Re: foundations of relational theory? - some references for the truly starving
Date: Tue, 21 Oct 2003 18:43:29 GMT
Message-ID: <l7flb.42132$Z86.5517_at_twister.socal.rr.com>
"Bob Badour" <bbadour_at_golden.net> wrote in message
news:ie6dna18NqYv2giiU-KYhQ_at_golden.net...
>
> <snip>
>
> The fatal error in Mr. Murthi's logic is the assumption that the theorists
> must have ignored Pick if they have nothing positive to say about it.
>
>
<more snip>
>
> The fatal flaw in Mr. Murthi's logic is the assumption that advocates of
the
> relational model somehow fail to understand Pick. I understand it more
than
> well enough to reject it, and I observe that Pick advocates have very
little
> understanding of AQL as evidenced by the whole red]blue car exchange.
>
>
Bob,
In all honesty, I realize you need no prompting, but please feel free to
correct me where I'm wrong. If this is to be an intelligent discussion and
not just a bunch of children throwing tantrums and invectives, please stick
to the point and show me where I'm wrong.
According to Marc Light of University of Iowa, School of Library and
Information Science, Linguistics Department (I paraphrase to avoid copyright
infringement):
1st normal form: No multivlaues; all rows contain the same number of
columns.
2nd normal form: 1NF & no data element depends on data in composite keys
No problem. We've been translating (what Pick calls a join) to create
relational data for years and if we're going to flatten out our tables
anyway, and since the number of translates is insignificant in normal
numbers, we can store atomic data and fetch it on demand.
The example on the web site was this:
Part
Since Address depends on warehouse and therefore the composite key
part/warehouse, this is NOT 2NF, so they move the address to a separate
table and make the 2NF table:
Part
We would do that naturally, because it's stupid to store the address or any
such static data depending on warehouse in every part record.
3rd normal form: (1NF) 2NF & No non-key field is a fact about another
non-key field.
Their example was
Emp
Since location depends on department they move it to another stand alone
table. And, again, we would probably have done that anyway, because, again,
it's stupid to repeat static information that's longer than a key to point
to it.
So I guess my point is that Pick's strength and value isn't in that it's
RDBMS compliant, though it seems like it could be, but that it's flexible
enough to take advantage of "natural relationships." In a 2NF table I could
never have a part item with multivalued warehouse and quantity. But there's
a natural relationship implied in the structure that says this part resides
at these warehouses and has these quantity respectively.
http://que.info-science.uiowa.edu/~light/courses/autoSys/lectures/kentfirst.pdf
No problem. we just flatten out our file structure. Every data element in
Pick is a variable length variant so we don't have to worry about data
typing and columns may contain different data types (numbers, dates,
strings); but we never saw that as a problem. So every "row" can have the
same number of columns, no matter if they exist or not.
Warehouse
Quantity
Address
Warehouse
Quantity
Again, flat tables and translates and we fit right in.
Department
Location
Maybe I'm wrong here and it's no about Normal Forms. Maybe I got the wrong source off the web...
Maybe it was the drugs from the 60's. I could use a good flashback right about now.
Please show me where I went astray, but please show me the courtesy of adult dialog.
Mark Brown Received on Tue Oct 21 2003 - 20:43:29 CEST