"Jonathan Hull" wrote:

> The only other approach that I can imagine (after a strong
> coffee) would be to tokenise the text on the way in so each BLOB
> has a corresponding text field containing keywords, but this seems
> a pretty high overhead.

If you are going into hardcore free text search, something
along this lines is your only option, as otherwise you will
end up scanning your entire hard disk over and over again.

Regards,
Peter Jacobi