On Mar 11, 2006, at 7:58, Andrew Piskorski wrote:
On Fri, Mar 10, 2006 at 04:37:36PM -0800, Darren Duncan wrote:
3. There is no such thing as a NULL.
3.1 All logic is 2VL (true, false) not 3VL (true, false, unknown).
There is no such thing as null, really? So, when you do an outer join
between two tables, which in SQL would produce null columns in the
result set, what do YOU propose producing instead of those nulls?
Perhaps I missed it, but in my brief reading of some of Date's work, I
never saw him answer that question.
I never understood that restriction. I read in the books: "since we
have defined things this ways from a formal point of view there's no
room for NULL". And my question is well, why don't you change the
definitions to augment the datatype sets with a special constant NULL
which is by definition not present in any datatype? Wouldn't that
give an analogous theory more aligned with real world?
The formalism in the relational model looks so-so to me (with due
respect), starting from the fact that "tuples" are _sets_ instead of
elements of a Cartesian Product, which by the way is what relations
are defined from in Set Theory, they are subsets of Cartesian
Products, not some kind of ad-hoc object. Sounds souspicious the
claimed math foundation with so fundamental deviations from basic,
standard math conventions. More than math foundation it is in my
view, I don't know, just some stuff presented formally and with some
degree of rigour.
-- fxn
