You better take care of your transactions explicitly for a test like this.
It is likely to influence the speed quite a bit, e.g. if you commit for
every record you will experience very slow inserting - and the same goes
for inserting 1000000000 records within a single transaction.

Set

At 13:42 11.02.2003 +0000, you wrote:

>I'll give it a shot in an upcoming nightly test. I had a couple crash
>on the client side and I noticed, I did not loose all the inserts.
>So I assumed there were some automatic commits.