Jeff Moden (11/23/2007)
DATETIME is really a FLOAT in the background (although it is a "fixed point float", if you can imagine that). DATETIME uses the same BINARY math as FLOAT.
You are right that datetime is some sort of a "mule". It takes 8 bytes, the first four is the number of 1/300 seconds since midnight, and the other 4 bytes are the days since 1/1/1900 (this can be negative, so it can express dates before 1900). But these could be calculated more precisely (were there a business case for it :)) (and then we have the nice new date and time, and datetime2 🙂 in 2008 :))
Regards,
Andras