Hello Michael, Nicholas et list,
I hope you don't mind me jumping in here with a few comments on selected highlights of this thread.
Taking /music as an example...
I wonder if you have looked at book data? I am working on issues to do with linked (open?) book data and it would be useful to compare notes.
wikipedia tends to conflate... composition with recording with release...
On the other hand, data does exist that separates these (and more!) entity types out very clearly, and it's potentially highly *linked* but it's unlikely to be *open*. See:
http://www.ddex.net/ddex-present - ddex descriptive data schemas, but also note the links there to IDs for
-names (ISNI) -compositions (ISWC) -recordings (ISRC) -releases (GRid)
These are all industry-standard IDs, and thus pretty stable. Maybe a starting point?
...domains where there's no established (open) authority (eg the equivalent of musicbrainz for films)...
EIDR? http://eidr.org/ - " EIDR is operated on a non-profit cost-recovery basis..." but maybe you get the stability and granularity you pay for? Plus; "... EIDR is founded on the principle of open participation and welcomes all ecosystem players (commercial and non-profit) to join the Registry as registrant, lookup user or even a promoter. The Registry is intended to provide a foundational namespace for A/V objects that can be leveraged by participant in the eco-system to further their own business needs and offerings." - http://eidr.org/resources/
Cheers,
Michael