Re: fulltext search in versioned content

My idea was to store complete revisions (along with properties) in a
database for retrieval, not much different from the way this is done in
catacomb. The database would handle those portions of a search request
that require matching in properties.

As for those portions of a search request that require matching in a
fulltext field, I would think an inverted index provides superior
performance to that of a database, given as the focus of search
optimization is different. In other words, you would maintain a fulltext
index alongside the database, so that each system handles those portions
of the quries that it is by design good at handling.

Upon reflection and thanks to your comments I see the drawbacks to the
idea... the merging of results from the inverted index and from the
database, the delegation of tasks to the two systems etc would be
cumbersome. The approach also prevents optimizations that can be made by
the database if it were to handle the query by itself.

back to the drawing board. thanks for the feedback :)

cheers
gerret

Received on Thursday, 26 February 2004 17:59:38 UTC