>Cetin,
>
>own problem is not on the time when the SELECT start, but while the SELECT execute, DATETIME() change, and this change is usefull to you.
>
>The process of call DATETIME() is random for the first call, but all next call are correlate with this first event;
>for remove this, is need that the execution process have a white spectre.
>
>For get 5 randomly picked customers, you need SELECT duration is more 5 seconds, and implement one randomization function RANDOM(DATETIME()).
>
>The naturally function for do this task is RAND() function.
>
>I need one exclusive example, that is one SELECT impossible to implement without the DATETIME() dinamically variations.
>
>Fabio

Fabio,
It seems you're fixed to datetime() function. The only usefull dynamic usage "I" can think of that is to generate test data. Apart from that I even find it dangerous to use 'datetime datatype' in comparison parts.

Back to picking 5 customers, fortunately this is easy both in VFP and SQL server :

VFP :
select top 5 rand(),customerID from customers order by 1

SQL server :
select top 5 NewID(),customerID from customers order by 1
* rand() in SQL server behaves as one of those evaluated once at top
* Fortunately newid() doesn't

PS: Don't tell me in VFP top 5 really doesn't mean top 5 all the time. I know, would only help to get the discussion further (in SQL server add 'with ties' and not there too).

Cetin