Mathieu,

> As a librairian, I think the "any" index defined in record.abs often return too much results, and inappropriate.
> It is partly caused by "any" index indexing all fields, including all notes fields (5XX in marc21, 3XX in unimarc), and of a lot of fields that could be usefull for staff but not for our public (internal item notes, statistic codes, date/place of publication etc).
> I want to submit you an idea :
> - creating a new index in zebra, matching only "usefull" fields and subfields (and it will be editable by all libraries...). For example we could call it "opac-any"
> - modifying the templates so that this index may be used instead of "any" when somebody makes a search in opac.
>
> What do you think of that?

I like the idea, though I would prefer to use the bib-1 attribute intended for that (there is one, though I forget which and can't check just now). Logistically speaking, however, this is quite difficult. Record.abs and its DOM equivalents are files on a file system, and cannot be edited from the staff client. This is not an insoluble problem, but making this possible in a way that could be integrated with solr without forcing us to go back to primitive GRS-1-style indexes would take some doing.

As an initial step, you might consider submitting a patch that adds the hardcoded index and allows librarians to choose which keyword index to use on the OPAC and staff client. This shouldn't be too hard.

Regards,
Jared