Probably because of some cache memory that is cleared automatically after a certain period of time.

Vlad

>>1. Because the index is "unique", ie not all records are indexed. If a index doesn't include all records, it cannot be used in Rushmore optimization. There are 2 types of such indexes: filtered and unique.
>>
>>2. Because in subsequent queries data is already extracted.
>>
>The file in question is over 30MB in size with a 7MB index file. The subsequent queries can be other keys or combinations. If the Web page is left in "idle" for a while then the next query is just as bad as the first. Why??
>
>Rgds
>Petras