>Well I'll try to pull together a few thoughts scattered throughout earlier
>messages, for a more coherent purpose:
>These would be a set of search points that:
>1. would have the best chance of corresponding cleanly to indexes when the
>database is mods, 
"when the database is mods..." Here we get into the different ways that 
one can implement a record format as a database. There are times when a 
database design and a record format are closely aligned, such that the 
fields in the record are pretty much the searchable fields in the 
database. I wouldn't expect this to be the case with mods due to 1) the 
general complexity of fields and subfields 2) the re-use of data 
elements in different contexts (e.g. name as subject v. name as creator, 
or the whole related record structure that consists of the mods record 
inside itself). My assumption is that indexes are not equal to fields in 
a bibliographic database using mods or marc as their record formats, for 
these reasons. With dublin core, because of its flat structure, I can 
imagine an implementation in a small database that treats each of the 
fields as an index, but my experience with more complex bibliographic 
data is that indexes are constructed and do not correspond directly to 
the record's fields. I guess that's what has confused me about this 
exercise -- because I see indexes as being a database implementation and 
not a function of the record.

But, you asked for comments on your list.... so I would propose that 
these elements are NOT useful for searching:

issuance (enumerated)
location URL
part- detail - number
part- detail - caption

part - extent - start
part - extent - end
part - extent - total
part - date

And I'm not clear on the element "part- detail - title". I believe that 
we have only "part-title" and that it refers to a host item, not a 
detail of the item in the bibliographic record. This would suggest that 
you would need an element "title-host" because "part-title" would seem 
to be a reference to the title of the part, and that is in the main 
title data element (equivalent to the MARC 245). I also note that there 
are no search fields here for the "related item" fields, like earlier 
and later titles for serials.

In terms of the identifiers,


... does it make sense that the search term and/or the index will be 
identifier specific, or that the identifiers themselves will carry their 
own identification, like "doi:10.1016/j.acalib.2004.12.001"? Should 
there be a generic identifier element that can carry the many other 
identifiers that can be found in a mods record?

I'll see if other things strike me when I look over your list again.


> but would also be mapable to other formats, perhaps not
>quite as cleanly; and
>2. are implementable; and
>3. are considered useful for searching.

Karen Coyle / Digital Library Consultant
[log in to unmask]
ph.: 510-540-7596
fx.: 510-848-3913
mo.: 510-435-8234