Hello Arofan,
XML is not a hype anymore. One of that reason is the complexity of
the business documents. The most of the business documents have too much
hierarchies. More than EDIFACT or ANSI ASC X.12 ever have. These many
hierarchies makes the business documents very heavy or not at all
processable. I know some projects, which had some big problems with the many
hierarchies.
-
It was not possible to put this documents or part of these documents into
the databases
-
It was a big effort to realize any application interface for processing
these documents
-
Additionally, the effort to define a XSLT rule was much more higher as for
defining any mapping rules for EDIFACT and Inhouse
formats.
I
guess that the instances itself do not need any containers, because it is
easier to define any interfaces for business objects.
My
suggestion is that we have partial object classes instead of unstructured
containers. This partial object classes makes the building of structure much
more controllable and the partial object classes are reausable,
too.
Otherwise, if we use "containers", than we need some simple rules for
the definition and processing in a automatic manner. One another import
thing is, that the names containers impacts the dictionary entry name. If
you use a container with the name "header", than the dictionary entry name
will be "order.header.id" instead of "order.id". This representation is not
ebXML CCTS compliant.
Kind regards,
Gunther
Gunther:
I have a fairly basic question for you.
Today, most business applications do not view XML business
documents as a direct serialization of the objects that they use internally.
Rather, the business document represents a somewhat truncated version of the
object hierarchy, designed for exchange between applications. While I see
the value of having a formal relationship between the model on which the
object hierarchy is based, and on which the business document definitions
are based, I have always thought that the serialization from one to the
other was a very useful level of indirection - different apps may actually
want to take the same object structures and treat them differently, given
that they generally perform different functions.
You approach removes this level of indirection (as well as
all the containers, making it difficult to use for non-object-based systems,
a different issue) - are you sure that this is the best approach?
What I am trying to get at here is the middle ground between
a strictly object-based approach, and one that will be suited for use by
other types of applications, while not losing the merits of what you have
outlined.
Cheers,
Arofan
-----Original Message-----
From:
Stuhec, Gunther [mailto:gunther.stuhec@sap.com]
Sent: Tuesday, August 20, 2002 1:57 PM
To: 'ubl-lcsc@lists.oasis-open.org';
'ubl-ndrsc@lists.oasis-open.org'
Subject: [ubl-lcsc]
Newest version of the OO-design position paper.
Hello all,
this is our newest
version of the OO-design position paper. It describes generally the
definition of aggregate core components as well as the aggregated business
information entities.
In this paper all aggregates representing an object. The
subelements which are basic business information entities representing the
attributes of an objects. An the aggregates insided of an objects are
further objects or parts of an objects. The object itself refers to these
furhter objects.
In addition all containers are avoided, consciously.
Because it is not possible to represents containers as objects with
attributes and influence the representation of a conclusive class diagram
unfavorably.
Please review this document and send me your comments and
remarks.
Kind regards,
Gunther