Ontology Driven Architectures and Potential Uses of the Semantic Web in Systems and Software Engineering

Abstract

This note outlines the benefits of applying knowledge representation languages common to the Semantic Web, such as RDF and OWL, in Systems and Software Engineering practices.

Target Audience

This note is aimed at industrial professionals, tool vendors and academics with an interest in applying Semantic Web technologies in Systems and Software Engineering (SSE) contexts. These may include:

• Software Engineers who are interested in the benefits and potentials of Semantic Web technologies
• Members of the Semantic Web community
• Members of the Model Driven Architecture (MDA) community who may be interested in ontologies as models
• Members of the the automated software engineering community who are interested in formal methods for [partially] automating the development of software systems
• Knowledge Representation and Inference researchers interested in practical applications

Objectives

• To enthuse SSE practitioners about the potential uses the Semantic Web and Semantic Web technologies in SSE
• To outline the benefits of applying Semantic Web technologies in SSE contexts
• To encourage collaboration between the Semantic Web and Systems and Software Engineering communities
  

Status of this document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at http://www.w3.org/TR/.

This is a public (WORKING DRAFT) Working Group Note produced by the W3C Semantic Web Best Practices & Deployment Working Group, which is part of the W3C Semantic Web activity.

Discussion of this document is invited on the public mailing list public-swbp-wg@w3.org (public archives). Public comments should
include "comments: [SE]" at the start of the Subject header.

Publication as a Working Group Note does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress. Other documents may supersede this document.

1. Introduction

Until recently work on accepted practices in Systems and Software Engineering (SSE) has appeared somewhat disjointed from that breaking ground in the area of formal information representation on the World Wide Web (commonly referred to as the Semantic Web initiative). Yet obvious overlaps between both fields are apparent and many now acknowledge merit in a hybrid approach to IT systems development and deployment, combining Semantic Web technologies and techniques with more established development formalisms and languages like the Unified Modeling Language (UML). This is not only for the betterment of IT systems in general, but also for the future good of the Web, as systems and Web Services containing rich Semantic Web content start to come online.

This note attempts to outline how Semantic Web technologies can be applied in Systems and Software Engineering, as well as the benefits such applications could bring. Firstly, we will review the rise of ontology driven architecture (ODA) in Section 2. We then, in Section 3, propose some ideas based on ODA and, more generally, on how Semantic Web technologies can be applied in Systems and Software Engineering. Some of these ideas are illustrated by the examples presented in Section 4. Finally, Section 5 outlines possible next steps in the near future.

2. Background

2.1 History

Those who are familiar with the heritage of software development might correctly suggest that the application of formal logic and declarative knowledge representation in Systems and Software Engineering is not new, holding up much good work in the areas of Automated Software Engineering, Formal Methods, Domain Theory, Relational Algebra and real-time and embedded systems engineering. Indeed, related approaches have seen much research for many decades now. The following list therefore aims to provide a brief overview of established academic work and industrial practice on which further discussions in this note are based.

• Knowledge-based Software Engineering
  A field of endeavour established nearly twenty years ago, with its annual conference (the International Conference on Software Engineering and Knowledge Engineering [SEKE 2005]) already in its 17th year. Relevant topics include AI approaches to Software Engineering, automated reasoning and software design, knowledge representation, retrieval, visualization and many more related topics.
  
  
• Goal-Driven Requirements Engineering
  Since the early 1990s, research in this area has considered UML-based domain models as parts of software architectures. [Pohl 1999]
  
  
• Faceted Software Classification
  The work on faceted software classification also goes back to the late 1980s. Its goal is to facilitate reuse of software components by attaching keywords to them, then are organized these into facets. [Pietro-Diaz 1991]
  

Following on from such works, modelling a common understanding of domains through formal and semi-formal methods has proven itself essential to advancing the practice. Modelling as a discipline for architecting, developing, communicating, and verifying abstract designs is now an integral best practice in Software Engineering, as in other engineering communities. Computer Aided Software Engineering (CASE), the Zachman framework for enterprise architecture, and, more recently, Model Driven Architecture® (MDA®) are prominent examples of this approach. We will highlight the latter in the subsequent section.

2.2 Model Driven Architecture (MDA)

In the Object Management Group (OMG)'s MDA initiative, models are used not only for design and maintenance purposes, but as a basis for generating executable artifacts for downstream use. The MDA approach grew out of much of the standards work conducted in the 1990s in the Unified Modeling Language (UML). It encourages model-based Software Engineering through

• specifying systems independently of the platforms that they will ultimately be deployed on,
• specifying the platforms themselves and enabling generation of the software artifacts specific to those platforms (e.g., J2EE, CORBA IDL, WSDL),
• supporting platform selection, and
• enabling transformation of system specifications into specifications for particular deployment platforms.

Critical to the success of such an approach are firstly the notion of separation of architectural concerns at every step in the design process, and secondly well-designed models, models that can be used to orchestrate understanding of a design in all phases of the development lifecycle.  MDA provides a methodology and the basic requirements for tools that prescribe the kinds of models that may be used at each design phase, how they should be developed to maximize reuse, portability, and interoperability, and the relationships among them to support such code generation.  It insulates business applications from technology evolution through increased platform independence, portability and cross-platform interoperability, encouraging developers to focus their efforts on domain specificity.

The OMG's Meta Object Facility (MOF) standard defines the metadata architecture that is central to MDA-based computing.  MOF tools use metamodels to generate code that manages metadata -- as XML documents, CORBA objects, Java objects, WSDL services, and so forth. This generated code includes access mechanisms, or application programming interfaces, to read and manipulate, serialize and transform, and abstract the details of various interfaces based on access patterns.  From a knowledge representation perspective, however, MDA only gets part of the job done.  The suite of standards supporting the MDA initiative streamline the mechanics of managing and integrating models of various aspects of a system's design, but say little about the underlying semantics of the domain being modeled.

Formal representation can help to limit ambiguity and improve quality not only in automation of business semantics but in overall engineering of complex systems. Yet, with increasing formality, tool support can be increasingly abstract, which makes methods more difficult to implement and may limit the freedom of expression available to the engineer. On the other hand, less formal approaches can be notoriously ambiguous: an isolated model, no matter how well specified, can be open to misinterpretation by those who are not familiar with the problem space. Supporting documentation can help alleviate such issues, but in practice, outside of the MDA community, development of documentation is frequently independent from actual model development.  Further, MDA does not currently support automated consistency checking or validation, although much has been accomplished in related disciplines including Automated Software Engineering. While MDA provides a powerful and proven framework for Systems and Software Engineering, Semantic Web technologies can naturally extend it to enable representation of unambiguous domain vocabularies, model consistency checking, validation, and new capabilities that leverage increased expressivity in constraint representation. Semantic models, or ontologies, will augment the OMG standards and methodology stack, hence giving rise to Ontology Driven Architecture (ODA).

2.3. Recent Developments

Over the past two years there has been significant work to bring together Software Engineering languages and methodologies such as the UML with Semantic Web technologies such as RDF and OWL, exemplified by the OMG's Ontology Definition Metamodel (ODM). While this work has been largely motivated by an interest to exploit the popularity and features of UML tools for the creation of vocabularies and ontologies, some have also advocated the potential benefits of applying Semantic Web concepts to model validation and automation, as well as to enable new Software Engineering capabilities.

The relatively recent introduction of Web Service concepts and technologies also adds compelling reason for the drive to use web-friendly ontologies in Systems and Software Engineering. Such concepts allow declarative functionality to be deployed, discovered and reused over the web to obvious advantage. Given the old computing adage that "all the software functionality needed in the world has already been written somewhere," it theoretically follows that if all this functionality were made openly available via Web Service interfaces, software construction would become a radically different and simplified activity. That is so long as Web Service metadata is accurate, complete and easy enough to use - and that's where formal ontologies and Semantic Web languages come into play. Indeed one could now consider that, given the vastness of the Web and the communal culture it promotes, the future of software development may well not actually lie in the construction of new functionality, but rather the discovery and gluing together of existing functionality to achieve all the desired aims of the solution in mind.

It may be fair to argue that the Semantic Web brings little that is new to Software Engineering. Formal methods, rigorous domain definition and knowledge based approaches have been explored with limited success outside the real-time, embedded software community to date. So what is it about the amalgamation of OWL, UML and the Model Driven Architecture (MDA) that will make a difference, and why now? Given that the rate of increase in scalability, performance, distribution, and interoperability requirements demonstrated over the last decade remains relatively constant, the mandate for new capabilities that further abstract and automate Software Engineering processes is clear.  Even small-scale, incremental improvements in low level capabilities have historically led to enormous gains at higher levels.  Advances internal to the Systems and Software Engineering community have not been sufficient to tip the scales thus far, but multidisciplinary approaches, such as bridging Semantic Web and MDA technologies in novel ways, may enable such significant improvements. 

As simple as it may sound, the Semantic Web brings one huge advantage - the Web itself - which has previously proved critical to the acceptance of at least one, somewhat less than original, idea in IT.  Java has gained widespread adoption in global software development in recent years, yet its main features are far from different to those of dozens of earlier programming languages. What is unique, however, is that it is specifically targeted at Web-based systems and is standards-based - both properties also common to the Semantic Web.  For this reason alone it is compelling to think that a combination of OWL, UML and MDA might indeed make a real difference.

Additionally, as mentioned above, MDA mandates separation of concerns at many levels; some MDA proponents advocate not only abstracting implementation characteristics from business logic but taking significant care in defining the agreements that software components expose via their interfaces. A critical component of this practice is to carefully represent the preconditions, post-conditions, and what are called invariant rules (similar to necessary and sufficient conditions in description logics), to specify the behaviors of these components under various circumstances.  The aim here is to make such rules unambiguous, enabling increasing automation based on the models, composition of components, and limiting misunderstanding of the design.  Primary limitations to widespread adoption of this practice, known as Design-By-Contract, include scalability as the number of rules increases, particularly in large-scale, complex designs, the ability to do more than cursory consistency checking, and unambiguous separation of the domain terminology from the rules themselves.  Semantic Web technologies can dramatically improve this discipline by (1) enabling unambiguous representation of domain terminology, distinct from the rules, (2) enabling automated consistency checking and validation of invariant rules, preconditions, and post-conditions, and (3) supporting knowledge-based terminology mediation and transformation for increased scalability and composition of components.  While collaboration between the Semantic Web Services and MDA communities is just beginning to occur, we see this as a tremendous opportunity for significant productivity and software quality improvement.

Regardless of the benefits that Semantic Web technologies can bring to Systems and Software Engineering, it is still worth noting that metadata ontologies can be difficult to construct, maintain and use. As such a number of tools have been produced to facilitate the engineering and maintenance of RDF and OWL ontologies. These range from text-based editors/validators through to experimental graphical environments. Further tools have also been created to enable inferencing over such ontologies. References to relevant tooling can be found at sites like semanticweb.org.

Related research is listed in Appendix.

3. Proposed Ideas

3.1 The Semantic Web in Systems and Software Engineering

Having raised the idea of using of the Semantic Web in Software Engineering, a commonly asked question arises; namely, how does one broadly characterise the Semantic Web in terms of Systems or Software Engineering use? In attempting to answer this question, two loose definitions are apparent:

• As a 'classification', merely to group together related tools and techniques for modeling rigorous semantics during specification and design stages of the Software Lifecycle.

  Primarily such tools and techniques should be viewed as being formally descriptive in character, but there appears little reason to restrict this definition other than standards alignment. Therefore, it may also be relevant, at some appropriate point in the Semantic Web's future, to include prescriptive, invasive and/or other types of approaches under this heading.

• As a 'mechanism' for rigorously describing, identifying, discovering and sharing artifacts amongst discrete subsystems, systems and systems' design teams both during design and at runtime.

  In such circumstances the Semantic Web can be viewed as a set of formalised corpora of interrelated, reusable contents, which can further be classified as being either:

  • Passive (data in the form of):
    • Flat documents and data (HTML, XML etc).
    • Dynamically generated documents and data (via JSP, PHP etc).
    • Metadata (RDF, OWL etc).
    • Media - pictures video, music etc.
    • Databases.
  • Active (functionality presented as):
    • Web Services (inc Semantic Web Services).
    • Functional components referenced as fragments within passive content (JavaScript, Java applets etc).

The presence of both of these types is highly important, as is the fact that they can be indistinctly described using ontological techniques and technologies. Yet their distinction is paramount in the Systems and Software Engineering space. On their own they are of limited from an automated systems engineering standpoint (requiring intelligent agents like humans to cater for the missing type), but combine then together using technologies that can treat them equally and all the ingredients needed to automatically compose fully fledged systems are present. In such situations, the Semantic Web can be likened to a catalyst; It may not be the most powerful element in a particular system, but without it the power of any of the other elements may never be fully realised.

Two examples on the proposed idea of 'The Semantic Web in Systems and Software Engineering' are presented in Section 4.1.

3.2 Ontologies as Formal Model Specifications and the Incorporation of Such Models in Semi-Formal Languages

Despite its importance, software development still lacks adequate technology in many areas. The development of software systems has been recognised as an important and complex activity that requires the participation and collaboration of physically distributed teams of people during systems development. Therefore, a large amount of skilled manpower, time and effort is necessary when creating, designing, writing, testing, evaluating, and maintaining software systems. The multiplicity of stakeholders and development participants, the addition of new features to existing services, and the heterogeneity and complexity of products being developed require software development to be a very knowledge and communication intensive process. Existing software development tools offer very good support for document management and modelling, but lack of support for communication and knowledge management. Therefore, a driving force to enhance existing tools with communication and knowledge management is necessary.

In many respects ontologies can be simply considered as rigorous descriptive models in their own right, being akin to existing conceptual modeling techniques like UML class diagrams or Entity Relationship Models (ERM). As such, their purpose is to facilitate mutual understanding between agents, be they human or computerised, and they achieve this through explicit semantic representations using logic-based formalisms. Typically, these formalisms come with executable calculi that allow querying and reasoning support at runtime. This adds a number of advantages, specifically in the areas of:

• Quality
  • Requirements conformance and consistency checking.
  • Rigorous typing, catogorisation and identification.
  • Ease of formal specification. With the aid of graphical modeling tools (as commonly favoured in MDA), ontologies can be built using much less abstract user interfaces.
  • Communication of requirement and intent to domain experts and developers through:
    • The ability to capture, relate and manage models of systems and information at multiple levels and from multiple standpoints.
  • Increases in semantic expresivity through coverage of concpets not embodied in current se tools.
  • Reductions in design ambiguity through.
  • Unified syntax for tooling. Standard Semantic Web languages (such as OWL) can provide unified syntax for existing formal method languages, providing the potential for open and cooperative tooling environments for formal methods (see Example C presented in Section 4.2).
  • Facilitation of decoupling between abstraction layers needed for effective modelling.
• Cost
  • Reduce maintenance overhead through increases in consistency.
  • Increased potential for reuse, substitution and extension via accurate content discovery on the Semantic Web.

Hence, given the semantically rich, unambiguous qualities of information embodiment on the Semantic Web, the amenable syntax of Semantic Web languages, and the universality of the Semantic Web's XML ancestory, there appears a compelling argument to combine the semi-formal, model driven techniques of Software Engineering with approaches common to Information Engineering on the Semantic Web. This may involve the implanting of descriptive ontologies directly into systems' design models themselves, the referencing of separate semantic metadata artifacts by such models or a mixture of both.

What is important is that mechanisms are made available to enable cross-referencing and checking between design descriptions and related ontologies in a manner that can be easily engineered and maintained for the betterment of systems' quality and cost.

3.3 Software Lifecyle Support

Some activities in the software development life cycle like requirements elicitation, analysis, design, and testing require intense stakeholders interactions. The use of ontologies, name spaces, and metadata can assist with these interactions by providing standardisation of terminology, relationships, and rules for a specific domain. This standardisation can

• Support the generation of domain models employed in different phases of the software development life-cycle,
• Reduce problems of inconsistencies between different artefacts generated during the development life cycle that normally occur due to the multi-perspective of involved stakeholders,
• Facilitate the enforcement of traceability of these artefacts, which improves quality of the software system (e.g. change impact analysis, maintenance and evolution, component reuse, inspection, understanding),
• Augment sharing and interoperation of models created by stakeholders,
• Assist with interface semantics and software architecture descriptions; and
• Help identifying and reusing existing components.
  

3.4 Corpora of Reusable Contents and the Use of Metadata as Relational Data

Given that The Semantic Web technologies uses triple-based data representation and that this is merely a minimization of the representation employed in relational database technologies, the attraction of considering the Semantic Web as a specialised relational framework has been understood for some time. Even so, by recognising such potential it is important to be pragmatic and consider the term 'Semantic Web' as a generalisation, having at least two specific interpretations from a relational data standpoint. These being:

• A collection of Semantic Intra-Webs: Distinct formal ontologies, relating to an enterprise or community of interest e.g. various collections of researchers who take responsibility for the care and managed evolution of such Webs in ways analogous to the Python or Linux communities. These groups, by their very nature, will be very concerned about ontology consistency and data quality, but may not necessarily wish for their information to be shared as part of some larger structure.
• A singular Semantic Web: A global collection of as many Semantic Webs (as defined above) as are public plus anything else marked up appropriately in referencible Web space.

At current levels of Web maturity it appears reasonably plausible that strong reliable inference could be achieved using ontologies that fall under the former of the two above definitions. Nevertheless for ontologies associated with the latter definition, less reliable approaches to knowledge extraction, such as data exploration, heuristics, and 'mining', appear more realistic. This is not to say that the concept of a global Semantic Web is not useful from a relational data perspective, it is just that data provenance and quality issues will undoubtedly lead to lesser degrees of information credibility across its entirety. Even so, regardless of the fact that a singular Semantic Web might be difficult to achieve in the short term, from a SE point of view the first viewpoint is already highly attractive.

When debating the potential for relational data schemes under any such definition of the Semantic Web, a significant benefit for SSE is often overlooked: If you can describe something sufficiently well (as is the ultimate aim on the Semantic Web), and that thing exists on the Semantic Web with a similar level of descriptive clarity, the chances of you finding precisely that thing are greatly increased. So, setting out to describe things better is indirectly rewarded by the increased ability to be [semi-automatically] discovered and to find content that is clearly equivalent or related. To be direct, rich descriptions empower discoverability.

By suggesting use of the Semantic Web as a system for runtime information and artefact sharing there is also an implicit need to provide means for clearly identifying participating objects based on composites of characterising semantic properties (metadata in the form of name-pair/predicate-object values), and this differs from current Semantic Web schemes for unique identification, such as FOAF sha1 (i.e. unique identification based on a single Inverse Functional Property (IFP)). By using composite identification the Semantic Web can be seen as a truly global relational assembly of content and, as with every relational model, issues dealing with composite forms of unique identification have to be addressed (i.e. identification via composite IFP schemes)

Such unique identification schemes should be capable of supporting both the interlinking of broadly related ontologies into grander information corpora (thereby implying formal similarities and relationships between discreet ontologies and/or systems through their classifying metadata), and the transformation of design time component associations into useful runtime bindings. This will, therefore, provide greater potential for metadata use across a broader spectrum of the Software Lifecycle. In so doing, this approach carries a number of obvious implications for systems employing such techniques:

• That Semantic Web technologies could be used to formalise associations between sub-components within a given system.
• That the Semantic Web could be used as a framework for design-time data and component sharing, and this includes the concept of design models being considered as valid, sharable artefacts in their own right.
• That the Semantic Web could be used as a framework for runtime data and component sharing between discreet and disparate systems.
• That new forms of systems could be created through the integration of discreet and disparate information and functionality with semantically similar metadata. This appears especially appealing given current advances the areas of Web Services and Service Oriented Architectures.
• That the Semantic Web could provide mechanisms for enabling re-usable software to work together.
• That the Semantic Web could provide 'richer' information on which systems and systems' components can work

In System Software Engineering it is also important to support systematic validation methods such as symbolic execution, term simplification, theorem proving, and model checking. In such settings, ontologies could be used to support declaration of automated reasoning knowledge in addition to type and signature information. Moreover, ontologies can also be employed to standardise the terminology used in different models and ultimately support the sharing and interoperation of models between different stakeholders, help model developers to choose appropriate level of abstractions to express models of various components of a system, and allow validations to be executed in terms of the domains of the systems represented by those ontologies. Examples of the use of ontologies to support the above tasks can be found in [HZ 2004a, HZ 2004b].

Semantic Web technologies can also be used to address the challenge of interoperability in distributed systems and heterogeneous data sources such as data warehouses [HNVV 2003]. In previous work [ZK 1999], we have seen the use of hierarchical terminology structures (i.e. metadata) to assist with the discovery of relevant data when dealing with heterogeneous database systems. More recently, the work in [ZCDKS 2002] has proposed the use of an 'information bus' that handles metadata in a generic way and enables metadata to be used as exchanged models and ontologies, in order to support interoperability between heterogeneous data sources. The use of metadata and ontologies preserves the autonomy of the existing systems, facilitate data search represented in heterogeneous data sources, provide classification and relationships between data, and allows evolution of the system by adding and removing data sources.

4. Examples

In this section we present examples for the proposed ideas presented in Section 3. The examples are grouped in subsections corresponding to the subsections of Section 3.

4.1 The Semantic Web in Systems and Software Engineering

This section discusses two examples for the proposed idea of 'The Semantic Web in Systems and Software Engineering'. In both cases, the applied characterization of the Semantic Web is that of 'classification', i.e., Semantic Web technologies and tools are used to facilitate the management of application servers and Web services middleware, respectively.

Example A: Developing and managing software components in an ontology-based Application Server

Application servers provide many functionalities commonly needed in the development of a complex distributed application. So far, the functionalities have mostly been developed and managed with the help of administration tools and corresponding configuration files, recently in XML. Though this constitutes a very flexible way of developing and administrating a distributed application, the disadvantage is that the conceptual model underlying the different configurations is only implicit. Hence, its bits and pieces are difficult to retrieve, survey, check for validity and maintain.

To remedy such problems, the ontology driven architecture (ODA) approach can support the development and administration of software components in an application server. The ODA approach is supplementary to MDA, where models abstract from low-level and often platform-specific implementation details. While MDA allows to separate conceptual concerns from implementation-specific concerns, currently MDA has not been applied for run-time relevant characteristics of component management, such as which version of an application interface requires which versions of libraries. MDA requires a compilation step preventing changes at runtime which are characteristic for component management. Besides, an MDA itself cannot be queried or reasoned about. Hence, there is no way to ask the system whether some configuration is valid or whether further elements are needed. This is explained by the lack of formal semantics of the notoriously ambiguous UML. In ODA, however, an ontology captures properties of, relationships between, and behaviors of the components that are required for development and administration purposes. As the ontology is an explicit conceptual model with formal logic-based semantics, its descriptions of components may be queried, may foresight required actions, e.g., preloading of indirectly required components, or may be checked to avoid inconsistent system configurations - during development as well as during run-time. Thus, the ODA approach retains the original flexibility in configuring and running the application server, but it adds new capabilities for the developer and user of the system.

Figure 1 shows how an ontology-based application server could be designed. The left side outlines potential sources, which provide input for the framework. This includes Web and application server configuration files, deployment descriptors or metadata files. With a standardized annotation syntax, Java 5.0 facilitates obtaining of sourcecode annotations as well. There are also first attempts to learn ontological descriptions out of Javadoc comments [SCGM 2005]. This information is parsed and converted into semantic metadata, i.e. metadata in terms of an according ontology. Thus, this data is now available conforming to a harmonizing conceptual model, weaving together so far separated aspects like security, component dependencies, version or deployment information. The semantic metadata and the ontology are fed into the inference engine which is embedded in the application server itself. The reasoning capability is used by an array of tools at development and at run time. The tools either expose a graphical user interface (e.g., security management) or provide core functionality (e.g., the dynamic component loader) [OESV 2004]. Concrete examples along a multimedia demonstration of reasoning with software components can be found at [OSE 2005].

Example B: Semantic Management of Web Services

Different Web Service standards, WS*, factorize Web Service aspects into interface (WSDL), workflow (BPEL) or security (WS-Security, WS-Policy) descriptions. The advantages of WS* are multiple and have already benefited many industrial cases. WS* descriptions are exchangeable and developers may use different implementations for the same Web Service description. The disadvantages of WS*, however, are also visible, yet: Even though the different standards are complementary, they must overlap and one may produce models composed of different WS* descriptions, which are inconsistent, but do not easily reveal their inconsistencies. The reason is that there is no coherent formal model of WS* and, thus, it is impossible to ask for conclusions that come from integrating several WS* descriptions. Hence, discovering such Web Service management problems or asking for other kinds of conclusions that derive from the integration of WS* descriptions remains a purely manual task of the software developers accompanied by little to no formal machinery.

As an example for a conclusion derived from both a BPEL and WS-Policy description, consider the following case. Let's assume a Web shop realized with internal and external Web Services composed and managed by a BPEL engine. After the submission of an order, we have to check the customer's credit card for validity depending on the credit card type (VISA, MasterCard etc.). We assume that credit card providers offer this functionality via Web Services. The corresponding BPEL process checkAccount thus invokes one of the provider's Web Services depending on the customer's credit card. The example below shows a snippet of the BPEL process definition.

Suppose now that the Web Service of one credit card provider, say MasterCard, only accepts authenticated invocations conforming to Kerberos or X509. It states such policies in a corresponding WS-Policy document like the one sketched below. The invocation will fail unless the developer ensures that the policies are met. That means the developer has to check the policies manually at development time or has to implement this functionality to react to policies at runtime.

As we may recognize from this small example, it would desirable to automate - or at least facilitate - such manual management tasks. Therefore, we postulate semantic management of Web Services to support developers and administrators who must cope with the complexity of Web Service integration and WS* descriptions. These two groups of users have the need to predict or observe how Web Services interact, (might) get into conflict, (might) behave, etc. It will be very useful for them to query a system for semantic management of Web Services that integrates aspects from multiple WS* descriptions. Ontologies are the obvious choice for the conceptual information integration because of their formal semantics. In addition, corresponding inference engines allow for reasoning and querying with semantic descriptions. [OLES 2005]

Coming back to our example, we could harvest available information for each of the services and construct two semantic service descriptions. For our internal service, we parse and integrate the workflow graph of the BPEL file. Existing OWL-DL ontologies (e.g. the Ontology of Plans cf. [GBCL 2004]) can be reused for that purpose. The ontology allows to associate invocations of external services with their respective semantic description. In the example, validateCardData is associated to the Mastercard service description via externallyInvokes. It suffices to state that there exists a policy for the Mastercard service in its respective description. That enables us to pose a simply query and infer that there is one externally invoked service that cannot be used without conforming to a policy. Thus, the developer is alerted to this fact and required to react (at development or deployment time). Two OWL-DL snippets of the service descriptions are depicted below:

It is possible to consider further advantages and more complex examples: checking for cycles in workflow, situations where large invocation sequences occur, or consideration of further aspects like the WSDL interface description. In further improvements it may be possible to consider formalization of the policy itself to support conformity checking. A target platform for integrating could be an application server, a workflow engine or a software IDE. In principle, ontology infrastructure has to be added and relevant information obtained from WS* descriptions.

The approach of semantic web service management is supplementary to the OASIS Web Services Distributed Management (WSDM) recommendation which defines how to manage Web services as resources and how to describe and access that manageability.

4.2 Ontologies as Formal Model Specifications and the Incorporation of Such Models in Semi-Formal Languages

Example C: A cooperative Semantic Web-based environment for semantic link among models

According to [Brown 2004], a formal underpinning for describing models facilitates meaningful integration and transformation among models, and is the basis for automation through tools. However, in existing MDA practice, semi-formal metamodels instead of formal specification languages are used as such formal underpinnings for describing models. An obvious reason is that, unlike UML, the industrial effort for standardising diagramatic notations, a single dominating formal specification language does not exist. Furthermore, different specifaction languages are designed for different purposes; e.g., B/VDM/Z^* are designed for modelling data and states, while CSP/CCS/π-calculus^** are designed for modelling behaviors and interactions.

To solve the problem, we can use ontologies as formal metamodels to describe various formal specification languages; moreover, the standard Semantic Web ontology language OWL can provide unified syntax. Semantic links among different models are explicitly specified by ontologies and form the basis of automation through tools in a Semantic Web-base enviroment. Based on these semantic links, various existing proposals of integrating formal specification languages can be supported in such enviroments.

[Wang 2004] briefly describes such an Semantic Web-based enviroment for semantic link among models, using DAML+OIL (instead of OWL). Examples of semantic links include assertions that Object-Z classes are (treated as) equivalent to CSP processes and that Object-Z operations are equivalent to CSP events.

Example D: Verifying Feature Models using OWL

Feature modeling plays an important role in domain engineering, which is a software reuse approach that focuses on a particular application domain. Examples of different domains are word processing, device driver for network adapters, inventory management systems, etc. Features are prominent and distinctive user visible characteristic of a system. Systems in a domain share common features and also differ in certain features. A feature model consists of a feature diagram and other associated information (such as rationale, constraints and dependency rules). However, the lack of a formal semantics and reasoning support of feature models has hindered the development of this area. Industrial experiences show that methods and tools that can support feature model analysis are badly appreciated.

A feature diagram provides a graphical tree-like notation that shows the hierarchical organization of features. The root of the tree represents a concept node, the instances of which are systems support the feature; the other nodes are sub-features of the root. There are four kinds of sub-features (see Figure 2).

• Mandatory: The feature must be included into the description of a concept instance.
• Optional: The feature may or may not be included into the description of a concept instance.
• Alternative: Exactly one feature from a set of features can be included into the description of a concept instance.
• Or: One or more features from a set of features can be included into the description of a concept instance.

A feature diagram itself cannot capture all the inter-dependencies among features. We have identified two additional relations among features: requires and excludes.

• Requires: The presence of some feature in a configuration requires the presence of some other features.
• Excludes: The presence of some feature excludes the presence of some other features.

We can use OWL to represent all the them (see the OWL codes below). For example, we can represent "F is a mandatory sub-feature of C" with the following OWL DL axiom

and represent "F is a optional sub-feature of C" with the following OWL DL axiom.

5. Next Steps

It is apparent that the descriptive advantages of an ontological view of the world, as outlined by the Semantic Web, are appealing to the field of Systems and Software Engineering. The challenge now is to move from research towards adoption, both in tooling and practice. This will undoubtedly be an arduous journey and may mean some significant concessions from both sides to produce what will almost certainly be a composite approach, combining elements from both existing Software Engineering practice and emerging Semantic Web fields.

Stronger semantics should, quite correctly, act as a catalyst in Software Engineering's advance. For real progress to be made, however, those involved must recognise tools like the UML to be evolving entities if their ultimate survival is to be assured. New ideas and significant challenges to current thinking are therefore right and proper and should be seriously debated in the appropriate manner. Such new ideas include Design by Contract and new variants of the UML in which its Object Constraint Language (OCL) is strengthened, or even replaced, by Semantic Web compliant languages or Simple Common Logic. Even more radical ideas exist relating to the role of abstraction or ‘meta-levels’ in the practice of modelling complex multidimensional systems. Furthermore, there are those of freely talk about the concept of meta-systems; huge semantically rich composites of systems in their own right, rigorously defined and interconnected, that will autonomically form and self-regulate for a recognised purpose(s). Much more work is needed in such areas, but their potential has already been recognised and the volume of related publication is most certainly on the rise. What remains now is to flesh out the detail behind such ideas through strong academic, standards and industrial liasons.

6. Acknowledgments

7. Normative References

[AXML 2003]

Active XML

[Brown 2004]

An Introduction to Model Driven Architecture - Part I: MDA and Today's Systems. Alan Brown, IBM.

[GBCL 2004]

Task Taxonomies for Knowledge Content. Aldo Gangemi, Stefano Borgo, Carola Catenacci, Jos Lehmann. Technical Report. Metokis Deliverable D07. June 2004.

[HNVV 2003]

Thesus: Organising Web Document Collections based on Semantics and Clustering, M. Halkidi, B. Nguyen, I. Varlamis and M. Vazirgiannis, Journal on Very Large Databases, Special Edition on the Semantic Web, November 2003.

[HZ 2004a]

Behavioral Models as Service Descriptions, 2nd International Conference on Service Oriented Computing , R.J. Hall and A. Zisman, ICSOC 2004, New York, November 2004.

[HZ 2004b]

OMML: A Behavioural Model Interchange Format . R.J. Hall and A. Zisman,12th IEEE International Conference in Requirements Engineering (RE'2004), Japan, September 2004.

[JMZZ 2005]

How Service Centric Systems Change the Requirements Process. S.V. Jones, N.A.M. Maiden, K. Zachos, and X. Zhu. In Proc. Of REFSQ'2005 Workshop, in conjunction with CAiSE 2005, Portugal, 2005.

[OESV 2004]

Developing and Managing Software Components in an Ontology-based Application Server. Daniel Oberle, Andreas Eberhart, Steffen Staab, Raphael Volz. In Hans-Arno Jacobsen, Middleware 2004, ACM/IFIP/USENIX 5th International Middleware Conference, Toronto, Ontario, Canada, volume 3231 of LNCS, pp. 459-478. Springer, 2004.

[OLES 2005]

Semantic Management of Web Services. Daniel Oberle, Steffen Lamparter, Andreas Eberhart, Steffen Staab. 3rd International Conference on Service Oriented Computing, ICSOC 2005, Amsterdam, December 2005.

[OSE 2005]

Towards Semantic Middleware for Web Application Development. Daniel Oberle, Steffen Staab, Andreas Eberhart. IEEE Distributed Systems Online. February 2005.

[Pietro-Diaz 1991]

Implementing Faceted Classification for Software Reuse. R. Pietro-Diaz, Communications of the ACM. Special issue on software engineering, 34(5):88, May 1991.

[Pohl 1999]

Requirements Traceability - Models of the Development World. Pohl, K.: In: M. Jarke; C. Rolland; A. Sutcliffe: The NATURE of Requiremtents Engineering. Shaker Verlag, Aachen 1999.

[SCGM 2005]

Learning Domain Ontologies for Web Service Descriptions: an Experiment in Bioinformatics. Sabou M., Wroe C., Goble C. and Mishne G. In Proceeedings of the 14th International World Wide Web Conference (WWW2005), Chiba, Japan, 10-14 May, 2005

[SEKE 2005]

17th International Conference on Software Engineering and Knowledge Engineering, http://www.ksi.edu/seke/seke05.html.

[SZK 2005]

A Service Discovery Framework for Service Centric Systems ,G. Spanoudakis, A. Zisman and A. Kozlenkov, IEEE International Conference on Services Computing (SCC 2005), Florida, USA, July 2005.

[Wang 2004]

Semantic Web and Formal Design Methods. Hai Wang. PhD thesis, National University of Singapore.

[WFSZP 2005]

A Semantic Web Approach to Feature Modeling and Verification. Hai Wang, LI Yuan Fang, Jing Sun, Hongyu Zhang and Jeff Z. Pan. In Proc. of the International Workshop on Semantic Web Enabled Software Engineering (SWESE). 2005.

[ZCDKS 2002]

Using Web Services to Interoperate Data at the FAO , A. Zisman, J. Chelsom, N. Dinsey, S. Katz and F. Servan,DC-2002 - Metadata for e-Communities: Supporting Diversity and Convergence, Florence, October 2002. (This paper was also presented in XML Conference and Exposition 2002, Baltimore, USA, December 2002).

[ZK 1999]

An Approach to Interoperation between Autonomous Database Systems ,A. Zisman and J. Kramer, Distributed Systems Engineering Journal 6, 1999, 135-148.

Appendix A. Informative References

[AJWH 2003]

Managing Software Engineering Knowledge. A.Aurum, R. Jeffery, C. Wohlin, M. Hadzic. Springer Verlag, 2003.

[Atkinson 2004]

On the Unification of MDA and Web-based Knowledge Representation Technologies. Colin Atkinson. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[CGL 2004]

Issues in Mapping Metamodels in the Ontology Definition Metamodel. Robert M. Colomb, Anna Gerber, Michael Lawley. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[ChKe 2004]

Major Influences on the Design of the ODM. Daniel T. Chang and Elisa Kendall. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[ChKe 2004b]

Metamodels for RDF Schema and OWL. Daniel T. Chang and Elisa Kendall. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[CrGe 2005]

Situation and Identity - A Generalisation of Inverse Functional Properties. Tom Croucher, University of Sunderland and Joe Geldart, University of Durham.

[EmHa 2005]

A Description Logic for Use as the ODM Core. Patrick Emery and Lewis Hart. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[EmHa 2005b]

Including Topic Maps in the Ontology Definition Meta-Model. Patrick Emery and Lewis Hart. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[FHKM 2004]

The Model Driven Semantic Web. David Frankel, Pat Hayes, Elisa Kendall and Deborah McGuinness. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[FHKM 2004b]

Simple Common Logic: A Constraint Language for the ODM . David Frankel, Pat Hayes, Elisa Kendall and Deborah McGuinness. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[GaTi 2004]

An MDA-Based Approach for Facilitating Adoption of Semantic Web Service Technology. Gerald C. Gannod and John T.E. Timm. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[GSNGW 2004]

Model Driven Middleware. Aniruddha Gokhale and Douglas C. Schmidt and Balachandran Natarajan and Jeff Gray and Nanbor Wang. In Q.H. Mahmoud (ed.): Middleware for Communications, Chapter 7, 163-187, Wiley, 2004

[Guha 2004]

Object Co-identification on the Semantic Web. R. V. Guha, IBM Research, Almaden.

[HS 2004]

Tools for design of composite Web services, by Richard Hull and Jianwen Su. SIGMOD 2004.

[Knublauch 2004]

Ontology Driven Software Development in the Context of the Semantic Web: An Example, Scenario with Protégé/OWL. Holger Knublauch. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004) Enabling Knowledge Representation and MDA® Technologies to Work Together.

[MSUW 2004]

MDA Distilled - Principles of Model-Driven Architectures. Stephan J. Mellor, Kendall Scott, Axel Uhl, Dirk Weese. Addison-Wesley, 2004.

[Oberle 2005]

Semantic Management of Middleware. Daniel Oberle. In A. Sheth, R. Jainn (eds.) The Semantic Web and Beyond, volume I, Springer 2005.

[PeSc 2003]

Curing the Web's Identity Crisis. S. Pepper and S. Schwab. Technical report, Ontopia, 2003.

[SKC 2001]

Handbook of Software Engineering and Knowledge Engineering, Vol. 1, S.K. Chang (editor),World Scientific, (ISBN: 981-02-4973-X), 2001

[Tetlow 2004]

SOA, Glial and the Autonomic Semantic Web Machine - Tools for Handling Complexity? Philip Tetlow, IBM, UK.

[Wallace 2004]

Experiences and Issues in Converting Data and Object Modeling Languages to OWL. Evan Wallace. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

[XLZMP* 2004]

SemanticWare: An EMF-Compatible RDF Infrastructure. Guo Tong Xie, Shixia Liu, Zhuo Zhang, Li Ma, Yue Pan, Li Zhang, Zhong Su, Li Qin Shen. 1st International Workshop on the Model-Driven Semantic Web (MDSW2004).

Appendix B. Related Topics

• Specification and design representation schemes, both formal and informal;
• Descriptions and models of the development process;
• Tools and environments to support software development;
• designers and implementors, and cognitive properties of representation schemes, programming and programming languages;
• Software development methods, analysis and validation;
• Requirements elicitation,
• Acquisition and formalization;
• Software quality and metrics;
• Software reuse and adaptation;
• Animation and execution of specifications and designs;
• Domain modelling and analysis;
• Software visualization;
• Software object management;
• Development of user interfaces;
• Group and team work in software engineering;
• Development of distributed, real-time, embedded and composite systems;
• Systems integration;
• Software evolution and maintenance;
• System testing;
• Reverse engineering and program understanding; documentation and program explanation;
• Software complexity and emergence;
• Network Theory
• Web Bow Tie Models

^* VDM is an abbreviation for "Vienna Development Method" specification language.
^**CSP is abbreviated for "Communication Sequential Processes", CCS is abbreviated for "Calculus for Communicating Systems".