ego.istic.org

Benjamin Mako Hill of the Free Software Foundation (FSF) has written convincingly about what he calls “antifeatures”. Antifeatures are features that the user doesn't want, like limiting the number of concurrent users on a server, or showing adverts or splash screens, or anti-piracy limitations. The developers of the product have spent extra resources implementing these antifeatures, because they help to prop up the developer's business model, to the detriment of the user.

In the case of adverts, they allow the vendor to sell the user's time and attention to third parties. In the case of arbitrary limits, they allow the vendor to charge users to take out the antifeature. In the case of splash screens, they take the user's time and attention to fulfil the vendor's marketing goals. In the case of anti-piracy limitations, they make the product less useful to the customer by restricting his fair-use rights or making him authenticate himself to the software with a CD key or an internet connection, because the vendor believes all its customers are criminals.

Mr. Hill claims that free software can't have antifeatures. Because the user has the freedom to change the product, and because it is easier to disable the antifeature than to enable it, he will always be able to publish a modified version without that feature.

In fact, publication of modified versions is number three of the FSF's four freedoms, as listed in their definition of free software. That article also uses the often-heard gloss that the “free” in “free software” is as in the phrase “free speech”, not “free beer”: that is, it describes the user being free to act rather than the software being free of charge. Not all free-to-use software is free in this sense, and not all free software is free-to-use.

The FSF and its GNU project often say that “free software needs free documentation”. So, it's quite odd that the FSF should create the GNU Free Documentation License (FDL), of which version 1.3 has recently been announced. On the face of it, a free license for documentation sounds like a good idea, and perfectly fits the phrase “free as in speech”. But GNU FDL contains some very controversial “optional” features.

It allows the author of a work under such a license to declare that one or more sections are “Invariant Sections”, which means that users do not have the freedom to modify or remove these sections. The author may make similar declarations for “Front-Cover Texts” or “Back-Cover Texts”. Richard Stallman gives some advice on these features, which lists examples of their use.

He mentions that the intended use of Invariant Sections is so that you (or rather he, as the maintainer of GNU Emacs) can include a copy of the GNU Manifesto: that is, marketing material for the original vendor of the product. He suggests that Cover Texts be used to advertise the commercial sponsor of authors of the work.

I say that both of these uses are antifeatures. They are additional work the vendor performs that make the product less useful to the end user. (If you think they don't make it less useful, consider that if this were the case, there would be no need for special licensing provisions preventing their removal.)

A truly free documentation license would prevent the existence of antifeatures by the mechanism Mr. Hill describes. Clearly the GNU Free Documentation License, when used with Invariant Sections or Cover Texts, is not a free documentation license. It removes the guarantee that users can make improvements and distribute those improvements to benefit the community.

I'm not the only one who says this. The Debian Project has resolved that GNU FDL-licensed works with Invariant Sections or Cover Texts do not meet the [Debian Free Software Guidelines][dfsg] (DFSG), which broadly agree with the FSF's four freedoms. This entails that such works (including the GNU Emacs manual) cannot be distributed in the main Debian archive.

But the FSF's attitude to non-free speech doesn't end with Back-Cover Texts. The FSF believes that software should be free, but that this doesn't apply to works of art, essays, scientific papers, &c. In fact, the FSF recommends non-free licenses be used for “works of opinion and judgement”: its own website, and most (if not all) of Mr. Stallman's essays and speeches, are published under a license that permits verbatim copying only, forbidding the user to make modifications. In his justification for this stance, Mr. Stallman claims that as the point of such works is to report a belief or event, any modification to such a work must necessarily misrepresent the author, so such an activity cannot possibly benefit the community.

In my opinion, this belief is completely flawed in two ways. If you consider this position as applied to software instead of written works or speeches, it's like saying that any patch to a program must necessarily pervert the original author's aim for the program.

But even if you don't accept that the situations are analogous, you can't ignore the fact that this practice goes on every day without the claimed effect. In fact, I did it just a moment ago: I reported my impression of Mr. Stallman's belief in my own words, with the intent that you can gauge his opinion without having to read the whole article now. In fact, the verbatim-copying-only license forced me to paraphrase in this way instead of creating a derived work of his speech.

When a reporter interviews an eyewitness to some event, he doesn't quote him verbatim, with all the “umms” and “aahs”, and the grammatical errors intact; no, the reporter's job is to put the eyewitness account into a better order, more accurately reflecting what the interviewee intended to say. Certain publications are renowned for misreporting their quoted sources, but to most reporters ‘cleaning up’ an interview transcript is an everyday task with a clear benefit to both readers (who can follow the story without distraction) and interviewees (who can read their quotes without feeling embarrassed at their conversational slips).

The point is that it is possible to make improvements of benefit to the community, even to a highly subjective work. You can fix spelling errors or distracting punctuation. You can translate it into a different language. You can incorporate it into a larger work like a collection of essays on the same subject. None of these need misrepresent the original author.

Even so, the question has to be asked whether allowing derived works to pervert the author's aims is a problem that restrictive licensing should be used to prevent. Again, let's look at the corresponding situation for software.

Suppose that as a programmer I write a program that looks for security vulnerabilities on my network and automatically installs the appropriate patches. I make this program free software, to benefit myself and the community. It's entirely possible for malware authors to use this software for their malicious aims, to the detriment of the community. I'm tempted to say that my software is under a license that permits modification but not to produce malware.

This situation has in fact come up many times, and the FSF has been asked about including anti-malware clauses (also anti–human-rights-abuse clauses, anti–global-warming clauses, and many others) in licenses. But their response has always been that the social cost of including such a clause, and the associated reduction of freedom, would be less than the benefit of including it.

It's understandable that Mr. Stallman might not want people who are opposed to free software—mostly non-free software vendors whose flawed revenue model can't compete in a free market—to create a modified version that inverts his arguments, making it look like he does not support free software. But there is already much existing precedent in free software licenses for provisions that prevent authors of modified versions from misrepresenting their modified version as the original, or from using the name of the original author to promote their modified version: even the GNU FDL has such provisions.

Thus I must confess myself unable to comprehend Mr. Stallman's objection to free speeches (and essays, papers, articles, and so on). He claims that modifying such works necessarily misrepresents the author; on the contrary, this takes place all the time. He claims that modifying such works does not benefit the community; I have given several ways in which it can provide benefit by making the work more accessible or more readable. He implicitly claims that a work's license has the purpose of protecting its author from misrepresentation or from people using his work for purposes he disagrees with; the FSF's official policy casts serious doubt on all but the narrowest interpretation of this belief, and the licenses he rejects out of hand can offer such protection anyway.

It's really quite amazing how much the FSF and the GNU project have done for computing in general. Even if you tend to avoid directly using free software, you can't visit many websites without seeing one that was created with or is served using free software, and you can't read many emails without reading one that was sent via a mail server running free software. The development and community tools and processes used by GNU and others to create, distribute, and support free software have been widely adopted by commercial software vendors too, because of the benefits they offer to developers and users. For me, it's especially sad to see an organisation like the FSF, which has long been a pioneer in the free world, take such a backwards step on freedom with its non-free documentation license and its non-free speech; so, it's with quite some disappointing that I have to recommend that authors of documentation avoid the non-free features of the GNU FDL, and that authors of other verbal works, be they books, essays, papers or speeches, seriously consider releasing those works under free licenses.