[mb-style] RFC: Works lists (and other related changes then implied)

[Brian Schweitzer]

On Tue, Mar 4, 2008 at 5:22 AM, Olivier <viapanda at gmail.com> wrote:
> 2008/3/4, Brian Schweitzer <brian.brianschweitzer at gmail.com>:
>
> > Was: Proposal: Add pre-NGS works lists, change the intent of CSG,
>  >
>  > modify the "version of " AR into an 'instance of' AR,
>
>  I'm not sure I understand what you propose here...
>
>  Do you mean we could skip entirely the "later version" AR, because
>  it's "implied" as soon as two different recordings from the same
>  artist are instances of the same work?
>  If so, what we still loose is the availability of such an information.
>  Right now, you can directly find out all the different versions of a
>  song by a given artist (assuming the data is present - by checking the
>  earliest version and see to what it links).
>  If I understand well your suggestion, that would be replaced by a work
>  linking to all "instances" of it, for *all* artists.
>
>  Actually, that would pretty much turn this AR into the "cover AR"...

I don't quite agree.  We have two separate ARs for the process you
describe: earliest version of and earliest release of.  (I'll get to
parody/cover in a moment).

Artist John writes a song.

Artist John releases the original recording of the song on album X.
Artist John rereleases a new recording of the song on album Y.
'Best of songs from artists named John' releases the original
recording of the song on comp album Z.

All three are instances of the same song.  All three a "versions of"
the work, linked from the specific instances back to a singular point
(thus avoiding AR clusters), the work listing.

There also are two different recordings there for that same song.  The
album Y recording would stand alone, until it perhaps is rereleased.
But the album X version still has an AR linking it as an "earlier
release of" the song on album Z.

This is why I suggested replacing the version of AR, rather than
creating a new one - it's not only (better) doing the same thing as
the version AR, it's actually less redundant compared with the
"earlier release of" AR than the earliest release/version ARs
currently are to each other - now each AR would have a singular
defined target, rather than simply the "avoid clustering" principle as
a means by which we avoid clustering.

Regarding covers, it's again quite similar - artist Cindy may record a
cover, but her cover still has it's own earliest release of chain
within Cindy's recordings, while the version of would point back to
John's work listing.  If Cindy didn't sing a cover, but wrote a
parody, then that parody is itself a work:

John gets the work listing for the original
Cindy gets the work listing for the parody
An AR links the two work listings, showing Cindy's parody to be a
parody of John's original
An AR links John's work to later versions he records
An AR links Cindy's work to later versions she records
(etc. with earliest release, etc)

Thus, actually, why I'd opposed the implementation of the parody AR as
an attribute to the cover AR.  The cover AR is distinctly a "song
instance"-"work listing" AR, linking from one artist's recording back
to another artist's work listing.  A parody, on the other hand, links
two work listings, and not a work and an instance.  It's descriptive
of the nature of the new work, whereas "cover" describes an action -
playing a song originally played by someone else - and not something
inherent to your new performance of that work.

Brian