License for JSON database that prevents proprietary use

Question

I'm building a database consisting of large JSON files, based on sources that are licensed under the public domain or CC-BY. I would like this database to remain open but I would also like to prevent people from creating proprietary software that they sell to others essentially milking my work while not giving back to the public.

I considered 3 licenses but all of them seem to have drawbacks:

• CC-BY-SA: Not good for databases or source code, not really compatible with other licenses, loopholes around derivative works like binaries.
• ODbL: Great for databases but the "share-alike" side seems to be only in effect when the dataset itself is changed.
• GPL: Great for code and has the most strict "share-alike" policy but seems like it can still be circumvented by downloading the data from an external source. Not recommended for databases.

Maybe my understanding of "open data" is not in line with its definition. Seems like downloading the data file from and external source (just for reading, not for modifying) can't be prevented through these licenses. Should I just give up this requirement?

Answer 1

You have a dataset which you wish to release under an open licence. You also wish to encourage people who write software to work with this dataset to release their software under a copyleft licence. Firstly, good for you for wanting to free your data.

That said, you have a problem. The licence on software does not generally affect the licensing on the output of the software; that is generally only a function of the licensing on the inputs. Copyright does not normally extend from software to data, or vice-versa.

Leaving aside unenforceable encouragements, you could distribute your database under terms that required any software written to work with it to be distributed under a copyleft licence, and required the database to be re-distributed under the same terms. Such a licence is unlikely to be seen as being open, much as the SSPL - another attempt to extend the reach of copyright - has not been so seen. So forcing open others' analysis software closes your database.

Assuming you can live without that, the main contender for opening your database is, as you say, the ODbL. You say that

the "share-alike" side seems to be only in effect when the dataset itself is changed

but that is not entirely true. It is true that the share-alike conditions only cut in for modified versions, but the attribution and openness requirements apply even to verbatim redistribution. Anyone receiving a verbatim copy from a third-party will know that they have an ODbL right to reuse it, and cannot be technologically prevented from extracting the database from whatever they received. Even if they find they can't do that, the ODbL requires attribution; the sample attribution given within the text of the ODbL spcifies the inclusion of "the name of the Database and a hyperlink to the URI of the Database", so any such recipient can easily find your original copy of the database, and work from there.