Re: Boolean as only necessary built-in type - was What is an "Extensible Database"

Troels Arvin wrote:
> I've often seen people claiming that Codd's only demand for built-in types
> in relational databases is the boolean type - and that any other type
> needed, could simply be added as user defined types. Does someone know
> where Codd wrote that, i.e. does someone have a reference?

I've not encountered a Codd statement to that effect - the index of his RM/V2 (1990) book lists 'Boolean Extension of Theta-Join (RB-24)", "Boolean Extension of Theta-Select (RB-13)", and "Boolean logic" only.

C J Date has said that the only mandatory built-in type is boolean (though a practical system will, of necessity, include other types, usually including a number and a string type) in a number of places, including the Third Manifesto. I rather suspect that his review of Codd's RM/V2 book also notes this [...time passes...] actually, it's the precursor chapter "Notes towards a reconstituted definition of the Relational Model Version 1 (RM/V1)" in Date and Darwen "Relation Database: Writings 1989-1991". I don't recall an earlier reference than that, but I may easily be misremembering and/or ignorant of the relevant location.

-- 
Jonathan Leffler                   #include <disclaimer.h>
Email: jleffler_at_earthlink.net, jleffler_at_us.ibm.com
Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/