Which License for Free Documentation?

Posted 6 Jul 2003 at 06:44 UTC by MichaelCrawford Share This

The Debian community has decided that the GNU Free Documentation License is not actually a free license at all, and is starting work to remove GFDL-licensed work from its main distribution. Such will only be available in non-free, which some people never install out of principle and CDROM vendors cannot distribute in whole because the licenses on some packages forbid it. Their objections stem from restrictions on modification of invariant sections and cover texts, and are neatly summarized in Why You Shouldn't Use the GNU FDL.

I wrote some GFDLed publications. I am considering changing their license, as I would like my documents to be included with Free Software distributions, including Debian. My question is, which one should I use? Should I change my license at all? I'm not actually convinced that my own desire for how my work is to be used actually meets Debian's requirements.

I have several documents under the GFDL, specifically the articles at the Linux Quality Database and The ZooLib Cookbook. I have quite a few proprietary writings, and have been considering relicensing some of them to make them Free.

Debian's objections stem mainly from the GFDL's allowance for the copyright holder to specify invariant sections (text which cannot be altered) and cover texts (text which must appear literally on the cover of printed publications). They feel this makes the license non-free because this prevents others from modifying all of the text. Even if the original author has no invariant sections, those who add to the document later on, who must keep it under the same license, may choose to add invariant sections.

There are significant problems with invariant sections - they may become inaccurate over time, but cannot be removed. They cannot be translated into foreign languages. They may be offensive. The cover texts may be so large, or so many may be added by subsequent authors as to use up all the space on a book's cover.

The GFDL requires that the license be reproduced in whole in the body of the publication. That prevents its use for quick reference cards, which are often printed on a single sheet of paper.

The Free Software Foundation's position on free documentation is summarized in Free Software and Free Manuals. Regarding invariant sections:

As a general rule, I don't believe that it is essential for people to have permission to modify all sorts of articles and books. The issues for writings are not necessarily the same as those for software. For example, I don't think you or I are obliged to give permission to modify articles like this one, which describe our actions and our views.

That reflects my own feelings - I write about my opinions quite a bit, and even when I want others to copy what I write I don't want them to be able to legally alter the text to make it unfaithful to my real opinions. The GFDL is in fact attractive to me specifically because of its allowance for cover texts. One that I use gives the simple requirement that printed publications give credit on the cover by simply saying "This contains material from the Linux Quality Database at http://linuxquality.sunsite.dk".

It would seem that I could use just about any of the popular documentation licenses, yet I find myself somehow uncomfortable with giving up this simple requirement for credit on the covers of printed books.

I don't think most Linux distributions would object to the GFDL this way, but at the core of the dispute is that the Free Software Foundation and Software in the Public Interest (the organization behind Debian) have different concepts of what Free Software means.

The FSF's position is given in some detail in Philosophy of the GNU Project. Debian's position is given in The Debian Free Software Guidelines, part of the Debian Social Contract. The DFSG specifically says "The license must allow modifications and derived works, and must allow them to be distributed under the same terms as the license of the original software."

That seems to make it pretty clear that a license that allows invariant sections is non-free as far as Debian is concerned.

I think you can understand the disagreement by considering the different purposes of Debian and the Free Software Foundation - their purposes have much in common, but aren't quite the same. Debian's is to serve the community by producing Free Software (and Free Documentation to go with it). The FSF's is to produce social change, as is I think is apparently from much of what Richard Stallman has written and said, and part of that process requires the ability to publish opinions, opinions which must not be misrepresented in derivatives of documents.

Some folks are also concerned with the GFDL's distinction between Transparent and Opaque formats. The license requires that documents be distributed in a format which "is suitable for revising the document straightforwardly with generic text editors". I think this reflects Richard Stallman's preference for editing with Emacs - it also prevents distribution of GFDL documents in formats used by many modern word processors, even those word processors which are themselves Free Software, in formats which may not be human-editable text but are openly documented.

I'll let you in on a little secret. I didn't consider the consequences of licensing my work with the GFDL very carefully at all when I first started using it for my writing. I used the GNU Free Documentation License because it was the one documentation license that came from the Free Software Foundation. I used it because I thought that was just the license one was supposed to use. I suspect many authors of both documentation and software use the GNU licenses for the same reason, without carefully considering the consequences or whether the licenses really represent their own wishes for the future of their work. I think the Free Software Foundation bears a heavy responsibility because of that. While certainly every author must take responsibility to choose his license carefully, I feel the Free Software Foundation bears responsibility to make the consequences of its licenses more clear.

I have not yet begun to consider what other license I might use for my work. I'm not yet convinced I want to change the license. I don't personally have any objection to either cover texts or invariant sections, but I'm considering dropping my cover text requirement to make my documentation acceptable to Debian. I am ambivalent about this though, as at times I feel that my desire to have the cover texts I ask for in the event my work is published in dead-tree form is enough that I'm willing to forgo having my work included in Debian.

Another course of action, although a lengthy and difficult one, would be to work to convince the Debian community that the requirement that one's opinions not be misrepresented in derivative works should not make the document non-free. That wouldn't satisfy my desire for cover texts, but would satisfy a strong objection I would have to placing some of my other writings under a license that would satisfy the Debian Free Software Guidelines.

I think it is important to understand that software and documentation, while they may share many qualities, are really different things, and deserve to be treated differently. The argument for allowing copying stems from the social benefit of the trivial cost of duplicating computer files. The argument for allowing the modification of software stems from maximizing the social benefit of source code - a "virtual machine" which can be easily modified by humans. But there must be a distinction between prose which is strictly utilitarian in nature, like a manual page, and writing which seeks to express an author's deeply held idea. Unfortunately, the flexible nature of human language makes it all too easy to mix the two kinds of writing in a single document, and our licenses must reflect that.

I do specifically object to the definition of Transparent vs. Opaque formats though. I have a very low regard for text format files for any complex use. In general I favor binary formats over text formats, as long as they're exactingly specified and the specs are freely available. So Debian's objection to invariant sections isn't enough to convince me to change my license, but the GFDL's prohibition of binary formats is. This is the twenty-first century; we should allow writers the full power of the software available to them, not the limited capabilities of human-editable text formats.

Thank you for any advice you can give me.

*sigh* Oh great..., posted 6 Jul 2003 at 07:33 UTC by tk » (Observer)

...now even the "Free as in Freedom" folks are fighting among themselves.

Possible licenses other than the GFDL include the Creative Commons Attribution-NonCommercial-ShareAlike 1.0 license and the other Creative Commons licenses. I've not yet studied these closely for problems, but maybe someone else has.

Factual correction, posted 6 Jul 2003 at 08:05 UTC by ElectricElf » (Journeyer)

With respect to:

I don't think most Linux distributions would object to the GFDL this way, but at the core of the dispute is that the Free Software Foundation and Software in the Public Interest (the organization behind Debian) have different concepts of what Free Software means.

I'd just like to point out that this is factually incorrect. I'm heavily involved in both Debian and SPI (as well as another SPI member-project). SPI has not had any discussions, either official or unofficial, about the GNU FDL. All the discussion I have seen has been within Debian, or within other Free/Open Source Software projects.

Also, Debian and the Free Software Foundation largely agree on what Free Software means. There is no notable difference of opinion there. The "core of the dispute" is that the Free Software Foundation (or at least Richard M. Stallman) feels that software and documentation should be treated differently. More specifically, that the author of technical documentation may wish to make an unrelated (likely opinion/political) statement within the work, which he feels shouldn't be tampered with.

I can't say I agree with Mr. Stallman, but it's important that the issues are represented clearly.

Also, as an aside, the other serious issue that many have with the GFDL aside from Invariant Sections is section 2 of the afore-mentioned license. Specifically, it reads in part:

You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute.

While on the outside it appears to be an attempt to stop people from publishing the document in a proprietary format (for good or ill), it also results in day-to-day restrictions. For instance, you wouldn't be able to view a document licensed under the GFDL over an SSL connection. Or store it on an encrypted hard disk. Or any number of other things - it's a very broad clause which can be manipulated by lawyers to mean anything.

Freedom to Derive vs. Freedom to Misrepresent, posted 6 Jul 2003 at 16:54 UTC by vorlon » (Master)

I won't try to address all of the points you bring up — to some degree you're right that you need to weigh the trade-offs for yourself and decide whether inclusion in Debian is important to you — but I do think the question of misrepresenting opinions warrants comment. You talk about the importance of not misrepresenting an author's opinion in derivatives of documents; but the fact is, not only is this not one of Debian's objections to the GFDL (the DFSG include an explicit provision for preserving the integrity of an author's source code), by and large no license restrictions are necessary to prevent people from misrepresenting your opinions: in all jurisdictions I'm familiar with, there are penalties for both libel and slander that protect you from someone misrepresenting an opinion as yours.

If licenses like the GFDL really only protected against misrepresentation of another's views, Debian wouldn't have nearly the problem with them that it does. In reality, the GFDL has two other effects where Invariant Sections are concerned: it prevents someone from reusing this text in the expression of their own opinion, and it prevents someone from removing these opinions from the technical matter of the work. While both of these limitations are commonly considered acceptable, neither is particularly Free as Debian uses the term. And just as I may find a program that does almost exactly what I need and modify it according to my needs, I might also read an essay that I almost entirely agree with and wish to use it as a basis for expressing my own opinion — since, indeed, my opinion is based on what I've read in the essay. Why should Debian distinguish between these two forms of "code reuse"?

Thank you for the corrections, posted 6 Jul 2003 at 21:25 UTC by MichaelCrawford » (Master)

I'm sorry, I didn't mean to spread any misinformation with the factual errors I made in my article. I did the best I could considering what I read on debian-legal and other mailing lists. I understand someone is drafting some sort of policy statement that will clearly state Debian's position - it will be important for it to be clear and complete.

You may wonder why I didn't just bring this up on debian-legal. I did ask about it there, but it seemed appropriate that developers who aren't part of Debian, but who have packages that Debian carries (or that they would like Debian to carry) should discuss it outside of debian's list, in a publicly visible way. Advogato seemed to me to be an appropriate forum for that.

I think that in general it is important for everyone to understand licensing more clearly than I think most people do.

Forums for discussion, posted 7 Jul 2003 at 01:06 UTC by slef » (Master)

Why is advogato a more appropriate place than a debian list to discuss debian's development? If you want people to discuss the GNU FDL in general, as most of your article seems to, then go ahead and post to advogato, but this is not a good place to discuss debian's development.

You do seem to misunderstand the concerns over the GNU FDL, but I think others have corrected the worst of them. I urge people reading this article to go read the original discussions themselves. Searching on lists.debian.org for FDL in the legal list should find most of them.

Re: Forums for Discussion, posted 7 Jul 2003 at 01:15 UTC by MichaelCrawford » (Master)

It's just because I'm not trying to discuss Debian's development. I'm suggesting the discussion should be opened up to upstream developers.

Debian is involved, but this is not primarily about Debian. I have some documentation that I'd like included with Debian, but I'm not a Debian developer, and I didn't write the documentation for Debian.

Re: Forums for Discussion, posted 7 Jul 2003 at 08:40 UTC by slef » (Master)

If you are interested in discussing the inclusion of your documentation in Debian, then debian-legal is the best place to have that discussion.

To discuss the FDL in general, then (as I said) this is as good as anywhere, but please don't start drawing imaginary battle lines between FSF and Debian. Debian has some concerns about FDL, sure, but I don't think it's half as bad yet as your article sounds.

Re tk's suggestion: I think the non-commercial creative commons licences would not be acceptable to Debian because they discriminate against commercial use. People sell boxed Debian sets, for example, which is commerce. Personally, I also dislike the "share alike" way of requiring the exact same licence, rather than the same rights to be preserved. Finally, notice that Creative Commons recommends the GNU FDL for software manuals.

GFDL with no invariant sections, OK with Debian?, posted 7 Jul 2003 at 23:13 UTC by atai » (Journeyer)

Would Debian accept GFDL documents with no invariant sections? Other people can add invariant sections to derived works, but the original should still be OK? Then, there is no need to change the license.

GNU FDL without invariants, posted 8 Jul 2003 at 09:57 UTC by slef » (Master)

The current opinion seems to be that FDL without invariants still is not acceptable to Debian, but I think the reasons are not quite as basic. This was recently discussed on debian-legal, so please read that.

You should also note that FDL-covered GNU manuals ("the original") normally include invariant sections, so even if FDL without invariants were acceptable, some important manuals would still not be included.

What about the ranomd pre-FDL Gnu files?, posted 8 Jul 2003 at 10:38 UTC by dan » (Master)

In my emacs/21.2/etc directory, for example, the following files don't allow for modified copies to be distributed:

  • GNU
  • MOTIVATION is "reprinted with permission of the author from the [...] Boston Globe" - was permission for modification also obtained?
  • copying.paper
  • gfdl.1

Clearly this makes the overall GNU Emacs package non-free. Should I report this to the Debian BTS? Really, would it help anyone if I did?

Random emacs files, posted 10 Jul 2003 at 09:20 UTC by dark » (Journeyer)

Speaking for myself (I haven't gotten agreement about this on debian-legal :-), I don't mind those files because they can easily be removed. They don't "stick" to the rest of emacs the way an invariant section does. Thus, they don't impact the freedom of the important parts, so I don't mind leaving them in. I still consider them a waste of space, but I'll only worry about that if I ever become the emacs package maintainer.

invariant sections., posted 10 Jul 2003 at 14:39 UTC by jamesh » (Master)

Well, the COPYING file isn't one that can easily be removed. It is essentially an "invariant section" of a GPL'd work. (and I am glad people can't modify that portion of my packages).

Licenses are not invariant sections, posted 11 Jul 2003 at 10:26 UTC by slef » (Master)

IIRC, nothing in the GPL prevents you from removing it. It's copyright law that does that, because the GPL is your copyright licence. Also, I don't think the GPL is normally regarded as part of the GPL-covered material. Not the same situation as FDL at all.

Use the BSD license, posted 11 Jul 2003 at 11:41 UTC by nik » (Master)

If it's good enough for code, it's good enough for documentation.

cf. http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/ln17.html


How can be invariant sections an issue, posted 15 Jul 2003 at 12:22 UTC by yeupou » (Master)

First, I would like to say that I perfectly understand why MichaelCrawford picked advogato to speak about the GFDL issue. Indeed, it's related to Debian development. But he also said that he's thinking about changing his project license to make it DFSG compatible, while he's ok himself with the GFDL. As he's probably not the only person in this case, it's clearly related to the whole free software community, inside or outside Debian.

Aside from that, I clearly do not understand the whole issue. In fact, basically, I understand why some people consider the invariant part as "non-free". Just I would be able to understand why they may even consider the GNU GPL as non-free because you're not free to include GPL covered code in a proprietary software. But is it a problem? We speaking of free software in the FSF idea, we are not speaking about anarchy. You are not free to do everything you want. Just like in a democracy : you are not free to do everything you may want to do (like raping a girl in the street, to get big picture of problem). You are free to do what the society consider as good for the whole society. There is a point when you're asking for access to the source code of a software, the right to modify it and to redistribute it - without these freedom, you would have to reinvent everything. But with a text, what the point of asking the right to modify everything? You already have the right to access the source code (there's no binary form), you can already modify a text and redistribute it: it's called writing (when you write, it's generally about something you thinked about, something you maybe got from a book).

These invariants sections, how can it be a problem for you?

When you read a book of History, or Sociology (whatever), you cannot alter it's content. But is it the point? The point is to get the content, and you can share it, even if you cannot alter this book itself. The book is only a medium, what matter is the idea behind, which one you can use even if you cannot modify the book.

To me, the whole issue is a joke. Free Software as been "invented" to establish in computing the pratices known and followed in most of the other scientific areas. These practices were ignored in the software industry because it's possible to provide only binaries, it's possible to proprietarize the ideas behind a software, which is almost unique.

So indeed, the software is something special, not just like a text. And frankly, what matters for me when dealing with a text (a book, documentation or whatever) is the fact you can redistribute it, not especially modify it. If I want to modify completely a text, I write a new one. If I want to take some parts of a text, I quote it.

About the invariant section, someone talked about the translations problem: well, if the upstream author makes a part of his text invariant, it means he do not want that part to be translated (and he must be aware of it). In fact, it even means that he want that part to be there as he wrote it but a translation can be added anyway, along with the invariant text. Someone also said that a part of the GFDL ("You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute.") "appears to be an attempt to stop people from publishing the document in a proprietary format (for good or ill), it also results in day-to-day restrictions. For instance, you wouldn't be able to view a document licensed under the GFDL over an SSL connection. Or store it on an encrypted hard disk. Or any number of other things - it's a very broad clause which can be manipulated by lawyers to mean anything.". I fail to see the logic here: having a GFDLed text on your user home do not mean that your user home must be publicly readable.

As conclusion, I think that this whole issue enlight different understanding of the purpose of the right to modify.

Right to modify, posted 15 Jul 2003 at 13:41 UTC by slef » (Master)

I could rephrase a question from the above comment as: with source code, what is the point of asking for the right to modify everything? You already have the right to access the source code, you can already modify a program and redistribute it: it's called writing your own.

Maybe I've been unfair, but that's an illustration, I think.

Comparing with most existing books is nor really valid: we know they are broken, in a similar way to proprietary software is broken. Yes, you can sometimes quote from them, but that is a very limited ability (more limited in some places than others) and a poor substitute for modification.

Why do we want the right to modify software? So that we, as users of that software, people who want to use it to do tasks, can adapt it to make sure it can still do our tasks. Why do I want the right to modify other works? So that I, as a user of that work, someone who wants to use it to do a task, can adapt it to make sure it can still do the task.

Other people have different views on this, such as those advocating freedom of expression and so on.

&quoCommon sense&quo logic vs. &quoThe Law&quo logic, posted 15 Jul 2003 at 14:46 UTC by tk » (Observer)

[yeupou] I fail to see the logic here: having a GFDLed text on your user home do not mean that your user home must be publicly readable.

You're applying "common sense" logic here. But as we all know, in law there's no such thing as common sense. If the law says that cryptographic software is munition, then as far as the law's concerned, it is munition, even if our "common sense" tells us it isn't.

Thus, even if we "know" by our guts that it isn't the FSF's intent to expose everyone's home directories, yet if the license forbids us to "obstruct" the reading of GFDL documents in any way, and placing GFDL documents in an rwx------ directory does indeed hinder others from reading them, then open up our home directories we must.

Until the FSF gets around to fixing this, I think this hole is enough reason for people to consider using other licenses.

GFDL and RMS, posted 15 Jul 2003 at 20:50 UTC by atai » (Journeyer)

I think one thing the GFDL issue demonstates is that RMS is not as extremeist as some people think, that RMS does not try to push his idea as is to all digital information, outside software. He said that he is no expert on books and music, and he does not insist on them being modifiable like software. So now we have a group of people, taking RMS's original idea on software as is and applying to other formats of information, and now saying RMS's license for documents is not free. There may be problems with the GFDL, but I hope at least people who criticize RMS for his "extremeist" views, like tk, can recognize that RMS is not as extreme as you think.

Re: GFDL and RMS, posted 16 Jul 2003 at 06:53 UTC by tk » (Observer)

[atai] He said that he is no expert on books and music, and he does not insist on them being modifiable like software.

But note that the GFDL is meant for software documentation, not general artistic works. And on the subject of software documentation, the FSF does make a stand:

While a blanket prohibition on modification is unacceptable, some kinds of limits on the method of modification pose no problem. For example, requirements to preserve the original author's copyright notice, the distribution terms, or the list of authors, are ok. [...]

However, it must be possible to modify all the technical content of the manual, and then distribute the result in all the usual media, through all the usual channels; otherwise, the restrictions do block the community, the manual is not free, and so we need another manual.

And strangely enough, that these are exactly the same concerns outlined at the beginning of Debian's article -- which means that the GFDL may be non-free even by the FSF's own standards!

I think the GFDL issue has to do more with its lack of polish of the terms, than any inherent "extremism" (or lack thereof).

accuracy: "Debian's article", posted 16 Jul 2003 at 10:00 UTC by slef » (Master)

The article is the work of a Debian developer and not an official publication. It is a Debian developer's article, not Debian's article. *sigh*

right to modify, common sense and logic, posted 17 Jul 2003 at 20:03 UTC by yeupou » (Master)

slef, when you say "You already have the right to access the source code, you can already modify a program and redistribute it: it's called writing your own", you're wrong. Writing your own software without access to the source code of others softwares means finding new ideas, not expressing in your way ideas you find elsewhere in the preferred form for modification. Because an idea expressed in a text is in the prefered form for modification: you have already everything needed to adapt it to your needs. It's absolutely not the case when you have to write some code from scratch. With a binary, you hide ideas, which is impossible with a text. That's why... the preferred form to modify a software is the text form... the source code.

<person>tk<person>, it can be discussed. When I read "You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute", I understand that I cannot block access to a copy I gave to someone. But the "or" ("make or distribute") seems weird: does it mean that I must distribute every copies I can make? I'll ask.

Right to modify, posted 18 Jul 2003 at 10:04 UTC by slef » (Master)

yeupou, I am already comparing text to source code, not to software in general. You are talking about using binary-only to hide the expression. I did not. I have had software source code which came with a licence that did not allow even local modification. That's what I'm trying to draw a comparison with.

In any case, you may not necessarily have the source for a text. You may have some object that lets you reconstruct the source code fairly accurately, but it's much easier if you have the marked-up source copy. That is tricky to specify well in the licence and I believe some people have questions about the way the FDL tries to do it. Just like the other questions about FDL, good answers would really help.

Confusion, posted 29 Jul 2005 at 16:07 UTC by yeupou » (Master)

« You may have some object that lets you reconstruct the source code fairly accurately, but it's much easier if you have the marked-up source copy. »

It matters to me to be able to access the source code of a software I use. It's convenient when I'm able to do it with my favorite editor. But what matters is the fact that I can access it.

And to access the content of a philosophical text, you only need a brain. What's the point in altering such text ? It would be necessary to achieve which purpose? What the point for instance in being able to edit the GNU Manifesto?

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

Share this page