Hi Nando,

> I couldn't really understand the meaning of Carlos' benchmarks,
> since it looks like he did other changes (like deactivating indexes
> and committing every x inserts) other than switching to IBO.

When Pelle first post his message I didn't know what scenarios he
wanted to target so I post a number of possible ways to improve
insert statement. Of course disabling indexes and committing for each
1000 records, it's only useful for certain specific implementations,
like massive inserts (and this is really useful in such conditions)
but this wasn´t the main change I pointed to him, which was:
replacing TIBOQuery for TIB_DSQL.

> In addition, a meaningful benchmarks should act on the same
> database each time and Carlos didn't say whether this one has been
> conducted this way or not.

Of course I did, when I say a repeated the tests more then once,
maintaining the environment conditions is always a must.

> Anyway, I am particularly concerned by the poor performance of
> IBODataSet.

That´s a fact and comparing with other TDataset layers like FIB
plus
you reach the same conclusions. Why? Only Jason can tell us.

Best regards,
Carlos Macao