At 05:39 AM 21/12/2004 +0000, you wrote:

>The lack of support for milliseconds in the timestamp (especially
>since the storage is available) is definitely an issue for me. We
>write an application that catalogs images, which produce dates with
>millisecond precision. The key isn't the actual value per se, but
>the relative value versus other images. If a digital camera can take
>10 pictures within a second, we need to use the millisecond portion
>to define the order.
>
>If the database can sort the dates using the milliseconds, that will
>be part of the battle, but we will still need to show the
>milliseconds in the application.

There is no problem at all *storing* milliseconds, nor with indexing to
that level of distinction (provided you store it!)

If you want an exact system timestamp you can get it using the UDF
GetExactTimestamp() in fbudf. Takes no inputs, returns system time to
ten-thousands of a second.

./hb