Hi,

On Fri, May 3, 2013 at 6:20 AM, Mathieu Saby
<mathieu.saby@univ-rennes2.fr> wrote:
> Thank you.
> So, just to know, do you think it would be
> 1. - a total waste of time ?
> 2. - a strange idea ?
> 3. - a good idea (for performance and code maintenance) ?
> 4. - harmful ?
>
> to merge the biblio and biblioitem tables ?

My gut feeling is #3, but a lot of testing would be needed to avoid
#4.  Possibly one way to reduce risk and improve testability would be
to move columns over to biblio one at a time.

Here are a couple additional thoughts if we undertake this:

[1] We should decide whether we want to keep both bibilioitems.marc
and biblioitems.marcxml.  I think we can get rid of the former.

[2] Even if we merge the tables, I think we should still keep the
marcxml column in its own table, e.g., one called biblio_metadata:

CREATE TABLE biblio_metadata (
   biblio_metadata_id INTEGER NOT NULL AUTO_INCREMENT; -- for the future!
   biblionumber INTEGER,
   metadata_type VARCHAR(10), -- for the future!
   blob LONGTEXT NOT NULL,
   -- and relevant FK contraints
);

The bits about slipping in multi-metadata-schema support aside, for
many queries there's no reason to pull in largish XML columns.

[3] We should see if we can drop some columns.  As a rule of thumb, I
suggest that if a biblio/bilbioitems column isn't referred to in the
code. any of the default MARC frameworks, or the SQL reports library,
it's probably not being used.