At 11:51 23-05-2003 -0700, Lennie Stovel wrote:
> >.. this looks like a killer argument against it. At least it has me
> >convinced :) I can think of plenty of times when I'd want to know how
> >many results came from each database.
>
>
>Rob,
>Yes, you can think of them -- but what are the requirements of the
>metasearch
>products?
I believe that the requirement of the metasearch products (at least the
ones I make these days) would be to be able to execute a query (or possibly
different queries) against multiple databases, and get back multiple hit
counts/result sets. Now as a metasearch provider, I am perfectly happy
launching multiple, parallel queries against the databases and not care too
much about whether they happen to be on the same server or not, or indeed
if they happen to be all logical subdivisions of the same physical database.
However, the content providers (like EBSCO, Gale, and Elsevier) do *not*
like the metasearch engines doing this because it potentially overloads
their systems and screws up their statistics gathering. So the requirement
for an IR protocol to address multiple databases on the same server, in a
flexible way, comes from the content providers, not from the metasearch
engine developers... I believe that the NISO meeting is somewhat different
from our typical ZIG discussions in that it very prominently included
commercial content providers... it is only to be expected that their
requirements would sometimes be different from that of the users or
metasearch engine people.
--Sebastian
--
Sebastian Hammer, Index Data <http://www.indexdata.dk/>
Ph: +45 3341 0100, Fax: +45 3341 0101
|