What prevents this same line of reasoning from being used to justify that SQL does not need a DATETIME primitive type? Just as you showed how a sum type could

What prevents this same line of reasoning from being used to justify that SQL does not need a DATETIME primitive type? Just as you showed how a sum type could be represented with component relations, a similar example could be constructed showing how any needed attribute of a datetime could be normalized to various tables combining the features of dates and times.