Re: Reminder, blatant ad

From: dawn <dawnwolthuis_at_gmail.com>
Date: 10 Feb 2006 17:17:58 -0800
Message-ID: <1139620678.347430.274910_at_g43g2000cwa.googlegroups.com>


Jan Hidders wrote:
> dawn wrote:
> > Jan Hidders wrote:
> >
> >>dawn wrote:
> >>
> >>>I'd like to better understand data independence in terms of
> >>>functionality.
> >>>
> >>>Does it mean that if a new version of a DBMS tool takes the same
> >>>logical specification it can lay the bytes down differently on the disk
> >>>without application code changing?
> >>
> >>No.
> >>
> >>>Or does the decoupling of physical and logical mean that if I want to
> >>>store attribute data physically "along with" a different entity (table)
> >>>than where it is now (and rework the logical model), no application
> >>>software needs to change?
> >>
> >>No.
> >>
> >>It means that you can make some changes in the physical model w/o having
> >>to change the logical model. Note that this is different from what you
> >>just said because you were talking about the logical model.
> >
> > I thought that was what I was saying in the first answer. How is this
> > different?
>
> The version of the DBMS tool doesn't change. You should be able to
> change the physical model w/o changing the tool.

Surely some software somewhere (the OS, the DBMS?) is getting a new version in order to make for this "change"?

> >>>I'm likely confused on this topic as I have not studied it in depth.
> >>
> >>You probably should. It is the singlemost important notion for
> >>understanding what the relational model is about.
> >
> > It is likely because I have read a bit that I'm confused. Why would
> > you think that a change in the physical model of a non-relational DBMS
> > (e.g. PICK) would require reworking the logical model?
>
> This is not necessarily true for all such systems but in many of those
> the logical and physical model are closely tied together and their
> efficiency in fact often depends on this close relationship.

What is the test for determining whether they are too tightly coupled? What is it that changes (the "physical model" doesn't resonate with me, I want to know what software components would change) that should not require a change in the logical model but does in some non-relational products? I do not recall ever changing the logical data model in any software I have written when a new version of anything was released. I do recall changing COBOL code with a new release of Primos in 1977, but it was not the logical data model for the indexed sequential files that had to change.

Still confused. Thanks for your help. --dawn Received on Sat Feb 11 2006 - 02:17:58 CET

Original text of this message