Non-copyrightable work with non-free license.
in at least two of my packages, bioperl and emboss, the test suite contains
protein sequence files from the UniProt database, which is distributed under
the non-free CC Attribution-NoDerivs license. Through a private discussion I
had with their helpdesk, my understanding of their position is that they are
trying to enforce the equivalent of a trademark.
Nevertheless, facts such as protein sequences are not copyrightable. This is
somewhat confirmed by the UniProt consortium itself on their website
(http://www.uniprot.org/help/license), and my conclusion is that, in isolation
from the rest of the UniProt database, the records in the test suites of
BioPerl and EMBOSS are not copyrightable.
I have asked the FTP team on that matter and did not got positive or negative
comments, so I think I will like to go ahead and simply dismiss the license of
these files, on the grounds that copyright does not apply to them. But given
that this is a potentially sensitive issue, I prefer to send a last message to
a broader audience before considering the problem solved.
The UniProt consortium is non-profit and funded by public money.
Note that this is not related to my past proposal of keeping non-free files in
the source packages when we do not use them. Here I state that these files are
free because they are not copyrightable, and moreover they are used at build
time, in the regression tests. Actually, this is one of the resaons that I do
not want to remove them: if on my machine I test the pacakge with the full test
suites, I can not "unlearn" that the UniProt parsers are working correctly,
even if I remove the files and break the test suite before upload.
Have a nice day,
Debian Med packaging team,
Tsurumi, Kanagawa, Japan