[
Lists Home |
Date Index |
Thread Index
]
Michael said:
"I think Date (if I recall his books correctly) would probably agree
with the conventional wisdom that there are three stages of database
design: information modelling, logical design, and physical design.
Information modelling is certainly independent of whether you use
relations, objects, or XML. Mapping the conceptual model to XML schemas
is logical design, and is very XML-specific. Designing indexes,
allocating storage, and so on is physical design, and is very
product-specific.
You seem to be using the word "design" to mean only the first of these
three stages."
No, I don't think so. I think I meant the second, logical design.
Certainly, that's what Codd was referring to in the quote. The
relational model of data explicitly has nothing to say about physical
design. It explicitly makes that the responsibility of the implementer
of the DBMS and not the designer of the database. (Clearly, current
DBMSs only partly support this aim). So I didn't mean physical design.
I don't think I meant information modeling either. On page 435 of his
book Date discusses E/R modeling. He agrees with Codd that it isn't
possible to design a database without both structure and operations (and
constraints). If information modeling is just modeling structure, then
it is not a sufficient first level of design. I agree with this, at the
business level you must model both structure (entities and
relationships, say) and process and then you must relate them together
(the 'CRUD' matrix). Clearly I wasn't trying to make a point about
creating this matrix.
I don't think Date does share your conventional wisdom either. On page
14 of my copy of Date's An Introduction to Database Systems (7th
edition), he says:
"In a nutshell: The model is what users have to know about; the
implementation is what users do not have to know about".
Users include programmers here, of course. On page 33, Date uses the
ANSI/SPARC 'architecture for a database system', having three levels:
Internal (aka Physical); External (aka User Logical); and Conceptual
(aka Community Logical or Logical). Note, confusingly, the external
level is between the internal and the external. The external is an
individual user's view of the conceptual. By way of illustration, he
shows the COBOL layout as an external view, the DDL as the conceptual
and a storage description as the internal. The conceptual view has all
the integrity and security constraints.
Where does that leave XML? Well, I don't think designing an XML document
by specifying a schema is designing a database. Clearly, it is related -
the domains used in the message should correspond to domains in the
applications it applies to. But there is no modeling of operations on
the schema. There are constraints, but they are constraints on unrelated
instances of the schema (a database constraint is for transforming one
instance into another).
Codd in his Turing address said:
"It is therefore no surprise that attempts such as those of CODASYL and
ANSI to develop data structure definition language (DDL) and data
manipulation language (DML) in separate committees have yielded many
misunderstandings and incompatibilities."
It looks like XML is making the same mistake (XML Schema and XQUERY).
Yours,
John F Schlesinger
SysCore Solutions
-----Original Message-----
From: Michael Kay [mailto:michael.h.kay@ntlworld.com]
Sent: Wednesday, May 01, 2002 4:03 AM
To: johns@syscore.com; 'Ronald Bourret'; xml-dev@lists.xml.org
Subject: RE: [xml-dev] Normalizing XML [was: XML information modeling
best practices]
> Codd and Date would say (I know, I'm putting words in their
> mouths, but
> maybe Date will see this and tell me if I'm wrong) that it makes no
> difference to the design of the database whether the data, when
> manipulated, is represented as XML or anything else.
>
I think Date (if I recall his books correctly) would probably agree with
the
conventional wisdom that there are three stages of database design:
information modelling, logical design, and physical design. Information
modelling is certainly independent of whether you use relations,
objects, or
XML. Mapping the conceptual model to XML schemas is logical design, and
is
very XML-specific. Designing indexes, allocating storage, and so on is
physical design, and is very product-specific.
You seem to be using the word "design" to mean only the first of these
three
stages.
Michael Kay
Software AG
home: Michael.H.Kay@ntlworld.com
work: Michael.Kay@softwareag.com
|
