> -----Mensaje original-----
> De: Geoff Hutchison
[snip]
> In some sense, it already is decoupled in the 3.2 code--the
> scores aren't stored in the database but calculated by htsearch
on-the-fly. The only
> "scoring" that must be done with retrieval is marking fuzzy
> queries with the appropriate weighting from search_algorithm.
>
Right. So we're ready to implement an eventual distributed architecture, in
which we could have a 'broker' process asking for results to several
'collection' servers (not necesarily in the same machine) and
merging/sorting and them as required by the display module.
[snip]
> ... there's a lot of
> information in the template variables which may not be available from
> other search engines.
>
Sure, and thinking of 'meta-searching' using different sources, one must
take in account the different scoring mechanisms and query syntaxes.
But all the above is anyway feasible if one adheres to some kind of standard
interface. Such thing would allow to negotiate the details with each
retrieval server. Just like the flexible interface that has been proposed
between display and logic.
Personally, I'm not especially interested right now in integrating different
engines, even if I find it most practical.
I just suggest that the ability of distributing retrievals (i.e. storage and
CPU) can give a terrific boost to the performance of our beloved digger...
Regards,
Q
_______________________________________________
htdig-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/htdig-dev
