[OPEN-ILS-GENERAL] Toward a new Bibliographic-Item Model

[Dan Wells]

Hello all,

Discussions surrounding the best way to model serials have served to highlight
some clear weaknesses in the traditional Record->Call Number->Copy arrangement.
 This arrangement is time-tested, so it would have been foolishly brave to
abandon it too soon.  Still, perhaps it is time to reconsider our model from
the ground up.

What I am about to propose is not fully thought out, and may be unreasonable
in reality, but it could at least be the start of moving toward something
better, a long-term solution.  It will deal with only the last two layers of
the FRBR model, but could be expanded ‘upward’ to more abstract levels.  As
it stands, bibliographic records in use today are generally at the
“manifestation” level, so it makes the most sense to start there.

First, bibliographic records are, by design, singular.  They fully represent a
manifestation of a given “expression” of a given “work.”  The
manifestations themselves, however, can have one or more parts.  This is one
missing link in the current chain.  We can partially address it at the call
number level, but that is ultimately overloading the meaning of call number and
is a workaround we should address.  Instead, we might introduce a new concept
to the model.  For lack of a better term, we will call it a
“manifestation-part”.

Next we have the instance layer, called the “item” layer in FRBR.  Because
we have added manifestation-parts, we also must add a concept of
“item-parts”.  This in turn implies that “item” is not a direct
representation of a single physical object, but a concept which can be split or
combined as reality dictates.  To accommodate this fact, we might introduce an
extra-item layer which can contain one or more item-parts.  For now, at least,
we will call this entity a “unit”.

What do we gain by this new abstraction?  Well, I hope a more accurate
representation of reality.  Bibliographic records map one-to-one with a
manifestation’s contents, not its containers.  Having split things as
described, a unit may contain just a portion of the contents represented by the
bib record (think volume, or accompanying materials like a CD), the entire
contents (the most basic unit, our current model), or even the contents of
multiple bib records (bound-withs).  This model attempts a complete and
deliberate separation of named content portions (-parts) from the local reality
of their containers (units), and the global splitting at the record level
allows accurate matching and comparison of contents, even when the containers
used locally vary.

You may have noticed at this point that call numbers are absent from the
model.  This is certainly intentional.  They should be nothing more than one
(possibly shared) aspect of the location information for any given unit. 
Assigning any other meaning invites problems.

Dan

Previous related threads:
http://markmail.org/thread/az3bvnbz7ziyyjfm 
http://markmail.org/thread/5yi5bf5hyqiwvpuw 



-- 
*********************************************************************************
Daniel Wells, Library Programmer Analyst dbw2 at calvin.edu
Hekman Library at Calvin College
616.526.7133