Licenses and contracts

Some days it seems that wherever two or more free-software enthusiasts gather together, there also shall be licensing discussions. One such, which can get quite heated, is the question of whether a given free-software license is a license, or whether it is really a contract. This distinction is important, because most legal systems treat the two differently. I know from personal experience that that discussion can go on, unresolved, for long periods, but it had not previously occurred to me to wonder whether this might be due to the answer being different in different jurisdictions. Fortunately, it has occurred to some lawyers to wonder just that, and three of them came together at FOSDEM 2018 to present their conclusions.

The talk was given by Pamela Chestek of Chestek Legal, Andrew Katz of Moorcrofts, and Michaela MacDonald of Queen Mary University of London. Chestek focused on the US legal system, Katz on that of England and Wales, while MacDonald focused on the civil law tradition that is characteristic of many EU member states. The four licenses they chose to consider were the "Modified" or "three-clause" BSD, the Apache License, the GNU General Public License (their presentation was not specific to GPLv3, but the passage they quoted to make a point was from GPLv3), and the Fair License. The first three are among the most common free-software licenses currently in use. The latter is the shortest license the Open Source Initiative has ever approved, and though it is used by hardly any free software, it was included as an example of the maximum possible simplicity in a license.

US considerations

Chestek, speaking first, said that in the US a license is a grant of permission, in this case permission to use copyrighted material. It can be "express" (explicit), either in writing or orally, or implicit in the rights-holder's conduct. It is a defense against a charge of copyright infringement; if you are accused of infringement, you can say "I have a license", and as long as your behavior is within any license conditions, you're not an infringer.

A contract, however, is different. In its basic form, it is an agreement between two parties, with an offer by one party, an acceptance of that offer by the other party, and mutual consideration (which means that both parties must gain a definable benefit from the relationship). Some people claim that a free-software license cannot be a contract simply because of the absence of any consideration to the licensor. Chestek, however, said there is jurisprudence in the US establishing that the increases in market share, reputation, and product quality that result from offering your software under a free license are all valid potential considerations.

Chestek introduced here the term "bare license", which was to crop up throughout the talk. While noting that it is not a term commonly in use in the US, it is helpful in comparing jurisdictions. She defined it as a license that contains a rights grant and nothing else, no language that might incline a court toward treating it as a contract. Typically, however, a license will have more than the mere rights grant in it. Often it will include some conditions; then the important issue is which of those are conditions on the license and which are other requirements in the agreement. Failure to meet the former means you don't have a copyright defense; the presence of the latter may push a court to treating the license as a contract. In deciding whether a particular requirement is a condition on the license a court will look carefully at the wording. So when section 5 of GPLv3 says:

These are clearly established as conditions on the license. A simple requirement not to disparage the licensor, however, is not such a condition.

It is possible, said Chestek, that the Fair License is a bare license. It says simply:

The warranty disclaimer worries her, though, as it may not be part of the rights grant. But, of the four licenses under consideration, it is the only one that she thinks might be a bare license and not also a contract. In the case of the three-clause BSD license, clause three may be troublesome: "The name of the author may not be used to endorse or promote products derived from this software without specific prior written permission." She thinks a court would most likely treat that clause as a contractual obligation rather than a condition on the license. This was, she admitted to a later questioner, a narrow reading, but clause three lacked the "magic language" that courts look for in license conditions. Furthermore, there is some law in the US that requires license conditions to apply to the copyright grant so, for example, a requirement not to use the author's name in advertising has nothing to do with the copyrights and thus cannot be a condition on the license. In response to a follow-up question, she felt that even the "FreeBSD" or "two-clause" BSD license, which removes the problematic clause three, would still be regarded as having contractual elements.

In short, under US law, with the possible exception of the Fair License, all the licenses under consideration are likely to be regarded as contracts.

England, Wales, and, perhaps, Northern Ireland

Katz then considered the question in the context of the law of England and Wales, noting that a colleague had opined that the analysis would likely apply in Northern Ireland also. Scotland is more of a civil law jurisdiction and the analysis would not apply there. As is so often the case in English matters, this presentation involved a certain amount of history.

A bare license, from an English law perspective, is a promise not to enforce certain rights, such as copyright, that the licensor may have. If this promise is not enshrined in a contract—for free software it usually is not—the person making the promise may withdraw it at any time. So English courts a long time ago came up with the principle of estoppel, which says that if someone makes a promise, and you rely on that promise, they may not later revoke that promise.

Katz then presented a Victorian case usually known as Carbolic Smoke Ball. The plaintiff, Mrs. Carlill, bought one of these balls, which were guaranteed to be so effective against influenza that anyone who used the ball as directed and got the flu would be paid £100. As Katz said, "on the upside, the chemicals in the ball failed to kill her. On the downside, they also failed to kill the virus, and she caught influenza. Probably even more unfortunately, they failed to kill her husband, who was a trial lawyer". She sued for her £100 ("which was a lot of money in those days, before the Brexit vote") and won, but it went to appeal. In a judgment which is still studied by law students today, the Court of Appeal upheld the verdict, and in so doing extended contract law. Because estoppel provides no basis for an action, the court developed the idea of a unilateral contract, where one side makes a promise and the other party acts in such a way as to make it binding.

So the courts have shown willingness to construe contracts: that is, to behave as if a contract existed where none was actually signed. Furthermore they have also shown a willingness to look at a contract that does exist, and behave as if it contained terms that it does not, again in the furtherance of justice. However, just because the courts have shown a willingness to do so when there is a need to, it does not mean they are willing to do it everywhere. The limits are shown in Robin Ray v. Classic FM, where the court said that if it was going to imply terms into a contract, it could do so only to the extent they were necessary, and no further. On this authority, Katz argued that the same principle applied to implying a unilateral contract: if a bare license would obtain the full effect required, there is no need to imply a contract.

On that basis, he was minded to say that English law would hold that the Fair License, Apache, and BSD licenses were just licenses, while the GPL might be held to be a contract.

Civil law in the EU

MacDonald then looked at the question from an EU perspective. The main difference is that software licenses, whether proprietary or free, would be interpreted as enforceable bilateral contracts. The idea of a bare license as something other than a contract is fairly alien to the civil law tradition. This means that a free-software license must meet all the requirements of a contract to be enforceable; it needs an offer and an acceptance, though consideration is not relevant. Unlike common law, the civil law tradition focuses on the obligations of the contracting parties, rather than their promises.

Her conclusion was that all four licenses would be interpreted as contracts in civil law jurisdictions. In more than one case, German courts have held that the GPL is enforceable as a contract.

Due to time pressure six slides were then very quickly presented showing how bare licenses, unilateral contracts, and bilateral contracts were treated in the three jurisdictions with regard to six specific legal aspects. It is the differences in how the law treats licenses and contracts in each jurisdiction that makes the question of whether free software licenses are also contracts so important.

The first aspect was taxonomy — principally, how acceptance is indicated. The second was the revocability of the license. Third-party beneficiary rights — whether someone other than the licensor can sue a licensee for failure to perform their obligations — was the third. The fourth was about specific performance; if a licensee fails to perform an act required by the license, whether you can ask a court to order them to perform it, or whether your only available remedy is damages in compensation. The fifth followed on, being about the award of legal costs: if you sue someone for violating your license and you win, can you get the other party to pay your legal costs? The sixth and last aspect was whether the licensor could lawfully exclude liability.

Given what conclusions were drawn about the status of free-software licenses in the various jurisdictions, the implications for the treatment of those licenses are as follows:

• In the US, at least for the frequently-used free software licenses, acceptance is by conduct. The license is irrevocable once even partial performance has occurred. Third-party rights depend on the wording. Orders for specific performance are rarely granted but are theoretically possible. You won't get legal costs awarded, and exclusion of liability is likely to be valid.
• In the UK, where many free-software licenses are probably just bare licenses, no acceptance is required for a defense against an infringement claim. The license is revocable by the licensor until the licensee has relied on it. No third-party rights exist. Orders for specific performance are not available. The award of legal costs is discretionary but may happen, and liability exclusion is generally enforceable.
• In the EU acceptance can be by any means, including conduct. The license is probably irrevocable. Third-party rights do exist, and specific performance is available. Nothing was said about legal costs, and liability exclusion is possible but is subject to the the requirements of proportionality and consumer protection.

I'm a system administrator. I know beyond all doubt that many systems questions that seem simple on the surface are complex in both depth and width when fully examined, though I don't generally have to live with the added complexity of my free software changing its behavior depending on which country I'm running it in. It was a useful lesson to be reminded that questions that leak out of my field of expertise into others' do not magically become simple in so doing, and that "common sense says ..." is not a good legal argument.

Video of the talk and the slides from it are available here.

[We would like to thank LWN's travel sponsor, the Linux Foundation, for travel assistance to Brussels for FOSDEM.]