Re: On view updating

"Laconic2" <laconic2_at_comcast.net> wrote in message news:<pvCdnRrp-JR6k8bcRVn-rw_at_comcast.com>...

> Nah! Just back off from the kludge of calling a NULL a "special value".
> Call it what it is: an alternative to a value.
>

I don't think I've ever referred to NULL as a "special value", or indeed a value of any stripe. Mainly because it isn't. NULL means (amongst other possible interpretations) "no applicable value from the specified domain". Or just "unknown". Or maybe "not known, but RSN". You don't really need NULL anyway; with a decent type system, if you need "out of band" markers you can define your own, sensible markers. Without even requiring 3VL to handle non-values. Although unknown information is always going to be a pain in the proverbial, it doesn't make sense to make the problem more difficult by using one (or more !) NULL and thence wandering beyond boolean logic.  

> You can call the intersection of a row and a column a "field", as long as
> you can keep Joe Celko from having apoplexy.

But why refer to it as a field ?

> Then you can say that every field contains either a NULL or a value from
> its domain.

But why ? (I'm thinking of the Information Principle.)

> Domains (or types, if that's better) don't need NULL included.
>

NULL isn't really part of any domain. If you need "unknown value" functionality, you should be able define your own, sensible, "unknown values". (I know of someone who can define nearly 50 interpretations of NULL.)

> A NULL is like the curious incident of the dog in the night.
>
>
> I'm not keen on "thou shalt not refer to Oracle RDBMS, DB2, or SQL Server
> as relational"!
> I like clarity in terminology, but this is pedantry!

One man's "pedantry" is another man's "call a spade a spade".

• Tony