The Grumpy Editor's guide to free documentation licenses

This article is part of the LWN Grumpy Editor series.

There may be little agreement on the question of what is the right license for free software, but there is, at least, a rough consensus on what the options are. There are two basic varieties: GPL-like (which require that derived products, if distributed, carry the same license) and BSD-like (which only require retention of copyright notices and credit). There are licenses which reserve special rights for the "primary contributor," and other variations exist, but the basic choice is clear.

This is not the case for licenses covering documentation. There seems to be little consensus on which rights authors need to retain, and which should be relinquished. Indeed, there is little agreement on why documentation should be free. Many of the reasons for keeping software free (ability to look at the source to see how it works, ability to correct misfeatures) do not apply to documentation. Obtaining a manual in "original source" form may be helpful for cranking out more copies, but that source will reveal little which is not already evident in the text. Software, when distributed in binary form, is a black box which hides its internal nature. Documentation, on the other hand, expresses its ideas on its face; it is transparent.

Or, at least, it should be. Certainly your editor has produced writings which fail on that front at times.

So why should documentation be free? Your editor has a renewed interest in free documentation licenses for a couple of different contexts. One is a longstanding item on LWN's "good intentions" list: putting our original content under a free license. The other is the almost-imminent publication of a book, which will certainly be released under free terms. In both cases, the motivations are similar:

• Free software changes rapidly; its documentation has, in rare cases, been known to lag a little behind. If the original author is unable or unwilling to update a document to match current reality, somebody else should be able to do so.

• Some readers never got the memo saying that English is The Language; they can have funny ideas about having manuals in their own tongue. It is rare that the original author can produce a translation in even one alternative language, but there are often people with the interest and skill who can do such translations. A free license should certainly enable that work to happen.

• Collections of documents can be good things. Consider the massive "All About Linux" books which were published in the mid-1990's, which were generally made of the Linux Documentation Project's output, combined with duct tape. Taking excerpts from free documentation can also be useful; a book on Python database programming could benefit from, say, Python and PostgreSQL introductions taken from other books.

• A printed book is unlikely to be available everywhere there might be an interested reader, but a free, downloadable book is available anywhere a net connection can be found.

For the purposes of updating and creating other sorts of derived works, having the "original" source of a free document is important - though not absolutely necessary. If nothing else is available, a free license, a scanner, and some sort of character recognition software can fill in. Translations and distribution do not necessarily require source; PDF files may be all that is required. Since not all free licenses are driven by the same goals, they do not all require the distribution of a machine-editable version of the text.

Documentation licenses address one other area which is typically not an issue with licenses applying to code: that of artistic integrity. Some authors feel that their words should be distributed intact, or not at all; others insist that certain types of material not be removed from their works. A survey of documentation licenses will find a number of "thou shalt not modify the text" terms. Such licenses will, for the purposes of this article, be considered non-free. A document which cannot be modified resembles a program which cannot be recompiled; it may have its uses, but it is also a dead end.

Creative Commons

The Creative Commons project is trying to address the current impoverishment of the public domain by encouraging the release of artistic works under any of a set of licenses. Many of the creative commons licenses forbid the creation of derived works or any sort of commercial use; they are thus, by this survey's standards, non-free. There are two licenses which lack those terms, however, being the Attribution 2.0 and Attribution-ShareAlike 2.0 licenses.

The Creative Commons licenses are explicitly written as contracts; they read:

The Attribution license allows the creation and distribution of derived works. Distributed copies must include a copy of the license - or at least a URL pointing to it; additional restrictions may not be imposed on the original work. Any distributed copy must include attribution giving credit to the original author, along with the author's URL pointing to the original version. The "ShareAlike" version of the license is GPL-like in that it requires derived works to carry the same license.

Interestingly, the Creative Commons 2.0 licenses explicitly disclaim any warranty or indemnification. Earlier versions of the license offered a warranty by the author that he or she was entitled to offer the work under those terms.

The Creative Commons licenses say nothing about the format in which works are distributed. By your editor's reading of the licenses, distribution of a derived work in PDF format, with no availability of the work in its original format, is allowed.

The GNU Free Documentation License

The GNU Project's recommended license for documentation is the Free Documentation License; it is complicated and, by some accounts, not truly free. In places, the FDL has clearly been written with the idea of furthering the Free Software Foundation's particular goals.

The FDL is GPL-like, in that it allows the creation and distribution of modified versions, but any derived versions must carry the same license. The FDL places limits on modifications, however. Any derived versions must carry the original's "History," "Acknowledgments," and "Dedications" sections, along with a full copy of the FDL. Beyond that, however, the FDL creates the concepts of "invariant sections" and "cover texts"; these features of the FDL are at the heart of the disagreement over its status as a free license.

An invariant section is not allowed to address the primary topic of the text. Instead, it deals with the "relationship" between the author(s) and publisher and the subject:

The FDL requires that all invariant sections be included in any derived work, and that, as indicated by their name, these sections not be modified. The purpose of invariant sections is clear: it enables the GNU project to include the GNU Manifesto (and related texts) in manuals and to forbid its removal. Thus, documents can be made to serve two roles: describing the subject matter of interest, and promoting the agenda of the group which created the document.

"Cover texts" are short passages which must, in some conditions, appear on the front and back cover of any distributed copy of the work. Use of cover texts is only required when over 100 copies are being distributed. Distributing large numbers of copies also obligates the distributor to make a "transparent" version of the document - one which is machine editable - available. The "transparent" copy need not be the original source; a plain text file stripped of markup will do. People who distribute a small number of copies can, if they wish, distribute them in an "opaque" format which does not allow editing.

An FDL-licensed work with no invariant sections and no cover texts is, by most peoples' reckoning, free. The inclusion of text which cannot be modified or deleted obviously changes the picture, and many people consider documents with those features to be non-free. Certainly the FDL makes certain types of derived products, such as those using an excerpt from an FDL-licensed work, difficult. An author wishing to take a few sections from the GNU emacs manual must drag along the entire FDL, the entire GPL, the GNU manifesto, the "Distribution" section, and the cover texts as well. In practice, these requirements will make that sort of use almost impossible.

The FDL makes no statement with regard to warranties or indemnification, other than to note that the document may carry warranty disclaimers outside of the license. It is also careful to note that warranty disclaimers cannot modify any other aspect of the license.

Open Publication License

The Open Publication License (OPL) dates back to 1999. Among other things, it is used for the Perens Open Source Series of books. The OPL is a relatively simple license; it allows redistribution of works, with or without modifications, in any format. The distributed copies must be licensed under the terms of the OPL, but nothing in the license requires that an editable version be made available. Modified versions must include a pointer back to the original, along with the usual notifications that changes have been made. The OPL includes a warranty disclaimer.

In its plain form, the OPL is a free license. It includes two "options," however, which can change the situation. "Option A" is a prohibition on the distribution of "substantive" modifications - essentially anything beyond reformatting or typo fixes. "Option B" is a restriction on commercial redistribution. If either of these options is exercised, the license becomes non-free. There does not appear to be anything prohibiting a person who distributes a derived work from adding options to the license, even if the original author chose not to use them.

The Creative Commons licenses and the FDL both include prohibitions on the use of "technical measures" to deprive recipients of the works of their rights under the license. The OPL, like many older licenses, has no such requirement. An OPL-licensed document could, conceivably, be distributed in some sort of DRM-infested electronic book format that, in practice, deprived the reader of the right to copy or modify the document.

Common Documentation License

The Common Documentation License was published by Apple Computer in 2001. It is a GPL-like license, requiring that all derived works carry the same license. It makes no requirement regarding credit to the original author beyond stating that copyright notices must be preserved. Derived works need not carry a pointer back to the original. Distribution in any format is allowed, with no requirement to make an editable format available. There is no restriction on the application of DRM schemes to CDL-licensed works. This license does carry a strong warranty disclaimer.

Design Science License

The Design Science License is, perhaps, the most direct attempt to translate the GPL into the world of text. It allows the usual freedoms, but requires that all derived works carry the same license.

The DSL takes a strong approach with regard to editable formats; it requires that any person distributing the document make it available in "the preferred form for editing." This requirement is rather firmer than the FDL's terms; a plain text file will not suffice unless that is how the work was created in the first place.

There is a warranty disclaimer in the DSL, though it does not explicitly disclaim warranties of noninfringement.

Conclusion

A significant amount of documentation has been released under the BSD license or the GPL. Putting a BSD-like license on a document makes some sense; it allows any sort of use as long as the copyright notices are preserved. Putting the GPL onto a document makes the author's intent clear in an informal sort of way, but the GPL was not written for this sort of application. The GPL refers explicitly to "programs" and acts like compiling and running programs; how such language applies to documents is unclear at best.

So which license would a grumpy editor use? Your editor co-authored a book which was released under the FDL. But the next edition is unlikely to go out under that license; the restrictions imposed by the FDL are simply too heavy. Any of the remaining licenses described above would probably be usable, though one of the licenses with a copyleft term looks preferable. No decision has been made on that subject; stay tuned.

---

(Log in to post comments)