>>Overall, I agree that SQL is a better solution. How are you going to cope with the delay if you have a large result-set (mentioned earlier in the thread with Ed Pikman)?
>
>That was for test performance. However, if I would have to offer a sub set to a user like this, I would probably force them to enter at least 3 characters as the sub set will not represent anything as there is about 1/2 million records.
>
>Also, on a funny note, as I was talking to Jean-René about, is to use the filter for less than 3 characters and use the SQL for 3 characters and more. At least, this is what we have found. Filtering on one or 2 characters is immediate. After that, it starts slowing down. Workaround for a bug with a funny coding. :)

Oddly, I ended up with the 3-character limit myself on a half million record table where the users were searching for all matches. It was a compromise with the users wanting 2, but they didn't like the delay.