>I am working on a system concept for technical information
>(documents, objects, ...). Briefly: files at specified locations
>would be scanned for a set of specified attributes, which would be
>pulled into an RDBMS server for querying (via HTML forms).
Do you want to search on the attributes and the full text? I'm afraid I
don't really know how other vendors handle this, but we build two indexes
to documents. One has attributes, which can be anything you want (though
it always includes two that we need) and the other is the full-word index
with a lot of information about words -- where they occur, density, zone,
But, if you really want to hook this into a relational database, we have
those hooks, too. You can use our keys to wire to SQL or whatever. I'm
still learning about what we have, too, but I think the basic difference is
that our attribute file is flat, so if you want the relational
capabilities, that's when you build a gateway via SQL or whatever.
The short, non-product-pitching answer is that it's certainly possible to
combine it all.
One reason you *wouldn't* want to stick attributes into a relational
database would be if you want to do text searching on the fields. An RDBMS
will get very, very slow if you are searching for text within a field over
lots of records. But it looks as though your application's fields would be
searched for exact or ordered matches, no?