What license is the data in `raw-data` available under?

[jordydickinson]

The data in data is under the MIT license, but raw-data seems to lack any license. The top-level license is GPL, which would suggest that the data in raw-data is also GPL. Is this intended?

[drammock]

Thank you for asking this question. To be honest, I don't think we thought hard about how the LICENSE file at the root would trickle-down to the sub-folders. As far as I recall, @bambooforest wanted the code to be GPL, and I wanted the aggregated data to be a non-viral license. More than a decade later, I now know that both GPL and MIT are not really suited to datasets, and (as I think you're implying) it would be incompatible for the raw data to be GPL and the aggregated data to be MIT-licensed.

So reasoning it through now, I would say that:

• scripts/ and tests/ we wrote and can pick the license. As the main author of the R code (cf the commit history of each file in https://github.com/phoible/dev/tree/master/scripts) I'd be fine with / prefer to (re)license as BSD-3 or MIT.

• mappings/ is data that @bambooforest authored/created. They're essentially database tables, so a CC license is more appropriate.

• raw-data/ are data mostly created/assembled by others, which we (mostly @bambooforest) reformatted to make them easy to ingest into PHOIBLE. I haven't checked each of them recently, but I spot-checked UPSID and SAPHON and neither one states a license, and the site for AA has disappeared.

  I am not a lawyer, but my limited understanding is that PHOIBLE can be considered a "compilation" and as such we are allowed to assert copyright over the collection itself, but not over the compiled items (cf. the section "Copyright Protection in Compilations and Collective Works" of https://www.copyright.gov/circs/circ14.pdf). I think one could argue that the files in raw-data/ have been sufficiently edited/processed/harmonized that it would be permissible for us to assert copyright over those specific files, and so it would be permissible for us to choose a license for them.

• the files in data/ are (to my mind) unambiguously something that we possess copyright for, so we can assign a license to those files. We made a more recent choice to license the dataset as CC-BY-SA 3 (see the website footer of https://phoible.org) and I think that should be reflected here.

So in sum, I'd propose a root-level LICENSE file that says something like:

Code in this repository is MIT-licensed. The data files in raw-data/ and data/ and mappings/ is CC-BY-SA licensed.

@bambooforest WDYT about all of that? @jordydickinson if you have relevant expertise, please also feel free to weigh in.