GPLv2 Libraries — Is There a Point?

PiSkyHi writes "I understand that if I build an application that links with a library that is licensed under GPLv2, I must also make my application GPL2. I can see that value in this for an application. But for a library, what's to stop me separating my program into a GPLv2-compliant client app that talks to the rest of my (choose my own license) application?"

Comments

Step 1: see GPL

[seanadams.com]

The GPL is not explicit on this, it's just that linking is the established rule of thumb. Last I checked, the specifics of multiple processes vs linking are only mentioned in the FSF's FAQ, so it's only a guideline for interpretation.

I think a reasonable test would be to ask: is my program still mostly useful even if the GPLd helper/plugin is removed (modulo the specific removed function)?. If so, then I think it could be argued that your program is NOT a derivative work and that the GPL helper is governed the same as, say, a GPLd user application bundled with a commercial Unix/OSX distro. Personally I don't think it should matter how exactly it communicates with your code - what makes launching a process any different from a function call here?

Conversely: artificially doing contortions with your software to move essential libraries out to a separate app is not only in bad faith, but it doesn't work around the license at all. And if you ever had to argue otherwise, anyone turning up your slashdot story would not probably work in your favor.

IMHO the GPL, even v3, needs some work to clarify this question and also to close the hole for the software-as-a-service industry to modify GPL code without reciprocating.

Re:Step 1: see GPL

[PhrostyMcByte]

IMHO the GPL, even v3, needs some work to clarify this question and also to close the hole for the software-as-a-service industry to modify GPL code without reciprocating.

I would like to see the GPL get clarified, too. Some companies play pretty loose with the GPL's definition -- like MySQL saying their protocol documentation is under GPL, and interpreting that to mean any code you write using that documentation must be under the GPL too.

FSF has done part of what you mention, though -- see the AGPL. It explicitly closes the SaaS hole.

Re:Step 1: see GPL

[mattbee]

Not sure about now, but a few years back, MySQL made this pretty explicit - they considered any kind of bundling of MySQL to be "linking". i.e. if your application could only work with MySQL, and you shipped it with your app, that was linking, even though they might have only communicated through a socket. The whole concept of program linking has come on quite a bit since 1990, so I agree with the parent, if there aren't alternative components that can be swapped out with a GPL'd one, that's linking.

Re:Step 1: see GPL

[Tony Hoyle]

MySQL also have this wierd clause that shipping an app that uses their library, even if you don't ship any of their code, is 'distribution' and subject to the Mysql license.

OTOH it did help us standardise on postgress for the OSS stuff.. Mysql wasn't worth the legal hassle to support any more.

Re:Step 1: see GPL

[Tony Hoyle]

FSF say that linking - even dynamic linking - creates a combined work (in memory) that is subject to the GPL.

Of course that theory has holes (eg. I'm not distributing that combined work.. it stays in memory, so the license it's under is irrelevant), but it's stuck for now.

A separate app is in no way a combined work as it doesn't share any memory space.. and it'd be hard to make a license that applied a restriction like that.

Re:Step 1: see GPL

[geminidomino]

And a good thing, too, or the Harry Potter books would have run afoul of the earlier series, The Worst Witch.

And so would a good number of the Discworld Novels! That would have SUCKED!

No Rincewind, no Orangutan Librarian, no parade of Arch-Chancellors... all because someone wrote a a series touching coming-of-age about an inept witch...

Yeah, I'd be pissed.

Re:Step 1: see GPL

[geminidomino]

As an example, you can't copyright the idea of a school for wizards and/or witches, although you can copyright a specific story built around that idea.

This isn't as clear as it sounds. If you write a story about a school called Hogwarts for wizards one of whom is named Harry Potter, you can expect to get sued for infringement under the theory that your story is a derivative work of J.K. Rowling's -- even though you've only copied ideas, not expressions.

I'm pretty sure that "Harry Potter," as a character, is an expression of the idea "student wizard".

Re:Step 1: see GPL

[mdwh2]

But as soon as you start copying names, I don't think that's merely what we're talking about "ideas" here, that would count as expressions. (In general, characters can be copyrighted AFAIK - "expression" is broader than simply "copying the exact book itself", it's just not as broad as concepts such as "a school for wizards". Or it shouldn't be, imo.)

Re:Step 1: see GPL

[sjames]

In general, the test would be "can I simply delete the GPL portions from the project and still have everything work (possibly by substituting in already existing commercial alternatives). If the answer is no, then you are at the very least violating the spirit of the license and are on shaky legal ground.

Another way to look at it would be if the proprietary portion of the project is deleted entirely, is the GPL portion useful standing alone (that is, is the proprietary portion merely a value add?). If so, release the GPL project separately on it's own merits, then re-iterate from there.

If the GPL portion is not independently useful, it is effectively a library of the proprietary app and so the whole thing is in clear violation of the GPL.

The problem with defining it better is that there are matters of degrees. I can easily drink two glasses of water. If I take two glasses of water from L:ake Michigan, nobody will even notice. A glass of water isn't much, so it stands to reason I can manage one more and that one more won't hurt the lake. That doesn't mean I can just one more my way to drinking Lake Michigan or that doing so would be harmless to the lake. Establishing where the thresholds are is a hard problem.

Re:Step 1: see GPL

[Abcd1234]

No. Standard Unix commands have had a stable interface since before many of you were born.

Uh, so what? libc has been standardized for at least as long, and yet if glibc, the GNU implementation of the libc specification, were under the straight GPL, then anything linked against it would fall under the GPL.

In short: API has absolutely nothing to do with it. The question is, what is the definition of "combined work", and that goes right to the heart of the GPL (and has been a point of controversy every since it appeared).\

So far, we know that linking directly against a library, such that program A calls library function B, results in a combined work. The question before us is, if program A sends a TCP messages to program B, and program B calls library function C, while we admit that program B must fall under the GPL, must program A?

I think it's clear the answer *must* be 'no', as anything else would be ridiculous (last I checked, my web browser didn't inherit the license of the web application it was interacting with). In which case, the OP's point is very much valid.

It depends on the API. As far as I know, no one is patenting APIs.

Patenting? What? What the hell does that have to do with the GPL, which is a license on a *copyrighted* work?

Can you copyright an API such that a black box implementation can not be made without violating the copyright? I don't know.

You can't copyright an API at all. Furthermore, reverse engineering is legal for the purposes of achieving compatibilty. There is, after all, a reason why the PC actually exists in the form we see it to day: Compaq reverse engineering the API of the IBM BIOS.

Umm...

[Phroggy]

what's to stop me separating my program into a GPLv2-compliant client app that talks to the rest of my (choose my own license) application?

Umm... nothing?

If you're writing your application from scratch without using anybody else's libraries, you're free to release it under whatever license you like, even if it happens to talk to a GPL'd client plugin thingie, and even if you wrote that GPL'd client plugin thingie around somebody else's GPL'd library.

Why do you imagine that somehow there's a problem here?

Re:Umm...

[seanadams.com]

If you're writing your application from scratch without using anybody else's libraries, you're free to release it under whatever license you like, even if it happens to talk to a GPL'd client plugin thingie, and even if you wrote that GPL'd client plugin thingie around somebody else's GPL'd library.

Which GPL are you reading that says this is OK? If the license doesn't permit it then it's a question for copyright law of whether you're making a derivative work. I would say that in this case you ARE, unless it's a bone fide plugin in the sense that it can be unplugged too.

Re:Umm...

[piojo]

If you write the Foo app on your own without any connection to a GPL licensed library, then Foo obviously cannot be said to be a derivative work of anything. What if you then add some capability for Foo to interact with a GPL'd library?--for instance, adding Readline capability to an app that previously used the BSD Editline library.

Would this make your program a "derivative work"? That's a question for our legal system--it's not the FSF's right to define what constitutes a derivative work (nor is it the right of the any copyright holder--this is a question of legal definition).

And... err... I'm obviously not a lawyer.

Re:Umm...

[SL Baur]

I would tend to agree. It's fairly well established that license does not extend across process boundaries - Clearcase interface code in emacs does not make emacs proprietary.

There's a lot of FUD at the library level. I think you're probably safe with dynamic linking/loading. Stallman hated me for introducing .so elisp modules in XEmacs, but it was on ideological grounds not legal grounds - it made it possible to distribute executable code in emacs sans source code.

I *would* recommend paying attention to discussion on lkml on the nuances between EXPORT_SYMBOL and EXPORT_SYMBOL_GPL. If your work is a derived work from a GPLed interface, then yes your work falls under the GPL.

One of the most significant events of 2009 in IT has been Microsoft releasing code under GPL. The intent of the GPL is that if I give you something under GPL and you modify it to do something interesting and distribute that, you have to share it back to the community. The tricky part is defining "modify". In US copyright law, that means creating a derived work. Is linking against a library a derived work? Sorry, that's above my pay grade.

I find it interesting that it is Microsoft that may validate the whole thing. They are certainly learning. Stallman's copyright assignment requirement for all GNU code, it gives the FSF total control over the work. That's more than Bill Gates managed to achieve. And dang it, I never received any of the $1 consideration payments I was supposed to receive when I signed FSF copyright assignment papers.

TL;DR Yes, I just likened FSF to Microsoft.

Disclaimer: I'm an ex-XEmacs maintainer, not a lawyer. Your mileage may vary. Offer void where prohibited. Contact your local dealership for more details.

Re:Umm...

[Anonymous Coward]

A history lesson: GCC includes an objective C front-end. This front-end was originally implemented at NeXT for their new ground-breaking computer in the 1980's, and published as a proprietary component of the NeXT system software. They did not link directly to the GCC backend, but implemented an intermediate file format that was written by the proprietary frontend and read by a patched GCC, and they only published the source to the latter. FSF argued that NeXT was in violation of the GPL, and after months of legal wrangling NeXT gave up, published the front-end source under GPL and the case did not go to court.

AFAIK the FSF still holds that the only way to use a GPL component in proprietary software is through an API that is supported by multiple implementations, not all of which are GPL, regardless of how the connection between the proprietary software and the GPL software is technically implemented.

Re:Umm...

[swillden]

For what it's worth, I agree with you.

FWIW, the IP attorneys I've discussed this with don't agree with you. They take the GPL at the FSF's word with regard to the meaning of "derivative work". I asked a couple about this and the response was that because the clearly-stated intent of the license was to preclude linking except to code distributed under compatible terms, that a court would most likely agree with the copyright holder. Basically, the "linking implies derivation" notion is reasonable and well-understood by all parties, so the court would accept it.

Of course, these were corporate attorneys looking to protect the corporations assets, and lawyers in such situations generally err on the side of an overabundance of caution.

Re:Umm...

[Jason Earl]

The idea that linking implies derivation stems from the days when static linking was the norm. It is hard to argue that your application isn't a derivative work if it actually includes a copy of someone else's library in the binary. Dynamic linking, especially for programs that are designed to use more than one compatible library are a slightly murkier case, but only slightly so. That's why the GNU software that clones other software libraries (like glibc) tend to be under the LGPL. After all, in those cases the FSF wants to encourage people to use their library instead of the library they are cloning. The few GPLed libraries tended to be things like readline, which were unique to GNU.

The reason that most lawyers are very careful about not crossing the line when it comes to copyright is twofold. First, most of the companies and people with skin in the copyright game are heavily invested in strong copyrights. That's why you'll never see Microsoft challenge the GPL in court, for example. Microsoft actually favors a far more restrictive interpretation of copyright, and the last thing that it wants is to see copyright weakened, even if it would mean getting access to piles of GPL software for use in their proprietary products.

The second reason that lawyers tend to be very careful about copyright is that the U.S. has ridiculously draconian copyright laws. You are free to disregard the GPL when creating your application, but if the judge disagrees with your interpretation you could end up in serious legal trouble. Not only would your profits be forfeit, but the owner of the GPLed code you used could press criminal charges. Unless you feel like following in the footsteps of Jammie Thomas it is simply far safer to take the FSF at its word when it comes to interpretting the GPL.

In fact, you should probably follow the advice of whoever wrote the software you are planning to interoperate with. For example, the folks at MySQL forward an alternate interpretation of the GPL that says that software that talks to their server needs to be GPLed unless you have a commercial license. Personally, I think that's a land grab, but I wouldn't consider testing it myself.

Re:Umm...

[slash.duncan]

That's the beauty of the GPL, in that if the opposition (defendant, perhaps plaintiff in a declaratory judgment) doesn't accept the license, everything defaults back to standard copyright protections, which don't allow use/copying/distribution in the first place, thus being far stricter than the GPL. They are thus free not to accept the GPL, but should they do that, they better make very sure they are not infringing standard copyright, either, because that's the fallback.

That's also why the legal system is likely to rely on the author's (or license author's) clearly laid out interpretation of his chosen license as well -- unlike EULAs and the like, which seek to impose additional restrictions on top of copyright as a condition of granting limited use under copyright, the GPL does not restrict what copyright law already permits. As such, the author's interpretation of the conditions under which he grants those additional permissions holds a lot of weight because the defaults would be far stricter.

IMNLO (NL=non-lawyer), this is why the very high majority of GPL violation cases ultimately settle out of court. Once it's pointed out that should the GPL NOT be valid, the fallback is far stricter copyright law, which would put the violator in a far worse position than they are in with the GPL violation, most reasonable people quickly see the light, and decide the chances of risking even WORSE penalties for full copyright violations are simply not worth it. Thus, they find that abiding by the GPL in the future is in their best interest, and all that must then be agreed is the penalty for past violation. The FSF and friends seem to be relatively flexible in that regard, since the object is after all to get the code out there, and a settlement for costs plus a normally undisclosed but apparently reasonable donation, plus some measures (license compliance officer or etc, agreement on release breadth and deadlines) to ensure the followthru, seems to be common. Compared to the penalties a company would pay for violating a proprietary license or raw copyright law, that's very reasonable indeed, and most violators quickly see how mutually satisfactory the offered settlement conditions actually are.

Re:Umm...

[Bluesman]

you don't make any use of GPL'd components if you're not prepared to put everything under GPL, that is the intent of the license

That's exactly right. Anyone who says differently, especially after reading the Clisp exchange, is deluding themselves.

That might be what you want (yay GPL), but if you're not thrilled about the GPL, you probably shouldn't be using any GPL software in your project, or you'll run into these issues. Take heed, original poster.

It's why I prefer the BSD license; I don't want my hobby to result in me talking to a lawyer.

Enhancements to the library stay LGPL

[poliopteragriseoapte]

The point is that enhancements to the library stay LGPL. I did this myself. I needed to modify LGPL libraries for the purposes of my application. I modified the libraries, and I am redistributing them under the LGPL. My main application is distributed under BSD license, and uses the LGPL libraries. The libraries make perfect sense separately (and in fact, I have people using them for reasons unrelated to the application I built). I think that, out of courtesy, and probably out of need, you need to make sure that the libraries are available separately.

Re:Enhancements to the library stay LGPL

[Anonymous Coward]

You're completely full of shit. Here's section 6 of the LGPL:
 
 

6. Revised Versions of the GNU Lesser General Public License.

The Free Software Foundation may publish revised and/or new versions of the GNU Lesser General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns.

Each version is given a distinguishing version number. If the Library as you received it specifies that a certain numbered version of the GNU Lesser General Public License âoeor any later versionâ applies to it, you have the option of following the terms and conditions either of that published version or of any later version published by the Free Software Foundation. If the Library as you received it does not specify a version number of the GNU Lesser General Public License, you may choose any version of the GNU Lesser General Public License ever published by the Free Software Foundation.

If the Library as you received it specifies that a proxy can decide whether future versions of the GNU Lesser General Public License shall apply, that proxy's public statement of acceptance of any version is permanent authorization for you to choose that version for the Library.

There's nothing in there about releasing applications that link to modified LGPL libraries under LGPL. Furthermore, the situation you describe is completely bizarre, because there's absolutely no way of determining what "upstream" is. The grandparent's post indicates that he forked the LGPL library. Which is a perfectly valid thing to do. And he's licensing the derivative work of the LGPL library under the LGPL. There should be no legal issues there.

Author's wish

[ianare]

This is exactly why the LGPL was created. Or sometimes you will have a GPL lib with the linking or classpath exception. You will find most libs are licensed under these, or even more permissive terms.

Therefore, if the lib in question is explicitly licensed under normal GPL, it's the author's wish that any apps that use it must be GPL compatible. I think it's only fair to follow the author's wish.

Re:Author's wish

[Bryan Ischo]

I can second this approach. I wrote a libary for some rather obscure functionality that took me about 120 hours to produce. I am a reasonably good software developer and it's not really possible for any organization to produce the same library for much less than 120 man-hours. A couple of companies found out about my library and decided that it was worthwhile to license my library from me rather than re-writing it themselves. They got a well-written, functional, well-tested library for much less than they would have had to pay their own developers to produce, and I got some "free money" for something I did for fun anyway.

I released my library under GPLv2 and NOT LGPL. I STRONGLY recommend that if you write your own library, that you release under GPL. You get the best of both worlds: you get to share your code with others who want to share their own code back, and you also get to charge $$$ to closed source companies who want you to re-license your code to them under a license that allows them to integrate it into their product(s) without having to change their own product(s) licenses.

It's a win-win. I can't even comprehend what the advantage of LGPL could be when releasing libraries. Unless, I guess, you really like just giving your stuff away for free when you could instead be paid ...

Of course, RMS would have not approve of using the GPL in this way; I am sure that he'd want the code to always be free rather than being incorporated into closed-source projects as well. And I absolutely respect his position on this, but as the creator of the code, I get to decide what I think are fair and moral licensing terms. Sharing with sharers and charging money to those who charge money, seems perfectly moral to me.

terms vs license

[NiteMair]

Doesn't the license basically stipulate that you must release your code under the terms of the license? That doesn't necessarily mean you have license your code as GPL.

This is actually a common FUD discussion that occurs between developers who use MIT/BSD license for their code, and are afraid to link to GPL libraries because it might force them to release their code as GPL. I'm pretty certain that as long as their code is available under the same terms as GPL code, they can license their code however they wish otherwise. It just means if a GPL nut comes a knockin', they'll have to provide the code as if it was GPL'd.

Correct me if I'm wrong (and I often am).

Re:terms vs license

[Hatta]

Doesn't the license basically stipulate that you must release your code under the terms of the license? That doesn't necessarily mean you have license your code as GPL.

What's the difference? If the original license is GPL, and you must release derivative works under the same terms as the original license, then you are releasing under the terms of the GPL i.e. you are licensing it under the GPL.

Re:GPL Fanatics

[msuarezalvarez]

There is nothing wrong with you not using my code if you do not like my conditions, either...

No, you misunderstand

[kripkenstein]

I understand that if I build an application that links with a library that is licensed under GPLv2, I must also make my application GPL2

That is NOT correct.

The criteria is 'derivative work', not 'link to'. Linking is sometimes a rule of thumb in this area, but it isn't decisive.

Note that 'derivative work' is a legal term, not a technical one. So before you try to circumvent the GPL in this way, consult a lawyer.

Talk to the authors

[minsk]

Quite aside from the legal meaning of the GPL, you could always ask the authors what they intended. If they want to prevent proprietary applications from using the library, as promoted by , that should influence your decision.

IMO, if the internals of your application are dictated by a GPL'd library -- *regardless* of how that library is linked, integrated, called, or plugged in -- your application is a derivative work. To separate them, put the library aside. Do a proper design. Gather scenarios you need the plugin to support. Document them. Write a new API. Document it. Collect some alternative libraries. Verify that you were not unduly influenced by the GPL'd one.

Or, you know, GPL the application.

Re:Talk to the authors (link)

[minsk]

Pardon, fumbled the link.

Why you shouldn't use the Lesser GPL for your next library

Seperate address spaces are the main test

[ChaseTec]

You can't make a linked wrapper library since the GPL wrapper would be GPL too. See: http://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#GPLWrapper However it sounds like you are talking about a service based wrapper. Then I'd say it depends on how integrated your service wrapper is with your main program. If you use separate processes but lots of IPC and shared memory then I'd think you app is to tightly integrated making it a derived work. If you make a network based service then I'd say you are legally clear even if you'll probably piss off the library author. See: http://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#NFUseGPLPlugins You could also make your app GPL but put a lot of the functionality in non-free plugs, see: http://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#FSWithNFLibs Of course you could also just release you app as GPL and not worry about it. What is stopping you from using the GPL?

Re:GPL Fanatics

[nmb3000]

There is nothing wrong with you not using my code if you do not like my conditions, either...

So much for being open and free.

I tend to find myself in agreement with the OP. More and more it seems like the GPL is not a license designed to promote free and open source software but instead comes across as a "I'll scratch your back but then you are legally obligated to scratch mine" contract. I realize this probably stems from an irrational fear of proprietary software -- fear that F/OSS code will get eaten up and essentially stolen if there wasn't the backing of the GPL. That combined with the attitude of "for-profit companies shouldn't get something for nothing" isn't very endearing and it's no surprise Microsoft can sell the idea that "the GPL is a virus". The truth is that they really aren't hitting too far from the mark, except, perhaps, with respect to intent. I do think the GPL authors mean well and lack the malice that Microsoft seems to be suggesting exists.

It seems like anyone that really values and supports F/OSS would prefer LGPL or BSD style licenses. Allowing derivative works to remain closed source isn't a detriment to the open components and it really does look better to closed source companies. This should lead to more people adopting and using F/OSS, both open and otherwise. That's good, isn't it?

Re:The GPL Problem

[minsk]

I've always found that position funny...

Some people share something freely, with the requirement that anyone building atop it does likewise. They're zealots.

Other people require you to agree to a contract before looking at their program. Disallow reverse engineering. Disallow redistribution. Disallow extension, enhancement, improvement, porting, circumvention. They're normal.

Dude, what?

Re:GPL Fanatics

[Anonymous Coward]

but it is free...free from abuse by those who want to take and not give anything back.

Re:GPL Fanatics

[msuarezalvarez]

'Free' is not a magic word that means something absolutely, there is no $DEITY-given dictionary which provides a canonical meaning for the word. Whether I call it 'free', 'open', 'shared', 'blue' or 'pretty' is completely irrelevant: what is relevant is what users and developers are allowed to do with the code and under what conditions.

This silly bickering about what kind 'free' is freer was boring already decades ago...

Re:GPL Fanatics

[Knuckles]

I realize this probably stems from an irrational fear of proprietary software -- fear that F/OSS code will get eaten up and essentially stolen if there wasn't the backing of the GPL.

What's irrational about that?

"Linking" is the weakest link of GPL

[ingo23]

The whole idea of "linking" is too specific to a technology or architecture. It assumes that some code is "compiled" into some form of library and the functions become available through a "linking" process during the build.

What if I call functions in my library over SOAP? The application is still "linked" to a service ("library"), but using that logic a web browser would have to be GPLed if you access any GPLed web server.

What's to stop you?

[mysidia]

RMS and others, (The community) possibly are there to stop you.

Probably some people have some creative interpretations of what it means to link against a GPL library, also. To call such things at least a gray area if not outright violations.

If/when they find out, they might point to your software as an example of bad practice, or put you in the "GPL Violations hall of shame", or some such.

This is especially interesting with respect to scripting languages... If I write a PHP script that utilizes mysql_* function calls of PHP, which can only be invoked when PHP has been linked against the Mysql Client library....

Is my PHP script essentially linked against the GPL'ed mysqlclient library, and therefore, (if I distribute my PHP script commercially without source) a GPL violation?

Re:GPL Fanatics

[mrsteveman1]

What's your definition of "use" here?

Say there is a nice GPLv2 library i know is distributed with every Linux distribution, something small. I write a closed source app that dynamically links to this GPLv2 library at runtime and calls a few functions from it. Who is violating the GPLv2 here? The developer? They haven't actually distributed the GPLv2 code, how can they be bound by its terms?

Is it the user? The GPLv2 explicitly says that the end user does not need to agree to the license at all just to run the program, so how can THEY be bound by it either?

Re:GPL Fanatics

[Trepidity]

the attitude of "for-profit companies shouldn't get something for nothing" isn't very endearing

I don't really see what's wrong with that. Isn't the normal way economic transactions work in a market economy with copyright laws something like: if you want to use use a part of my copyrighted work in your own, you have to get a license from me, usually involving payment?

I see using the GPL as sticking with that as the default, but making a special exception that if you give blanket permission to the general public to use and distribute your own code, royalty-free, in both original and modified versions, then you may use my code royalty-free under the same terms. But if you want to stick to the normal copyright model, then I will also, and we can agree on terms in the usual manner.

Basically I don't see how someone who uses the normal approach to copyright licensing in their own products could possibly object to me asking them to negotiate a license in order to use my code as part of their product.

Re:GPL Fanatics

[Draek]

It seems like anyone that really values and supports F/OSS would prefer LGPL or BSD style licenses. Allowing derivative works to remain closed source isn't a detriment to the open components and it really does look better to closed source companies. This should lead to more people adopting and using F/OSS, both open and otherwise. That's good, isn't it?

In theory, yes. In practice, however, it seems volunteers and companies contribute a lot more to GPL'ed projects than to BSD'ed ones, otherwise we'd be arguing for the "Year of FreeBSD on the Desktop" instead of Linux.

Personally, I like the idea of a BSD kernel with LGPL libraries and GPL applications, it seems to me as the best balance between freedom and widespread usage I can see. However, if devs believe otherwise and GPL everything I have no problem with it, the market will decide which option is better and, so far, it seems they're right.

Re:GPL Fanatics

[Jurily]

There is nothing wrong with you not using my code if you do not like my conditions, either...

Use GPL for Free Software. If you want to give us Code With Conditions, I recommend the MS-PL.

I looked into this

[dowlingw]

I was recently looking into this using the concept of dual-licensed GPL/LPGL 'glue' connectors. This is the response I got back from the Free Software Foundation:

You are correct. The answer is no. If the LGPLed library dynamically calls the GPLed library, then it is the FSF's position that the LGPLed library is a derivative of the GPLed library, and thus the work as a whole may only be distributed under the GPL. Please see this section of the FAQ:http://www.gnu.org/licenses/gpl-faq.html#GPLWrapper.

For your specific requirements, I'd advise touching base with them - they have an advice service for these types of questions.

Re:GPL Fanatics

[beelsebob]

The fact that it can't happen. If you BSD license project A, and company M takes it and makes closed source project B with it, project A is *still* available from your site for free and open. Your code has not been magically closed, the only thing that is closed is *their* code which they used to make their project different from yours.

If that's not very much, well, they've not got very much to sell then, have they.

Re:GPL Fanatics

[beelsebob]

In theory, yes. In practice, however, it seems volunteers and companies contribute a lot more to GPL'ed projects than to BSD'ed ones, otherwise we'd be arguing for the "Year of FreeBSD on the Desktop" instead of Linux.
Really? Every company I've ever worked at has used and contributed to BSD and LGPL projects, but not touched GPL with a barge pole.

Personally, I like the idea of a BSD kernel with LGPL libraries and GPL applications, it seems to me as the best balance between freedom and widespread usage I can see. However, if devs believe otherwise and GPL everything I have no problem with it, the market will decide which option is better and, so far, it seems they're right.
Really? The one with the closed kernel (Windows) has the biggest share, the one with the BSD'd kernel (Mac OS) has the second biggest, and the one with the GPL'd kernel (Linux) has about 1%. Doesn't sound like they're right to me.

Mysql, anyone?

[sanermind]

This reminds me of our experience with mysql.. they changed the license of their drivers (the jdbc driver, in this instance) to GPL from LGPL a while ago, and also affirmed a legal position (at least in their marketing materials) that using the now GPL library would require any application dynamically linking/loading it to also be GPL, with the exception of being able to buy licenses to a separate non-GPL'd commercial instantiation of the library. (They also conveniently hired the guy who had been maintaining the previously freely useable LGPL version).

In theory, of course, someone else could re-invent the wheel and create another fully LGPL JDBC driver for mysql.. (I was ever-so-mildly tempted to do start work on that myself, out of spite and irritation, but didn't have the time).

Needless to say, after talking to the lawyer we had to switch our entire infrastructure over to postgresql. On the bright side, I found that I rather preferred postgres as a more comprehensive and functional sql anyway. (No more having to do manual tricks for multi-merges, etc). Still, it seemed like very bad faith to switch the license like that right out from under people who had already been using it, all in the hopes of further monetizing mysql. (Yes, the earlier license applied to the old versions of the driver... which happened not to work properly with newer versions of mysql).

I once talked to someone else else who makes a a bit of money on the side selling some piece of booking software. Unfortunately for him, almost half his clients were deploying his software in mysql shops. His software was a drop-in .war archive that was trivial to install (just copy it into tomcat/webapps or whatever (well, and run the database creation script to set up the tables for it) and it was good to go. After the license change he was unable to distribute a fully packaged pre-working .war file with the necessary drivers for client databases. (Oracle was propriatary, but free to distribute).. he had to write a specialized install script that the customer could use to unpack the .war, install the mysql jdbc driver, and repackage it, where the script required the client to affirm that they were using a legally licensed non-GPL version of the jdbc driver from mysql to be able to install/link it. This took the liability off of him ...The liability for having the temerity to want to talk with a basic software infrastructure service (such as a sql server), without GPL'ing his little niche app.

I'm all for the GPL. I love it! I've used it myself in a few things I've developed. I love all the good it's brought, etc, the idea of openness, a publishing and peer-reviewed manner of developing code, all that.

But I still find it very obnoxious to make a piece of library glue GPL. That's exactly what the LGPL is there for. Logically, it's not a derivative work of mysql if you use it for processing sql transactions (making no modifications or redistributions of it of any kind) any more than a file you create in the gimp is a derivative work of the gimp.

Re:Mysql, anyone?

[jeremyp]

When you added up the man-time required to move everything to Postgres, did it come to more or less than the cost of buying a commercial licence for the MySQL JDBC drivers?

Re:Mysql, anyone?

[julesh]

When you added up the man-time required to move everything to Postgres, did it come to more or less than the cost of buying a commercial licence for the MySQL JDBC drivers?

MySQL don't appear to publish the prices of their commercial licenses, telling you instead to "contact sales". My experience has generally been that when this happens, the answer is "more expensive than you can afford."

In a well engineered system, the cost of changing the database should be less than a man week of work (likely cost, around $2,000). In a particularly-well-engineered system, it should be less than a man day (around $400). I'd be surprised to find the cost of licensing MySQL's drivers for any realistic number of installations comes to less than $10,000.

Re:GPL Fanatics

[onefriedrice]

Exactly. The reason for the GPL is not to "protect" code or keep code free. Code will be free forever under any open source license. Big Bad Corporation can't take your code and lock it up so nobody else can use it; it's still free! The only reason for the GPL is to "stick it" to corporations and anyone else to whom it isn't convenient to fully comply with all the strings (restrictions) attached to and imposed by the GPL. In my opinion, GPL code is counter-productive because of the complexity and ambiguity of the license itself, as well as the incompatibilities that come along with a "viral" license. Case in point: How stupid is it that ZFS can't be incorporated into the Linux kernel, not because of any technical reason, but because two "free" license aren't compatible with each other. Seriously? That doesn't sound like freedom to me.

Re:GPL Fanatics

[Trepidity]

but remember, you were meant to be being "open" by using the GPL

Who "meant" that? The reason I'd use the GPL isn't for some generic notion of openness, but more specifically as a way of allowing multiple copyright/reuse models to exist in a mutually agreed fashion. If you agree to license your code under the GPL, you can have mine under the GPL too. If you prefer a standard copyright/licensing model for your own software, then fine, let's do that for mine too. I'd consider adding more options of that sort too if any seemed particularly compelling. What I don't see is why I should license you my code under terms that you aren't willing to reciprocate.

Re:GPL Fanatics

[Knuckles]

The fact that it can't happen. If you BSD license project A, and company M takes it and makes closed source project B with it, project A is *still* available from your site for free and open. yada yada.

We've been through this a million times before. The company has now stolen the time of the guy who wrote the free code in the first place, or at least the guy may feel like that. It's also possible that they extended his file format or network protocol with proprietary parts, harming interoperability and the users of the free version by impeding the network effect.

Some developers care, some don't, some care or not depending on circumstance and goals of the software project. None of these positions is more irrational than the other.

Re:GPL Fanatics

[beelsebob]

We've been through this a million times before. The company has now stolen the time of the guy who wrote the free code in the first place
No it hasn't. That guy was kind enough to give it away in an open way. This is the *point* of open source software, to be generous, and let everyone benefit from your work.

or at least the guy may feel like that.
Then he (a) is not an open source developer (b) should use a closed license like the GPL, and stop claiming to be open.

Re:GPL Fanatics

[beelsebob]

That's entirely fair enough then. I have nothing against people using the GPL because it's genuinely the model they want. I have something against people claiming to be open when doing so.

Re:GPL Fanatics

[Draek]

Ever heard of "Embrace, Extend, Extinguish"? It's not limited to Microsoft.

Re:GPL Fanatics

[Chrisq]

No it hasn't. That guy was kind enough to give it away in an open way. This is the *point* of open source software, to be generous, and let everyone benefit from your work.

The guy might quite reasonably want to restrict use to other people who want to share code for free. Like if I run a "bring a bottle" party I am not too upset if everyone brings something, even if some drink more than they bring. I would not like it if someone went round collecting the bottles and selling them at the local market though. If you don't want to come because "it's not really a free party then that's up to you - but don't complain about me making my own rules.

Re:GPL Fanatics

[onefriedrice]

In theory, yes. In practice, however, it seems volunteers and companies contribute a lot more to GPL'ed projects than to BSD'ed ones, otherwise we'd be arguing for the "Year of FreeBSD on the Desktop" instead of Linux.

That's just because we've already seen the "Year of FreeBSD on the Desktop." It's called Mac OS X.

But seriously, you'll have to provide a citation with your assertion that companies contribute more to GPL'd projects than projects of other licenses. The only major GPL projects with corporate contributions that come to my mind immediately are Linux and MySQL. On the other hand, many, many non-GPL projects with corporate contributions immediately come to mind. All Apache projects, postgresql, ffmpeg, BSD, memcached, OpenOffice.org, QT, postfix, bind, all Mozilla software, webkit, need I go on?

The sad truth is, the actual value of the GPL is a lot less than everybody thinks. It doesn't protect code from suddently becoming non-free because 1) corporations can't do that anyway, and 2) companies that do use non-GPL code tend to give back their improvements anyway, even though they are not legally obligated to do so. They may do it to keep good karma amongst developers rather than a strong believe in Free software, but who cares? What then is the value of the GPL? I propose it has negative value, considering the incompatibilities it introduces (ZFS anyone?) and the needless complexity it brings, but of course that's just my opinion.

Re:GPL Fanatics

[MadnessASAP]

Linux is only popular because of the GPL.

That's a pretty stupid statement don't you think? I (and I'm sure I'm not the only one) use Linux because it WORKS and does what I want it to in the situations that I apply it to. I also use Windows 7, Windows Server, Apache, Python, Java, Opera and piles of other products under just about every license under the sun. And you know what? I don't give a rats ass about the license so long as it doesn't hinder me they can shove it under any license they want.

Re:GPL Fanatics

[koxkoxkox]

What he was saying is that Linux is working now because of the GPL during its development.

Of course most end users don't give a rat's ass about the GPL, but it is this GPL which allowed one modification made to Linux to be used anywhere. Without it, we would have an IBM Linux, a Red Hat Linux, etc. In this case, why not stick to Unix ?

Re:GPL Fanatics

[man_of_mr_e]

The GPL does not guarantee that someone has to give you anything back. In fact, there is no reciprocation requirement at all.

I can, for example, take your GPL'd work, make some changes, then sell it to my customer. I am required, by the rules of the GPL, to give my customer the source, and they are allowed to do whatever they want, including give it to others, but if they paid $1,000,000 for it, chances are they're not going to do that. You have no right to demand the changes I made, or even a copy of the program from me.

This is a common myth around the GPL, that it enforces a reciprocation agreement. It doesn't. In fact, licenses that have such agreements are not compatible with the GPL.

Re:GPL Fanatics

[Yfrwlf]

They were saying that Linux got to where it is today because of the GPL license. The argument is that Red Hat, IBM, Novell, and other companies wouldn't have contributed to Linux if they weren't forced to by the GPL. Instead, they would have kept the code only in their own software offerings instead of sharing it, and Linux wouldn't have gotten as far along as it is today.

The GPL basically forces collaboration and sharing. I agree that that isn't true freedom in the truest sense, definitely, but certainly it did help get it this far, and has allowed more computer uses to be more free. However ironic that may seem, I believe that's the truth.

Now with the Internet being as developed as it is, it is much easier for collaboration to occur. In my opinion this needs to be much more fortified than it is now though, but at least things are getting there. There needs to be more marketplaces for everyone to buy into collaborative software projects. These need to be made easily available for gamers and desktop users, businesses, education systems, governments, etc etc. I.e. better bounty/support/development systems are needed IMO. Any way, the point I was making was if these systems can blossom, the GPL license won't be really needed as most everyone will understand what "collaborative software" is and the value in having that openness, so that you open things up to anyone contributing to help keep it afloat. Relying on closed commercial software is just dumb and dangerous.

Re:GPL Fanatics

[init100]

This is a common myth around the GPL, that it enforces a reciprocation agreement. It doesn't.

No, but in practice, that is usually what happens in the end.

Re:GPL Fanatics

[Eivind Eklund]

What he was saying is that Linux is working now because of the GPL during its development.

Of course most end users don't give a rat's ass about the GPL, but it is this GPL which allowed one modification made to Linux to be used anywhere. Without it, we would have an IBM Linux, a Red Hat Linux, etc.

Actually, we do. And it doesn't matter, since it's just minor patches.

It is not at all clear whether the difference in forking behavior is due to the license or the external development velocity / source code organization. We don't have a large scale IBM BSD, a Red Hat BSD, etc - we have a few open source BSDs and Mac OS X, and IBM choose to contribute to an open source BSD (FreeBSD through Whistle) rather than create their own pure fork. Mac OS X also contribute changes back.

I'm attributing the major differences in fork behavior mostly to source code organization (BSD is a large tree containing the entire OS, "GNU/Linux" is a bunch of small trees that are aggregated to form a bunch of different operating systems which are close enough that people think of them as they were the same rather than forks.)

Re:GPL Fanatics

[eugene2k]

Here's another case in point:
Being the free enthusiast that you are you develop a popular program. Say an open source word processor. It's free, and is quite popular in its niche. Now if you have created a program for yourself that fully serves your needs, you can release the code with a 2 clause BSD license.

If, however, you want to compete with other software makers, and you release the code with a BSD license, then any of your competitors is able to take parts or all of your code and use it in their products without ever releasing anything of theirs to the public. In the end, their software will always be better than yours because they build upon your work. Since their software is better, more people will be using it instead of your software, especially if the software is cheap (and your software will be of great aid to that). More users of your competitors' proprietary products will mean less testers and less possible developers for your open sourced software, which in turn means lower quality. Now how will this help spread free software?

Also take a look at the number of embedded devices and the number of super computers that run BSD (the FreeBSD kernel is released under a two-clause license, that is the highest permissiveness). Linux runs on many more embedded devices and on many more super computers, because the companies that embraced open source, are able to use the contributions of others, contribute themselves and make sure that nobody will turn their contribution against them by developing a closed source product with their code.

And about the ZFS being released under CDDL which in turn is incompatible with GPL: ever wondered how companies like Qt software were able to release Qt under a dual license and why Sun didn't? Or why Sun didn't release ZFS under a 2 clause BSD license, which is compatible with both, the CDDL and the GPL The correct answer is: Sun wanted to gain a marketing edge. It's no different than releasing the ZFS as closed source (from the marketing point of view).

Re:GPL Fanatics

[man_of_mr_e]

They might. Or, they might consider the code to be critical to their business and not want their competitors to have it.

The point is, the GPL does not guarantee what many people seem to think it guarantees, particularly in the "must give changes back" situation.

Re:GPL Fanatics

[man_of_mr_e]

Software is worth whatever someone will pay for it. There is no fraud involved in charging whatever someone will pay.

And the GPL puts no limit on what you may charge. The section you quote above is in regard to the fee for transfering a copy of the source code (upon written request). It is not a requirement for how much you can sell the binary for (with the source).

Many businesses pay in excess of $1,000,000 for custom software.

I'm not referring ot simply filing off the serial numbers. I'm talking about investing many 1000's of man hours of development work into something that could be considered a derivitive work of a GPL'd application.

Re:GPL Fanatics

[man_of_mr_e]

"usually". Not always. My point is simply that the original author claimed it was "free from abuse by those who want to take and not give anything back". That's simply not true.

Re:Step 2: lawyers

[Insanity Defense]

RMS uses the GPL because he thinks its got the right idea of "freedom"

Not quite right.

RMS created the GPL because he thinks its got the right idea of "freedom"

Now it's right.

What exactly are you asking?

[itsdapead]

Are you actually asking "can I circumvent the GPL on someone else's library using a wrapper"? or was the question "Why should I bother licensing my own library under the GPL, when others could circumvent it using a wrapper"?

The first of these has been pretty much answered elsewhere.

The answer to the second is to relax. If Big Evil Software Corp abuses your work this way then they're heading for a lot of bad publicity, even if you don't have (or can't afford) a legal case.

If someone gets away with this, shame on them, but what have you actually lost? Would Big Evil Software Corp have paid you (or even have heard of you) if you'd kept the code closed?

Licenses will have loopholes, and the more loopholes you try to close the more you introduce, and the more work you make for lawyers. How do you legally define the difference between a wrapper and a server so that a jury in Utah can tell the difference?

The free softwar community shouldn't fall into the trap of thinking like the RIAA and confusing "intellectual property" with real property. Violating the GPL isn't big, isn't clever but is no more "theft" than copying a movie, and the damages caused are intangible and hypothetical.

Re:GPL Fanatics

[photon317]

It may not require direct reciprocation, but there is an element of preventing abuse involved. In the case where a company extends GPL software and then only sells the resulting binaries commercially, they are still required to provide any customers with the resulting source code under the GPL license, which further allows them to re-distribute it if they see fit. That means there's no effective way to hoard this code or lord it over others. If your derivative product sucks so bad or has such limited applicability that only a handful of people ever buy it and thus get access to the derived code, perhaps they will never re-distribute it, and perhaps the world could care less. But for any derived code with wide applicability or value, enough people will purchase the product and gain access to the code that it's virtually inevitable that someone will legally "leak" it to the rest of the un-paying public, who can then compile it for themselves (again, legally) without paying for the product. This makes it pretty unrealistic to build a business plan around selling software derived from GPL software commercially.

That is the GPL's intent: it is a license for authors who wish to say: "I give this work to the world, and I don't want some greedy person down the line to be able to make money by locking up a modified form of my code from the users of said code, which is something I don't approve of. If you want to derive your work from mine, you must adhere to what I consider to be basic principles of software ethics, the main ones being that you can't sell people binaries and refuse them access to the source code, and you also can't tell them who they can or can't give that source code to themselves".

That's not to say you can't build a successful business around reselling (slightly modified or not) GPL software, it's just that you can't hoard the code in order to force people to pay you. You have to make your money by offering something else: support, services, hardware, etc.

Re:GPL Fanatics

[True Grit]

otherwise we'd be arguing for the "Year of FreeBSD on the Desktop" instead of Linux.

That's just because we've already seen the "Year of FreeBSD on the Desktop." It's called Mac OS X.

Except as Daengbo points out above us, OSX doesn't use the BSD license for the kernel part, it uses the APSL, which more resembles the GPL than the BSD, e.g. its a stronger "Copyleft" style of license thats more "restrictive" than the BSD. The rest of Mac OS X, the part that really counts!, of course, is proprietary, so the GP's point remains: "Year of *BSD on the Desktop" hasn't happened yet.

The only major GPL projects with corporate contributions that come to my mind immediately are Linux

In the context of the GP's "Year of XXX on the Desktop" reference, this is exactly the point: BSD-licensed OS kernels (*BSDs) vs. Copyleft-licensed OS kernels (Linux). The latter is seeing more corporate interest/support than the former, so far.

On the other hand, many, many non-GPL projects with corporate contributions immediately come to mind. All Apache projects, postgresql, ffmpeg, BSD, memcached, OpenOffice.org, QT, postfix, bind, all Mozilla software, webkit,

Yes, the GP did use the word "projects", thus opening up the whole "open" or "free" software world for inclusion as well, besides OS kernels, and that obviously means there are plenty of examples of BSD-licensed software out there. No one disputes that.

On the other hand, your use of phrase "non-GPL" here plus your particular list of examples is misleading:

ffmpeg = LGPL
OO.org = LGPL
QT = GPL/LGPL (dual-licensed)
Postfix = IPL
Mozilla = MPL/GPL/LGPL (triple-licensed)
Webkit = part BSD, part LGPL

where IPL = IBM Public License, and MPL = Mozilla Public License.

If you're trying to argue *for* the BSD license, then you haven't actually helped your case since just over half of your examples are using some form of a "strong" (or at least "stronger") Copyleft license. LGPL, IPL, & MPL are all Copyleft licenses that are more "restrictive" than the BSD.

If you're trying to argue *against* the GPL license, then first, most people don't think of the LGPL as being "non-GPL", but merely "not quite as strong as", and second, 2 of your examples are actually dual-licensed under the GPL, and third, anyone who wanted to go to the trouble could come up with a list of projects that *are* GPLed.

So your list doesn't really *prove* anything, and as far as the larger context of this debate is concerned, I think you're confusing/conflating two different issues: strong Copyleft (GPL) vs. weaker Copyleft (LGPL/MPL/IPL/APSL), and any kind of Copyleft vs. *non*-Copyleft (BSD). At a minimum, your mistaken assumption that Max OS X is an example of a BSD-licensed project at the outset damages the rest of your response.

The sad truth is, the actual value of the GPL is a lot less than everybody thinks.

If an astute reader interprets your reference to the GPL as "strong Copyleft" and to the BSD as "non-Copyleft", then the real sad truth is that judging by the licenses used by more than half of your own example projects, a lot of people seem to disagree with you!

Re:GPL Fanatics

[DrLang21]

There are so many examples for this I don't even know where to start. Lets say that I'm a manufacturing corporation and I have a database system and interface that monitors all of my manufacturing processes automatically and warns the appropriate people when a process starts to go out of a controlled state. I don't want my competitors to have this because it's just a huge boon to my output quality and reduces cost by millions every year.

Do NOT see the GPL

[Cajun Hell]

IMHO the GPL, even v3, needs some work to clarify this question and also to close the hole for the software-as-a-service industry to modify GPL code without reciprocating.

I've seen this kind of thing before, and it surprises me that so many people have such a deep misunderstanding of copyright law.

The GPL ***CAN'T*** define what a derived work is; it lacks the power to do that. Congress (not FSF and not Microsoft), through copyright law, does that.

Reading pages at FSF may clarify the issue, simply because those pages are written by presumably well-versed people (though beware: they may try to persuade you, rather than inform you).

Re:GPL Fanatics

[jedidiah]

In other words, MacOS is not open in any meaningful sense of the term.

It is exactly the type of "walled garden built with the help of free labor"
that GPL zealots always have whined about with regard to the BSDL.

What's the point of MacOS if I just end up turning it into a Linux clone?

Really? I have 3 Macs so this isn't just an academic/rhetorical question.

If the first thing I am going to do to overcome the inherent limitations
of MacOS is to start installing Linux software then what's the point really?

Re:GPL Fanatics

[spitzak]

He can continue to merge in any changes that the community makes, and certainly can take advantage of them.

Also if there is some bug or feature in mysql they need, they can make a stripped-down version of their application, such as an otherwise useless page of code that demonstrates the desired feature or bug, submit that, and hope the mysql guys fix it.

They could also fix mysql itself, contribute the fix back, and if it gets incorporated into the main source they are in exactly the same state but their code is still secret.

That's a long standing concern

[real gumby]

I wrote the LGPL v1 and badgered RMS into allowing it (John Gilmore thought up the great dynamic linking clause). RMS's biggest objection back then was exactly the case you give.

I still think in the balance the library license is a net positive for the FSF's cause. But this hinges on the definition of a "derivative work"

Re:GPL Fanatics

[shaitand]

The community for a given derivative is its users. By definition the users will have a copy of the dervivative app and therefore the source. It was never intended that anyone and everyone should be required to have the changes.

In practice though, even private development will result in bug fixes and it will be in that private party's interest to pass those back to the original project so they are maintained by the community and don't have to be manually patched and merged in with each update.

I don't think you're reading the law right.

[Estanislao Mart�nez]

OK this means that in order for a new work to be affected by the copyright on an earlier work, the new work must actually **INCLUDE** "major copyright-protected elements of the earlier work". So, if you statically link to a GPL library, then you include that library's code in your own work. If you dynamically link to a library, you don't include the library's code in your own work. It would seem then that the "GPL inheritance" would ONLY apply in the case of a statically linked library, according to copyright law.

This is, at the very least, not the position that the FSF takes. And I agree with them.

It comes down to how you interpret the part that says "major copyright-protected elements of the earlier work." You're reading "elements" as actual pieces of code. Copyright law in general, IMO, doesn't read the term "elements" as narrowly as you're doing. For example, the elements of a novel isn't just the words and their arrangement; it also includes the characters and the plot. You can't publish your own Harry Potter novels commercially without permission because the character and tons of background plot elements are protected by copyright. That your novel shares no substantial amount of text with the originals doesn't come into the argument.

So for example, the FSF can take the position that the call interface of a GPL-licensed library is a major copyright-protected element of the library itself, and that by using the library in your application, you're incorporating that element of the library into your work. Or more generally, if your application was written to use the GPL-licensed library, and cannot work unless linked with that library, then your application incorporates the functionality of that GPL-licensed library, and that therefore to distribute it, you must license it under the terms of the GPL.