![](bg2.png)
2.2 Khodorkovsky’s ONF, 4ONF,
5ONF and 6ONF
The paper (Khodorkovsky, 2003) sets up the idea
of object normal form (ONF), which concerns the
“right” relation among data of object and methods
of object. Khodorkovsky’s rule is added to classi-
cal “relational” definitions 4NF, 5NF (and to the 6NF,
which is author’s enhancement of the 5NF). The au-
thor calls these additions to classical definitions as
4ONF, 5ONF and 6ONF.
The paper is considered to more qualified formula-
tion of similar ideas as the example above. It is valid
according to the author that 1NF, 2NF and 3NF are
common for relational and object databases.
2.3 “Chinese” ONF
The paper (Yonghui, Zhou, 2001) set up one object
normal form as substitute for all relational normal
forms. It considers the object data model to be simi-
lar to the tree-like data hierarchy in the way of XML.
We suppose that this approach does not concern with
object databases design, as we understand it.
2.4 “Australian-Swiss” ONF
The authors (Tari, Stokes, Spaccapietra, 1997) set up
one ONF by the help of more types of functional
dependences among objects. Concrete ”path depen-
dency” concerns composition of objects and naviga-
bility among objects, ”local dependency” concerns re-
lations of internal object components and ”global de-
pendency” concerns requirements on application. Ob-
jects structure is in ONF then, if users’ requirements
on applications are retrospectively deducible from the
relations among objects. In agreement with our opin-
ion it concerns very interesting contribution to prob-
lem of testing the accordance of suggested object ori-
ented application model with its requirements. We
think, that this problem is related to the other issues
of databases development, but it not solves the prob-
lem we are concetrated on.
2.5 Three Ambler–Beck’s Object
Normal Forms
Three object Ambler–Beck’s normal forms for object
oriented applications are set up in internet (Ambler,
2004) and in the book (Beck, 2003). These normal
forms are analogical with first, second and third rela-
tional normal form. The authors define these normal
forms 1-ONF, 2-ONF and 3-ONF themselves and talk
about them as a tool for objects classes’ normalization
complementary with technique of design patterns.
3 OUR INTERPRETATION OF
THE PROBLEM
All works mentioned here are very important contri-
bution into discussed area. When we resume, what
probably the community of analysts and designers ex-
pect from the technique of object normalization, we
can define the following conclusion:
1. It has to be very simple, clear, and understandable
and it should work with minimum of concepts sim-
ilarly as it is with ”classical” normalization. We
suppose that implementation of difficult definitions
distinctively exceeding the range of classical nor-
mal forms, a lot of types of concepts and relations,
is not the right way.
2. It should be focused only on database design, e.g.
on the structure of objects, which will serve as data
storage and manipulation in database systems. It
does not need to work with objects, which are re-
sponsible for ”operation” of applications. There are
design patterns for them and there is no need to sub-
stitute this technique.
3. It is possible, that in the future the object approach
will become universal approach to information sys-
tem analysis, and relational technology will limit it-
self to be only one of possible implementation vari-
ant. So present conditions can be turned to the op-
posite. It would be smart to have the new theory
analogical with entity-relation modelling concept
and relational normalization. In the best case, the
relational normalization (as a tool of the relational
technology) should be deducible from new theory
as its special kind.
We have to define first, what we understand by
database object. Because it “only” serves for data sav-
ing and manipulation. It is not object, which ensures
some behavioural aspects of the software application.
This is why we propose to not work with data and
with methods separately and define one concept of
“attribute”. We will not distinct, if the particular at-
tribute is implemented into object by its data or if it is
the result value of some method.
There is a question, if this simplification is not
large. For example Ambler-Beck’s approach works
directly with data and methods and uses them in its
definitions. But we think that we can afford this sim-
plification for the data objects.
We think that the modified form of Ambler-Beck’s
approach fulfils best all above described requests.
This version was already used in our lessons and prac-
tically in Czech companies, and we shall work on it
further. We suppose that this method should precede
during modelling all possible next thoughts about us-
ing of inheritance, composition and other relations
among objects.
ICEIS 2006 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
472