Joseph,

I knew it was a joke, and a pertinent one too. *s*

I'd just test for difference < 0 and add 86400 to it. seconds() can't handle multiday, so datetime() would be needed for that, of course the need for subsecond accuracy is way out of line in this case. *g*

>Of course. I was mostly joking. If someone were concerned with subsecond timing, I don't think they'd be doing something that took very long (ie. something that would have to run for a while and that might span midnight). So if they saw that it took -86398 seconds, they would know it was too late and that they should go to bed and try again later.
>
>>You can store the datetime() and the millisecond portion of seconds(), this eliminates any rollover issues.
>>
>>>Just don't span midnight.... ;~)