>>There is a difference in these two approaches if you have several same days. In my case it still returns just one record, in Sergey's it would return all of them.
>
>But as in all status history tables, we know that the status may change during the day (asleep, booting, 1st coffee, eating, siesta, nth coffee, drinking, crawling home), so there's no place for a date field - we need a timestamp or a datetime.

That's a really good point! But Mike is already aware of that. :)