I have a hunch that life isn't going to be that simple.

It might be worth reviewing the FRBR (Functional Requirements for
Bibliographic Records) (I attach an extract of a presentation I gave at
MPEG earlier in the year which provides a rough overview).

FRBR defines four entities for an object:

The work (e.g. the Chroniques of Jehan Friossart)
The expression (e.g. the Chroniques of Jehan Friossary edited and
translated by...)
The manifestation (e.g. the Chroniques of Jehan Friossary edited and
translated by... Online version)
The item (e.g. the Chroniques of Jehan Friossary edited and translated
by... On interactive DVD available at http://www....)

The fact is that the metadata record is itself is a work and therefore
the above categorisation is itself applicable to the metadata record


The work (MARC record for the above created by LC Cataloguer)
The expression (MARC record created by LC Cataloguer translated to
French by...)
The manifestation (MARC record created by ditto, translated by ditto,
supplied in OAI MARC XML)
The item (the MARC record downloadable from

So I'd argue that metadata versus data is a different axis for level
than work versus instance one.

However, I can't claim to be an expert in this area.

The idea of capturing this in CQL is probably worthwhile but I think we
need to get in some comments from cataloguers/metadata experts etc.
rather than half cook something.


> -----Original Message-----
> From: Z39.50 Next-Generation Initiative [mailto:[log in to unmask]] 
> On Behalf Of Robert Sanderson
> Sent: 18 August 2004 13:15
> To: [log in to unmask]
> Subject: (Meta)*Data in CQL
> As Mike says, "One man's data is another's metadata."
> In Eliot's GILS set he makes a threefold distinction for bibliographic
> data:
> The 'work'        (eg The Chroniques of Jehan Froissart)
> An instance       (eg Pierpont Morgan M.804)
> A catalog record  (a surrogate in the Morgan catalog for the instance)
> Which seems a very reasonable distinction to make, and could 
> easily apply to non bibliographic data as well.
> Each of these might have a date associated with it.  A date 
> for the original, a date for the printing, and a date for the 
> cataloguing.
> They all have different creators as well:  Froissart, an 
> unknown scribe, and the cataloguer at the Morgan who 
> described the manuscript.
> Currently in CQL we have the dc context set and a (somewhat 
> substandard) rec context set for metadata about records.  
> However this doesn't provide for anything other than 
> searching the catalog record and for information about that 
> record.  We still have a modeling problem when it comes to 
> distinguishing the 'meta'ness of our data.
> My proposal is to add a new relation modifier to the CQL 
> context set called, for example, 'level'
> This would be used to determine the level of metaness to 
> apply the search at.
> For example, to use a simple case of author vs cataloguer:
>     dc.creator any/level=work "tolkien asimov"
>     dc.creator any/level=metadata "sanderson taylor"
> Or:
> any/level=work "1336"
> any/level=instance "1420"
> any/level=metadata "2004"
> Then we can ditch a lot of the indexes from 'rec' which don't 
> sync well with the indexes from dc (for example)
> eg:  rec.lang       vs   dc.language
>         vs   dc.identifier
>       rec.createdBy  vs   dc.creator
> Thoughts?
> Rob
>        ,'/:.          Dr Robert Sanderson ([log in to unmask])
>      ,'-/::::.
>    ,'--/::(@)::.      Special Collections and Archives, extension 3142
> ,'---/::::::::::.    University of Liverpool
> ____/:::::::::::::.  L5R Shop: 
> I L L U M I N A T I