Should a SIL Font Derivate Be Released?

Albert Creus

I'm considering a project to make a derivate of an Open Font licensed under SIL, I understand the derivate should follow the same license, but does it have to be publicly published?

The SIL site gives contradictory messages, or at least contradictory to my limited comprehension on legal jargon:
https://openfontlicense.org/how-to-modify-ofl-fonts/
In the same page it first states:

You are making changes to an existing open font and creating a derivative which means that new version must also remain open and should be publicly released.

And later:

3.6  Do I have to make any derivative fonts (including extended source files, build scripts, documentation, etc.) publicly available?
No, but please consider sharing your improvements with others. You may find that you receive in return more than what you gave.

I understand it's best practice to publish it, just want to know if it's mandatory. Thanks!

John Hudson

The Open Font License itself does not include any requirement to publish derivative fonts. You can make a modified font for your own purposes only, without any requirement to distribute it or actively make it publicly available.

Disclaimer: I am not a lawyer and this is not legal advice.

As I read it, the license also includes no terms that prevent you from restricting access to the modified font, i.e. if someone were to ask you for a copy of the font file, you are not obliged to give it to them. If they can get their hands on it, they have a right to use it, redistribute it, publish it, etc. under OFL, but the license does not put a requirement on anyone to do those things.

Thomas Phinney

MUST remain open: must remain under the same open source license

SHOULD be publicly released: that is a should, not a must. The bit in 3.6 makes this explicit, that one does not have to do it.

John Hudson

Also note that whatever that page or the FAQ says is not legally binding, so 'should’ in this context is moot. Only the text of the license itself is binding.

Disclaimer: I am not a lawyer and this is not legal advice.

Daniel Benjamin Miller

You're not required to publish whatever derivative you make. (You are encouraged to do so on the theory that what your version may be useful to others.) However, if you do make it public, you have to release it under the OFL.

Thomas Phinney

Daniel Benjamin Miller said:

However, if you do make it public, you have to release it under the OFL.

No, the font MUST be OFL-licensed even if not public.

At least, that is my reading of the license.

If you are wondering how that could possibly matter, then imagine somebody (a disgruntled employee, or a hacker, or...) taking a copy of the font from the place that made the derivative version, and posting it publicly.

Maybe they are violating an employment contract, or illegally hacking into a system, but the posting of the OFL font would not seem to be any form of copyright infringement.

Like John, I am not a lawyer and not offering legal advice.

Dave Crossland

Disclaimer: I am not a lawyer and this is not legal advice, and only my personal opinion, not the view of any organization I might be associated with.

Thomas, I agree. A font made by copying any part of another OFL font, where that part is subject to copyright in the first place, is therefore OFL even if never redistributed to third parties, yes Thomas, same as a font made by copying any copyrighted part of a propriety retail font is violating the copyrights of its rightholders. If the former isn't OFL, it is likewise violating the same rights. 

Daniel, I basically agree given the above nuance. 

Albert, the purpose of libre licensing works in the libre culture movement is to give the recipient of the work the freedom to use, modify and redistribute it, if they wish. Therefore a license that required them to publish their modifications wouldn't be a libre license. However, licenses like the OFL require that all recipients receive the same freedoms as the first recipient. This should make the situation crystal clear. 

Daniel Benjamin Miller

Thomas Phinney said:

Daniel Benjamin Miller said:

However, if you do make it public, you have to release it under the OFL.

No, the font MUST be OFL-licensed even if not public.

At least, that is my reading of the license.

If you are wondering how that could possibly matter, then imagine somebody (a disgruntled employee, or a hacker, or...) taking a copy of the font from the place that made the derivative version, and posting it publicly.

Maybe they are violating an employment contract, or illegally hacking into a system, but the posting of the OFL font would not seem to be any form of copyright infringement.

Like John, I am not a lawyer and not offering legal advice.

No, that's not correct. It's not correct when it comes to the OFL, and it's not correct when it comes to the GPL or any other "copyleft" license.

Copyright licenses only exist with respect to grantors and recipients. The OFL is a public license (as is the GPL, and as are Creative Commons licenses), meaning that it is offered to all recipients (and can be passed on to the next recipient).

The OFL itself states that you receive a number of permissions ("to use, study, copy merge, embed, modify, redistribute and sell modified and unmodified copies"), subject to conditions, among which one is that the font "be distributed entirely under this license." This condition only applies if you actually redistribute the font. If your hard disk was stolen, then you did not redistribute the work in the sense that word has in copyright law.

Moreover, though — and this applies to all copyleft licenses — the license being conditioned upon you granting a license to use derivative works doesn't make that grant self-executing. The license is granted to you on the condition that if you offer the work to the next person that you offer the license to them as well. But what happens if you don't follow this condition? Well, the license specifies what happens: the license granted to you is void (and thus you may be liable for copyright infringement). Some licenses have only a voiding provision and others (such as the CC 4.0 licenses) have a curing provision. If you void the licensed you offered by failing to offer a license on your derivative work, though, all that has happened is that you have voided your own license (and may be liable for infringement); it doesn't mean you did grant the license in accordance with the condition.

Dave Crossland

Maybe I've misunderstood, but, are you referring only to the disgruntled employee etc scenario?

If you receive it under the OFL, in what sense is it then not OFL?

Daniel Benjamin Miller

Dave Crossland said:

Maybe I've misunderstood, but, are you referring only to the disgruntled employee etc scenario?

If you receive it under the OFL, in what sense is it then not OFL?

It definitely applies in the disgruntled employee scenario, but the principle also applies more generally.

If we are to get in the weeds here: the OFL is not really an inherent property of the material, but a grant of permission related to the font which is made between two parties.

To put it another way, the OFL doesn't govern the font — it governs people (recipients and grantors). Specifically, it governs the relationship between the grantor and the recipient(s) and makes a grant of permissions on the condition that the recipient follows certain conditions. But you're not a recipient unless you receive and you're not a grantor unless you offer.

If you have a font, and modify it, but never offer it to anyone, your modified version has not been distributed under the OFL, because it has never been distributed at all. While you are required to distribute it under the OFL if you do distribute it (or else you lose the permission you were granted), until you actually do distribute it, the OFL has not been applied to your new version, since the license can only be applied to something by the licensor offering it to a (potential) licensee.

Dave Crossland

But "all rights reserved" means even retaining a copy isn't allowed except under the license terms..?

Albert Creus

Thanks for the clarifications and no worry, not taking any bit as legal advice, the client has its legal team that can figure things out.
It is clear to me now that public redistribution is not mandatory.

But the disgruntled employee scenario is quite interesting, since the license doesn't specify the how is the font obtained, so it could be taken/stolen from a company server and could then be disgruntledly redistributed (as OFL):

Permission is hereby granted, free of charge, to any person obtaining a copy of the Font Software […]

John Hudson

But "all rights reserved" means even retaining a copy isn't allowed except under the license terms..?

Nuance: even retaining a copy isn't allowed except as permitted by the copyright holder (which might not be limited to the terms of a single license).

[In Canada, at least, the phrase ‘All rights reserved’ seems to have been dropped from copyright statements. As I understand, it is considered legally meaningless: failure to include the phrase in a copyright statement has no significance regarding what rights are actually covered by copyright.]

Again, I am not a lawyer...

_____

Has the disgruntled employee scenario, or similar, ever been tested in court? It seems widely assumed that once a derivative of an OFL-licensed font is loose in the wild, then it is automatically free to be distributed under OFL, but I wonder if that would stand up in court in the circumstance of a custom derivative that is not intended to be distributed, but reserved for the private use of the person who made the derivative?

Here’s an interesting scenario that actually occurs: a font maker publishes a font under OFL. Later, the font maker is contacted by someone who has made or wants to make a derivative but wants it under some other license, possibly a private use proprietary license, or an unrestricted free software license such as MIT or Apache 2.0. This is possible because the font maker, as the copyright holder, can grant any kind of legal license.

Does this suggest that in the disgruntled employee scenario, a decision on the license of the unauthorised distribution of the derivative font might revert back to the copyright holder. since that copyright extends to the derivative?

John Butler

I don’t know how closely the OFL relates to the GNU General Public License, but if it’s closer to GPL than MIT/Apache/BSD, then the license assumes a “viral” quality, attaching itself to all derivative works, and this has long been a source of friction between the GPL Stallmanists and the rest of the open source world.

Daniel Benjamin Miller

Dave Crossland said:

But "all rights reserved" means even retaining a copy isn't allowed except under the license terms..?

The phrase "All rights reserved" is only included in copyright notices for historical reasons. Under the Berne Convention, there are no requirements to include a notice of any kind (although including one may be useful, and not including one might lead to a stronger claim of innocent infringement). As of the past few decades, the Berne Convention rule against formalities has been adopted in virtually every country in the world. While the history of this is pretty long (and I can't really get into it here), it used to be the case that a copyright notice was required in certain countries, most notably the United States and most countries in Latin America. Until 2000 (when the Buenos Aires Convention became fully redundant), omitting this phrase would mean that a work would not receive copyright protection in certain countries. The only reason that the phrase is still included today is because people got in the habit of using it. Whether or not you include the phrase "All rights reserved" no longer has any legal significance in itself anywhere in the world.

John Hudson said:

Nuance: even retaining a copy isn't allowed except as permitted by the copyright holder (which might not be limited to the terms of a single license).

As for whether or not retaining a copy of something without the copyright holder's permission constitutes infringement, that depends. It's not an absolute.

You can always do things that you have been given permission to do, but you can sometimes do some things without permission (fair use/fair dealing exceptions are part of that).

To my knowledge, generally speaking, possession is not infringement (and possession is not one of the exclusive rights of the copyright holder; compare the first sale doctrine); copying is. Infringing copies might be recoverable by the copyright holder, though. In Canada, for instance, while the Copyright Act provides for the recovery of infringing copies, "Nothing in this Act entitles the copyright owner to damages in respect of the possession or conversion of the infringing copies or plates." Of course, these laws were all written for a pre-digital world, and a lot of concepts can get muddled by how computers work.

But that's got to do with how copyright law works generally (a rather complicated topic), rather than how the OFL works specifically.

Has the disgruntled employee scenario, or similar, ever been tested in court? It seems widely assumed that once a derivative of an OFL-licensed font is loose in the wild, then it is automatically free to be distributed under OFL, but I wonder if that would stand up in court in the circumstance of a custom derivative that is not intended to be distributed, but reserved for the private use of the person who made the derivative?

Here’s an interesting scenario that actually occurs: a font maker publishes a font under OFL. Later, the font maker is contacted by someone who has made or wants to make a derivative but wants it under some other license, possibly a private use proprietary license, or an unrestricted free software license such as MIT or Apache 2.0. This is possible because the font maker, as the copyright holder, can grant any kind of legal license.

Does this suggest that in the disgruntled employee scenario, a decision on the license of the unauthorised distribution of the derivative font might revert back to the copyright holder. since that copyright extends to the derivative?

I'm not aware of any case anywhere with these specific facts. But I am certain that a derivative of an OFL-licensed font is not automatically free to be distributed under the OFL simply by virtue of being out in the wild. The GPL works very similarly, in (as John Butler puts it, borrowing a turn of phrase used by many others) a "viral" fashion.

John Butler said:

I don’t know how closely the OFL relates to the GNU General Public License, but if it’s closer to GPL than MIT/Apache/BSD, then the license assumes a “viral” quality, attaching itself to all derivative works, and this has long been a source of friction between the GPL Stallmanists and the rest of the open source world.

The OFL is a "viral" license, much like the GPL. But what does it mean for a license to be viral? Well, licenses do not technically attach themselves to derivative works; they're attached to works by the author. The GPL and OFL both contain provisions which require that, if you distribute a derivative work, you must offer it under the GPL/OFL. The GPL is explicit in saying that various requirements only apply if you "convey" the copyrighted work, and the OFL (similarly) provides for conditions that relate to "distribution" of the work.

While the disgruntled employee or stolen disk scenario has never gone to trial, there have been a number of cases dealing with copyright infringement of GPL-licensed software. The sample principles would apply to the OFL.

In the GPL infringement cases, the facts were as followed. The original developers created some software and released it under the GPL. The infringers developed some derivative software which incorporated the original GPL-licensed software. Then they released that derivative software, but did not comply with the GPL terms (so, they released their derivatives under a proprietary license). Since they did not comply with the conditions of the GPL license grant, they had no license to use the original software and were liable for copyright infringement. However, their derivatives were not automatically released under the GPL as a result of this discovery; the consequence for not applying the GPL as you're supposed to is the revocation of your license (and ensuing liability), not the GPLing of your code. (Note: the GPL also contains a curing provision, where, if you are contacted by the copyright holder regarding your non-compliance with the GPL, and then become compliant within 30 days, your license is reinstated. This means that if you infringe, you can fix it by releasing your derivative under the GPL. But this cannot happen automatically.)

The disgruntled employee/stolen disk scenario is a bit different. In the GPL infringement cases, the infringers really did convey/distribute their derivative version to other parties (without complying with the requirements of the license). But there is a big difference between actually offering something to somebody and having it taken from you. (In fact, publication is defined in copyright law as to require that you actually offer or distribute something yourself; having something taken from you doesn't constitute you publishing it.) Here, you never conveyed/distributed the derivative version, so you didn't break the terms of the GPL/OFL. If the thief who took your derivative version purports to distribute it under the GPL/OFL, then that's unlawful in a number of ways. If they claim to have the copyright, that could be slander of title. If they claim you offered a license, that can be fraud. And, in general, one of the strongest rights of any copyright holder is the right to control first publication of a work.

Another interesting (and even less likely to be raised in the real world) question is this: if the original is offered under the GPL/OFL, and then the derivative is stolen without being published, and the thief purports to offer the item under the GPL/OFL, this surely violates the rights of the creator of the derivative, but does it also infringe the original unmodified work's copyright? The GPL, for instance, has an automatic licensing provision for downstream recipients, but it only applies when the work is conveyed. It could be argued that there is no license received by a thief (since there was no valid conveyance). This is probably not relevant if the thief can also get the original online (since then they'd be able to get a license with respect to the original material independently). It might be relevant for the theoretical scenario where the original licensor privately distributes the item under the GPL/OFL only to the recipient who creates the derivative: can a GPL/OFL licensing chain be extinguished (in the extremely rare case where the work is not published on the internet in a way that's accessible to the public)?

This is all a purely academic discussion, of course. It is interesting to examine the mechanisms used by these licenses (since they have to be about people and their responsibilities, even if they are often thought of as being about the works themselves independent of the authors and users).

André Simard

I have a serious question: Why somebody would like to make a derivate of an Open Font licensed under SIL? Personal use only or something else?

John Butler

I have a serious question: Why somebody would like to make a derivate of an Open Font licensed under SIL? Personal use only or something else?

The simplest example would be to add new glyphs, e.g. IPA glyphs that harmonize with the basic design, or math extension, or Greek, Cyrillic, etc.

André Simard

John Butler said:

I have a serious question: Why somebody would like to make a derivate of an Open Font licensed under SIL? Personal use only or something else?

The simplest example would be to add new glyphs, e.g. IPA glyphs that harmonize with the basic design, or math extension, or Greek, Cyrillic, etc.

Thank you for your answer. That’s clearest to me.

jeremy tribby

John Butler said:

I have a serious question: Why somebody would like to make a derivate of an Open Font licensed under SIL? Personal use only or something else?

The simplest example would be to add new glyphs, e.g. IPA glyphs that harmonize with the basic design, or math extension, or Greek, Cyrillic, etc.

personally I think this is not a great example because glyph additions to an OFL typeface would ideally be contributed back to the upstream project. a better reason to fork an OFL project might be if you plan on making changes to the design itself

Dave Crossland

André, individuals or organizations regularly make derivatives of OFL fonts. I see these as being in two categories, convergent and divergent. The reasons vary widely. 

A strong motivation for people to appreciate libre licensing is that they use a useful digital work like a font and find something "broken" for their use case, and roll up their sleeves to "fix" it. Say, the vertical metrics are producing undesirable spacing in some software environment in active use, and it's easier to change the font than the system. Or say, a Vietnamese accent is mispositioned by a typeface designer who isn't fluent in reading that language, and a native reader wants to reposition it. These are convergent changes to a typeface design. It is common that these fixes get added back to the original "upstream" design, but not always, if the fix is not generally desirable.

A subtle way that usual things can be broken is they can be missing something, such as missing glyphs from a typeface design. Such glyph set extension is also convergent: the additional glyphs harmonize with the original ones, which may not be altered at all. A classic example is Vietnamese people who want to use a Latin font that lacks Vietnamese characters at all. These can be ripe for integration upstream, and being maintained by the original font developer, but they can also be rotten and unacceptable to the original developer, and refused. Often integration of downstream work can be more effort than doing it from scratch, sadly. Better collaboration tools in font editors might help, but this is under explored imo. 

Divergent changes are "remixes", and these are inherently difficult or impossible to integrate upstream, and thus are typically "forks", where they are positioned as new upstreams - with new family names. This can be just one glyph or a few, such as the Elizabeth typeface which Gawker Media in house graphic designers derived from Merriweather to change the question mark and quote marks; and which Sorkin Type eventually integrated into Merriweather as a stylistic set. Or, in the case of Ysra, Rosetta Type remixed the entire Merriweather typeface into something recognizably different in overall appearance. They did a nice job documenting this at https://github.rosettatype.com/yrsa-rasa/ although the css seems to be broken on that site now; but it was nearly a decade ago. DM Sans and DM Serif were remixes commissioned from Colophon by DeepMind's branding team. Vernon Adams made Tienne by intermixing Droid Serif and Artifika, https://fonts.google.com/specimen/Tienne/about

Since the OFL is a "strong" copyleft license, not a weak "pushover" non-copyleft license, all derivatives, convergent or divergent, remain OFL, and the possibility to converge them into the original is never foreclosed. 

Since web fonts are inherently being redistributed, it's sometimes nice to come across a fresh cut derivative of an OFL font, and know that it's available to download and reuse - and even further remixing.

John Hudson

I agree with your characterisation of convergent and divergent derivatives, Dave, but isn’t this assertion what is at question in some of this discussion

all derivatives, convergent or divergent, remain OFL

I think the most that can be said is that all derivatives are covered by the original copyright—as commonly the case for derivative works under copyright law—, and that if 
distributed by anyone other than the original copyright holder must be done so under OFL. Daniel’s point about a license pertaining to actions, e.g, the act of distribution, not to objects, seems an important one to keep in mind even when we do sometimes use a shorthand terminology such as referring to ‘an OFL font’.

John Savard

Daniel Benjamin Miller said:

Moreover, though — and this applies to all copyleft licenses — the license being conditioned upon you granting a license to use derivative works doesn't make that grant self-executing.

This is a very important point. And some people might find it hard to understand it as well. To make it easier to understand, I think it would be useful to explain why it is true.

Someone makes a modified version of an OFL font. They then distribute it as a proprietary font. Why aren't you free to copy it, since their act of so doing was illegal, the only way they could legally distribute the modified OFL font was under the OFL license?

One reason is that violating the terms of the OFL isn't a severe criminal offence.

But there is a more specific reason. A modified OFL font contains... the modification. This is a creative work with some value of its own. Perhaps taking away the copyright for the modification might be too harsh a penalty for the violation of the OFL.

And now here's the kind of example that demonstrates why the law has this principle in it. Let's use the GPL as our example instead of the OFL.

Let's suppose the maker of a major commercial operating system includes some code from BSD Unix in it. This is legitimate, since it's under the BSD license, which is a different kind of license, allowing fully proprietary derivatives. Both the Macintosh operating system and Microsoft Windows do, in fact, include some code from BSD - as Apple openly admitted (which would have been hard not to do, as OS X is basically a BSD distro with KDE, but with some added Apple proprietary code), and as was discovered of Windows after a source code leak (in their case, it was just a minor I/O library).

Now, suppose, oops, somebody by mistake put some code from GPL-licensed Linux in it, rather than code from BSD.

Our legal and political system, claimed by some to be in thrall to the forces of Big Business, judges that putting the entirety of Microsoft Windows, say, under the GPL as the penalty for such an oversight to be overly harsh. I find it hard to disagree with this, no matter how I might try to be sympathetic to the concerns of the socially aware.

John Savard

John Savard said:

Now, suppose, oops, somebody by mistake put some code from GPL-licensed Linux in it, rather than code from BSD.

It turns out this isn't even strictly hypothetical. For Windows 7, Microsoft hired a third-party developer to prepare the USB/DVD Download Tool for installing the software.

It turned out that parts of that tool were written with a GNU GPL-licensed compiler-compiler by the name of BISON; and, as it happens, unlike GCC, the libraries of which are under the LGPL, the output of BISON, because of the large amount of provided code it relies on, is also protected by the GPL.

As it happened, since the Download Tool was a separate piece of software of little monetary value in itself, Microsoft did remedy that situation by placing it, in its entirety, in open source, rather than removing it and replacing it by something else.

Dave Crossland

John Savard said:

Now, suppose, oops, somebody by mistake put some code from GPL-licensed Linux in it, rather than code from BSD.

Our legal and political system, claimed by some to be in thrall to the forces of Big Business, judges that putting the entirety of Microsoft Windows, say, under the GPL as the penalty for such an oversight to be overly harsh. I find it hard to disagree with this, no matter how I might try to be sympathetic to the concerns of the socially aware.

Citations please

John Savard

Dave Crossland said:

John Savard said:

Now, suppose, oops, somebody by mistake put some code from GPL-licensed Linux in it, rather than code from BSD.

Our legal and political system, claimed by some to be in thrall to the forces of Big Business, judges that putting the entirety of Microsoft Windows, say, under the GPL as the penalty for such an oversight to be overly harsh. I find it hard to disagree with this, no matter how I might try to be sympathetic to the concerns of the socially aware.

Citations please

Here's a cite for the story in the post following the one you replied to:

https://bit-tech.net/news/tech/software/microsoft-confirms-gpl-code-use/1/

I'm not sure what in the expression of opinion you quoted needs a cite; I was just trying to explain and clarify what other people had already discussed here, and any facts stated were obvious ones.

Dave Crossland

I think you'll find judges do rule in favor of the gpl requiring the complete work to be distributed according to the license; see SFC vs Visio that's going down at the moment, and Microsoft's compliance in exactly this way in the link you provided.

Thomas Phinney

The thing that this part of the discussion is tripping over, is “what is the complete work?”

The open source license in this case was related to software that was part of the “Windows 7 USB/DVD Download Tool”—a completely separate program that functions to download a particular version of Windows to removeable media.

The fact that said issue did not affect Windows is in no way evidence for the proposition that: “Let's suppose the maker of a major commercial operating system includes some code from BSD Unix in it.… Our legal and political system, claimed by some to be in thrall to the forces of Big Business, judges that putting the entirety of Microsoft Windows, say, under the GPL as the penalty for such an oversight to be overly harsh.”

The case you are citing simply does not provide any evidence for this proposition, because the code in question was not in the operating system at all. Not even in something bundled with the operating system. Just in a separate tool that is used by end users to download an operating system. So Microsoft just made it open source. Leaving some of us wondering how that is evidence that the opposite would happen if the case was different?

I am not even arguing the OPPOSITE. I am just baffled by how you think you are providing evidence that supports your theory.

John Savard

Thomas Phinney said:

The thing that this part of the discussion is tripping over, is “what is the complete
I am not even arguing the OPPOSITE. I am just baffled by how you think you are providing evidence that supports your theory.

I hadn't seen the need to provide evidence, and the theory isn't mine.

Instead, this all was explained in a post by Daniel Benjamin Miller:

If we are to get in the weeds here: the OFL is not really an inherent property of the material, but a grant of permission related to the font which is made between two parties.

To put it another way, the OFL doesn't govern the font — it governs people (recipients and grantors). Specifically, it governs the relationship between the grantor and the recipient(s) and makes a grant of permissions on the condition that the recipient follows certain conditions. But you're not a recipient unless you receive and you're not a grantor unless you offer.

My intent was simply to offer an illustrative scenario that would make his explanation easier to understand, by putting the matter in more concrete terms.

I wasn't advancing a new theory of my own, I was merely referencing one that had already been adequately argued for here.

Dave Crossland

Kindly, I think it's adding details but without changing the substance: you're not a recipient unless you receive, but when you do, you are thus bound by the terms in the use of the work, and therefore treating the terms as a property of the work makes sense. 

John Hudson

We already know that that licensing chains can be forked, i.e. that the same font that is being distributed under OFL might also be distributed under another license if a) that other license originates with the copyright holder or b) is another license that is permited downstream of a license that originates with the copyright holder.

So it seems a reasonable question to ask whether licensing chains can be broken?

Dave Crossland

A separate chain isn't a break