In a recent Data Modeling Master Class, someone asked whether an indicator can be null. An indicator is an attribute containing only two values, such as True or False, Yes or No, On or Off, etc. If null is allowed as a value, that would be a third value.
My response was that on a logical data model, we need to show the rules, and therefore an indicator must be not null to show the rule that only two values should be allowed and not the third null value. On a physical data model however, where we frequently require tradeoffs such as loss of rules in exchange for performance or ease of development, an indicator might be null if the tradeoffs make sense.
What are your thoughts?