Steering Model-Driven Development of Enterprise
Information System through Responsibilities
Ming-Jen Huang and Takuya Katayama
School of Information Science, Japan Advanced Institute of Science and Technology
1-1 Asahidai, Nomi-shi, Ishikawa, Japan
Abstract. OMG proposes the MDA that promotes the ideas of modeling in
UML and transforming UML models to code. But UML is not universal for
every domain and the direct translation approach of the MDA is not adequate.
In this paper, we introduce REST, an idea of using responsibilities as contextual
information to instruct machines to generate software systems. First, we give an
overview of RESTDA - a software development architecture for business based
on the concept of REST. Then we describe a domain-specific language -
Business Models. It helps developers to describe a business from a document-
processing perspective. We also introduce a rule-based validation of
consistency within Business Models. Finally, we describe the transformation
mechanism of RESTDA. Our approach provides machines higher intelligence
to generate source code for different contexts.
1 Introduction
Model Driven Architecture (MDA) proposed by OMG [1] is a software development
approach that promotes defining platform-independent models in UML and having
machines to transform them into technology-specific code [2]. Its concept is based on
two assumptions. First, UML is precise and expressive enough to describe problems
we are interested in. It is also universal enough to describe any domain of problems.
Second, all problems defined by every kind of UML modeling constructs should
imply identical contextual information. For the MDA, UML becomes the master key
to open every door to any solution.
With regard to the first assumption, different domains have different requirements.
Thus, a domain-specific language (DSL) that is customized for a specific domain is
more realistic and more productive [3]. With regard to the second assumption,
considering the following example: does the case of implementing UML models of a
car having four wheels equal to the case of a teacher having four students? By UML,
they may be drawn identically in class diagrams or even sequential diagrams. For an
effective model transformation mechanism, we do not only have to give machines
syntactic and semantic information, but also the capability of reacting according to
different contexts. To that end, we propose a conceptual idea, Responsibility-Steering
Model Transformation (REST) to augment existing model-driven approaches.
Huang M. and Katayama T. (2005).
Steering Model-Driven Development of Enterprise Information System through Responsibilities.
In Proceedings of the Joint Workshop on Web Services and Model-Driven Enterprise Information Systems, pages 165-170
DOI: 10.5220/0002566201650170
REST is a conceptual idea of model transformation that is inspired by
Responsibility-Driven Design, which is proposed by Wirfs-Brock [4]. She promoted
the idea of designing a software system from responsibilities and devising role objects
to collaboratively work together to assume these responsibilities [5]. In REST, we
consider responsibilities of an abstraction level are realized by responsibilities of a
level beneath. And realization of all responsibilities of all levels, combining with
domain-specific languages, instructs machines to generate detailed implementation of
different technology-specific code. The advantages of our approach are: (1)
Responsibilities provide extra contextual information of domains under consideration.
The problems like the example of car and teacher can be avoided. (2) By defining
model transformation in terms of responsibilities of different levels, any change of
requirements can lead to easy and reliable modification to the target system. (3) By
formalizing responsibilities, the correct transformation can be ensured.
The purpose of our work is to devise a development architecture and to apply the
idea of REST to the development architecture to solve the problems of the MDA
mentioned above. In this paper, we introduce the development architecture for
business called Responsibility-Steering Development Architecture (RESTDA).
The remainder of this paper is organized as follows. Section 2 gives the overview
of RESTDA, the details of the DSL - Business Models, and the description of the
rule-based consistent validation of BM. Section 3 describes the details of REST and
its implementation in RESTDA. Section 4 gives the conclusions and future works.
2 Responsibility-Steering Development Architecture
Fig. 1. Architecture of RESTDA
Fig. 2. Structure of Business Models
RESTDA is composed of a DSL to model concepts of business world and a model
transformation mechanism between models and code. Fig. 1 shows the architecture of
RESTDA. The DSL, Business Models, describes different business scenarios from the
structural, behavioral, and constraint aspect. Definition of BM of a target system is
transformed into a technology-neutral object model - Collaborative Responsibility
Model (CRM) by machines with a business scenario as a unit. CRM does not contain
details of technology-specific implementation but generalized software objects and
responsibilities of these objects. By means of CRM, a system can be divided into
many vertical-sliced parts, and each part can be transformed into different
technology-specific code that is most suitable to a situation. Instead of direct
translation of meta-model to code, RESTDA applies the idea of REST, using
syntactic, semantic, and contextual information, to instruct machines to transform BM
to CRM and CRM to source code.
2.1 Business Models Description
BM defines the running of a business from three different views. Business entities
describe the structural view. Business activities describe the behavioral view.
Business rules describe constraints of business entities and business activities.
Definition of BM of a target system has one or more scenarios which describe a
possible situation of document processing (see Fig. 2).
Fig. 3. A Sample of Business Entity
Fig. 4. A Sample of Business Activity
Business entities are roles that participate in a scenario. They are described in
business entity diagram. A sample is shown in Fig. 3. Here, we borrow the drawing
conventions from UML. There are two types of business entity, actor and document.
Actor type entity represents human role in a scenario. In the diagram, it is displayed in
stereotype <<actor>>. Document type entity is what is usually printed out as a
formal or legal document in a business. In the diagram, it is displayed in stereotype
<<document>>. Between business entities, they may have relations.
A business activity is a sequence of operations on which business documents are
processed. A business activity has three parts, request, operations, and response.
Request describes how the request is sent (Channel), who makes the request (Actor),
and what information is carried by the request (Params). A single operation is an
action operating on a document. It describes what type the operation is (Type), what
document to operate on (Target), and what information to provide after completion of
an operation (Result). There are four types of operation: CREATE, RETRIEVE,
UPDATE, and DELETE. Operations can be linked sequentially to represent
sequential operations. A business activity is described in a business activity diagram.
As the exemplar Fig. 4 shown, the request is sent via HTTP and made by
WebCustomer. WebCustomer should provide information of Book in the request.
The business activity has a single operation to RETRIEVE information of Book and
return resulting Book. The response is sent via HTTP to the WebCustomer.
WebCustomer and Book are referred to the business entities of the scenario.
Business rules are constraints of business entities and business activities. For a
business entity, business rules define the possible range of values of its properties. For
a business activity, they define conditions of allowable activity requests or conditions
of allowable operations, among other things.
2.2 Formalization and Implementation of Verification
The semantics of BM is formalized as predicates and implemented in a rule-based
engine to verify validity of BM. These predicates are called verification rules. They
are defined in terms of three basic constructs,
be of business entity,
of entity
attribute, and
ba of business activity. The types of business entity and activity
, and
. Each construct has
an identifier
, and
. Relations (own and detailedBy) between
business entities are
wn(be ,be
etailedBy(be ,be
. Channel, actor, params
of request are
, and
. Type,
target, and result of operation are
, and
respectively (
denotes the sequence of operations). For
denotes the type of the first operation. Channel, actor,
params of response are
, and
There are five types of verification rules within BM. In this paper, we explain only
the first type of verification rules - structural relation. In BM, business entities have
two types of relation, own and detailedBy. For example, an actor type Manager
owns a document type MonthlySalesReport and MonthlySalesReport is
detailed by a document type WeeklySalesReport. Types of entity at two ends of
a relation should be correct and they are represented as two rules:
1. Only an actor type entity can own a document type entity
be ,be Own(be ,be ) ActorType(be ) DocumentType
212 1 2
2. Only a document type entity can be detailed by a document type entity:
be ,be DetailedBy(be ,be ) DocumentType(be ) Do
212 1 2
The verification rules are implemented in a rule-based engine, Jess [6]. Jess
contains facts and rules. The collection of facts is information Jess knows. The
collection of rules in Jess is a kind of actions that triggers under certain conditions [7].
Rules in Jess can be stated as “if
denotes a set of conditional facts.
denotes a set of actions.
is tested against all known facts. For example, if we
know (1) a verification rule states that only an actor type entity can own a document
type entity and (2) a fact states that SalesStaff (actor type) owns
PurchasingStaff (actor type). If a Jess rule states “if (1) is not satisfied against
all known facts, then displays a warning message.” Since PurchasingStaff of
(2) is not a document type, Jess would display a warning message.
3 Responsibility-Steering Model Transformation
REST is a conceptual idea of model transformation that uses responsibilities of
different levels as contextual information to instruct machines to transform platform-
independent models into technology-specific code. Real-world responsibilities of
structural and behavioral constructs and constraints of a DSL are
realized by
generalized object responsibilities and the generalized object responsibilities are
realized by responsibilities of technology-specific code, such as classes or
components. In RESTDA, the realization of generalized object responsibilities is pre-
defined. Developers only have to define (1) the responsibilities of BM and (2) how
generalized object responsibilities realize these responsibilities for each scenario.
First, developers have to define real-world responsibilities from BM. A
responsibility of any level always has a
holder and a receiver. A holder represents a
structural role which assumes the responsibility. A receiver represents a structural role
that is affected by the consequence of the responsibility. Responsibilities of the same
level are connected by holders and receivers. We use
Collaborative Responsibility
(CRD) to draw responsibilities, holders, and receiver as shown in Fig. 5. A
collaborative responsibility diagram shows the structural and behavioral aspect of
responsibility realization. To read the diagram, a rounded rectangle represents a
responsibility and a rectangle represents a role. The left-hand role of a responsibility
represents a holder and the right-hand role represents a receiver. A receiver of a
responsibility could be a holder of another responsibility. The responsibilities are
fulfilled from left to right one by one.
BW-1: Request for
searching book service
BW-2: Provide
request information
BW-3: Provide its
own information
BW-4: Process request
BW-5: Execute
operations in turn
BW-6: Provide its
own information
TargetSystem WebCustomer
Interfacer Interfacer Controller Controller
GO-1: Provide
GO-2: Decode
request message
GO-3: Validate
GO-4: Decide
next executing
GO-5: execute
next operation
GO-6: Search
book by
GO-7: Provide
GO-8: Render
resultant views
GO-9: Manage
life-cycle of
service provider
GO-10: Manage
life-cycle of
service provider
BW-7: Return executing
results to response target
Fig. 5. A Sample of Collaborative Responsibility Diagram
Second, developers have to define how generalized object responsibilities realize
the real-world responsibilities. It is a process of refinement by decomposing a real-
world responsibility into smaller chunks. For example, the real-world responsibility
“Process request” is realized by two generalized object responsibilities: “Decode
protocol-specific message” and “Validate WebCustomer’s security”. Again, a holder
and a receiver are assigned to a generalized object responsibility. They come from
generalized objects. We borrow the concepts of role stereotypes from Responsibility-
Driven Design. It defines six types of role: information holder, structurer, service
provider, coordinator, controller, and interfacer [5]. A generalized software object
represents a stereotype that assumes a set of generalized responsibilities. Developers
have to contemplate types of responsibility and types of generalized object
simultaneously for each scenario. Responsibilities of generalized objects and their
holders and receivers form CRM that are further transformed into Java code by Jess.
RESTDA predefines how a generalized object of CRM is transformed into one or
more Java classes. The generation rules are also implemented in Jess in a code-
template-generation fashion where the data for placeholders of code templates come
from definition of CRM. These rules also define how different source code to
generate for different responsibility definitions.
4 Conclusion and Future Work
In this paper, we introduced the software development architecture for business –
RESTDA which is based on the idea of REST. The significance of the research is that
domain experts can use BM to describe the running of a business without concerning
any technology details. Instead of direct translation approach, the combination of
syntactic, semantic, and contextual information of each level offers machines higher
intelligence to generate software systems from platform-independent models.
With regard to future work, one is to formalize the concept of responsibilities
Another is to use much expressive higher-order logic to quantify over predicates and
to apply automatic theorem provers, such as HOL, to verify consistency of BM and
responsibility realization [8,9].
This research is conducted as a program for the “21
Century COE Program” by
Ministry of Education, Culture, Sports, Science and Technology.
1. MDA Guide Version 1.0.1. OMG. (2003)
2. Frankle, D.S.: Model Driven Architecture : Applying MDA to Enterprise Computing.
Wiley, New York (2003)
3. Thomas, D.: MDA: Revenge of the Modelers or UML Utopia? IEEE Software, Vol. 21, No.
3, pp. 15 – 17 (2004)
4. Wirfs-Brock, R.: Object-Oriented Design: a Responsibility-Driven Approach. OOPSLA ’89
Conference Proceedings, pp. 71 – 75 (1989)
5. Wirfs-Brock, R., McKean, A.: Object Design: Roles, Responsibilities, and Collaborations.
Addison-Wesley, Boston (2003)
6. Jess v7.0a4.
7. Friedman-Hill, E.: Jess in Action. Manning: rule-based systems in Java. Manning,
Greenwich, CT (2003)
8. Aoki, T., Katayama, T.: Unification and Consistency Verification of Object-Oriented
Analysis Models. Asia-Pacific Software Engineering Conference, (1998)
9. Yatake, K., Aoki, T., Katayama, T.: Collaboration-Based Cerification of Object-Oriented
Models in HOL. Verification and Validation of Enterprise Information Systems (2004)