Re: Implementation of boolean types.

-CELKO- wrote:
> >> Where do you think the definition of natual join comes from? <<
>
> Getting serious for a minute, Tom Johnston had a paper and a
> conversation with Dr. Codd about problems with the Natural Join because
> we cannot be sure which of the two or more columns are or are not in
> the resulting relation.

If I understard what you're saying correctly (not sure) then I have to say that I buy into Date's viewpoint: relations are predicates, and attributes are the variables of the predicate.

> >> A language not having a clean treatment of the boolean type in version 1 is a sign of bad design. <<
>
> Procedural languages, not declarative ones. Dijkstra "only" worked
> with non-deterministic procedural languages. I think that we grew up
> with the Dijkstra/ Manna/ Gries model of programming that we had a
> bitch of a time with RM. I think I can say that with authority since my
> first Master's Thesis was on an algebra of Structured Programming :)

I must respectfully disagree. Haskell, for example, is as declarative as they come, and has the boolean type. (Not that I didn't have a "bitch of a time with RM" at first. But this is something that I think a clear algebra would help with a lot.)

Marshall Received on Fri Jul 15 2005 - 03:51:31 CEST