Hacker News Reader

Comment by deafpolygon | Hacker News Reader

> I dislike the GPL, and the stigmatization of proprietary software for any reason, and the notion that it's inherently evil and unethical.

A flaw in OP's assertion is that I feel that most of what we have today wouldn't exist if it weren't for the GPL licensing.

Edit: Despite RMS's unique bent on things, he hasn't been wholly wrong about the nature of free/nonfree software. If the Linux kernel, or the GNU software suite hadn't been GPL'd in the first place - we wouldn't be using it today. Instead, it would have died out or we'd be using some proprietary fork. I think, in the long term, GPL software benefits everyone because we can fork and share modifications. It fosters a mindset and community that's beneficial to a hacker culture (as in tinkerer, not as in black).

View on HN

> A flaw in OP's assertion is that I feel that most of what we have today wouldn't exist if it weren't for the GPL licensing.

That is pure RMS rethoric. The narrative that free software exists because of GPL is a compelling one (and sounds plausible), but its not based on reality. It was quite common in the end of the 70's and beginning of the 80's (specially after the creation of the C language) to distribute software in source form, instead of binary packages. Some early examples of this that still exists today are the BSD utilities (granted, now without the Berkeley clause), laying in the ground the notion that GPL is determinant in preserving OSS as an ecosystem.

The "trigger story" about not having access to a printer driver source code is also often simplified. The printer was a prototype of a commercial product, and he was offered the source under an NDA. More details on this in https://www.oreilly.com/openbook/freedom/ch01.html

> If the Linux kernel, or the GNU software suite hadn't been GPL'd in the first place

While there is no arguing that GPL is at least in part responsible for the success of Linux, it wasn't GPL until version 0.99. The major reason why you use Linux today is because of the "unix wars" (https://en.wikipedia.org/wiki/Unix_wars) - the lack of a free i386 unix system is what drove Linus to write one in the first place. If AT&T was less beligerant at the time, you'd probably be running some form of BSD :)

> you'd probably be running some form of BSD :)

You mean macOS? Or Playstation 3/4? Proprietary, non-free form of BSD, yes.

You use unfree Linux too...for example google. What people seam to forget, under GPL you don't need to give your code back if your DON'T redistribute it.

And that's why google forbids the use of AGPL internally.

>You mean macOS?

https://github.com/apple/darwin-xnu

Why dont you know that? If your a apple user and so interested in opensource?

Why wouldn’t you be using one of the many freely distributed BSDs if software freedom is so important to you? That’s what I do, anyway.

You mean macOS? > The relevant parts of macOS for this discussion are actually opensource (kernel & most userland utilities), albreit not updated often;

> Playstation 3/4?

Does the fact that product XYZ has a BSD-based product reduces or affect the continuity of the opensource project? No, it doesn't. And that is the beauty of BSD. Some companies build closed source products on it, some companies actually give back contributions (such as Netflix). Its fine either way.

> Proprietary, non-free form of BSD, yes.

Well, Microsoft used BSD code to some extent in previous versions of Windows (afaik tcp stack, services for unix, ftp client, etc). On the other hand, I do use Android-based devices - which are, for practical purposes, mostly a generic non-free form of Linux. In fact, not only the source code (when you can get it easily) is almost useless for practical use, but often bootcode and drivers are proprietary signed blobs - without which the device is just a fancy brick. Also, I'd argue there is little value brought into the linux project just because it is a core Android component. How is this scenario different from a "non-free form of BSD"?

> Does the fact that product XYZ has a BSD-based product reduces or affect the continuity of the opensource project? No, it doesn't. And that is the beauty of BSD. Some companies build closed source products on it, some companies actually give back contributions (such as Netflix). Its fine either way.

The irony of the GPL is that it does not require giving back. The only requirement is to provide source code when you do binary distribution. There is no provision about giving back to the upstream developers.

Forced code dumps are not giving back and projects like Linux do not consider them to be a form of giving back. Sending useful patches and responding to maintainer feedback would be giving back. The GPL is no better at obtaining that than a 2-clause BSD license.

> You mean macOS? Or Playstation 3/4? Proprietary, non-free form of BSD, yes.

As someone who has been dedicated to OSS since I started writing software and also wrote some proprietary code for the health care industry (which was not open sourced mainly because there was no point since it would have been a code dump useless to people because hardware patents were involved), I have this feeling:

I would rather good code be reused by proprietary stuff, because someone is going to put that code in systems where failures kill people and it is better that they reuse good code to prevent deaths. It is ridiculous to embrace a philosophy in software licenses where people dying because of your decision to deny good code to people writing proprietary software is okay. The OSS versus proprietary debate should not be done in a manner that causes fatalities.

It would be neat if GPL advocates only complained about proprietary software, but some of them are hellbent against other open source software. They bash BSD-style licensing and treat anything not under what are regarded to be GPL compatible OSS licenses even worse.

I have done a significant amount of work in OpenZFS and the hostility from certain GPL proponents has been ridiculous. Meanwhile, the way that Linux's kernel module interface interacts with the GPL means there is no problem for using ZFS with Linux since ports from other systems in the form of kernel modules are not derived works works. This has passed legal review by actual lawyers, one of which is a co-author of the GPLv3:

https://softwarefreedom.org/resources/2016/linux-kernel-cddl...

The view that the GPL is what enabled OSS to develop is also ridiculous. The original BSD license for example, predates the GPL. If anything, OSS developed in spite of the GPL, since controversies surrounding the GPL have been the single most counterproductive aspect of the OSS community. It is a wonder developers have any time to develop software given all of the headaches caused by GPL supremacists.

If the GPL had never existed, the community would have avoided GPL supremacists and we would all be happily using BSD/Apache/Mozilla licensed software.

I argue the opposite.

The GPL consumed all the (OSS developer) oxygen in the room in a number of useful spaces (gcc, for example), where it was far too expensive for far too long to start a non-GPL’d replacement.

This is still the case for things like GPG, where we have zero integration with mainstream e-mail clients, and zero penetration into the “normie” consumer market.

The GPL creates an inescapable sub-critical mass. You can almost never grow beyond the “hacker” market of other GPL software developers, because the people who cannot code — but can spend money — have no reasonable way to pay for development.

The GPL is a cancer on open source that has hobbled OSS (and the general progress of the industry!) at almost every juncture.

Imagine if BSD and Linux developers could cross-pollinate code, for example. How much more vibrant would OSS be for the interaction of those different branches of the UNIX tree?

GPL OSS that has succeeded has done so despite the GPL, not because of it. The GPL is viral, coercive, and captures and co-opts the value of OSS to further the aims of the GPL, rather than the OSS development itself.

> Imagine if BSD and Linux developers could cross-pollinate code

Nothing in the GPL prevents you from making your own derivative. Anyone can take that derivative and make their own. What it prevents is forking and closing the source of the software. We wouldn't have code to cross-pollinate at this stage if it were not for the GPL.

Of course we’d have code to cross-pollinate; an entire ecosystem of liberally licensed code exists.

The GPL absolutely prevents you from making your own derivative unless you adopt the same coercive, viral license.

> Of course we’d have code to cross-pollinate; an entire ecosystem of liberally licensed code exists.

> The GPL absolutely prevents you from making your own derivative unless you adopt the same coercive, viral license.

And what incentive I have to allow everyone to create closed derivatives of my code and use them to take freedom away from their users(most likely including me).

If they love closed code so much, they are free to hire as many developers that will write it for them. But it costs money while taking someone else code is cheap. I don't want to help people who want to enrich themselves on someone else work without giving anything back.

> a number of useful spaces (gcc, for example),

What is LLVM & Clang then, chopped liver?

It was literally decades before LLVM and Clang become a reality, and even longer before they could supplant the GNU toolchain for most popular architectures.

GCC’s dominance held back new languages, and language front-end development, for decades.

The LLVM project started in 2000, and the first release was 2003.

https://en.wikipedia.org/wiki/LLVM#History

The first release of GCC was 1987:

https://gcc.gnu.org/wiki/History

That's a 13 year gap, 16 if you count first release rather than the start of work.

That is not, to quote you, "literally decades". It is significantly less than 2 decades.

Secondly, there was lots of language development between 1987 and 2000 -- for convenience, we can say the 1990s.

That is the era when Python ('95), Ruby ('95), Java ('96) and Javascript ('96) and many others first appeared. Perl is contemporaneous with GCC.

It is not true to say that GCC stifled development. What is true is that this is the era when a lot of interpreted and VM/bytecode/runtime-based languages originated, including both the JVM and .NET CLR.

The focus of FOSS language development was arguably elsewhere, maybe _because_ GCC evolved to be a serious competent professional tool in that timeframe.

Additionally, there was a lot of R&D in compiled languages in that decade or so, including Delphi, VB6, the first native C++ compiler from Zorland, and so on.

It took time for FOSS code to catch up and surpass proprietary compilers, yes, but with one good compiler, there was less pressure to do so, and so people worked on scripting languages instead.

Given how poorly GCC's rust front-end development has been going, I doubt Rust would have been made without LLVM.

> with one good compiler, there was less pressure to do so, and so people worked on scripting languages instead.

See this comment regarding your idea that GCC is a good compiler:

https://news.ycombinator.com/item?id=34008284

It is a buggy compiler and it is a wonder that it works as well as it does. It is no surprise that LLVM is better.

As for people working on scripting languages instead, it is more that they were forced to work on scripting languages because the GCC codebase was designed to be incomprehensible to all but a select few. This is obvious if you try reading GCC. If you have not, I highly recommend that you pick some random files and begin reading. I can read random parts of the Linux kernel and generally follow along what is being done. I can do the same with LLVM to some extent, which I know since I have patched it locally in the past. I cannot do that with GCC. It is a mess. :/

I personally have no particular feelings either way about GCC; I don't write code any more, and when I did, I barely used compilers or xNix to do it.

All I am doing is responding to @catiopatio's comment: that apparently it was good _enough_ to inhibit competitors for a decade and a half.

The economics of FOSS are not yet well understood (and FWIW, I think @catiopatio is totally wrong about the GPL). There were lots and lots of compilers in the proprietary space, and aggressive competition was the norm.

The headhunting of Anders Hejlsberg from Borland to Microsoft shifted the entire competitive market of programming languages for _decades_, but FOSS people don't tend to notice such things.

(He developed Delphi. Without him, Borland floundered and died. With him, MS grafted a compiler onto VB, making a vastly successful product, and then abandoned it, and came up with .NET and C# instead, muddying the waters in ways that still shape the industry.)

The code quaity of GCC leaves much to be desired. Their mailing list has some figures from Coverity, where the defect rate in 2017 was at 1.67 defects per thousand lines of code for their 2.5 million lines versus 0.62 defects in LLVM per thousand lines of code for LLVM's 5.1 million lines:

https://inbox.sourceware.org/gcc/793886b8-2fd8-e593-f065-239...

At present, GCC is at 1.60 defects and LLVM is at 0.76. LLVM was at 0.32 in July, but it jumped to 0.76 in August around the time it would have branched to start LLVM 16 development and has stayed there since then.

GCC has a policy of hiding its coverity data contrary to the spirit of open source, so I have not been able to see its historical data. I could forking it on github and submit my own builds to coverity to see what the historical data likely was, but I do not see much reason to do that beyond knowing the current state is a mess.

Anyway, I felt like expanding on your GCC remark, since it is a pile of bad code that is definitely holding back OSS and the industry.

Reply on news.ycombinator.com