and therefore fine. That's something else that pisses me off, however: MicroSoft is trying its damnedest to spread confusion in this copyright laundering, but woe be to those whom MicroSoft can claim in court to have possibly, maybe, probably looked at a source code leak of its shitty software.

 

[Monopoly Supporter]

The only good thing that can possibly come out of this is the total destruction of copyright as a concept

My opinions on Copyright (I don't agree with the "le IP bad!"):

Copyright, patents, and maskwork protections exist to make the goods that they protect into club goods rather than public goods. A book's text for example is non-rivalrous as duplicating the text does not deprive someone else of the text. This works in the same way as a movie theatre. One person sitting in a (large enough) movie theatre does not deprive someone else of the ability to sit in the movie theatre. The issue is that you have to somehow make back the fixed costs (writing the book, running the projector, etc.). Tresspass laws keep people from sitting in the theatre without paying for a ticket (even though sitting in the theatre does not cost the theatre any money), and someone can physically prevent you from entering. However, it is not as simple as something like a book's text, which can be duplicated an infinite amount of times. IP laws provide the excludability to turn the naturally public good of book text into a club good, like a (big enough) movie theatre or Costco, where you pay a license fee for access to something. This is the American purpose of Copyright law (European law focuses more on moral rights but the American system is almost purely economic). IP Laws provide a way to convert the fixed costs of writing a book, making an invention, writing a program, or designing a chip, into marginal costs.

Let's take creating a chip for example. Many people are not aware of how cheap the physical silicon actually is. All the money spent is in R&D and licensing. IIRC, the marginal cost of an ASIC containing a 4-core ARM CPU is around $4. The cost per core of licensing the ARM CPU is $1-2 per unit. Then there is the $1MM-$10MM fixed tape out cost. Then there is the $100MM+ in R&D that your company spent on designing the ASIC. So the MSRP of the ASIC is going to be $20-50. Issue: Jing Ling Precision Electronic Co. Ltd. decides that it likes your company's design and takes a dieshot and duplicates the chip and tells it for $5. You are fucked now because someone who doesn't have to make back the fixed costs just stole your design. This is the rationale behind Patents and MaskWork Protections.

Copyright works the same way. If nobody is paying for the good (free riders) then WTP is stuck at $0 and the fixed costs cannot be made back.

The alternative to telling a product are the hellspawns known as the "as a Service" business model and the advertisement based business model. The advertisement model is typically associated with public goods (non-rivalrous and non-excludable) like OTA TV and radio. This is the only way they can make back their fixed costs. XaaS is a reaction to software piracy and the tendency of people to fall into spread out cost traps (for only pennies a day!) XaaS doesn't rely on copyright protections because it never lets the customer directly use the software. It always has to communicate back with the server.

Linux development is paid for by companies that use SaaS, Advertisement, or Copyright Protected business models. AMD, Linaro, Samsung, IBM, NXP, ARM Holdings, Renesas, NVIDIA, and TI all develop Linux as it is an investment in their (maskwork/patent protected) hardware. Oracle develops Linux because of their database products (that aren't free). Linux development is done to support proprietary software and hardware and it can be more economically viable to do it this way (ESR thinks this).

In countries without IP protections, (e.g. China), products are made in assholeish ways to deter reverse engineering, copying, or counterfeiting. The Chinese sand the labels off their chips. The XGecu EEPROM burners detect and brick suspected counterfeit devices. FTDI is counterfeited so much that they have TWICE put code in their drivers that changes the USD IDs to 0000:0000 and causes garbage to be read from the serial port.



But modern Copyright Laws are just total and utter bullshit. Look at the comparison in duration for IP:
Mask Works: 10 years (registration required)
Patents: 20 years (registration required)
Copyright: LIFE OF THE AUTHOR PLUS 70 YEARS (registration not required so that some dipshit can pretend that some memo on the back of a napkin or their Tweets are copyrighted material and you must pay royalties).

See an issue? Copyright should last for 14 years + MAYBE an extension for another 14 yearsJust like the founders intended. With mandatory (free) registration via. U.S. Mail after a 12 month grace period.

 

[9gfuwegw9j9]

Several more fuckwits don't understand the difference between a human being and a machine poorly pretending to be a human being through text alone; in all fairness, these people probably are very close to an inanimate object pretending to be a human being

I think there's a very distinct difference between humans reading text and actually understanding it in one piece and using gradient descent to effectively compress the text.

 

[AnOminous]

See an issue? Copyright should last for 14 years + MAYBE an extension for another 14 yearsJust like the founders intended. With mandatory (free) registration via. U.S. Mail after a 12 month grace period.

It's even more insane for corporations, where it's 120 years from creation or 95 years from publication, whatever is shorter. At least protecting something for the life of the author is rewarding the person who actually created it, and potentially allowing them to leave a source of income to their kids.

I think it should be 14 years or something like that for material authored by or PURCHASED by a corporation, to discourage rent-seeking scumbag leeches like RIAA from vampirizing creators, and force them to deal with the creators again in another 14 years if they're still around. Also drastically narrow work made for hire because that's some bullshit, at least for creative works.

 

[Monopoly Supporter]

On the topic of Walter Bright: this has to be my favorite thread (A) on the DLang Forums.

The artism is question:

Spoiler

OMGSISA

 

[9gfuwegw9j9]

On the topic of Walter Bright: this has to be my favorite thread (A) on the DLang Forums.
View attachment 5624888
The artism is question:

Spoiler

View attachment 5624890
OMGSISA

Walter Bright is a known CHUD and TRANSPHOBE

 

[Rusty Metal Skull Gun]

Yes. The ReactOS project was worried about this, because it turned out reverse engineering some part of Windows resulted in identical machine code, but it was determined to be the only way to do that, and therefore fine.

I'm fairly sure large chunks of the ReactOS code are straight-up stolen. Isn't a coincidence how they're targeting the obscure Windows Server 2003, which just happens to be the latest one with a substantial source code leak...?

 

[Monopoly Supporter]

I'm fairly sure large chunks of the ReactOS code are straight-up stolen. Isn't a coincidence how they're targeting the obscure Windows Server 2003, which just happens to be the latest one with a substantial source code leak...?

If someone could find non-exported, non-documented, non-included-in-the-pdbs functions that exist in both ReactOS and NT, that would be a clear sign that somebody was viewing leaked code. But unlike UNIX/Linux, NT is extremely well documented without needing source code. The PE32+ documentation is some of the best I have seen. But I have no doubt they used SoftICE and IDA to reverse engineer everything.

Reimplementing Win32 isn’t that novel. HX DOS Extender reimplemented USER, GDI, ADVAPI, WinSock, and DirectDraw to run on DOS. NT however is not reimplemented.

The reason they target NT 5.x and not NT 6+ is because Microsoft changed a lot of how NT works in the backend starting with Vista (see MinWin).

 

[The Real Slim Shady]

ReactOS is a very strange beast for my taste. Reimplementing Win Server 2003 as open source does not have any rationale in my opinion. It's over 20 years old and nothing special of an OS for today's standards. Reimplementing it as a research project does not make sense because research projects have to bring something novel to the table, and ReactOS aims specifically for being just like old NT, so there is no space for features that are not in NT.

Reimplementing it as the Glorious Russian National Operating System failed as they tried four times to get funding from the Duma, and does not make sense for all the previous reasons. I think the reason the russian government were uninterested was it is an OS for x86 and the architecture that Russia is developing for their internal use is SPARC (ELBRUS CPUs). And nowadays they just buy western ARM components for their weapons anyways.

Reimplementing it for business reasons is nonsense because of it's unclear legal status. In fact reimplementing it for most other reasons is pointless because Server 2003 ISOs and license keys are flying around the Internet in all shapes and flavors, so if you really need it and do not want to buy it you can just install it. This is also true for XP and it's derivatives like WMC or XP Embedded. If you need full NT compatibility, you just install NT. Nobody will go after a pirated XP install in this day and age.

Every other operating system tries to bring something new to the table. BSDs come up with OS architecture ideas or mechanism implementations. Haiku tries to build on BeOS performance and legacy. QNX is reasonably good RTOS that once powered car radios and BlackBerries. FreeRTOS and the other bunch are racing who is building a better real time OS. Plan 9 is attempting to reinvent UNIX concepts but up to 11. MINIX is designed to be compact, student friendly and found it's way in Intel ME.

All ReactOS meanwhile has to say is "We are NT (nearly), but open source!"

 

[Rusty Metal Skull Gun]

I don't think it's necessarily just copy-and-paste, but they're clearly running the code through a decompiler and then cleaning it up. I remember being confused because a Windows function was doing something unexpected, and looking at the ReactOS code - it was there, but there wasn't the // Yes, Windows is batshit here or whatever that you'd expect.

And they reimplement all the internal functions too, rather than just the syscalls or exports. There's no rationale for doing that unless you're doing a function-by-function decompilation.

 

[Monopoly Supporter]

I don't think it's necessarily just copy-and-paste, but they're clearly running the code through a decompiler and then cleaning it up. I remember being confused because a Windows function was doing something unexpected, and looking at the ReactOS code - it was there, but there wasn't the // Yes, Windows is batshit here or whatever that you'd expect.

And they reimplement all the internal functions too, rather than just the syscalls or exports. There's no rationale for doing that unless you're doing a function-by-function decompilation.

"It was a lucky guess"


_HalpClockInterrupt also has a FIXME comment about updating performance counters, something that ReactOS doesn't implement but NT does. A lot of ReactOS is like this. A bunch of stub functions that do nothing but return error codes. While they probably aren't viewing the source code from Microsoft, they aren't doing a "clean room reverse engineering" by just reading the documentation. _HalpClockInterrupt is NOT exported by hal.dll. Somebody is using DDK PDBs in conjunction with a debugger/decompiler/dissasembler which explains why ReactOS shares the names of labels that are public (as in, the public assembler directive) with NT but none of the local labels.

Reimplementing it for business reasons is nonsense because of it's unclear legal status. In fact reimplementing it for most other reasons is pointless because Server 2003 ISOs and license keys are flying around the Internet in all shapes and flavors, so if you really need it and do not want to buy it you can just install it. This is also true for XP and it's derivatives like WMC or XP Embedded. If you need full NT compatibility, you just install NT. Nobody will go after a pirated XP install in this day and age.

It makes even less sense given that nearly every OEM computer sold in the last 23 years has a Windows license attached to it that can do everything that ReactOS can do but better. There is literally no market for ReactOS beyond running it in a VM (because you are sure as hell not getting it to work on real hardware) out of curiosity. Businesses that need Server 2003 already have Server 2003 (ReactOS doesn't have any of the server components anyways). Businesses that need NT 4 for their sewing machines or whatever already have NT 4. New OSes don't just "become successful." The ONLY reason the Linux Operating System took off is because Net/2 BSD was in deep legal shit with USL, causing it to be delayed and replaced with 4.4BSD-Lite which still needed things added to it to boot on i386. If a complete bootable i386 BSD had been released before Linux was, BSD would be the primary UNIX-like operating system, not Linux.

 

[AnOminous]

While they probably aren't viewing the source code from Microsoft, they aren't doing a "clean room reverse engineering" by just reading the documentation.

That may bite them in the ass. They should really put more effort into separating the "writer" from the decompiler. The clean room method of decompiling and describing as some kind of pseudocode, then having someone who's never seen the original code turning it back into something functional pretty much guarantees no copying of nonfunctional code, duplication of variables, formatting, etc. all of which actually is copyrightable.

 

[UERISIMILITUDO]

The ONLY reason the Linux Operating System took off is because Net/2 BSD was in deep legal shit with USL, causing it to be delayed and replaced with 4.4BSD-Lite which still needed things added to it to boot on i386.

It's GNU/Linux. Linux is just a kernel, and not even a particularly good one.

 

[teriyakiburns]

I'm fairly sure large chunks of the ReactOS code are straight-up stolen. Isn't a coincidence how they're targeting the obscure Windows Server 2003, which just happens to be the latest one with a substantial source code leak...?

There was an issue in which part of the leaked windows 2000 codebase ended up in reactos. They had to audit their entire codebase afterwards, to ensure nobody else had contributed leaked code, which is one of the bigger reasons why development stalled for so long.

Their reasons for implementing like-for-like behaviour is because enough applications use "undocumented features" in the windows API, that not implementing them would be detrimental to the project's goals.

 

[Monopoly Supporter]

It's GNU/Linux. Linux is just a kernel, and not even a particularly good one.

Linux is the Operating System. GNU is the Operating Environment. The OS-kernel distinction only makes sense with microkernels. For example, GNU Hurd is often called a "kernel." This is false. Hurd is an operating system, Mach is the kernel. Specifically, it is a collection of microkernel servers and the kernel, which is GNU Mach (a really shitty kernel at that). On Operating Systems like NT the Kernel-OS distinction is greater as it is split into Subsystems ("personalities"), the Kernel (Ke), HAL (Hal), Executive (Ob, Mm, Reg, Io, etc.) (which while included in ntoskrnl.exe is considered separate from the kernel because of VMS/MICA heritage) and the drivers. With monolithic systems like Linux, everything is baked into the kernel (except for ld-linux which I consider part of the operating system).

This is why you will often hear that Windows 1.0 was not an operating system but was actually an operating environment or "shell." This is correct. Windows 1.0 is a collection of dynamic libraries that get linked against real mode executables. All the operating system functionality is handled by DOS. In the same way that GEM for DOS isn't an operating system.

Advocates of the OE as OS definition (which is limited to mainly Richard Stallman) will say that the operating system cannot do anything by itself. This is correct. Operating systems cannot do anything by themselves, that is the point. If you try to start up the MS-DOS OS without a command line interpreter, it will ask for the path of a command line interpreter, typically COMMAND.COM, but you can swap COMMAND.COM out for 4DOS. This is because the MS-DOS OS is contained in IO.SYS (the drivers) and MSDOS.SYS (the kernel), not COMMAND.COM, which is just a shell, like GNU Bash is for the Linux Operating System.

I use this definition because it is the most commonly used on in academia and in books. For example, The XINU Approach by Bell Labs defines what an operating system is not:

First, an operating system is not a language or a compiler. An operating system does not depend on an integrated language facility Second, an operating system is not a windowing system or a browser. Although windowing mechanisms rely on an operating system, a windowing system can be replaced without replacing the operating system. Third, an operating system is not a command interpreter. In a modern operating system, the command interpreter operates as an application program, and the interpreter can be changed without modifying the underlying system. Fourth, an operating system is not a library of functions or methods. Despite the close relationship, library software remains independent of the underlying operating system .

I consider coreutils to be a combination of methods packaged as programs due to the UNIX software tools philosophy. The book goes on goes on to define an operating system:

The essence of an operating system lies in the services it provides to applications. An application accesses operating system services by making system calls. In source code, a system call appears to be a conventional function invocation. At runtime, however, a system call and a conventional function call differ. Instead of transferring control to another function, a system call transfers control to the operating system, which performs the requested service for the application. Taken as a set, system calls establish a well-defined boundary between applications and the underlying operating system that is known as an Application Program Interface (API). The API defines the services that the system provides as well as the details of how an application uses the services.

Of course with Microkernels this definition gets washy because the microkernel's job is to provide IPC between servers that implement drivers from userspace. So In the case of microkernels we can go look at Andy Christmasstree's MINIX Book's definition of operating system:


If you go to a library and look at the selection of books about Operating System Development, they all discuss implementing operating systems, not operating environments like GNU. None of them discuss writing a windowing system, a stream editor, a shell, or a text editor.

Based on this, I take the view that the software that runs on a computer, can be split into four parts: the bootloader (GRUB, NTLDR, Syslinux, Limine), the operating system (NT, Linux, MINIX, BSD, GNU Hurd, VMS, OS/2), the operating environment (Win32, System V, GNU, BSD, OS/2), and add-on software (Microsoft Word, GIMP, Firefox, GTA V).

 

[skybutler]

Isn't this meant to be a drama thread and not tech sperging

 

[9gfuwegw9j9]

the idea of a monolithic kernel and a microkernel is pretty weird, most kernels use a combination of both. go and tell me which category the NT kernel falls into