the OFL-1.1 license violates 17 USC § 105

[mbutterick]

17 USC § 105 says that any work of a US government employee is not subject to copyright. The OFL-1.1 license, by contrast, depends on the assertion of copyright for its effect. Therefore, there is no way for the work of US gov’t employees to be released under OFL-1.1. (I take no position on whether the terms of OFL-1.1 have been correctly applied here, though I’m skeptical.)

[mbutterick]

[I am not anyone’s lawyer and nothing I’m saying in this thread is legal advice.]

A note that the project "might or might not be partially or wholly within the public domain in the United States" would be nice and completely sufficient for any lingering doubts in my mind.

Respectfully @malanthomasii, from a software-licensing perspective that would be the worst possible outcome. The whole premise of this issue is that the GSA has put Public Sans into an uncertain legal state. Papering over the problem with some wibbly-wobbly language about what “might or might not” be true will only multiply the doubts. This is why, as the OSI puts it, “certainty” is an indispensable ingredient of software freedom.

[malanthomasii]

@mbutterick: That statement was intended to be conditional on my other conclusions being true and the intervention of the rightsholder to heal the government's intervening copyright violation, but to address your point more directly:

Yes, the situation is uncertain. Many copyright situations are. It would not be wise or ethical to pretend otherwise. Until the uncertainty regarding § 105 and derivative works is clarified, pretending that something that might be public domain is definitely copyrighted is highly objectionable to me as a matter of clarity and as a matter of ethics with regards to the public domain. I would therefore like a disclaimer for so long as the situation remains uncertain.

Your link points out the benefits of having a fallback license where the public domain may be unavailable. I agree with that point. I am not calling for a waffling disclaimer as the sole solution to the problem; I think that if the government's violation can be healed so that the only remaining question is on what terms further use may be made, a license should remain a option for all of the reasons that I have previously stated.

As a practical matter, operating under a "worst-case" assumption is a regular practice in other areas of copyright uncertainty, and for someone who's uncertain to choose to abide by a license that might or might not be legally necessary or enforceable in all instances isn't impossible or even that odd to me. I think that we can render academic the uncertainty in this (highly fact-specific) case, allowing for practical use while the uncertainty remains unresolved.

[mbutterick]

[I am not anyone’s lawyer and nothing I’m saying in this thread is legal advice.]

operating under a "worst-case" assumption is a regular practice in other areas of copyright uncertainty, and for someone who's uncertain to choose to abide by a license that might or might not be legally necessary or enforceable in all instances isn't impossible or even that odd to me.

I see your point, though the worst case is quite a bit worse than you suggest. We’re not in a situation where “at best” the fonts are public domain, and “at worst” they’re under the OFL. If the Libre Franklin OFL was violated, then the Public Sans fonts are a unpermitted derivative work, and amount to one big copyright violation.

Still, if you think there’s some language that would “allow[] for practical use”, I encourage you to open a separate issue, since it concerns the current state of the fonts, whereas this issue concerns their future state.

[pbhj]

3. If the derivative work is completely public-domain in the U.S., then U.S. users would be free to ignore the OFL and use the derivative work. However, for reasons stated both earlier in this comment and in my prior comment, this doesn't require removing the references to the OFL. 

( @malanthomasii )

Surely though if a work is public domain then there are no copyright restrictions, in USA at least, and a USA based person can [re-]publish the work CC0, or public domain, and thus make it available globally from that source regardless of whether USA government retain some copyright on their version in other countries, it doesn't matter as the people aren't copying their typeface but are copying John Does' published work? With copyright it matters if you copy, if you don't copy from the USA owned work in some other jurisdiction but copy from the CC0, say, work of John Doe (which is really a re-publish of a "public domain" work) then there's no case to answer against USA government.

I realise it's a slight side-issue but it would circumvent some of the shenanigans (publishing a "public domain [but only in USA]" work as also OFL).

[malanthomasii]

@mbutterick - You can't separate the pre-existing copyright violation from the problem of whether people can make use of this distribution in the future; they both derive from the intersection of § 105 and the OFL, and the underlying violation in creating and publishing the derivative work is sufficient to create legal uncertainties in any "future state" of the fonts. I have suggested that @impallari can heal the underlying violation. If that is not how the law works, or you have a better fix in mind, I am open to being educated.

@pbhj - That will not work. If a work is out of copyright in Country A, and you make a copy of it in Country A, it is still the same work and not a new one. If that work is copyrighted in Country B, it is still copyrighted in Country B by the original copyright holder. You cannot distribute that work to people in Country B because you do not have that (copy)right; someone else has that (copy)right. You cannot give up the copyright in Country B because you don't own the copyright in Country B. You cannot hijack someone else's copyright in Country B by republishing the work in Country A. Someone in Country B generally does not have the right to make a copy of the work or any other use of it in Country B regardless of where they get their copy. (There are some slight exceptions that are irrelevant here.)

[david-a-wheeler]

Whew, a lot of discussion. Let me first scope this down: Outside the US, the US can assert copyright and release its changes under the OFL-1.1 like anyone else. The US appears to have done so, since it says that the whole thing is under the OFL-1.1. The complication discussed here only applies within the US, where the changes by US government employees as such cannot be copyrighted by law. So let's focus on that.

The complication is due to the OFL-1.1's license text, which is very different from licenses that are used by most programs (MIT, GPL, etc.). It says:

5) The Font Software, modified or unmodified, in part or in whole, must be distributed entirely under this license, and must not be distributed under any other license

Again, this text is totally different from commonly-used licenses. OFL-1.1 may be common for fonts, but in the wider world of software it's not a common license.. So it should be unsurprising that there are some oddities that create complications when examining an unusual license.

Personally I think this text is poorly drafted, because one obvious interpretation isn't actually possible. Namely, you can't say that every part that you distribute is copyrighted. For example, if I extracted one character from the software, that character would be a part, and yet any one character can't be copyrighted (it's in the public domain) & thus its distribution doesn't need a license at all. If parts were developed at different times, and they fall out of copyright at different times, this could be read as making distribution illegal after a while, and that doesn't seem to be the intent at all. So it seems to me that the "parts" considered here don't seem to have anything to do with what's being contributed. Instead, it has to do with what happens if someone extracts any portion of it that happens to also be covered by a license. In short, if you receive something that as a whole is licensed under the OFL-1.1, and it has a license at all, it must be licensed under the OFL-1.1. Saying "it must be under a license" is essentially inoperative: if you extract a part that isn't covered by copyright, there is no license that can be enforced to prevent it.

So now let's look at the phrase "distributed entirely under this license" and see if we can find a more reasonable interpretation of the clause, since the interpretation we just considered doesn't make sense. I think a better and more reasonable interpretation is that text means that the work as distributed is entirely under the OFL-1.1 license. That is, you can't add new clauses to the license to reduce the permissions granted to the recipient. It says "modified or unmodified" but that's straightforward - even if the resulting work is modified, you still get the rights & responsibilities of the OFL-1.1 (no more, no less). The oddity is when you see "in whole or in part", but I think there's a simple interpretation: if it's licensed at all, and you distribute a part, then that thing you distributed has to be under the OFL-1.1. But what if the part you extract is in the public domain? That's easy: If some part is in the public domain, and only that part is distributed, then no license (including the OFL-1.1) can change anything: that part is in the public domain. The license cannot prevent you from extracting parts that are not copyrighted, because once there's no copyright, there's no legal mechanism in copyright to prevent its redistribution.

I think the solution to understanding this license is that "distribution" is different from "modification". The US government additions do not have a copyright in the US, so those modified portions don't have a copyright in the US. But the entire modified work is under the OFL-1.1, because it includes parts that are under the OFL-1.1. That's no different than how it started: the original work (before US government modifications) included parts that were not under copyright, because you can't copyright single characters. Nothing in the license text forbids adding parts of uncopyrighted material, nor can it, since all copyrighted materials are built from combinations of uncopyrightable material. The license controls distribution of the result. The work as a whole (when combined) is still under copyright, so the OFL-1.1 is satisfied. When a part is extracted, it's either copyrighted (so the OFL-1.1 is satisfied) or it's not (in which case no copyright license can re-copyright it, and thus the OFL-1.1 cannot prevent it). The OFL-1.1 cannot imply that all parts of a work be copyrighted, because if you break anything into small enough parts then every single one of them is not subject to copyright.

This whole thing is very fact-specific to OFL-1.1. The widely-used OSS licenses don't have these odd text clauses that require puzzling through when you're combining public domain with copyrighted material.

Caveat: I'm not a lawyer, I'm certainly not a government lawyer, and I don't have a lot of time to devote to this discussion sadly. These are just my personal opinions, and I reserve the right to change my mind if someone provides me a good argument :-). That said, I think this resolves the issue.

[tssva]

@david-a-wheeler I don't know where you get the idea that you can't copyright a single character. That single character contains computer code as far as copyright law is concerned and is certainly still under copyright. Even the characters modified by government still contain code from the original font and therefore aren't completely public domain when extracted individually.

[malanthomasii]

@tssva I'm going to assume for the sake of argument that, because GitHub is a code repository and he's talking about other software licenses by comparison, that he means a single character of code. Take   as an example; I think he would mean the ampersand character in the code, not the non-breaking space character that it outputs when processed in context. The fact that we're talking about the code for a font (which produces specific character shapes as an output) is confusing the language here.

Unless he doesn't, in which case you're right.

[david-a-wheeler]

@malanthomasii - correct. What I mean is that you cannot copyright a single character in complete isolation.

[mbutterick]

[I’m not anyone’s lawyer and nothing I’m saying in this thread is legal advice.]

if you break anything into small enough parts then every single one of them is not subject to copyright … I think this resolves the issue

It “resolves the issue” only by detonating the very idea of copyright in literary works (a category that includes software code). By your argument, if we just zoom in far enough on any unauthorized copying of a literary work, we can claim that we’re merely copying “single character[s]” that are individually “not subject to copyright”. Poof — no copyright violation is ever possible.

(Leaving aside that this theory seems to contradict the factual representations the GSA has already made about the relationship of Public Sans and Libre Franklin.)

[david-a-wheeler]

By your argument, if we just zoom in far enough on any unauthorized copying of a literary work, we can claim that we’re merely copying “single character[s]” that are individually “not subject to copyright”.

Not at all. A single character in isolation is not subject to copyright, because any one character in isolation cannot be original. Combinations of characters of course potentially are subject to copyright. My point is that claiming that any possible part must be subject to copyright is manifestly absurd, because there is a single character that is "part" of a larger work, and can be extracted out of the larger work, yet a single character in isolation is not subject to copyright. Therefore, interpreting the license clause that "all possible parts are subject to this license" is absurd.

A better interpretation is that the "entirely" in the clause "must be distributed entirely under this license, and must not be distributed under any other license" means "this work (modified or not) must be under this license and not a different license". That seems far more logical, since "the entirety of every possible part no matter how small must be under this license" is not actually possible. Under this reading of "entirely" there is no problem. No license at all is not a different license - it's not a license at all. Therefore, there's no conflict.

[mbutterick]

[I’m not anyone’s lawyer and nothing I’m saying in this thread is legal advice.]

interpreting the license clause that "all possible parts are subject to this license" is absurd.

i don’t recall anyone making this argument. The principal question in this thread pertains to the OFL requirement that the modified work “must be distributed entirely under this license”, which is incompatible with the fact that the GSA’s modifications are public domain, and therefore cannot be placed under any license.

A better interpretation is that the "entirely" in the clause "must be distributed entirely under this license, and must not be distributed under any other license" means "this work (modified or not) must be under this license and not a different license".

I don’t recall anyone seeking to debate the meaning of “entirely”.

No license at all is not a different license - it's not a license at all. Therefore, there's no conflict.

Correct, there is no apparent conflict between public-domain modifications and the part of the OFL that requires that the modifications “must not be distributed under any other license”. But again, that’s not the troublesome phrase, as mentioned above and elsewhere.

In any case, if the GSA agrees with your argument, I’m sure we’d all be interested to hear that from them, since they’re the only ones who can rectify the problem.

[tssva]

In any case, if the GSA agrees with your argument, I’m sure we’d all be interested to hear that from them, since they’re the only ones who can rectify the problem.

@mbutterick - Given the lack of any response I assume the plan of the USWDS at this point is to ignore the issue and hope it goes away. Any real movement would probably require reporting the issue to the GSA OIG.

[david-a-wheeler]

@mbutterick:

The principal question in this thread pertains to the OFL requirement that the modified work “must be distributed entirely under this license”...

My current thinking is that it is, because no other license is being used. What's more, the work in its entirety is being distributed under exactly that license without change to the license (which I think is the intent of the "no other licenses" part). Done and done.

In addition, I suspect a court will try to move forward by trying to figure out intent of the license (not just its raw text) when trying to interpret it. The intent of the license is stated in the preamble: "The goals of the Open Font License (OFL) are to stimulate worldwide development of collaborative font projects, to support the font creation efforts of academic and linguistic communities, and to provide a free and open framework in which fonts may be shared and improved in partnership with others." It also says that the " The fonts and derivatives, however, cannot be released under any other type of license."... and since the work as a whole continues to be under the same license, that also is met.

So I think there's an interpretation of the license text that has been met (and I think it's quite plausible), and the overall goal of the license is being met. That doesn't sound like a problem.

In any case, if the GSA agrees with your argument, I’m sure we’d all be interested to hear that from them, since they’re the only ones who can rectify the problem.

Clearly people disagree on whether or not there is a problem, and the issues are extremely specific to the OFL-1.1. The question would not even come up with more widely-used licenses such as MIT, BSD-2-Clause, BSD-3-Clause, Apache-2.0, LGPL, or GPL, because none of them have this unusual (quirky!) clause.

It'd be awesome if USWDS released a formal opinion of counsel (or similar) that explains why they believe it's fine. I do not have the power to force that :-).

[thisisdano]

Addressed in https://github.com/uswds/public-sans/pull/123