Re: Stored fields ordered left to right
Date: Fri, 09 Jan 2004 10:57:49 -0500
Message-ID: <3ffecf81$0$6735$61fed72c_at_news.rcn.com>
Dawn M. Wolthuis wrote:
This practice came about because it was easer to remember
meaningful names, e.g. Salary, that it was to remember memory
locations, e.g. 0357321. Codd did nothing more than follow
common practice by assuming the use of names as aliases for
positions and the fact that he did so should not be taken to
mean that he removed the ordering requirement.
> Sorry for all the typos -- saw them while clicking to post, argh. Read this
> instead.
>
> Yes, the info is in Codd's 1970 paper, but for those who do not prefer to
> read more mathematical jargon than necessary for this particular point, it
> is this:
>
> A mathematical relation is a set of ordered tuples {a1, a2, ..., an} with
> certain characteristics.
>
> In Codd's definition of a relation, however, he removes the ordering
> requirement (by using names instead of positions) so that a "Codd relation"
> is a set of unordered tuples. Therefore, relational database theorists
> often think that relations MUST be unordered, when in fact it is they who
> opted not to use the correct mathematical definition of the term.
>
> My point in stating this is that I'm writing up responses to a set of
> questions that Chris Date put out regarding the MultiValue model and he
> claims that MultiValue must not be relational (and I'll admit it isn't by
> his definition) because it has ordered tuples. That argument always just
> sounds out-n-out wrong to me since mathematical relations are ordered
> tuples, dag nab it! So, I just had to clear this little matter up.
>
> MultiValue uses mathematical relations which are, in fact, also functions
> (they map a unique key to an ordered tuple) but it does not use Codd's
> definition of a relation (although a developer can use names for locations
> in the tuple, thereby using the database as if it were unordered tuples)..
>
> I wasn't trying to say "nah nah na boo boo, relational databases are not
> mathematical relations like MultiValue databases are" (I'll leave that for
> others), but was simply trying to get a sound, logical, and fair response
> for Mr. Date. Cheers! --dawn
>
>
From the perspective of an assembly language software developer
(a rapidly disappearing breed) Codd's use of names rather than
positions is immaterial. This is because in assembly language
development it is common practice to assign names as aliases to
locations rather than use the location itself as an identifier.