XOCL - an XML Language for Specifying Logical Constraints
in Object Oriented Models
Franklin Ramalho
(Universidade Federal de Campina Grande, Brazil
Universidade Federal de Pernambuco, Brazil
franklin@dsc.ufcg.edu.br,
fsr@cin.ufpe.br)
Jacques Robin
(Universidade Federal de Pernambuco, Brazil
jr@cin.ufpe.br)
Roberto Barros
(Universidade Federal de Pernambuco, Brazil
roberto@cin.ufpe.br)
Abstract: In this paper, we present XOCL, an XML-based language
to represent OCL (Object Constraint Language) constraints in UML models.
XOCL was designed in two steps from the UML meta-model and OCL EBNF grammar
published by OMG: (1) construction of a simple OCL meta-model and (2) derivation
of an XML Schema for this meta-model. XOCL applications include full interoperability
among UML modelling tools as well as finely grained structured input for
automatic behavioral code generation and model checking.
Keywords: OCL, UML, XML, Meta-Modelling, XMI, XMLShema.
Categories: D.1.5, D.1.6, D.2.12, D.3.3
1 Introduction
UML (Unified Modeling Language) [Booch et al. 1998]
has emerged as the main standard language for pre-code software engineering
artifacts for several reasons:
- It is able to model all aspects of a system: structural, behavioral,
distributed, etc.
- It provides a simple and concise graphical notation;
- It provides mechanisms that allow its extension for several problem
domains;
- It is supported by a wide variety of CASE tools.
With its use progressively outgrowing its original high-level object-oriented
modeling purpose, new requirements emerged for UML. One of them is the
ability to specify integrity constraints and application domain rules within
models in a detailed, structured, unambiguous and semi-formal way. To address
this requirement, OMG incorporated from version 1.4 of the UML standard
the semi-formal textual model annotation language OCL (Object Constraint
Language) [Warmer and Kleppe 1999].
Specifying constraints and rules as OCL expressions instead of natural
language notes brings numerous benefits:
- It avoids ambiguity and therefore misunderstandings between developers;
- It allows for much more thorough automatic correctness and consistency
model checking [Brucker and Wolff 2001];
- It greatly facilitates automated behavioral code-generation from models
[Mellor and Balcer 2002];
- It widens the applicability of UML beyond software engineering to new
fields in which detailed, rigorous constraint or rule specification is
crucial. These fields include data and metadata integration in data warehouses
[Poole et al. 2001], disparate federated information
systems [Purvis et al. 2000], semantic web ontologies
[Cranefield and Purvis 1999], knowledge engineering [Schreiber
1999] [Devedzic 2001] and multi-agents systems
engineering [Bergenti and Poggi 2000].
Several UML modeling tools such as Poseidon [Poseidon 2003] and Rose
[Rose 2003] already support OCL constraints. However,
each of these tools relies on an internal UML model storage format that:
(1) is proprietary, and (2) mixes model content with its graphical layout.
This virtually prevents reuse and integration of models developed with
different tools. However, such reuse is a key aspect of the most modern
trends in software engineering, information systems, databases and artificial
intelligence.
To address this problem, OMG put forward XMI [OMG 2003],
an XML based standard for representing UML models in textual format. XMI
defines one XML element or attribute for each element of the UML graphical
notation. Therefore, XMI encodes UML model content and structure separately
from their visual presentation, much like XML does with web pages. Being
based on XML also makes the XMI codification of an UML model both legible
by humans and an appropriate input to model exchange, model checking and
code generation software.
Although XMI has elements and attributes for all UML diagrams, its current
version does not cover OCL constraints. This gap has prevented UML, OCL
and XMI to reach their full potential as an integrated standard set for
a variety of emergent applications. These applications include reuse of
pre-code development artifacts in modern software engineering processes
based on component markets, product lines, design patterns, frameworks
and software architectures [Atkinson et al. 2002],
automated model checking and fully automated code generation.
In this paper we propose XOCL, an XML language for detailed, fully structured
encoding of OCL constraints and rules that completes the OMG and W3C standard
puzzle to support such advanced applications. An XMI document with embedded
XOCL expressions captures the complete structure and semantics of a UML
model detailed with OCL constraints in a format that is adequate for both
interchange and sophisticated processing.
Consider for example an automated behavioral code generator. An OCL
constraint might serve as the basis for generating exception-raising code
to call upon violation of the constraint. With current UML editors, the
OCL constraint will appear in the XMI document exported by the editor and
input to the code generator as a long, unstructured string inside a single
<constraint>tag. Before tackling its generation task, the
code generator will first need to parse this string into semantically relevant
syntactic constituents and hence incorporate an OCL syntax grammar. But
this defeats the very purpose of XML mediation among components, namely
to avoid encapsulating knowledge about upstream component languages into
downstream ones to achieve a more modular and extensible task distribution.
2 Technological Background
2.1 OCL
OCL permits the specification of three kinds of constraints: invariants,
which are static, and pre and post-conditions, which are dynamic. An invariant
is associated to a class, an interface or a type. It specifies a condition
that must be true for all instances of the associated class, interface
or type, at any time. In contrast, a pre or post-condition is associated
to an UML operation and specifies a condition that needs to be verified
only immediately before or immediately after its execution (respectively).
Figure 1 shows an excerpt from the class diagram
of a simple company UML model. The diagram includes two OCL constraints.
The first one, managerConstraint is an invariant that requires all
instances of the Company class that constitutes its context to have
its manager association filled with an instance of the Person
class with the age integer attribute value within the [18,65] interval
and the isUnemployed boolean attribute set to false. The
second one, resultOkConstraint is a post-condition that requires
the method income of the Person class that constitutes its
context to always return a value greater than 5000.

Figure 1: Example of an UML model with OCL constraints
These two example constraints illustrate the design rationale underlying
OCL: to achieve the most practical balance between formal precision and
intuitiveness for the widest possible user base.
2.2 XML and XMLS
XML (eXtensible Markup Language) [W3C 2003] was
initially created for publishing documents on the Web. Its main goals were:
(1) divide semantic structure of document content (in XML) from its visual
rendering by the browser (in HTML); and (2) codify the semantic structure
in a format adequate for being used as both documentation (human legible)
and data (input prepared for automated processing). These two features
combined with the fact that XML is an open and general purpose standard,
have progressively turned XML the choice for interoperability among databases,
programming languages and intelligent agents in heterogeneous distributed
applications. In these multiple roles, XML has blurred the distinctions
between documents, concepts, data, information and knowledge and between
areas such as information systems, distributed system, databases and artificial
intelligence.
This new role for XML as a universal platform for interoperability made
necessary to address two new requirements: (1) schema specification (or
metadata) for XML data and (2) data and metadata distribution among multiple
web sites. The W3C put forward XML Schema (XMLS) to address the first of
these requirements and XML Name Spaces (XMLNS) to address the second one.
The <xmlns>element of an XML document generally indicates
the URL where to find the schema that specifies the structure pattern to
which the document must conform. This use of namespaces and schemas allows
establishing, in a non-centralized way, a common understanding about the
validity and the meaning of the XML data interchanged between components
of an open system.
XMLS allows defining:
- The elements that can appear in a XML document;
- The attributes that can appear in each of these elements;
- How these elements can be nested;
- How these elements must be ordered at a specific nesting level;
- The data types of the elements and attributes, as well as its default
or fixed values.
Before the XMLS definition, an XML document schema could be specified
only using a DTD (Document Type Definition). Although handy for the initial,
restricted, document representation purpose of XML, the DTD formalism revealed
too limited for adequately specifying database schemas and the intentional
part of knowledge bases. XMLS improves on DTD for such purposes thanks
to the following features:
- Support for a richer set of built-in data-types than DTDs, including
byte, date, integer, and SQL and Java primitive data-types;
- Use of an XML-based syntax, thus preventing the user to have to learn
yet another syntax (that of DTD) and allowing reuse of all tools and languages
available to create and manipulate XML documents to work with XMLS documents;
- Extensibility through XMLS constructs that allows reusing of parts
of schemas in other schemas;
- Object-oriented user-defined data types with inheritance of element,
attribute, and data-type definitions;
- Integration with XML namespaces.
2.3 Models, meta-models, XML and UML
Three trends recently emerged in computational language design. The
first one is the use of meta-circularity to specify language parts or extensions
using the language itself. The relation between XML and XMLS exemplifies
this trend: XML provides a syntactic standard for XMLS, whereas XMLS provides
a structural standard for any documents, data or knowledge about a specific
application domain codified in XML syntax. The second trend is the use
of UML class diagrams together with OCL constraints to specify object-oriented
models for representing basic types and language constructors. Such diagram
is frequently called a language meta-model [Carlson 2001].
The third trend is the use of XMI for representing such meta-models in
textual format with XML syntax. This approach permits using XLST [Kay
2000] to declaratively implement software that process artifacts written
in the projected language (in our case, XOCL). Declarative programming
within the XML paradigm allows faster prototyping and maintenance for a
wide range of services such as automated code-generation, language translation,
compilation and compression [OMG 2003].
Cooperation between such services in an open, component-based, heterogeneous
system requires the definition of a common XMI meta-model schema. For a
language meta-model is essentially a conceptual ontology of what the language
can express, the use of a DTD to specify a meta-model schema is as problematic
as its use to specify an artificial intelligence application knowledge
base. XMI meta-model schemas are thus preferentially defined in XMLS.
3 XOCL Design Methodology
The three trends mentioned in the previous section can be combined in
a two step language design methodology: (1) define an object-oriented meta-model
of the language using a visual CASE tool that supports UML and OCL; and
(2) codify in XMLS the schema for the XMI textual format of this meta-model.
This is the approach we followed to design XOCL. We first defined a meta-model
of OCL using UML and OCL itself. From this meta-model, we then specified
an XML Schema for XOCL expressions to be embedded in XMI encoding of domain
application UML models.
3.1 An OCL Meta-Model
Our OCL meta-model consists UML class diagrams with OCL constraints
that represent the OCL constructs and their associated UML constructs.
Although being much simpler than the current draft meta-model under construction
by OMG [OMG 2003], our meta-model covers all OCL constructs.
We developed it from three sources: (1) the OMG UML 1.4 meta-model [OMG
2003] to guarantee its integration with the latest UML standard; (2)
the OCL EBNF (Extended Bacchus-Naur Form) grammar [Reisner
1984], and (3) the OMG natural language OCL specification.
Figure 2 shows a small excerpt of our proposed OCL
meta-model. Our OCL construct classes are shown in gray whereas the reused
UML OMG meta-model classes are shown in white. An OCL constraint (Constraint
class) consists of one or more expressions (Expression class) and
is associated to a context (Context class).
This context either is structural (StructuralContext class) for
invariant constraints or OCL variable definitions (respectively Invariant
and Definition subclasses of the StructuralConstraint class),
or behavioral (BehavioralContext class) for pre and post-condition
constraints (respectively PreCondition and PostCondition
subclasses of the BehavioralConstraint class)1.
In addition, a structural constraint applies to either a Feature or an
AssociationEnd, two classes reused from the UML meta-model, while a behavioral
constraint only applies to a BehavioralFeature another UML meta-model
class that subsumes UML methods and operations.

Figure 2 : Part of the OCL meta-model
3.2 XOCL Schema
With this schema, applications can interchange XOCL messages with the
guarantee that: (1) the receptor will have the same understanding about
the data sent; (2) these data will not present any syntactical error; and
(3) any piece of OCL code included at the XML document will not be incomplete.
Figure 3 shows an excerpt from the XOCL schema written
in XMLS. This schema specifies that an XOCL constraint be identified through
an <OCLConstraint>root element (line 1). According to the meta-model
showed in section 3.1, a constraint is classified
as StructuralConstraint or as BehavioralConstraint. A StructuralConstraint
can be further classified as a Definition or an Invariant, whereas a BehavioralConstraint
can be classified as a PreCondition or as a PostCondition. An <Invariant>element,
specified between the lines 13 and 24 in Figure 3,
has two mandatory child elements: <Context>(line 17) and <Body>(line
19) in order to validate only XOCL constraints with a specified context
and a specified OCL code body.
[1] This cross association
between constraint subtypes and their respective associated context subtype
is actually represented in the complete OCL meta-model as an OCL constraint
that we omitted from Figure 2 to avoid cluttering it.

Figure 3: XOCL elements definition in XMLS
Each OCL construct of our complete OCL meta-model has a corresponding
XML construct, either an element or an attribute, in XOCL. In addition
to define such constructs, the XOCL XMLS also specifies valid nesting among
these constructs. There are many different possible ways to encode such
constructs and nesting constraints in XMLS. The most straightforward codifications
turned out to be very verbose and thus hard to understand by human users.
There are several XMLS design issues involved making XOCL a concise and
clear language. They are discussed in the next section.
4 Issues in Mapping Object-Oriented Models to XML
Schema
The design of XOCL raised many issues concerning the relative conciseness,
genericity and extensibility of alternative codifications and the resulting
compromise between human legibility and machine processing simplicity.
For lack of space, we focus our discussion here on only two key issues:
(1) usage of inheritance and (2) usage of XMLS grouping.
4.1 Usage of Inheritance
In XMLS, inheritance is provided by the <extension>constructor,
which allows a type to extend another already declared. This is illustrated
in Figure 3: <OCLConstraint>is declared
as a ConstraintType (line 1). This type is complex, in the sense
that it may have at least one child element or attribute, and abstract,
in the sense that it cannot be instantiated as an XML element. Two others
abstract types that extend ConstraintType are StructuralConstraintType
(line 3) and BehavioralConstraintType (line 8). While InvariantConstraintType
(line 13) extends StructuralConstraintType, it is not abstract and
can type any <OCLConstraint>element. PreConditionConstraintType
(line 25) similarly extends BehavioralConstraitnType. Given this
type hierarchy, declaring in a XML document that an OCL constraint is an
invariant, only involves adding the single element: <OCLConstraint
xsi:type="InvariantConstraintType">2.
Using inheritance, one of the features that make XMLS more expressive than
DTD, allows an XOCL specification to be concise without leaving out details.
This carries over to the code of processing applications taking as input
XMI documents with embedded XOCL. For example, an XSLT application can
have different rules for processing invariants and for processing pre-conditions.
Such differentiated processing would not be straightforward were all constraints
identified only through the <Constraint>element.
As mentioned in section 3.1, while the context
of an invariant is only an UML classifier (e.g., a class), the context
of a pre or post-condition consists of a pair <UML classifier, UML operation>(e.g.,
a class and one of its methods). In our OCL meta-model, this is accounted
for by a constraint context type hierarchy that reflects the constraint
type hierarchy. How XMLS type inheritance allowed us to encode this directly
in XOCL is shown in Figure 4. The StructuralContextType
has just two child elements: (1) <Package>(line 6) that identifies
the package where the UML artefact is; and (2) <UMLClassifier>(line
8) that identifies the UML artifact whose constraint is applied. On the
other hand, the BehavioralContextType has three additional child elements:
(1) <OperationName>(line 22) that identifies the method name where
the pre or post-condition is applied; (2) <FormalParameterList>(line
23) that contains a list of all parameters of the method identified by
<OperationName>tag; and (3) <ReturnType>(line 25) that declares
the return type of the method.
[2] The xsi:type
attribute is a XML keyword for typing the current element, in the case,
<OCLConstraint>.

Figure 4: Type inheritance in XOCL schema
4.2 Usage of Element Groups
In our XOCL schema, all operators are defined through the <:xs:group>XMLS
constructor. This allows defining element groups to be reused in various
complex type definitions, so as to avoid repetitions of the same element
definitions within several other different complex type definitions. An
example of the combined usage of element grouping and type inheritance
in XOCL is given in Figure 5. An expression type hierarchy
first distinguishes between atomic and compound expressions. For example,
'A > B' is a compound expression of subtype comparative. Its components
are the comparative operator '>' and its operand A and B, both atomic
expressions. Lines 8-23 in Figure 5 define the group
of comparative operators (ComparativeOperators group). This group can then
be referenced anywhere as a substitute for the repetition of the all the
comparative operators definition. Lines 3-5 contain an example of such
reference. Figure 6 shows the XOCL <GreaterThan>element
encoding the 'A > B' expression following the Schema fragment of Figure
5.

Figure 5: ComparativeOperator group and ComparativeOperatorType
composed type specification

Figure 6: Use of <GreaterThan>element
5 Example of OCL constraint codified in XOCL
Having discussed some aspects of the XOCL schema, let us now look at
an example XML element that follows this schema. Such element is given
in Figure 7. It codifies in XOCL the first example
OCL constraint of section 2.1 reproduced below:
Context Company inv managerConstraint:
self.manager.age > 18 and Self.manager.age < 65 and
self.manager.isUnemployed = false
In Figure 7, the <OCLConstraint>root
element, at line 1, just identifies the element as specifying an OCL constraint.
Lines 3 to 5 declare the context associated with the constraint. Line 4
specifies that this context is an UML class identified through an xsi:type
attribute of the <UMLClassifier>element. At line 4, the class associated
to the constraint is identified via a name attribute, in this example with
value "Company".
Next comes the <Body>element that contains the OCL expressions
specified in lines 6-51. Line 6 indicates that the constraint body expression
is of type NaryLogicalExpressionType. In this example, the expression is
a Boolean conjunction encoded by an <And>element in line 7. This
element has three operands, each one being a comparative expression (indicated
by an <Exp>element). They are respectively encoded in lines 8-20,
21-34 and 35-49.
The first comparative expression specifies the "self.manager.age
> 18" constraint consisting of the '>' operator and its two
operands: 'self.manager.age' and '18'. This constraint is encoded as a
<GreaterThan>element with one child element per operand. The first
one, in lines 9-15, is an <Exp>element, with the xsi:type attribute
filled with "UMLModelPropertyCallExpressionType" because it references
a role (manager) and an attribute (age) of a class (Company). <PropCall>elements
appear as child elements which nesting reflects the construct nesting of
the OCL expression. Therefore, we have one <PropCall>element (line
10) for the 'self' OCL keyword, another (line 11) for the 'manager' role
and another (line 12) for the 'age' attribute. The second operand, in lines
16-18, is an expression element of type StringLiteralExpressionType with
one child <Literal>element that contains the operand value specified
through the value attribute.
The two other comparative expressions of the constraint, "self.manager.age
< 65" and "self.manager.isUnemployed = false" that appear
next as child elements of the <And>element are encoded following
the same conventions.
6 Conclusions
In this paper we presented XOCL, an XML language to codify OCL constraints.
XOCL elements can be embedded within XMI documents to provide the same
detailed XML structuring for constraints as for the other parts of a UML
model. To design XOCL, we started by constructing a complete, but simple
OCL meta-model using UML and OCL itself. From this meta-model we then specified
an XOCL document schema using XMLS.
1 <OCLConstraint xsi:type="InvariantContraintType"
2 name="managerConstraint">
3 <Context>
4 <UMLClassifier xsi:type="ClassType" name="Company"/>
5 </Context>
6 <Body xsi:type="NaryLogicalExpressionType">
7 <And>
8 <Exp xsi:type="CompartiveExpressionType">
8 <GreaterThan>
9 <Exp xsi:type="UMLModelPropertyCallExpressionsType">
10 <PropCall name="self" type="Object">
11 <ProCall name="manager" type="Associatin">
12 <ProCall name="manager" type="Attribute"/>
13 </ProCall>
14 </ProCall>
15 </Exp>
16 <Exp xsi: type="ArithmeticalLieteralExpressionType">
17 <Literal value="18"/>
18 </Exp>
19 </GreaterThan>
20 </Exp>
21 <Exp xsi:type="ComparativeExpressionType">
22 <LessThan>
23 <Exp xsi:type="UMLModelPropertyCallExpressionType">
24 <PropCall name="self" type="Object">
25 <PropCall name="manager" type="Association">
26 <PropCall name="age" type="Attribute"/>
27 </PropCall>
28 </PropCall>
29 </Exp>
30 <Exp xsi:type="ArithmeticalLiteralExpressionType">
31 <Literal value="65"/>
32 </Exp>
33 </LessThan>
34 </Exp>
35 <Exp xsi:type="ComparativeExpressionType">
36 <Equal>
37 <Exp xsi:type="UMLModelPropertyCallExpressionType">
38 <PropCall name="self" type="="Object">
39 <PropCall name="manager" type="Attribute">
40 <PropCall name="isUnemployed">
41 type="Attribute"/>
42 </PropCall>
43 </PropCall>
44 </Exp>
45 <Exp xsi:type="BooleanLiteralExpressionType">
46 <Literal value="false"/>
47 </Exp>
48 </Equal>
49 </Exp>
50 </And>
51 </Body>
52 </OclConstraint>
Figure 7 - Example of code in XOCL language
Our main concern when faced with alternative design options for this
schema has been to make XOCL elements as concise as possible to facilitate
their understanding by both human readers and software applications. Concision
also save communication bandwidth for model interchange over the network.
The methodology we outlined in this paper for the design of XOCL is
general. It can be used to design XML codification of any language. This
is an interesting contribution given the ubiquitous need for such codifications
triggered by the growing success of XML mediated multi-language development
process and platform such as the MDA architecture [OMG
2003] and .net.
The XOCL language presented in this paper provides a key missing piece
of the OMG and W3C standard puzzle, that includes XML, XMLNS, XMLS, XSLT,
UML, OCL, XMI and CWM, to support advanced applications and development
practices within such open, heterogeneous paradigms. We intend to submit
it as a proposal to the appropriate OMG working group. It is available
for inspection and feedback on the web at www.cin.ufpe.br/~fsr/xocl.
The purpose of XOCL is to fill a gap in XMI. To be useful in practice,
it needs to be integrated XOCL with XMI. This not an immediate step because
the current XMI specification provided by OMG is encoded as a DTD. For
the same reasons discussed in sections 2.2 and 4
which motivated specifying XOCL as an XMLS, the expressive limitations
of DTD turn this XMI specification under-constrained and the documents
that conform to it extremely verbose. To address this problem, an OMG working
group is currently drafting an XMLS for XMI.
XMLS suffers himself from some limitations. For example, it does not
allow expressing validity constraints that involve elements located in
different branches of an XML document viewed as a tree. To address this
limitation and others, Schematron [Schematron 2002]
and XCSL [XCSL 2001] have been proposed to complement
XMLS to encode more sophisticated XML document schemas. Theses languages
are based on arbitrary document fragment matching and transformation patterns
using the W3C standards XPath [Kay 2000] and XSLT.
They can be easily integrated with XMLS (e.g., Schematron expression can
even be embedded within an XMLS <appinfo>element). As future
work, we intend to choose one of these constraint specification languages
to refine and more concisely and generically express the most complex constraints
of our XOCL Schema.
Acknowledgements
This research was supported by grants from CNPq (Conselho Nacional de
Desenvolvimento Científico e Tecnológico) of the Brazilian
Federal Government.
References
[Atkinson et al. 2002] Atkinson, C., Bayer, J.,
Bunse, C., Kamsties, E., Laitenberger, O., Laqua, R., Muthig, D., Paech,
B. Wust, J. and Zettel, J.: "Component-based product line engineering
with UML"; Addison-Wesley, Reading/MA (2002).
[Bergenti and Poggi, 2000] Bergenti F. and Poggi
A.: "Exploiting UML in the Design of Multi-Agent Systems"; Proc.
ECOOP - Workshop on Engineering Societies in the Agents' World, Sophia
Antipolis and Cannes (2000), 96-103.
[Booch et al. 1998] Booch, G., Jacobson, I. and
Rumbaugh, J.: "The Unified Modeling Language User Guide"; Addison-Wesley,
Reading/MA (1998).
[Brucker and Wolff 2001] Brucker, A. D. and Wolff,
B.: "Checking OCL Constraints in Distributed Systems Using J2EE/EJB",
Technical Report 157, Albert-Ludwigs-University at Freiburg, Germany (2001).
[Carlson 2001] Carlson, D.: "Modeling XML
applications with UML: practical e-buisiness applications"; Addison-Wesley,
Reading/MA (2001).
[Cranefield and Purvis 1999] Cranefield and Purvis,
M.: "UML as an ontology modelling language"; Proc. Workshop on
Intelligent Information Integration, 16th International Joint Conference
on Artificial Intelligence, Stockholm (1999).
[Devedzic 2001] Devedzic, V.: "Knowledge Modeling
State of the Art"; Integrated Computer-Aided Engineering, 8, 3 (2001),
257-281.
[Poseidon 2003] Poseidon for UML, May 2003, http://www.gentleware.com
[Kay 2000] Kay, M.: "XSLT Programmer's Reference";
Wrox Press, (2000).
[Mellor and Balcer 2002] Mellor, S.J. and Balcer,
M.J.: "Executable UML: a foundation for the model driven architecture";
Addison-Wesley, Reading/MA (2002).
[OMG 2003] OMG - Object Management Group; Link:
http://www.omg.org
[Poole et al. 2001] Poole, J., Chang, D. Tolbert,
D. and Mellor, D.: "Common Warehouse Meta-model: an introduction to
the standard for data warehouse integration"; Wiley & Sons (2001).
[Purvis et al. 2000] Purvis, M.K., Cranefield, S.,
Bush, G., Carter, D. McKinlay, B., Nowostawski, M. and Ward, R.: "The
NZDIS Project: An Agent-based Distributed Information Systems Architecture";
Proc. Hawaii International Conference on Systems Sciences, Big Island (2000).
[Reisner 1984] Reisner P.: "Formal grammar
as a tool for analyzing ease of use"; Proc. Human Factors in Computing
Systems, (Thomas J.C. & Schneider M.L.; eds.), New York (1984), 53-78.
[Rose 2003] Rational Rose, May 2003, http://www.rational.com/uml
[Schematron 2002] Schematron Assertion Language,
Version 1.5, October 2002, http://www.ascc.net/xml/resource/schematron/Schematron2000.html
[Schreiber 1999] Schreiber, G.: "Knowledge
engineering and management: the CommonKADS methodology"; MIT Press,
Cambridge/MA (1999).
[Warmer and Kleppe 1999] Warmer J. and Kleppe A.;
" The Object Constraint Language: Precise Modeling with UML",
Object Technology Series, Addison-Wesley, Reading/MA (1999).
[W3C 2003] W3C - World Wide Web Consortium; Link:
http://www.w3.org/
[XCSL 2001] XML Constraint Specification Language,
Version 2.2, May 2001, http://www.di.uminho.pt/~jcr/PROJS/xcsl-www/
|