Huge datafiles in free culture projects under GPL

4 ways how large raw artwork files are treated in free culture projects to provide the editable source.¹

In the discussion about license compatibility of the creativecommons sharealike license towards the GPL, Anthony asked how the source-requirement is solved for artwork which often has huge raw files. These are the 4 basic ways I described in my answer.

1. The Wesnoth Way

“The Source is what we have”

The project just asks artists for full resolution PNG image files (without all the layering information) - and only uses these to develop the art. This was spearheaded by the GPL-licensed strategy game Battle for Wesnoth.

This is a viable strategy and also allows developing art, though a bit less convenient than with the layered sources. For example the illustrator who created many of the images in the RPG I work on used our PNG instead of her photoshop file to extract a die from the cover she created for us. She took the chance to also touch up the colors a bit - she had learned some new tricks to improve her paintings.

This clearly complies with the GPL, because the GPL just requires providing the file used for editing published file. If the released file is what you actually use to change published files, then the published file is the source.

2. The External Storage

“Use the FTP, Luke”

Here, files which are too big to be versioned effectively or which most people don’t need when working with the project get version-numbers and are put into an external storage - like an FTP server.

I do that for gimp-files: I put these into our public release-listing via FTP. For example I used that for a multi-layer cover which gets baked into our PDF.

3. The Elegant Way

“Make it so!”

Here huge files are simply versioned alongside other files and the versions to be used are created directly from the multi-layered files. The usual way to do that is a Makefile in which scripts explicitly define how the derived file can be extracted.

This is most elegant, because it has no duplication of information, the source is always trivial to find, it’s always clear that the derived file really originated from the source and it is easy to avoid quality loss or even reduce it later.

The disadvantage is that it can be very cumbersome to force new developers to get all huge files and then create them before being able to really start developing.

The common way to do this is a Makefile - for example the one I use for building my PhD thesis.

4. Pragmatic Elegance

“Hybrids win”

All the ways above can be combined: Huge files are put in version control, but the derived files are included, too, to make it easier for new people to get in. Maybe the huge files are only included on request - for example they could be stubs with which the version control system can retrieve the full files when the user wants them. This can partially be done with the largefiles extension in Mercurial by just not getting the large files.

Also you can just keep separate raw files and derived files. This is also used in Battle for Wesnoth: Optimized files of the right size for the game are stored in one folder while the bigger full resolution files are stored separately.

If you want to include free art in a GPL-covered work, I hope this article gave you some inspiration!