Re: So what's null then if it's not nothing?

michael_at_preece.net wrote:
> Well Jon - I'd say they're both NULL. NULL is type-less.

So any attribute can store a "value" outside its intended domain? I thought you mentioned earlier that NULL is an empty set - is it actually of type E, where E consists of the empty set?

> In the case of
> "no value at all" for an integer variable we can't store an empty
> string so we would store NULL.

This makes no sense in other domains either - what about "NULL dates"? What sort of date manipulations (e.g. "same date next month") can one do on it?

> I know a lot of educated people will
> tell me that NULL, when compared to any other value, results in
> Unknown. I'd agree - if I agreed that NULL was unknown - or missing -
> or absent - or not, in fact, data. I believe that NULL is, correctly,
> "no value at all".

I keep reading this, over and over, and it still makes no sense. Can you define some semantics of a null when it occurs in a date attribute, a string attribute, an integer or real attribute, custom typed (e.g. "object") attributes? How do they "behave" in the presence of operations over the attribute's domain?

> We store something on a database because we know
> something. We can know something is empty.

Maybe if you define "empty." Does "empty" mean the same as "no value at all"? "Empty" isn't any more helpful than "NULL" to me.

> We can't know anything about something that doesn't exist.

So why is it in the DB?

> We have no facts about it. We have no data.

So we know the existence of something non-existent? I've just read a large chunk of this discussion, and have seen nothing but tail-chasing so far. Can you define, with any precision, the semantics of a "NULL value"?

• Eric