Text search takes advantage of the fact there are relatively few words
in English.  You can enumerate the words, and enumerate the documents
assigning each an integer.

Then for each word you have a list of document numbers sorted in order
by the number of hits of that word in the document.

To find documents containing two words, you read the list for the two
words looking for duplicates discarding everything else.  So you can
pull this off with two physical reads, and none if the words are
common ones cached in RAM.

You could do this is a clumsy way by having millions of SQL records of
the form  word, document, count, probably converting word and document
to a dense int first.
--
Roedy Green Canadian Mind Products
The Java Glossary
http://mindprod.com

If you want to create a search engine matching capabilities of the
real world ones, you can look at lucene (lucene.apache.org) and use
its API to index the database content. It is pretty straightforward to
use, and should get you a very good search engine. However, you'd have
to reindex every time the db content changes.

If the goal is to actually create a search engine, you'd have to read
up on indexing and related algorithms.

-cheers,
Manish

Since you are using MySQL, look at this:

http://dev.mysql.com/doc/refman/5.0/en/fulltext-search.html

Jeff Coffield