Dawn M. Wolthuis wrote:
> "Joe "Nuke Me Xemu" Foster" <joe_at_bftsi0.UUCP> wrote in message
> news:1072496216.517226_at_news-1.nethere.net...
>
>>"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message
>
> <news:bsira5$jlt$1_at_news.netins.net>...
>
<snip>
>>>Date writes "...MVS fields are ordered left to right (and so MVS files
>
> are
>
>>>certainly not relations, and the system is certainly not relational)."
>>>
<big snip>
Allow me to play the role of the fool jumping in where angels
fear to tread.....
Two points which may clarify RDBMS implementation (as opposed to
theory).
- The relationships are imposed externally to the data in the
form of indexes and/or foreign keys. The data itself is
unordered. If I remember correctly Codd specifically stated
that the data was not ordered. From an implementation point of
view (circa 1970 when the largest mainframes weren't much faster
than my Palm IIIc) this allowed signifiantly better performance
when adding rows that should logically be anyplace other than
the end of the table.
- Within a table row the physical order of the columns as
stored on disk need not conform to the logical order of the
columns as specified in the CREATE TABLE statement. Again
looking at the mainframe computers of 1970 when all but the very
largest had less than 16MB of RAM, the highest capacity disk
drives only had 33MB of storage and there were arcane rules
about floats and integers starting on word boundaries while
short integers and strings could start on any byte in a word it
became desirable to store all of the floats and integers at the
beginning of the physical disk record and the shorts and strings
after them in order not to waste any space in either memory or
on disk.
HTH
Received on Fri Dec 26 2003 - 22:58:50 CST
