>I would think the ultimate solution is to have an index of the documents ( and
>their associated attributes ) be a completely seperate querriable entity.
>This would allow users ( consumers, other servers!! ) to query the index and
>get the index results back. Then local relavance and cost based processing can
>be performedon on the returned data. Once the consumer is satisfied with the
>potential list of candidate hits they may retrieve them ( or just keep a link
>to their existance ). Using this approach the current infastructure for
>document retrieval can remain the same.
Our engine builds a query-able index of the attributes, as well as a
full-word index. The question I was asking was how we tag attributes in
the HTML documents so that they can be captured by a spider or other
indexing tool.
I'm not sure that being able to exchange the indexes is the ultimate
solution, but it's a very interesting one that we'd like to enable.
Instead of retrieving a number of indexes to one's local machine, it will
sometimes be more efficient to send a number of queries to search engines
where the indexes live, searching all of them in parallel.
I suspect that both models will co-exist.
Nick
--------------------------------------------------------------------------
Nick Arnett "We are surrounded by insurmountable opportunity."
Verity Inc. -- Pogo (Walt Kelly)
narnett@verity.com