Comment by matthewmacleod | Hacker News Reader

Comment by matthewmacleod | Hacker News Reader

Parent Full thread

Imagine Swift becomes a wild success … Will you brush it off again as an understandable move by Apple?

I'm a bit unclear what you seem to think is lost under these circumstances versus copylefted software. Let's say Apple released Swift under the GPL. They would still be free to do all of the things you describe.

So as a developer right now, given that you have a plethora of open tools and an open ecosystem, why would you choose to use a language where a company has the opportunity to manipulate you in the future?

This is, as above, no different from the copyleft case.

So as a developer right now, given that you have a plethora of open tools and an open ecosystem, why would you choose to use a language where a company has the opportunity to manipulate you in the future?

It's extraordinary difficult to argue that Apple would be in a position to manipulate me as a developer if I started using Swift. For one thing, once the code is open-source – that's it. I can continue developing software using that language. If Apple decide not to release new features, and instead keep them proprietary, then the community is free to fork development.

More importantly, if it's important to members of the free software community, they are entirely able to develop capital-F-free software on top of Apple's release!

I'm really thankful for copyleft licensing and the amazing tools it's given us. But it's a vary complex issue, and I just don't agree that release under a license with fewer restrictions is necessarily a bad thing!

EDIT: I'm wrong see below

> "I'm a bit unclear what you seem to think is lost under these circumstances versus copylefted software. Let's say Apple released Swift under the GPL. They would still be free to do all of the things you describe."

Absolutely not. GPL means anything you put out must be published. If LLVM and Swift are GPL'd they can't develop additions to them, ship them on their Macbooks, and not publish the changes to the rest of the internet.

This is what's happening right now with LLVM.

If LLVM and Swift are GPL'd they can't develop additions to them, ship them on their Macbooks, and not publish the changes to the rest of the internet.

This is true if you redistribute the GPL code of others.

However, as the copyright holders, Apple are free to release their code under whatever license they wish, or to refrain from releasing it. They can ship it on MacBooks, and are under no obligation to release the source code.

You could argue that they couldn't accept GPLd submissions anyway, but since they'd require copyright assignment in any case it's a moot point.

Oh crap, you're completely right. Good point.

The bit about copyright assignment is significant though. There is a way for Apple to ensure the community that they will not screw them over for embracing their language. I think it would be prudent for the community to ask for that.

I guess one aspect is that LLVM (which Apple doesn't completely own, as far as I know) should have been copylefted to begin with, then any changes (like adding a language) would need to be shared with the community.

The exact same situation happened with Objective C and GCC. In that case, the Objective C frontend was released as free software.

> which Apple doesn't completely own, as far as I know

Apple don't own LLVM at all - they hired Chris Lattner who was one of the original directors of the project when it started at the University of Illinois.

> should have been copylefted to begin with

Why? To prevent the vast majority of companies with the money to actually support and grow such a project the way LLVM has, from doing so?

> Why? To prevent the vast majority of companies with the money to actually support and grow such a project the way LLVM has, from doing so?

You mean the ones that haven't provided LLVM with changes for the compilers they are selling?

They might not have "provided LLVM with changes for the compilers they are selling" but they have provided TONS AND TONS of changes and fixes and new features for the core LLVM infrastructure.

I could not care less about having the changes to their compilers, and I would even dislike it if LLVM's license required them to do so if that meant they wouldn't touch LLVM in the first place -- in which case we wouldn't get all these core LLVM improvements.

Replying to stephenr:

> Why? To prevent the vast majority of companies with the money to actually support and grow such a project the way LLVM has, from doing so?

No, because the freedom of new languages in LLVM would be guarenteed.

Again, you can of course have the opinion that this is not important, or not as important as the success of LLVM, but that's your decision.

>the freedom of new languages in LLVM would be guarenteed.

No, the GPL does not work that way.

If LLVM was GPL it wouldn't make any difference. LLVM is modular. You can write a front end that generates LLVM IR, then pass that LLVM IR file to LLVM to be compiled. It would be no different then a program that generates bitmaps and then passes them to another program to be jpeg encoded. The licence of the jpeg encoder doesn't affect the licence of your bitmap generating code.

This is why GCC is not modular.

Yes, they can. The GPL is a software license, that is, it authorizes the use of the software by persons or entities other than the legal copyright holder under certain conditions. The legal copyright holder can do anything they want with the software under US law, and any provisions of the GPL that purport to bind the legal copyright holder are not enforceable. I believe Richard Stallman even acknowledges as much.

LLVM is not GPL'd because the FSF didn't want it.

I got downvoted for this but: https://lists.gnu.org/archive/html/emacs-devel/2015-02/msg00...

Very interesting, I did not know that!

Maybe a better link (RMS's response when he recently discovers that Apple did indeed offer LLVM copyright assignment to FSF in 2005):

https://lists.gnu.org/archive/html/emacs-devel/2015-02/msg00...

Also note that 2005 is pre-GPLv3 and that Apple to this day dislikes GPLv3 so much that it continues to ship antiquated GPLv2 versions of GNU utilities in MacOS. So, my reading of the tea leaves is that the offer is not likely to be resurrected.

I'm not sure you're wrong. I was under the impression that contributors retain copyright over their portions under GPL.

This would require the copyright holder of the larger piece to get permission from all contributors before releasing under a different license, and I thought Ubuntu's CLA with Mir was to ensure they weren't trapped by this.

Nearly all companies who release stuff under the GPL require copyright assignation on contribution, actually giving the company far more control than if they simply used a liberal license. Using the GPL is more about making sure third parties can't lock up derivatives than that _you_ can't.

I've read below and I see where you've changed your thinking. That said, I've seen many people (not necessarily on HN) espousing the belief that once the copyright holder releases, they're now legally committed to releasing all their future changes as well.

Where does this line of thinking originate?

In a lot of GPL projects, this is essentially the case, because they either do not have copyright assignation, and have lots of external contributors (Linux, for example) or do have copyright assignation, but to a non-profit dedicated to the principles of the GPL (FSF stuff). People tend to forget about the third case, where a company has copyright assignation for contributors (Oracle stuff), or doesn't take contributions (common in dual commercial/GPL stuff).

> Absolutely not. GPL means anything you put out must be published.

Assuming that you don't own all the copyright. If you do (and Apple presumably would, in this instance), you're under no obligation to release source, and can produce alternate non-open-source versions. See MySQL for an example of this in action; also some non-GPL extensions to OpenJDK in Oracle JDKs, and Virtualbox (Oracle likes this model).

Reply on news.ycombinator.com