Mitzi! I am honored to have you drop in on this thread.
Lots of big picture arguments from Mitzi. Let’s see if I can drag things back into the details :)
Let me address two of Mitzi’s claims:
Claim 1: We use OSS practices, so our use of the CC-BY-ND license for documentation is on a sustainable legal footing.
If we were discussing an OSS license, then I would agree. However, CC-BY-ND is not an OSS license because it is not for software and it is neither OSI nor GNU.org approved. But perhaps we are fine because CC-BY-ND is commonly used in OSS projects for documentation? I am not aware of a major collaborative OSS project that uses the CC-BY-ND license for documentation, so I did some research:
I checked wikipedia’s list of major works using a Creative Commons license. I found a few single author books that use an ND clause. None of the works listed under the following sections use an ND clause: Educational Resources; Games; Knowledge, Research and Science; Technology, blueprints and recipes; Websites (except for two anomalies that I think are errors). Only one of the works listed under Video Games uses an ND clause. That game was released as abandonware under the ND clause when the studio owning the rights to the game went out of business.
But we are discussing OSS projects, so I looked into some big ones. The Python developers created the Python Software Foundation in order to “own Python-related intellectual property”. The python documentation is licensed under the Zero Clause BSD License. The R core team created the R foundation to “Hold and administer the copyright of R software and documentation”. The R manual R-intro is released under a BSD-like license that allows derivative works. “The Rust Programming Language” book is hosted on github just like the stan docs and uses the MIT and Apache2 licenses for contributions.
But this investigation would not be complete without the answer to the question: What Would Hadley Do? Hadley Wickham is about to publish the second edition of a book called R for Data Science. The open-source website is available under the CC-BY-ND3.0 license. He asks contributors making substantial contributions to the book/website to assign their copyright to Hadley Wickham so that he can publish his book.
I am happy to be proven wrong, but to my eyes, the CC-BY-ND license is an exotic choice for the documentation of a collaborative OSS project, with the notable exception of Hadley’s book, although I think Hadley may be aware of the same issues that we are discussing here because he asks for a transfer of copyright upon contribution. I therefore don’t see how the stan community’s choice of licenses is “standard across OSS projects”.
Claim 2: The work is not a derivative of itself, it is a single collaboratively edited thing
While in our hearts we all know this to be true, legally, this does not match anything I have ever read about copyright. It also does not match my reading of the stan docs contributor agreement. When a group of copyright holders gets together to publish something, an academic article say, then all copyright holders need to consent to the publication of their joint work by issuing a license or transferring their copyright. Absent a transfer of copyright to the publisher, this consent needs to be obtained every time a revision of that work is published. However, with code that is publicly hosted on github, every commit is a revision of the publication. Absent a permissive license, every copyright holder needs to consent to the hosting of every commit publicly available on github. The CC-BY-ND license does not avoid this issue because it can only be issued for currently existing work. The ND clause even explicitly witholds consent for future work derived from the current work. I suppose reconsenting everyone could be done while the copyright holders are still able and willing to consent, but there is the Mitzi-getting-hit-by-a-bus-problem.
This issue could be avoided if we took the route that Python, R and Hadley took by asking contributors to transfer their copyright to a legal entity such as a non-profit or Hadley Wickham. The entity could then choose to make the documentation available under any license that it chooses. Alternatively we could take the route of the Rust book and allow all contributions to be turned into derivative works through a permissive license.
Please don’t let the stan docs become abandonware ;)