[mb-style] RFC: "Changed name to" for labels

[Chris B]

On 09/02/2008, Brian Schweitzer <brian.brianschweitzer at gmail.com> wrote:
> (withdrawing this RFC btw)
>
> > >, yet still lets us document
> > > the interesting bits about who owned who
> >
> > We hit it.
> > That's interesting, but for either annotations or a financial database
> > (which we are not)...
>
> If so, why do we bother to have LabelOwnershipRelationshipType?  If
> you read the example, they have this same confusion.  "Label" is being
> used interchangibly to refer both to "the company" and "the imprint".
>
> While you can say that it is always "the imprint" referred to, I think
> the perception, and a lot of the confusion, has come from people
> putting information about *the company* in, not realize that it's *the
> imprint (originally) by that company* which the listing intends.
>
> This is even more true when you get to the
> LabelDistributionRelationshipType, which is entirely about financials
> - company Foo can quite easily distribute the imprint Xan for company
> Bar - but if we're only talking imprints - what's on the record itself
> - then how can one imprint distribute another imprint?  Wouldn't that
> latter pretty much be, by definition, LabelReissueRelationshipType,
> and not LabelDistributionRelationshipType?
>
> <snip>
> > The current version of the documentation IMO is already quite clear about that:
> > http://wiki.musicbrainz.org/Label
> >
> > "Record labels in MusicBrainz refer mainly to Imprints
> > And to a *lesser extent*, to Record companies, [which] you *may*, in
> > *some specific cases*, want to use [...] rather than an imprint"
> >
> > Am not too sure how to make that more straight-forward :-)
> > This just works 99% of the time.
>
> In theory, yes.  In reality, no.  Were it being done in this way, EMI
> might not have 4 pages of different labels in the database.  I really
> don't think labels have to be as confusing as you make them sound.
> Can they be confusing?  Sure - but the way we're doing it pretty much
> guarantees that confusion.
>
> Reviewing the "Label" page, as it says, we store three very distinct
> different things and call them all the same thing:
>   * imprints
>   * companies
>   * music groups
>
> Yet when you create a new label entry, this isn't how we break it
> down.  There, we instead break it down to:
>   * distributor
>   * holding
>   * production
>       - original
>       - reissue
>       - bootleg
>   * publisher
>
> Even the label page linked from the editing page,
> http://wiki.musicbrainz.org/LabelResource , is principally about
> companies (EMI Group, etc) and not imprints.
>
> So right from the start, there's a clear mismatch between the Label in
> theory and the Label in implementation.  If I'm adding, to grab a
> random CD:
>
> NIN: Further Down the Spiral
>
> There's a Nothing logo, an Island logo, and a Warner Records logo each
> equally sized.  The text itself reads:
>
> 1995 tvt/interscpose (C) tvt/interscope. the copyright in this sound
> recording and artwork is owned by tvt/interscope and is exclusively
> licensed to island records ltd. in the uk. [...don't copy CDs
> text...]. LC 0407 PY 899 - biem/stemra
>
> Now, clearly, one single imprint actually released the CD.  But to be
> quite honest, adding this, is is then TVT, TVT/Interscope, Interscope,
> Nothing, Island Records Ltd., Island, or Warner Records?
>
> Most of those are companies, not imprints.  In our current way of
> entering and listing labels, though, we don't differentiate - in
> implementation - between those concepts.  So all of them end up as
> just "label"s.

nah, all of those can be imprints and companies depending on the
context. you get releases on warner with no other imprint, same with
interscope and island. nothing recs is reznor's vanity label - they're
currently 'owned' by interscope AFAIK, so it's an imprint of
interscope, but i imagine nothing recs probably released stuff on its
own when reznor was between labels, or 'unofficially'.

(TVT i don't know of)

the problem is (both here and discogs), is that we have one 'label'
for each release event, so we have to pick the lowest level imprint
(in this case nothing recs), but that's not the whole story. we can
show the whole story with the label-label ARs (so in theory you can
see the whole organisation structure of a label for such and such a
date), but unfortunately labels can release different things in
different ways at any given time (eg not involving the whole
hierarchy), and also this doesn't show all the cat#s that each level
use on one release.

discogs is moving to a multi-field label system, so you can represent
all the various different involved parties on a given release, but
that means things get very complicated. personally i don't think
there's a nice way of doing it, but finding out the lowest level
imprint of a release is 'easiest' and is normally the most useful
element anyway.