A Reactionary Rant by The Famous Brett Watson, 16-Aug-2004.
I first pause to consider whether I should address this issue at all, and if so, on what basis. The issue in question is a column by one Stephen H. Wildstrom appearing in BusinessWeek online, entitled A Big Fly in the Open-Source Soup. My usual policy is, "don't feed the plants, trolls, or convicted monopolists", and Wildstrom could easily be a member of the second category and/or operating on behalf of a member of the third. Still, pointing out sophistry doesn't count as "feeding", as far as I can tell, so I'll use his column as an exercise in criticism: identifying weak or specious arguments and rhetorical devices (used to make an argument persuasive without lending it any rational weight).
At the outset, Wildstrom uses the rhetorical device of sensationalism to spice up his column. After giving the "open-source movement" some praise as a build-up, he warns that "the movement is now facing a crisis". What is the nature of this alleged crisis? Summarising in advance, Wildstrom puts it this way: "How does software owned by everyone and by no one survive in a world where copyrights and patents shape the legal landscape?" But any open source software which becomes successful enough to be worthy of note attains that status while carrying this alleged disadvantage, so where's the crisis? There isn't one, at least not on the basis of the summarised argument. The opening paragraph of the article wouldn't pique any interest if he told it like it is, however, so call "crisis" and hope nobody notices the lack of smoke and/or fire.
In paragraph two, we see the real target of this article: Linux. This isn't about an alleged problem with open source software in general, but Linux in particular. (As we'll see later, it's not even Linux so much as the licensing terms of Linux — the GPL — that is the core concern of this article.) Wildstrom once again lifts up his target (as a precursor to tearing it down) by offering some praise, calling Linux "the most important piece of free, open-source software". In what sense is Linux "the most important"? Why is it more important than BSD, or BIND, or Apache, or Mozilla, or OpenOffice, to name but a few? Well, if you happen to be Microsoft, then you may consider it to be the most significant threat to your control of the marketplace, and thus the most important, but I rather doubt Wildstrom intended to convey that particular meaning. A more generally accurate turn of phrase would have been to describe Linux as "important", "significant", or "well known", but these lack the sensational value of "the most important".
Take stock of the rhetoric so far: he's alerting us to a "crisis" in "the most important piece of free, open-source software", and claiming that the matter is sufficiently serious that "if open source is to play an important role in software's future, the issue will have to be dealt with decisively." Stirring stuff, but can the facts live up to this level of hype?
Note also in passing that the reference to SCO is factually inaccurate: Wildstrom reports that SCO "acquired the Unix trademarks", whereas those trademarks are the property of The Open Group. This falsehood (that SCO owns the Unix trademark) has been reported with sufficient frequency that The Open Group has been prompted to address the matter directly. Wildstrom's perpetuation of the error is sloppy reporting at best; a cynical disregard for truth at worst.
Next, Wildstrom raises the spectre of legal action by patent holders, although not in an alarmist manner. Indeed, he says that patent infringements "pose no immediate threat to Linux". He does, however, point out that "the uncertainty is taking a toll", citing delays in the widely publicised Munich Linux migration. I'll assume that submission deadlines prevented him from providing the most up-to-date news on this front, since The Register reported (one day prior to the dateline on his article) Munich OSS switch to go ahead, patents or no patents. So far, at least, the toll taken on the Munich project has amounted to no more than a brief pause.
Wildstrom doesn't mention that patent issues aren't specific to Linux or free software. In fact, they are likely to be a far bigger problem for smaller proprietary software vendors (and their users, if the vendor goes out of business as a consequence of a patent suit) than for users of free software. Patents on software are most annoying for the fact that the protected methods are (in the vast majority of cases) likely to be independently reinvented by other parties. The most profitable (and, coincidentally, cynical) tack that an owner of a software patent can take is to sit on the patent and hope that someone else does invent the same idea independently, and that it subsequently becomes widely incorporated into proprietary software. Once the technique is ubiquitous, the patent-holder can spring the trap, threatening legal action against those proprietors that do not license the technology. Smaller vendors will be most threatened by this, since they have less negotiating power and a smaller legal budget.
It's not at all clear that free software projects (or users thereof) present, in actual practice, an attractive target for this kind of legally-sanctioned extortion, but Wildstrom assumes the worst and advises accordingly. His thinking on the matter, however, seems rather locked into the proprietary software paradigm, where licenses are typically issued on a per-user or per-computer basis. Specifically, he sees the need for (and absence of) a central authority to pass licensing fees on to end users. This whole approach is quite impracticable where free software is concerned: free software is characterised by (among other things) a complete lack of per-anything license costs and central points of authority. If any kind of licensing deal is ever struck between a patent holder and a consortium of commercial free software interests, it's likely to be a class license covering all free software implementations. Few other options are likely to make any sense, and patent holders have yet to demonstrate (so far as I am aware) that they can obtain meaningful legal injunctions against free software "infringements", so a class license may be their best practical option.
I should mention at this point, that the best of Wildstrom's argument has now been presented and considered. From here, it goes into a bit of a decline, as you will see.
The next problem that Wildstrom raises is that "a slew of backers see open-source as part of a social and political movement that's frankly anti-corporate." I find this assertion unreasonable as it stands, and he presents no evidence for it. I'd be willing to believe a slightly modified version: to wit, "a slew of potential backers see open-source as anti-corporate, and consequently choose not to back it." My version is more credible and shows why we might want to view this state of affairs as problematic (i.e. the missed potential investments). Rather than summarily dismiss Wildstrom's point as "incredible", "unsupported", and "not problematic in any case", let us assume that my reformulation reflects his intended message and criticise that instead.
Even if we grant that corporate backing is entirely a good thing (a rather generous assumption, in my opinion), my reformulation of Wildstrom's point still fails to express a problem. The fact that some assumed potential backers hesitate to finance free software projects (citing the GPL as the problematic factor) is hardly a compelling reason to drop the GPL, because there are quite enough backers who are willing to play the game as it stands, GPL and all. Is free software suffering for a lack of backing, corporate or otherwise? I'm far from persuaded that this is the case, and even if I granted that it were, I'm even less persuaded that the GPL bears any relation to the matter. Richard Stallman's ethical proclamations may well be considered unrealistic ideals at IBM headquarters for all I know, but there's nothing in evidence to suggest that this is an impediment to their involvement in Linux, GPL notwithstanding.
Wildstrom now returns to the realm of factual error, as he offers the following summary of the GPL.
"The GPL not only requires that any programs licensed under it be freely distributed but also that any modifications made to the software, or any other software derived from it, are themselves automatically covered by the GPL."
The GPL (version 2) forbids the licensor from charging a license fee (section 2b), but does not require that anything be freely distributed. Indeed, section one explicitly permits the charging of a fee for the act of distribution, or for a warranty. Section three does mandate that a distribution of the software be accompanied by the source code or an offer to supply the source code (charging only the cost of distribution). Modifications made to the software are not automatically covered by the GPL, but a modified version of the software can only be distributed under terms dictated by the GPL. The modifications themselves can be made available under any license, so long as they are offered on their own.
But even if Wildstrom's summary had been an accurate reflection of the GPL, then so what? A typical proprietary "End User License Agreement" (so-called) disallows derived works, reverse engineering, and slew of other legally dubious restrictions which nobody ever bothers to read because they have no choice but to click "I Agree" if they want to use the program they've just bought. Compared to this, even Wildstrom's distorted rendition of the GPL is a model of saintly charity. You don't even have to agree to the GPL in order to use programs licensed under its terms. So what, exactly, does Wildstrom imagine to be the problem with the terms of the GPL? He fails to complete the argument here, but we'll see a strong hint in his conclusion.
The next charge against the GPL is that it's unclear. I believe Wildstrom is correct that there is a lack of case law relating specifically to the GPL — so far as I'm aware, every case involving it to date has been settled out of court — but this is a consequence of the GPL's clarity, not ambiguity. It's legal ambiguity that causes cases to go before a judge, and the willingness of parties to settle out of court in GPL-related matters is testament to the fact that no such ambiguity exists. There's little point in taking a matter to court when it's obvious that the judge would rule summarily one way or the other. Case law is lacking primarily because there's been no need for it.
To add further worries, Wildstrom cites an article in ACM Queue (although he gets his facts wrong again, citing the June issue, when he's actually quoting from an article in the May 2004 issue) in which the definition of "derived work" is examined. Wildstrom is actively misleading here: his selective quotation gives the impression that "derived work" is a vague term. Granted, there are boundary cases where the question "is this a derived work?" has no clear answer, but these are exceptions rather than the rule. The author of the ACM Queue article claims, "I believe that device drivers, however and whenever they are loaded, are derivative works not only of the operating system kernel but potentially of other modules that they 'elaborate.'" If this (ludicrous) opinion is shared by the courts, then authors of device drivers for Windows have much more to fear than those who do so for Linux, unless they've had license from Microsoft to create "derived works" all this time and I haven't heard about it.
The quotation from ACM Queue also contains the following gem: "Some people argue that the GPL as a whole isn't even enforceable..." Well, some people argue that the Earth is shaped like a flat disc, but this doesn't make the proposition credible in and of itself. Neither the author of the ACM article nor Wildstrom offer a reputable name to go with this claim, nor do they offer the actual argument. For all we know, the argument goes, "the GPL as a whole isn't enforceable because my knees are yellow."
And now, we reach the momentous conclusion, where Wildstrom engages in further rhetorical devices which flatly contradict his earlier ones. As the article opens, Linux, "the most important piece of free software" is facing a "crisis", and "if open source is to play an important role in software's future, the issue will have to be dealt with decisively." Well, the crisis has spontaneously dissolved somewhere along the road, because Wildstrom now considers the future bright! But "bright as it is, the future of commercial open source might be considerably brighter if Linux and other programs went to a more commerce-friendly license with fewer complexities and ambiguities than the GPL."
So here is the conclusion in a nutshell: GPL bad; other free software licenses good. If he really wanted to reach that conclusion, then an argument which actually supported it may have come in handy. He baselessly accuses the GPL of ambiguity without showing that the other licenses are clearer. He asserts without evidence that adoption of the less restrictive licenses would encourage commercial participation, whereas Linux appears prima facie to attract more commercial investment than its BSD-licensed peers. He never even specifies what the supposed benefits of commercial investment will be, simply asserting that it "would be good for all concerned."
The key difference between the GPL and the other free software licenses is that the GPL requires a quid pro quo. Most other licenses are rather generous in that they grant rights to modify and redistribute without benefit to the original copyright holders. The GPL extracts a cost, offering the software for use as a base to further work only if the offer is reciprocated. So Wildstrom's idea of a "commerce-friendly license" is one where he gets something for nothing: quid pro nihilo.
Well, Mr Wildstrom, here's my conclusion. If you want to persuade me that the open source community will benefit from making something-for-nothing parasites feel more welcome, you'll have to try much harder than that.