On Thu, 2005-11-24 at 12:48 +0000, Mike Taylor wrote:
> > From: Rob Sanderson <[log in to unmask]>
> > What is the benefit of having a marc context set, rather than a
> > profile which maps marc fields to semantically named and defined
> > indexes?
> > 
> > eg 245 -> dc.title
> > 999 -> yourPrivateContextSet.yourPrivateField

> Surely it's obvious.  If you already have a MARC database which you
> search in other ways, and want to add an SRU interface to it, then the
> obvious way to do that is by using a set of indexes that let you
> trivially translate all the searches you're used to doing.  There
> really is nothing to be gained in defined private context sets with
> pricate indexes for such a database.

If all you are doing is searching your own database, why do you need
SRU/CQL at all?  If you want others to search your database, then how
will they know what your 999 field means?  Equally, how will they know
if your database is UKMARC, USMARC, FINMARC etc with all of their cute
little variations? How are they going to know that you've put notes in
521a and 520d, but not 500a ?  Or are they going to search ALL of the
possible note fields in the query to be sure that they find the note?

marc.500a any foo or marc.500b any foo or marc.500c any foo or ...
marc.535z any foo

If you include a-z0-9 subfields, and all combinations of indicators
that's thousands of possible indexes, just to find a lousy note.

Or you could send 'xxx.noteText any foo' and do the right thing of
leaving it up to the database to interpret that index against its own
data.

So no, it's not obvious to me that there's any benefit :)

>Otherwise I'm going to have to set an alias of marc.245a on all my
> > dc.title indexes, and no one wants to do that.
> 
> Not at all.  The marc.245a index would only be used on database of
> MARC records.  (And a _polite_ MARC database exposed through SRU would
> accept dc.title and the others as well as the MARC indexes, so that it
> can also participate in metasearching.)

How do I know that any given SRU interface is to a MARC database?  It's
not especially difficult to translate between DC or MODS and USMARC on
the fly ... does that make my database a MARC database?

What about MARC authority databases?  Do they use the same indexes to
refer to different semantic concepts to their bibliographic brethren?
I think I'm searching a "GENERAL NOTE" field with 500a, but actually
it's "SEE ALSO FROM TRACING--PERSONAL NAME" (whatever that means :)) And
if so, how does a machine know if it's searching a bibliographic
database or an authority database?

Also, SRU doesn't allow for the concept of a MARC database any more than
it allows for the concept of an XML database, a PDF database, a JPG
database or any other format.

In short, if there's nothing to be gained by using semantically aligned
searches, what's the gain of using CQL and/or SRW/U at all?

Rob

-- 
Dr Robert Sanderson
Dept of Computer Science, University of Liverpool
Home:     http://www.csc.liv.ac.uk/~azaroth/
Cheshire: http://www.cheshire3.org/