On 12/07/13 18:02, Víctor Rodríguez Doncel wrote:
Hi Barry,
First, the obvious: the MusicBrainz core data is in the Public Domain,
and you can freely publish its transformation as you like, but the
MusicBrainz supplementary data is CC-BY-NC-SA and its transformations
must be shared "alike".
Agreed. More as an academic curiosity we were wondering whether data
produced through mappings could be constrained by the licensing applied
to those mappings.
So, if you have created a dataset, this time in RDF, containing the
"user ratings" (supplementary data) in MusicBrainz, you have little
choice: CC-BY-NC-SA. Yet DBpedia requires CC-BY-SA, so in theory this
downgrade would not be acceptable: you should not include in DBpedia
any piece belonging to the "MusicBrainz Supplementary Data".
I was only thinking about the linkset (sameAs and seeAlso
relationships), but now you've made me wonder whether even the
Wikipedia-targeting advanced relationships aren't actually supplemental
data - I will check this with the MB folks.
Next, about the granularity level for licensing: individuals of
rr:TriplesMap look a perfect target to be attributed and licensed.
Yes, I think that would be neat.
[...] A single triple pointing to a license ( <myTripleMap>
dc:license <license_of_your_choice>) should suffice.
Agreed, though the attribution (dc:creator, I guess - plus potentially
dc:contributor) could go over subsets.
Finally, I see the question "triples that are created using it" as
very conflictive one. Are those triples transformed versions of the
mapping?
Yes, as above. It's a bit of an academic question whether mappings can
constrain the license of data created using them (as certain software
affects products run through it - cf. Apple's)
Barry
El 12/07/2013 14:20, Barry Norton escribió:
I'd like to publicly release R2RML mappings for the MusicBrainz
dataset. DBpedia has shown interest in including the subset that can
be used to create a linkset.
Any idea what (kind of) licence could/should apply? (To be clear, to
the mappings, as opposed to the dataset)
I'd also like to attach, since R2RML is RDF, a licence and
attribution on a per rr:TriplesMap basis. (The mappings are hosted on
github and contributions will be accepted as I'm never going to get
through all of the MB Advanced Relationships, a moving target, myself
and I'm being a bottleneck.)
The question's also been raised on whether a given licence can in
turn impose conditions on the triples that are created using it (as
derivative works)? Does that sound feasible?
Any input appreciated.
Barry
