>Hi Dragan,
>
>>I guess that's because VFP used to store milliseconds once upon a time - in VFP 5 and maybe 6 - but displayed only seconds, so we had the same problem in bare dbfs.
>
>Sorry, should have expressed myself better. VFP does store the millisecond part. It's 0 now, in VFP 6 it used to be 997, 998, etc. with different millisecond values each time the field was written. Strangely, the milliseconds value was always close to the second values differing by like a maximum of 50 ms. I haven't seen date time fields in VFP 6 where the millisecond part was in the .400 to .600 range, for instance.
>
> In memory however, the datetime value is only precise to the second.

Sure ?

X=DATETIME()+0.123 && add milliseconds

CREATE CURSOR DT (F T)

INSERT INTO DT VALUES (m.x) && VFP truncate here

? "on variable ",(DATETIME()-m.x)*1.0000  && -0.123
? "on field ",(DATETIME()-DT.F)*1.0000     && -0.122