Some things to bear in mind:

1.  SRW/CQL is not a relational system.

We cannot express a search like:
    Find all items of 1st edition expressions of works that have
    manifestations which are catalogued by Rob.

2.  Our 'atom' for searching is a record.

When we create a result set from a search, we create a list of records.

If you want to search for records of Works:
   dc.title all "lord rings" and frbr.entityType = work
But you can't search for:
   dc.creator any sanderson and xxx.yyy = metadata

because the 'metadata' applies only to the previous index.
Hence for this we need a relation modifier for that axis.

The other axis is the one expressed by FRBR.  You can have a creator for
any of their Group 1 Entities [Yes, I went off and read the full PDF, I
still think it's insufficient for correctly modeling my PhD] and we don't
want to have different index names for each different level.

You could have a search:

   dc.creator =/frbr.entity=work froissart and
   dc.creator any/frbr.entity=item "scribe1 scribe2" and
   dc.creator =/cql.metadataLevel=metadata "sanderson"

Records created by me, about manuscripts created by scrib1 or scribe2,
which are (exemplars of embodiments of realisations of) works of Froissart.

You could (though I don't see any reason to do so) FRBRize the record.

   dc.creator =/frbr.entity=work/cql.metadataLevel=metadata "sanderson" and
   dc.creator =/frbr.entity=manifestation/cql.metadataLevel=metadata "taylor" and
   dc.creator = froissart

would mean that you wanted to find records where I conceived a metadata
record about this book, but Mike actually created the record.
(As it's an electronic record, we need to use manifestation, because it
could be copied from some other catalog, which would make the copier the
item's creator...)

Right? :)

On Wed, 18 Aug 2004, Peter Noerr wrote:
> In modeling terms what Rob is proposing is both simpler and more complex
> than described. Ray's Godfather description is all within a "bibliographic
> plane" - the various objects (real and abstract) are all related to each
> other within that plane. The cataloguing record (a surrogate) is orthogonal
> to that plane. So in Rob's original example the first two objects are within
> the plane (the work and the instance), but the cataloguing record is outside
> the plane. There are, however, two howevers.

> So we end up with a situation where we may be creating an extension which
> nobody wants to use. I think the extension is a good idea ('level' is the
> wrong word though as it is used in FRBR), and potentially useful in a

Any appropriate name is fine by me! :)

While creator isn't an actually useful distinction in classical terms (but
see Matthew's example), and neither is title, there are other fields which
are useful.

Creation Date is particularly useful, especially OAI wise.  Size (length
of book vs length of metadata record).  Identifier (isbn vs docid)  ...
Language (of book vs of record) ...

Rob


       ,'/:.          Dr Robert Sanderson ([log in to unmask])
     ,'-/::::.        http://www.o-r-g.org/~azaroth/
   ,'--/::(@)::.      Special Collections and Archives, extension 3142
,'---/::::::::::.    University of Liverpool
____/:::::::::::::.  L5R Shop: http://www.cardsnotwords.com/
I L L U M I N A T I