Wright State University CORE Scholar

The Ohio Center of Excellence in Knowledge- Kno.e.sis Publications Enabled Computing (Kno.e.sis)

1-2003

Web Service: Been There, Done That?

Steffen Staab

Will van der Aalst

V. Richard Benjamins

Amit P. Sheth Wright State University - Main Campus, [email protected]

John A. Miller Wright State University - Main Campus

See next page for additional authors

Follow this and additional works at: https://corescholar.libraries.wright.edu/knoesis

Part of the Bioinformatics Commons, Communication Technology and New Media Commons, Databases and Information Systems Commons, OS and Networks Commons, and the Science and Technology Studies Commons

Repository Citation Staab, S., van der Aalst, W., Benjamins, V. R., Sheth, A. P., Miller, J. A., Bussler, C., Maedche, A., Fensel, D., & Gannon, D. (2003). : Been There, Done That?. IEEE Intelligent Systems, 18 (1), 72-85. https://corescholar.libraries.wright.edu/knoesis/189

This Article is brought to you for free and open access by the The Ohio Center of Excellence in Knowledge-Enabled Computing (Kno.e.sis) at CORE Scholar. It has been accepted for inclusion in Kno.e.sis Publications by an authorized administrator of CORE Scholar. For more information, please contact [email protected]. Authors Steffen Staab, Will van der Aalst, V. Richard Benjamins, Amit P. Sheth, John A. Miller, Chistoph Bussler, Alexander Maedche, Dieter Fensel, and Dennis Gannon

This article is available at CORE Scholar: https://corescholar.libraries.wright.edu/knoesis/189 Trends & Controversies

Steffen Staab University of Karlsruhe [email protected] Web Services: Been There, Done That?

What are Web services? The Stencil Group defines them as benefits might even seem to waste away, once we touch on “loosely coupled, reusable software components that semanti- the nitty-gritty details, because Web services per se do not cally encapsulate discrete functionality and are distributed and offer a solution to underlying problems such as programmatically accessible over standard protocols” (see www.stencilgroup.com/ideas_scope_200106wsdefined. • How can I effectively and efficiently distribute computa- html). Does this sound familiar? Long ago, people learned how tion efforts? to send software components and RPCs over HTTP (for an expla- • How can I make a software component really reusable? nation of RPC and other terms, see the “Glossary” sidebar). Mes- • How can I control and monitor these processes? sage exchange is increasingly based on XML, but so what? • How can I effortlessly integrate the components? Do you recognize hype when you see it? Let us measure fame as a Google count of Web pages that include a specified term. The following contributions delve into some of these issues. For example, measure the Semantic Web—now the topic of an Wil van der Aalst describes the pitfalls of workflow issues, many established department in IEEE Intelligent Systems—against of which we’ve encountered before. V. Richard Benjamins points Web services. (The Semantic Web department in this issue to the various research in the 1990s into structuring procedural also covers Web services.) The first Semantic Web language knowledge into problem-solving methods. Amit Sheth and John document (a working draft on the Resource Description Frame- A. Miller discuss how a low initial entry barrier and simple tech- work) was issued as part of the World Wide Web Consortium nology are balanced against the long-term goal of easy integra- metadata activity in October 1997 (www.w3.org/2001/sw). Using tion. Christoph Bussler, Alexander Maedche, and Dieter Fensel Google’s exact search, the term “Semantic Web” yields approxi- strongly support this idea—in particular, by including semantics mately 136,000 Web pages. in their Web Service Modeling Framework. Finally, Dennis Gan- Now consider “Web services.” The first UDDI document I non argues that Web services should not follow the well-worn could find dates from 1999, and the first W3C XML Protocol and unsuccessful paths of other distributed-object technology. Activity preceding the Web Services Activity started in Septem- Rather, they should build on new kinds of applications, such as ber 2000 (see www.w3.org/2002/ws).1 The corresponding grid enterprises, which are only possible using technologies such count delivers a whopping 3,210,000 pages. Compare these as Web services. Have we been there before? Let’s see. numbers with “artificial intelligence,” whose count ranks at —Steffen Staab 1,410,000, “Corba” (Common Object Request Broker Architec- ture) at 1,650,000, and “TCP” at 7,640,000. Even given the inevitable fallacies of these numbers, the Reference overall result is clear: Web services have received a lot of hype, 1. A. Preece and S. Decker, “Intelligent Web Services,” IEEE Intelligent the reasons for which are not easily determined. Some of their Systems, vol. 17, no. 1, Jan./Feb. 2002, pp. 15–17.

bar.) BPEL4WS allows for a mixture of block- and graph- Don’t Go with the Flow: Web Services structured process models, thus making the language Composition Standards Exposed expressive at the price of being complex. Although Wil van der Aalst, Eindhoven University of BPEL4WS is not such a bad proposal, it is remarkable Technology how much attention this standard has received, while more fundamental issues and problems such as semantics, The recently released BPEL4WS is said to combine the expressiveness, and adequacy have not received the atten- best standards for Web services composition, such as tion they deserve. IBM’s WSFL and Microsoft’s XLANG. (For an explana- Of course, having a standard is a good idea, but there are tion of these and other acronyms, see the “Glossary” side- too many of them—and most die before they mature. Con-

72 1094-7167/03/$17.00 © 2003 IEEE IEEE INTELLIGENT SYSTEMS Published by the IEEE Computer Society Glossary sider the growing list of acronyms: PDL, B2B Business to Business XPDL, BPSS, EDOC, BPML, WSDL, BPEL4WS Business Process Execution Language for Web Services WSCI, ebXML, BPEL4WS—and these are BPMI Business Process Management Initiative just some of the acronyms referring to vari- BPML Business Process Modeling Language ous standards in the domain. Another prob- BPSS Business Process Schema Specification lem is that these languages typically don’t Corba Common Object Request Broker Architecture have any clearly defined semantics. The ebXML Electronic Business Using Extensible Markup Language only way to overcome these problems is to EDI Electronic Data Interchange critically evaluate the so-called standards IBROW3 Intelligent Brokering Service for Knowledge-Component Reuse for Web services composition. In other on the World Wide Web words, don’t just go with the flow. IDL Interface definition language OMG Object Management Group Web services composition PSM Problem-solving method Two trends are coming together in e-busi- RPC Remote procedure call ness that are creating both opportunities and SOAP Simple Object Access Protocol pressures to automate business processes UDDI Universal Description Discovery and Integration across organizational boundaries. One is WfMC Workflow Management Coalition the technology push created by enabling WSCI Web Service Choreography Interface technologies taking XML-based standards WSDL Web Services Description Language and the Internet as a starting point. The WSFL Web Services Flow Language other trend is improving the efficiency of WSMF Web Services Modeling Framework processes from a business perspective. XLANG Web services for business process design After the dotcom crash, there was a XPDL XML Process Definition Language pressing need to use Internet technology’s potential by automating business processes across enterprise boundaries. Web services aim to exploit XML technology and the Internet by integrating applications than can Web services composition: be published, located, and invoked over the Business Process Execution Web. A typical example of a Web services Language for Web Services, Publication and discovery: application is the Galileo system, which XLANG, Universal Description Discovery connects more that 42,000 travel agency Web Services Flow Languages and Integration locations to 37 car rental companies, 47,000 hotels, and 350 tour operators. Service description layer: To truly integrate business processes Web Services Description Language across enterprise boundaries, merely sup- porting simple interaction using standard XML messaging layer: messages and protocols is insufficient. Simple Object Access Protocol Business interactions require long-running interactions that are driven by an explicit Transport layer: process model. This raises the need for Web HTTP, SMTP, FTP services composition languages such as BPEL4WS,1 WSFL,2 XLANG,3 WSCI, and BPML. These languages are also known as Figure 1. Overview of Web services technology. Web services flow languages, Web services execution languages, Web services orches- • An envelope that defines a framework saging layer such as SOAP. A WSDL docu- tration languages, and Web-enabled work- for describing what is in a message and ment defines services as collections of net- flow languages. Before discussing such lan- how to process it work endpoints, or ports. WSDL separates guages, I focus on the typical technology on • A set of encoding rules for expressing the abstract definition of endpoints and which they are building. instances of application-defined data messages from their concrete network Figure 1 shows the relation between Web types deployment or data format bindings. This services composition languages and other • A convention for representing RPCs and lets us reuse abstract definitions, including standards such as SOAP, WSDL, and UDDI. responses (SOAP can potentially be built messages, which are abstract descriptions SOAP is a protocol for exchanging informa- on top of any transport layer, such as an of the data being exchanged, and port tion in a decentralized, distributed environ- HTTP-based infrastructure) types, which are abstract collections of ment using typed message exchange and operations. The concrete protocol and data remote invocation. It is an XML-based pro- WSDL is an XML format for describing format specifications for a particular port tocol that consists of network services based on a standard mes- type constitute a reusable binding. Associ-

JANUARY/FEBRUARY 2003 computer.org/intelligent 73 ating a network address with a reusable Unlike XLANG, WSFL is not limited to ware, XPDL is consistent with BPEL4WS binding defines a port, and a collection of block structures and allows for directed but goes far beyond the standards for Web ports defines a service. graphs. The graphs can be nested but must services composition. Clearly, many people UDDI defines a set of services supporting be acyclic. Iteration is only supported working on standards for Web services com- the description and discovery of businesses, through exit conditions—that is, an activity position did not benefit from the experiences organizations, and other Web services pro- or subprocess iterates until its exit condition in the workflow domain. Therefore, “been viders; the Web services they make available; is met. The control flow part of WSFL is there, done that” comments are justified. and the technical interfaces we can use to almost identical to the workflow language However, workflow vendors clearly have access those services. Simply put, we can use that IBM’s MQ Series Workflow uses. This not adopted WfMC standards. Some systems UDDI to build “yellow pages” for Web ser- might be surprising, given that this work- can export to XPDL, but none can import vices. Consensus currently seems to exist on flow language is very different from most XPDL from another system and still produce the use of SOAP,WSDL, and UDDI, so I languages. For example, the “Death-Path meaningful results. This is partly because assume these standards will remain in place. Elimination” allows for the so-called “Syn- after work on workflow standards for more Web services composition languages chronizing merge pattern.” This way, rout- than a decade, still no consensus exists on the build directly on top of WSDL. A language ing is not restricted to explicit AND-joins workflow constructs that must be supported such as BPEL4WS both provides and uses and XOR-joins, as in most workflow prod- and their semantics. It is remarkable how one or more WSDL services. A WSDL ucts. This is a nice feature, but it is quite many different interpretations of a join con- service is composed of ports that provide exotic and most systems don’t support it. struct exist in contemporary workflow lan- operations. Each operation receives a mes- Although the correspondence between guages: “Wait for all (AND-join),” “Wait for sage (one way), receives and sends a mes- the WSFL standard and IBM’s workflow first and reset (XOR-join),” “Wait for first sage (request response), sends and receives product might surprise people not involved and block until all have arrived,” “Wait for all a message (solicit response), or sends a mes- in the standardization process, we can easily to come,” and so forth. sage (notification). WSDL services and the explain it by the fact that the same set of corresponding operations are glued together people (most notably, Frank Leymann) have Comparing BPEL4WS, XLANG, to provide composed services. To glue such defined both languages. We can make simi- WSFL, XPDL, and WFM services together, a process model must lar comments for XLANG and Microsoft’s products specify the order in which the operations BizTalk Orchestrator. XLANG is based on With respect to Web services composition execute. A Web services composition lan- Microsoft’s current middleware solution and languages, software vendors such as IBM, guage provides the means to specify such a therefore hardly qualifies as a “standard.” Microsoft, Sun, BEA, SAP, and Intalio have process model. Unfortunately, BPEL4WS, WSFL, and been the main drivers of development. This An important difference between WSDL XLANG are not the only recently proposed has resulted in an abundance of standards and a language such as BPEL4WS is re- standards. Sun, BEA, SAP, and Intalio have having overlapping functionality. When you vealed when we consider the states. WSDL introduced another candidate for Web ser- look at the standards in more detail, you can is essentially stateless because the lan- vices composition: WSCI. Intalio also initi- see clearly that they are often based on exist- guage is unaware of states between opera- ated the Business Process Management ing products, just as WSFL is almost a copy tions. The only state notion supported is Initiative (BPMI.org), which developed of the MQ Series Workflow language. Stan- the state between sending and receiving a BPML. OASIS and UN/CEFACT support dards that involve multiple software vendors message in a request-response or solicit- ebXML. Part of ebXML is BPSS, which is are often a compromise between competing response operation. Any technology sup- yet another standard similar in scope to viewpoints. Consequently, such standards porting a Web services composition lan- BPEL4WS, WSFL, XLANG, WSCI, and tend to be imprecise or unnecessarily com- guage will have to record states for processes BPML. The abundance of overlapping stan- plex. WfMC’s XPDL is an example of a that are more complex than a simple re- dards for Web services composition is over- standard that is imprecise, thereby letting quest response. Only by recording the state whelming. Some people refer to these com- vendors have their own interpretation of it can we determine what should be done, peting standards without clear added value (and making it useless). BPEL4WS joins thus enabling long-lived business transac- as WSAH—Web Services Acronym Hell. viewpoints from both WSFL and XLANG, tions. This has triggered the development Outside the Web services domain, other thus making the language complex. of languages such as BPEL4WS, WSFL, initiatives are attempting to standardize Given these observations, looking for XLANG, WSCI, and BPML. the specification of executable business objective measures for comparing Web ser- processes. Most notable is the initiative of vices composition languages is useful. For Overview of so-called the Workflow Management Coalition. Since the control-flow aspect of such languages, we standards 1993, the WfMC has been active in standard- can use some of the results from workflow The BPEL4WS specification1 builds on izing both a workflow process definition lan- research.5 One way to compare standards WSFL2 and XLANG.3 XLANG is a block- guage and the interfaces between various such as BPEL4WS, XLANG, and WSFL is structured language with basic control flow workflow components. In August 2002, to use the set of workflow patterns available structures such as sequence, switch (for condi- the WfMC released XPDL4 to support the from www.tm.tue.nl/it/research/patterns.6 tional routing), while (for looping), all (for par- exchange of workflow specifications between Each of these patterns corresponds to a rout- allel routing), and pick (for race conditions different workflow products. According to ing construct often required during workflow based on timing or external triggers). Jon Pyke, WfMC Chair and CTO of Staff- design.7,8 The whole set of patterns has been

74 computer.org/intelligent IEEE INTELLIGENT SYSTEMS Table 1. Comparison of BPEL4WS, XLANG, WSFL, XPDL, and four workflow products. MQ Series Panagon Pattern BPEL4WS XLANG WSFL XPDL Staffware Workflow eProcess FLOWer 1 Sequence + + ++++++ 2 Parallel split + + ++++++ 3 Synchronization + + ++++++ 4 Exclusive choice + + ++++++ 5 Simple merge + + ++++++ 6 Multichoice + – ++–++– 7 Synchronizing merge + – +––++– 8 Multimerge – – –––––+/– 9 Discriminator – – –––––+/– 10 Arbitrary cycles – – –++–+/–– 11 Implicit termination + – +++++– 12 Multiple instances without + + +–––++ synchronization 13 Multiple instances with + + ++++++ a priori design time knowledge 14 Multiple instances with – – –––––+ a priori runtime knowledge 15 Multiple instances without – – –––––+ a priori runtime knowledge 16 Deferred choice + + –––––+/– 17 Interleaved parallel routing +/– – –––––+/– 18 Milestone – – –––––+/– 19 Cancel activity + + +–+––+/– 20 Cancel case + + +–––++/–

used to evaluate and compare about 20 work- Support for such features is rated “+/–”. flow management systems. Without going into details, we can make sev- The Web site of BPMI.org, one of the Table 1 compares three Web services com- eral observations. organizations proposing a Web services position languages, XPDL, and four concrete First, BPEL4WS is indeed a combination composition standard, states that workflow management systems.6 The first of XLANG and WSFL when it comes to sup- five patterns correspond to the basic routing porting the patterns. Second, WSFL and MQ BPMI.org defines open specifications such as constructs you can find in any language. The Series Workflow are indeed identical when it the Business Process Modeling Language (BPML) and the Business Process Query Lan- other patterns refer to more advanced con- comes to process specification. Third, XPDL guage (BPQL) that will enable the standards- structs that most standards and products don’t seems less expressive than BPEL4WS. (In based management of e-Business processes support. Every “+” refers to direct support— a way, we can view XPDL as the greatest with forthcoming Business Process Manage- meaning the language has a construct that common denominator of existing workflow ment Systems (BPMS), in much the same way SQL enabled the standards-based management directly supports the pattern. A “–” refers to languages rather than the least common of business data with off-the-shelf Database no direct support. This does not mean that multiple.) Finally, Web services composi- Management Systems (DBMS). realizing the pattern through some work- tion languages and workflow management around is impossible. (For example, any of systems have relevant differences when it The goal of obtaining standards similar to the constructs can be realized using a stan- comes to supporting routing constructs. Of SQL for Web services is ambitious. dard programming language, but this does the four workflow management systems As history shows, such standards do not not imply that there is direct support for all listed, only FLOWer (a workflow and case- originate from vendors pushing their own workflow patterns.) Sometimes there is a handling system) is block structured like products. Recall that the Entity-Relationship feature that only partially supports a pattern, XLANG. The other three systems (Staffware, model by Peter Chen and the Relational such as a construct that imposes certain MQ Series, and eProcess) are graph-based Model by Eduard Codd enabled languages restrictions on the structure of the process. like WSFL and XPDL. such as SQL. Although there are well-estab-

JANUARY/FEBRUARY 2003 computer.org/intelligent 75 lished process-modeling techniques combin- the Web that people can access through a Web Services Solve Problems, ing expressiveness, simplicity, and formal standard protocol and execute remotely. and Problem-Solving Methods semantics (such as Petri nets and process Furthermore, when used together, Web ser- Provide Services algebras), the software industry has chosen vices can deliver a complex functionality. to ignore these techniques. So, the world is V. Richard Benjamins, Intelligent Software The three major ingredients for Web ser- confronted with too many standards, mainly Components, S.A. vices are a description of the service, which driven by concrete products or commercial uses the WSDL; the XML protocol SOAP, interests. The only way to stop this is to In the fall of 1996, on a flight home from through which users access the services; and ignore standardization proposals that are the Knowledge Acquisition Workshop, some a UDDI directory, where you can find what not using well-established process-modeling colleagues and I were outlining a new re- Web services are offered and where. (For an techniques. This will force vendors to address search project (an FP4 Esprit project) that explanation of these and other acronyms, see the real problems rather than create new ones. aimed to link knowledge technology to the the “Glossary” sidebar.) With those ingredi- Web. The Web was gaining importance, and ents in place, anybody can use a Web service, knowledge technology was looking to join regardless of the programming language in Acknowledgments the bandwagon. We were talking about the which the service was originally defined. IBROW3 project, an Intelligent Brokering Web services thus tackle the problem I thank Arthur ter Hofstede, Bartek Kiepu- Service for Knowledge-Component Reuse of heterogeneous sources and make them szewski, Marlon Dumas, and Petia Wohed for contributing to the results mentioned in this essay. on the World Wide Web (see www.swi.psy. interoperable. Technologies such as Corba, uva.nl/projects/IBROW3/home.html). RPC, and EDI had the same objectives, but The project envisioned reusable heteroge- those solutions needed their proper (that is, neous components located at repositories on specific or dedicated) infrastructure, with References the Web, which a broker (or software agent) the corresponding costs and implementation would configure into a distributed program efforts. We can thus explain the success of 1. F. Curbera et al., Business Process Execution Language for Web Services (Version to solve a particular problem. The broker Web services by viewing them as a technol- 1.0), IBM, July 2002, www-106.ibm.com/ therefore had to know about the problem ogy based on maximal decoupling (and thus developerworks/webservices/library/ws-bpel. and the individual components’ capabilities. maximal reusability) available over an exist- The brokering process’s result would be a ing economic infrastructure (the Internet). 2. F. Leymann, Web Services Flow Language distributed system that could, for example, Their power is not so much in their technol- (WSFL 1.0), IBM, May 2001, www-3.ibm. com/software/solutions/webservices/pdf/ classify edible and poisonous mushrooms ogy (the idea of RPC is nothing new) but WSFL.pdf using a public mushroom database and a rather that they offer a Web-native XML- classifier, both available on the Web. based solution. So, we can rapidly design, 3. S. Thatte, XLANG: Web Services for Business The components that the IBROW project implement, and deploy Web services. Process Design, Microsoft, Redmond, Wash., 2001, www.gotdotnet.com/team/xml_wsspecs/ considered were problem-solving methods These characteristics make Web services xlang-c/default.htm. and ontologies. A PSM is a generic descrip- an interesting candidate for integrating het- tion of a reasoning process, independent of erogeneous enterprise applications. Organi- 4. Workflow Process Definition Interface—XML the domain to which it is applied. A PSM zations spend a lot of money making their Process Definition Language (XPDL), might solve classification tasks, for exam- internal systems interoperate with their WFMC-TC-1025, Version 1.0 Beta, Work- flow Management Coalition, 2002, www. ple, by classifying mushrooms, diseases, or partners’ and providers’ enterprise applica- wfmc.org/standards/docs/TC-1025_ financial transactions. Before we can apply tions—and sometimes even with their inter- 10_xpdl_102502.pdf. a PSM to a particular domain, we have to nal applications. Web services offer an inter- check its assumptions, which describe the esting alternative because they wrap existing 5. W.M.P. van der Aalst and K.M. van Hee, Workflow Management: Models, Methods, domain model’s required characteristics. (legacy) systems and let them communicate and Systems, MIT Press, Cambridge, Mass., For example, the establish-and-refine through SOAP over existing infrastructures 2002. PSM, which B. Chandrasekaran introduced (extranet or intranet). This is probably why in the late 1980s, requires hierarchically orga- research analysts at companies such as the 6. W.M.P. van der Aalst et al., Workflow Pat- nizing the domain knowledge. The PSM’s Gartner Group and McKinsey are monitor- terns, QUT tech. report FIT-TR-2002-02, Queensland Univ. of Technology, Brisbane, assumptions capture the interaction between ing developments in this area. Australia, 2002; www.tm.tue.nl/it/research/ the PSM and domain knowledge explicitly patterns (also to appear in Distributed and such that reusing each of them in different sit- The Semantic Web Parallel Databses). uations becomes less troublesome. One main Having said all these nice words, I must 7. W.M.P van der Aalst et al., Pattern-Based motivation behind research on PSMs and admit that current Web services technology Analysis of BPML (and WSCI), QUT tech. ontologies in the 1990s was to increasingly solves only part of the problem. Distributed report FIT-TR-2002-05, Queensland Univ. of reuse and share code rather than develop code systems developers now have a common Technology, Brisbane, Australia, 2002. from scratch for each new problem. platform for rapidly developing or integrat- ing complex software by configuring exist- 8. P. Wohed et al., Pattern-Based Analysis of BPEL4WS, QUT tech. report FIT-TR-2002- Web services ing services. Through UDDI, they can look 04, Queensland Univ. of Technology, Bris- We also see this motivation in Web up the basic services they need expressed bane, Australia, 2002. services—pieces of software available on in WSDL, which they then can access by

76 computer.org/intelligent IEEE INTELLIGENT SYSTEMS Table 2. Resemblance between IBROW, on the one hand, and the Semantic Enabled Web Services project and the Semantic Web in general, on the other hand. IBROW3 (Esprit FP4); IBROW (IST, FP5) The SWWS project, Semantic Web Comments The Web is changing the nature of software Web services are orchestrated into In IBROW, problem-solving methods development to a distributive plug-and-play process. complex services. (PSMs) and ontologies were the The components concerned are problem-solving components being configured, versus methods (generic algorithms) and ontologies. Web services today. PSMs and ontologies Web services PSMs and ontologies, when connected to the Web, deliver services. PSMs deliver reasoning services, while ontologies enable intelligent query answering. IBROW integrates research on heterogeneous Semantics and enterprise application Web services can wrap heterogeneous sources databases, interoperability, and Web technology integration to make them interoperable. with knowledge-system technology and ontologies. Semantic agreement is currently hard coded in the wrapper. The project aims at providing intelligent From information overload to task The current paradigm of information retrieval reasoning services on the Web as opposed delegation causes much information overload. to the more common information services. We need to delegate tasks to agents such that people focus on the interesting information. The broker needs to reason about Annotation of services, WSMF, UPML is used to characterize ontologies and characteristics of PSMs, for example about DAML-S PSMs in terms of their competence (capabilities) their competence: Unified Problem-Solving and requirements. Method Description Language (UPML)

programming SOAP. However, a person semanticweb.org). They would need a and Web services. However, IBROW has must do all this work, and without much semantic characterization of the services played a crucial role in identifying relevant support. As more Web services become they offer as well as a middleware for service issues and major challenges and in bring- available, information overload will make discovery and orchestration, to hide complex ing together the key people and groups that finding the service you need difficult. technology from users. A user can thus con- nowadays lead the European effort in To overcome these problems, we must sume a high-level functionality without Semantic Web services research. first recognize that current Web services noticing that a complex process is going on technology is basically a syntactical solu- under the hood. This is a main objective of tion and that the semantic part is still lack- the Semantic Enabled Web Services project ing. A Web service is described in WSDL, (SWWS, swws.semanticweb.org)—to The notion underlying Web services is outlining what input the service expects develop such intelligent middleware along nothing new. In this sense, it is “been there, and what output it returns. To exploit their with a language to semantically express done that.” However, the infrastructure we potential (beyond enterprise application Web services capabilities. currently count on is much more advanced integration), Web services must be able to If we look back at the original IBROW than several years ago. In IBROW, we spent orchestrate themselves into more complex project (www.ibrow.org), we see that it much effort making the underlying infra- services. Thus, we need ways to combine resembles the SWWS project in particular structure work and creating the (heteroge- individual Web services into a distributed, and the Semantic Web in general. Table 2 neous) content. This hampered us from higher-level service. illustrates this, showing some phrases and focusing completely on the real issues, such The Web Service Flow Language, which terms from the IBROW project and the cor- as semantic interoperability. In the current can express the sequencing of individual responding terms used in current Semantic technological situation, coupled with a strong services, is taking the first steps. WSFL lets Web research projects. business pull (PSMs were more of a tech- the user decide which Web services to We are now entering FP6, the Sixth nology push from the knowledge-engineer- combine and in what order. However, we European Research Framework Program. ing community), we soon will have Web- still need a framework that semantically In addition, the IBROW projects (one in accessible methods that solve real business describes services, such that software FP4 in 1998 and the current one in FP5 problems through knowledge-intensive or agents can locate, identify, and combine from 2000–2003) are about to finish. Since intelligent Web services. the services. these projects began, many things have This is where the Semantic Web comes changed. The Web has drastically changed in. It semantically describes (annotates) how we communicate, do business, and content available on the Web such that soft- obtain information (where Google is the Acknowledgments ware agents can understand and process the ruling champion). A revolution of new I would like to acknowledge the IST projects information. In this manner, Web services technologies has overtaken the IBROW IBROW (IST-1999-19005), Esperonto (IST- should form part of the Semantic Web (www. project, such as Semantic Web technology 2001-34373), and SWWS (IST-2001-37134).

JANUARY/FEBRUARY 2003 computer.org/intelligent 77 starting pilot projects—something large see Web services as a natural follow-up to Web Services: Technical businesses always do before adopting any Electronic Data Interchange and ebXML, Evolution yet Practical new technologies—difficult. which already have significant commercial Revolution? However, like the Web, the basic compo- success.2 Amit Sheth and John A. Miller, nents of Web services are relatively simple, University of Georgia and the learning curve is low. XML is Loosely coupled architectures already known to the likely adopters and The fourth consideration is that Web ser- Web services are here to stay. Both busi- accepted as the data exchange standard by vices are a good choice for loosely coupled ness and technical considerations will pro- enterprises. Standards such as SOAP and architectures.3 This means not allowing vide them with the staying power and tail- WSDL are simple to use and learn. (For an object references or requiring statefulness. wind to survive the hype. Despite their lack explanation of these and other acronyms, Although the Corba architecture was more of a major technical breakthrough, Web see the “Glossary” sidebar.) Although suitable for intra-enterprise environments, services will initially succeed owing to the UDDI as a whole (with all the details of the technical features and choices of Web right confluence of evolutionary tech- white, yellow, and green pages) is not sim- services make them more reusable and thus nological choices, low barriers and entry ple, the basic parts needed to get started are more appropriate for interenterprise and costs, and their strong standards-based relatively easy to learn and use. Also, Web global environments. approach. In the long run, however, they services can be developed with essentially must be empowered by semantics and cou- no initial technology cost. Furthermore, Languages pled with Semantic Web technologies to just as the Web is used not only for publish- Finally, Web services do not use languages fulfill broader, longer-term promises. ing or sharing data but also as an applica- that look like programming languages (that tion infrastructure, the complexity of Web is, there’s no interface definition language). Evolution in the right services can be incrementally increased as This is what the industry is trying to accom- direction: Reasons for likely more functionality is added. plish with Web services. success Progress in software componentry has The hype Incremental advances in been ongoing. The Web is becoming more The second consideration is the hype sur- software componentary than just a collection of documents; appli- rounding Web services, and the industry’s The introduction of the object-oriented cations and services are coming to the fore- ability to sustain interest over time. Consider programming paradigm in the 1980s front. History shows that incremental tech- XML. Although it has been significantly promised to make reusable software compo- nological progress can produce dramatic hyped—for example, it only addresses the nents a reality. Although there were certainly effects, such as the advancement from FTP syntactic interoperability issues, not seman- some gains in software reuse and program- to Archie to Veronica to the Web. Web ser- tic ones—it has succeeded. This is because ming productivity, the change was not sub- vices might have a similar fate. it solves a limited yet important and well- stantial. The 1990s saw further attempts to Several technical and technology-driven understood business problem, and there is a increase software reuse, focused on the idea business advantages present themselves, groundswell of commercialization efforts of software components and extended to dis- boding well for a quicker and more success- around the use of XML technology. Some tributed components. A notable effort in this ful adoption of Web services. enterprise software such as Enterprise area was Corba. It let components on differ- Resource Planning quickly adopted XML, ent machines, using different operating sys- Low initial complexity broadening its adoption by businesses. In tems and even different languages, commu- The first consideration is the low initial comparison, corresponding commercial nicate. From one viewpoint, Web services complexity (although the complexity of activities did not support the hype of expert represent an evolution of Corba. We contend fully mature technology might be substan- systems and other artificial intelligence that Web services are more convenient than tially higher), with low entry barriers and technologies. Corba and that eventually incremental costs. Although the Web now consists of In the case of Web services, various soft- improvement will lead to widespread use. many components and technologies—sev- ware segments, including application servers One difficulty in using distributed tech- eral of which were added as the Web gained and Enterprise Content Management,1 have nology such as Corba is dealing with lan- wider acceptance and matured—it indeed found them worthy enough to adopt. This guages. Corba tried to eliminate this problem started with a remarkably small, simple gives the hype of Web services an ability to by moving the language into the back- core technology. That simple core—free sustain with follow-through development ground by defining an IDL. Although this is availability and low- or no-cost initial adop- and adoption. a good idea, an IDL looks much like C++, tion—played an important role in its growth and developers must also understand lan- and acceptance. Standards guage bindings to use Corba. If there was a Unlike the Web, the Corba architecture The third consideration is the use of more way to replace language-like IDLs with was complex from the start (considering its widely accepted standards. XML already higher-level specification, surely communi- various services and facilities) and required enjoys wide adoption in enterprises. Stan- cation could be simplified using meaning- seasoned developers to use it. Also, most dardization efforts are well aligned and are ful messages. Fortunately, Web services, businesses had to rely on object request structured to act much faster than what hap- defined as “self contained, self-describing brokers that were not free, which made pened with Corba. Some in industry also modular applications that can be published,

78 computer.org/intelligent IEEE INTELLIGENT SYSTEMS located, and invoked across the Web,”4 this area is ebXML, which lets enterprises involve substantial further research and engi- automatically get the Web-wide scope conduct business over the Internet using an neering that consider both functional and rather than primarily the enterprise-wide open XML-based infrastructure. operational perspectives. The Large Scale scope (at least initially) for Corba. In the In contrast to these commercial XML- Distributed Information Systems Lab’s near term, we see the use of a few well-tar- based standards, researchers are developing a Meteor project is one of the projects looking geted Web services within and across enter- unique Web service markup language called at developing semantics-based solutions to prises for well-targeted B2B applications DAML-S.10 According to the group of re- QoS, discovery, and composition issues in and as integration points between enterprise searchers working on DAML-S, it supplies the context of Web processes (see http:// software. Examples include integrating a Web service providers with a core set of lsdis.cs.uga.edu/proj/meteor/SWP.htm). document management system and portal markup language constructs for describing management system within an enterprise, or the properties and capabilities of their Web interenterprise interfaces between different services in unambiguous, computer-inter- References parts of a supply chain or ERP solution, pretable form. DAML-S markup of Web ser- which already uses XML for data exchange. vices will facilitate the automation of Web 1. R. Perry and R. Lancaster, Enterprise Con- service tasks including automated Web ser- tent Management: Expected Revolution or Vendor Positioning, The Yankee Group, Intermediate to long term: vice discovery, execution, interoperation, Boston, 2002. Processes, QoS, security, and composition and execution monitoring.10 semantics If Web services are to provide not only 2. A. Kotok, “ The E-Business Continuum: Web For the intermediate term, we predict an enterprise-wide but also a global archi- Services, ebXML and EDI,” WebServices. Org, June 2002, www.webservices.org/index. distinct advances in tecture for application interoperability and php/article/articleview/479/1/24. integration, we will need to enhance Web • Web processes services and processes with semantics. 3. D. Austin et al., “Web Services Architecture • Technical support for quality of service Many applications will require a Web Requirements,” World Wide Web Consor- • Technical support for security process created from composing several tium, 2002, www.w3c.org/TR/wsa-reqs. Web services. In the intermediate term, we 4. D. Tidwell, “Web Services—The Web’s Next Although we expect researchers to address can create enterprise-scale Web processes Revolution,” IBM tutorial, 29 Nov. 2000, QoS and security in the intermediate term, we (or service compositions) by adopting www-105.ibm.com/developerworks/ expect support for Web processes to take workflow management technology. How- education.nsf/webservices-onlinecourse-bytitle/ BA84142372686CFB862569A400601C18? longer, because initial research is just starting ever, to achieve Web services’ full Web- OpenDocument. on these topics.5 Because Web services are wide and global potential, semantics holds components, if individual services are limited the trump card. Using semantics involves 5. J. Cardoso and A. Sheth, Semantic e-Work- in their capability, we can compose existing describing resources with formal, machine- flow Composition, tech. report, Large Scale Web services to create new functionality in readable description. Resources in this case Distributed Information Systems Lab, Dept. of Computer Science, Univ. of Georgia, the form of Web processes. Web service com- are Web services, and given that they wrap Athens, Ga., 2002. position is the ability to take existing services applications, they must be described both (or building blocks) and combine them to functionally and operationally. 6. G. Piccinelli, Service Provision and Composi- form new services.6 In carrying out this com- As already recognized by DAML-S,10 tion in Virtual Business Communities, tech. report HPL-1999-84, Hewlett-Packard, Palo position task, we should be concerned about Web Service Modeling Framework ini- Alto, Calif., 1999; www.hpl.hp.com/techreports/ 11 the efficiency and the QoS that the composed tiatives, the Work Group on Web Services 1999/HPL-1999-84.html. process will exhibit when it executes. This at a recent Semantic Web workshop,12 and task of composing services to create efficient others, semantics can play a critical role in 7. F. Leymann, “Web Service Flow Language Web processes is analogous to designing a developing Semantic Web services. This can (WSFL) 1.0,” IBM,Armonk, N.Y.,2001, www- 4.ibm.com/software/solutions/webservices/pdf/ workflow. lead to better discovery of Web services for WSFL.pdf. Web service composition is an active area reuse in a global, Web-scale environment of research, with academic and industrial that is not limited to one enterprise or a static 8. S. Thatte, “XLANG: Web Services for Busi- research groups proposing many languages. collection of enterprises and where current ness Process Design,” 2001, www.gotdotnet. com/team/xml_wsspecs/xlang-c/default.htm. IBM’s WSFL7 and Microsoft’s XLANG8 syntax-based techniques do not work. Strong were two of the earliest languages to define support for is also a 9. F. Curbera et al., “Business Process Execu- standards for Web services composition. prerequisite to developing Web processes13,14 tion Language for Web Services,” 2002, www- Both extended WSDL, W3C’s standard lan- and addressing various issues of composi- 106.ibm.com/developerworks/webservices/ guage used to describe the syntactic aspects tion, interoperability, and execution.15,16 library/ws-bpel. 9 of a Web service. BPEL4WS is a recently There is demonstrable progress in develop- 10. A. Ankolekar et al., “DAML-S: Web Service proposed specification that represents the ing semantics-based solutions when dealing Description for the Semantic Web,” Proc. Int’l merging of WSFL and XLANG. It combines with data, such as to achieve better interoper- Semantic Web Conf., Springer-Verlag, Berlin/ WSFL’s graph-oriented process representa- ability and integration of information re- Heidelberg, 2002, pp. 348–363. tion and XLANG’s structural construct- sources. However, the challenges of dealing 11. D. Fensel and C. Bussler, “The Web Service based processes into a unified standard for with applications that are enabled by Web Modeling Framework WSMF,” 2002, http:// Web services composition. Another effort in services are even more difficult. They will informatik.uibk.ac.at/users/c70385/wese.

JANUARY/FEBRUARY 2003 computer.org/intelligent 79 12. A. Sheth and R. Meersman, “Amicalola more than happy to state). For the follow- problems, clearly showing the vast spectrum Report: Database and Information Systems ing reasons, we see Web services as being that Web services are said to solve. Research Challenges and Opportunities in at the crossroads: First, SOAP was originally the acronym Semantic Web and Enterprises,” ACM SIG- MOD Record, vol. 31, no. 4, Dec. 2002. for the Simple Object Access Protocol. • They have been touted as the silver bul- Nomen est omen (the term reveals its intent): 13. S. Narayanan and S. Mcllraith, “Simulation, let for a (too) wide array of integration the underlying paradigm follows the client- Verification and Automated Composition of problems, currently being addressed by server model. The SOAP specification in Web Services,” Proc. 11th Int’l World Wide Web Conf., W3C, 2002, pp. 77–88. complex integration solutions such as conjunction with WSDL (see www.w3.org/ B2B integration technology1 as well as TR/wsdl) clearly enables the definition of the 14. J. Cardoso et al., Modeling Quality of Service simple technology such as Java for external interface of a server in a particular for Workflows and Web Service Processes, remote server invocation. way but not of its clients, violating the gen- tech. report, Large Scale Distributed Infor- • Many competing, conflicting, and eral peer-to-peer approach that most integra- mation Systems Lab, Dept. of Computer Sci- ence, Univ. of Georgia, Athens, Ga., 2002. overlapping standards exist that address tion solutions require. The only pattern this Web service functionality (see www. supports is the one-way invocation with and 15. J. Cardoso et al., “Semantic Web Services and -open.org/cover/sgml-xml.html). without results between the defined roles of Processes: Semantic Composition and Qual- • Many research projects are repackaging client and server. ity of Service,” tutorial at Federated Confer- ences (CooPIS, DOA, ODBASE), 2002; http:// existing work as Web services work. In this sense, SOAP, in conjunction with lsdis.cs.uga.edu/lib/presentations/SWSP- • Many start-up companies, including Cape WSDL, clearly implements yet another tutorial-resource.htm. Clear (www.capeclear.com), Actional RPC model. This is also explicitly called (www.actional.com), and Intalio (www. out in the SOAP specification. Because it 16. S. Chadrasekaran et al., Composition Perfor- intalio.com), provide technology imple- is based on the ubiquitous HTTP protocol mance Analysis and Simulation of Web Ser- vices, tech. report, Large Scale Distributed menting Web services technology. in conjunction with XML as the message Information Systems Lab, Dept. of Computer • Major infrastructure companies (such syntax, computer system boundaries are Science, Univ. of Georgia, Athens, Ga., 2002. as BEA, IBM, Microsoft, and Oracle) easy to overcome and the notion of the already provide implementations of Web “better RPC” or even “better distributed- service technology. object-management technology” was born, • Broad journalistic coverage of Web ser- leading to the wide awareness of Web ser- Web Services: Quo Vadis? vices exists in the trade press, such as vices in the developer community. eAI Journal (www.eaijournal.com) and Second, because SOAP messages can be Christoph Bussler, Oracle Corporation ebizQ (www.ebizq.net). transported over HTTP to implement the Alexander Maedche, FZI Research • Web services have no underlying real runtime invocation, bridging computer sys- Center for Information Technologies conceptual integration model and are tem boundaries is clearly easy owing to the at the University of Karlsruhe only defined as an implementation tech- commodity of the HTTP protocol. If only Dieter Fensel, Leopold Franzens nology, such as SOAP (see www.w3. packaged applications like Enterprise Universität Innsbruck org/TR/soap12-part1/ and www.w3.org/ Resource Management systems, such as TR/soap12-part2). Oracle or SAP (www.sap.com), were to Web services have been touted for some • No one has analyzed available and proven expose WSDL interfaces, then integrating time now as the technology-based silver solutions such as EDI translation technol- them with Web services would be easy, bullet solution for many significant integra- ogy2 or reported on lessons learned from fast, cheap, manageable, and so on (better tion problems in information technology. deployments that could advance Web ser- along all possible dimensions). These integration problems are of so differ- vices technology functionality (more In such a scenario, Web services would ent a nature due to the specific required than 300,000 Electronic Data Interchange be the “better Enterprise Application Inte- functionality that we can’t avoid asking deployments exist worldwide—see www. gration solution,” overcoming the current right now, “Web services: Quo vadis?” disa.org/x12org/about/faqs.cfm). costly integration technology deployments. Here, we illustrate the immense discrep- This is because the whole integration world ancy between the magnitude of the integra- This situation cannot continue indefinitely. would be a nice homogeneous sea of SOAP tion problems and the simplistic functional- We must consolidate the Web services space, messages implementing WSDL operation ity that Web services provide. We propose focusing Web services technology on a invocations. the WSMF as the future direction of Web well-defined and well-manageable set of Finally, because we can easily exchange services so that they can cope even with the integration problems. This includes the SOAP messages over the Internet, the pro- most real complex integration problems. consolidation of the multitude of Web ser- posal of using Web services as the only and (For an explanation of this and other vice standard proposals. Current Web ser- superior way to implement B2B interactions acronyms, see the “Glossary” sidebar.) vices expectations are too high and too var- between companies (interenterprise B2B ied for one technology to solve all the communication) is not far off. The vision in Web services at the crossroads given integration problems. this case is to very rapidly replace EDI, Let’s spotlight the state of the art of SWIFT (Society for Worldwide Interbank Web services and the situation of the The silver bullet problems Financial Telecommuication, www.swift. “movement” (or “hype,” as some might be Here we discuss some of the silver bullet com), and other existing B2B standards

80 computer.org/intelligent IEEE INTELLIGENT SYSTEMS (some of which have existed for over 30 • The better RPC overlap, compete, or disagree with each years). • The better plumbing for application- other significantly. Another big “however” These three example integration problems to-application or B2B integration is that these efforts haven’t really considered alone span an impressive array of serious • The better intelligent agent platform semantic unification at all. Only a few con- requirements for an integration solution that • All of the above tributions (standards and research) highlight Web services would have to address: the need to solve the semantic mismatch There seems to be the implicit agreement problem that inherently exists in interenter- • Efficiency: To scale on an industrial in the community that Web services should prise integration and that no real integration basis, Web services execution must be be “the” new and “the” better technology for solution can do without. very efficient. integration (especially for B2B integration). So, we need a comprehensive conceptual- • Expressiveness: B2B interactions in sup- For instance, together, the vast array of integration model that combines all aspects ply chain scenarios are complex, requir- industrial-standard proposals appears to pro- of a solution for the interenterprise as well ing an expressive set of supported inte- vide an almost complete set of concepts as intra-enterprise integration problem. gration concepts. required for implementing complex B2B From that we can derive a comprehensive • Security: Interactions within as well as integration. Examples of such integration modeling and execution framework that can across enterprises must be secured to include supply chain management, health address even the most complex integration prevent security attacks of all types, and care organization integration, or financial problems completely. nonrepudiation must be provided for transaction processing. Such a framework must provide a holis- reliable record keeping. tic and general conceptual-integration • Reliability: Remote and distributed com- model that characterizes all integration munication must be reliable, and mes- problems. From that we can derive a set of sages must be sent exactly once to ensure expressive modeling constructs that model dependable interactions. Web services that can address Web services, their invocation, and other • Manageability: Interenterprise commu- requirements such as composition and nication changes frequently, requiring complex integration problems security, addressing even the most complex easily manageable technology. integration scenarios. Furthermore, such a framework must provide an expressive and These requirements pose a high demand require conceptual modeling scalable mediation approach that can tackle on a technology that addresses their the semantic unification problem through implementation. based on a well-defined set of powerful mediation. On the other side, we have the current state of Web services technology. The integration concepts. The future of Web services: following standards currently define Web WSMF services: SOAP, WSDL, and UDDI (see Web services that can address complex www.uddi.org). This standards triple, even if integration problems require conceptual fully implemented, is far from addressing modeling based on a well-defined set of even a subset of the requirements we’ve Zapthink (www.zapthink.com/reports/ integration concepts. This set of concepts listed, because it is based on the primitive poster.html#) has a poster that shows 135 of must be able to represent all relevant aspects notion of synchronous remote invocation the 450 (at its time of publication) industrial of integration, including document formats, following a simple client-server model. standards and standards proposals classified processes, security, trading partner manage- Complex integration problems, such as B2B and related to each other. They define busi- ment, mediation, and discovery. integration problems based on a peer-to-peer ness documents, security services, transport Because integration always occurs be- paradigm, are not at all in the scope of the protocols, packaging, trading partner man- tween two or more parties that are exposing standards triple owing to missing integration agement, and many other relevant standards. services, any integration solution is “in concepts, missing reliability protocol defini- However, it’s not just industry that is between” the parties. This calls for an inte- tions, or missing security concepts. working on a complete set of standards gration framework that clearly identifies Because Web services have caught on addressing the complete integration space. system boundaries and modes of interaction very impressively across the developer and Academic research is also tackling the vast between all elements. Furthermore, it must business community, it is worthwhile explor- array of problems in interenterprise integra- define the execution model for the concep- ing potential future developments to capital- tion, as we’ve seen at workshops such as tual model so that the application of concepts ize on the state Web services have reached. Technologies for E-Services (TES 2002) or is uniform. the Workshop on Web Services, E- Last but not least, we need a methodol- Directions leading from the Business, and the Semantic Web (WES ogy that suggests best practices for the crossroads 2002), or in certain research papers.3–5 modeling task when integrating services of We can only ask where to go from here if Even more important, the big “however” several parties. The methodology suggests there are alternative paths going forward. is that the need for 135 (or 450) standards appropriately using integration concepts. Possible directions for Web services include indicates a broken conceptual model, The WSMF is an important step toward it being because the standards conflict, contradict, such a conceptual model in conjunction

JANUARY/FEBRUARY 2003 computer.org/intelligent 81 with an integration framework and method- References highly extensible, message-oriented middle- 6 ology. It provides all the integration concepts 1. C. Bussler, “The Role of B2B Engines in B2B ware for building distributed, composable necessary for modeling even the most com- Integration Architectures,” ACM SIGMOD services. To illustrate this point, I look at plex integration scenarios. It also provides a Record, vol. 31, no. 1, Mar. 2002. how it is being used to build a component framework that defines the meaning of the architecture for applications and services that 2. M. Sherif, Protocols for Secure Electronic concepts and suggests a methodology. Commerce, CRC Press, Boca Raton, Fla., makes ubiquitous grid computing a reality. The major aspect of WSMF is, besides the 2000. mere definition of complex Web services, a Grid computing strong support for mediating data as well as 3. F. Casati, M. Sayal, and M.-C. Shan, “Devel- The concept of grid computing grew out oping E-Services for Composing E-Services,” interaction behavior (message exchange pat- Proc. 13th Int’l Conf. Advanced Information of attempts to build large-scale distributed terns). In general, different services expose Systems Eng. (CAISE 01), Lecture Notes in applications that group remote super- business data following different schemas Computer Science, no. 2068, Springer-Ver- computers, databases, and online instru- and expect different message exchange pat- lag, Berlin, 2001, pp. 171–186. ments into tools used by groups of collabo- terns. For two or more services to interact, rators. Examples include the Particle Physics 4. C. Bussler, “The Application of Workflow the differences in data structure and meaning Technology in Semantic B2B Integration,” Data Grid and GriPhyn (www.griphyn.org), as well as in message exchange behavior Distributed and Parallel Databases, vol. 12, which involves a widely distributed team of must be mediated. WSMF recognizes a sig- nos. 2–3, Sept.–Nov. 2002, pp. 163–191. physicists working to analyze the data nificant set of mismatch patterns and provides streaming out of large international particle 5. A. Maedche and S. Staab, “Services on the mediation support. Only after the mediation Move: Towards P2P-Enabled Semantic Web physics projects. Another example is the problem is solved does integration become Services,” to be published in Proc. 10th Int’l Network for Earthquake Engineering Simu- possible. Conf. Information Technology and Travel & lation Grid (www.neesgrid.org), which is a For Web service integration and Web Tourism (ENTER 2003), Springer-Verlag, national virtual collaboratory for earth- services mediation to scale, the integra- Berlin, 2003. quake engineering. A third example being tion concepts must be represented as a 6. D. Fensel and C. Bussler, “The Web Service organized is the Linked Environment for formal language so that particular Web Modeling Framework WSMF,” Electronic Atmospheric Discovery, which will provide services definitions are self describing, Commerce Research and Applications, vol. the tools to do “better than real-time” de- machine interpretable, and machine exe- 1, no. 2, 2002. tailed prediction of severe storms, such as cutable. Ontologies are a suitable tech- tornadoes, based on the adaptive coupling of nology for this purpose, owing to their instruments, databases, and massive super- ability to represent semantics in the form computer simulations. Disciplines such as of concepts. From Web Services to astrophysics, genomics and proteomics, and The formal representation of services, their Grid Services chemical engineering are building many data, and their behavior are a precondition for Dennis Gannon, Indiana University other grid applications and testbeds. automatic Web services registry, discovery, To enable a distributed team of collabora- and composition. Only a strictly formal There are two common criticisms of the tors to build these applications, grids need approach allows automatic and meaningful Web services model. The first is that Web an infrastructure of ubiquitous services that Web services composition. UDDI is far from services define a mechanism for doing RPCs manage the resources provided to them. providing such a formal mechanism today, on an implementation of some XML-speci- These services include and WSMF can serve as an important input fied interface, using a slow protocol called to this necessary development. SOAP. (For an explanation of these and other • Security services: Grid users need a way With WSMF as a foundation, service- acronyms, see the “Glossary” sidebar.) This to authenticate themselves, and services oriented architectures and computing be- concept appears to be a second-rate reinven- are needed that reveal what the user is come possible where all elements that tion of OMG’s Corba or Java remote method authorized to do. Privacy is also a serious require integration are represented as ser- invocation. concern for many grid collaborations. vices. Once a service is formally defined, The second criticism is that the World • Scheduling services and resource brokers: it is registered and can be discovered and Wide Web works, and Corba and the other If a grid application needs to use a dis- composed to achieve its integration for distributed-object technologies are all hope- tributed set of resources (such as advanced business goals such as supply-chain auto- less failures. Web services are just trying to instruments, supercomputers, and data- mation. WSMF provides the formal foun- drive the Web down the same path to doom bases) in a linked computation, the user dation that allows service-oriented archi- as these other RPC disasters. must schedule these resources for con- tectures to be executable in a complex As is the case with most folk wisdom current use. business world. (and other less benign forms of oversimplifi- • Information and data services: Many We therefore strongly suggest that Web cation), these statements contain some truth. large scientific applications require services will choose the path toward an In fact, attempting to use the Web services numerous data archives that are distrib- integrated, complete conceptual-integration framework as a replacement for Corba or uted around the world. These must be model supported by a framework defining trying to use it to redo what Web technolo- cataloged, indexed, and provided to the execution semantics and a methodology gies already do well is folly. What the Web the user to make information access for successful definition of integration. services architecture brings to the table is a transparent and ubiquitous.

82 computer.org/intelligent IEEE INTELLIGENT SYSTEMS • Workflow services: Grid applications and an operation with only an input message supports a set of Service Data Elements, usually require a complex set of interac- is a one way message. Web services also con- XML documents that describe the service tions between many services and re- tain port types, which group together a set of metadata and its state. The grid service port sources. For example, data must be operations; bindings, which are associations provides every grid service with two types extracted from instruments and then between port types and transport and encod- of operations. One operation lets clients farmed out to data mining services or ing protocols; and ports, which associate a search the service’s SDEs to answer ques- used as input to numerous simulations name, binding, and network location. tions such as whose output is correlated to other sen- WSDL provides an extensible schema sor data. Orchestrating these tasks is the for encoding all these elements to define a • What other port types do you support? workflow engine’s job. specific Web service (see www.w3.org/TR/ • What other SDEs do you support? • Monitoring, messaging, and logging ser- wsdl). A complete WSDL document of a • What status and state information can vices: Grids are complex and dynamic. service is usually enough to generate the you provide? At any given time, some component will code needed to access that service—that is, be down or temporarily unavailable. it is a service reference. The standard pro- The other operation involves service Hence their applications must rely on tocol most services use is SOAP over lifetime management. autonomic services that keep track of HTTP; however, that is not a requirement. In his PhD dissertation, Roy Fielding alternate resources, redundant computa- Because operations can be one-way and described the Representational State tions, and better network routes. Transfer model, based on the principles that have helped the Web achieve success.2 Many more standard grid services exist. REST emphasizes a few verbs applied to The research community defining these many nouns. The verbs in the Web are the standards is called the Global Grid Forum The Web Services Definition HTTP operations, such as GET. The nouns (see www.gridforum.org). Consisting of consist of the network of URLs that consti- about 30 working groups and research Language provides an extensible tute the Web. The REST model assigns groups and about 600 individuals, GGF most of an operation’s semantics to the meets three times a year. It occupies a place data, rather than to the operation’s name. in the grid-computing world similar to that schema for encoding all these The grid service specification uses the of the IETF for the Internet and W3C for message and document delivery model of Web standards. GGF depends on both the elements to define a specific Web services framework to achieve a simi- IETF and W3C because it builds on lar goal. The grid service port provides a standards both groups are developing. Web service. simple mechanism to convey information In GCF’s early days, the standards were through service data elements rather than considered in isolation and often imple- through complex interfaces of methods mented in very different ways by different and parameters. organizations. However, the emergence of messages can be almost arbitrary docu- Every grid service is represented by a Web service standards has changed that. Ian ments, it is possible to implement a service Grid Service Handle, which is a universal Foster, Carl Kesselman, Jeffrey M. Nick, and over just about any transport protocol such resource identifier that can be resolved into Steven Tuecke proposed the Open Grid Ser- as email, Corba Internet InterOperability a specific Grid Service Reference, which vice Architecture as an extension of Web ser- Protocol, or a peer-to-peer protocol such as might be the WSDL for the service instance. vices to build a true component framework Sun’s JXTA. Using a separate URI handle lets services be for the Grid.1 By looking at OGSA, we can Another advantage of Web services is that easily replicated or maintained and still have see where many of us believe Web services the specification says little about the way a globally unique identity. standards will evolve. services are actually implemented. Open- The other standard grid service port source and commercial toolkits exist for types include Web services building Web services from standard Java, A conventional Web service consists of Enterprise JavaBeans, C/C++, Python, and • HandleMap: This service provides the C# using .NET. mapping between the Grid Services Han- • A set of types, which are usually described The GGF Open Grid Service Infrastruc- dle and a current Grid Service Reference. with XML schema documents ture working group, led by Steve Tuecke • Registry: This service binds service instance • Messages, which are XML elements and David Snelling, has taken advantage of metadata to a registry. consisting of a set of typed and named WSDL’s extensibility to define a grid ser- • Factory: This can create instances of parts vice component model. OGSI defines a other services and generate stateful, tran- • Operations, which can have an input standard set of port types that grid services sient instances of grid applications from message, a response output message, support and that applications can expect. a stateless, persistent service. and possibly a fault message One port type that every grid service must • Notification: Several standard ports are support is called the grid service,which provided to allow asynchronous, publish- An operation with an input and output provides a basic, essential introspection and-subscribe notification between ser- message is called a solicit response operation, mechanism for services. Each grid service vices and clients.

JANUARY/FEBRUARY 2003 computer.org/intelligent 83 Wil van der Aalst is a full professor of infor- Christoph Bussler is a principal member of the mation systems and head of the Information and technical staff at the Oracle Corporation’s Inte- Technology section of the Department of Tech- gration Platform Architecture Group, based in nology Management at the Technische Univer- Redwood Shores, Calif. He is responsible for siteit Eindhoven. He is also a part-time full pro- the overall architecture of Oracle’s next-genera- fessor at the Computing Science faculty at the tion integration platform technology addressing Department of Mathematics and Computer Sci- B2B, A2A, and ASP integration. He received his ence at the same university. His research inter- BA and MS in computer science from the Tech- ests include information systems, simulation, nical University of Munich, Germany, and a PhD Petri nets, process models, workflow management systems, mining, ver- in workflow management from the University of Erlangen-Nuremberg, ification techniques, enterprise resource planning systems, computer- Germany. He is a member of the IEEE Computer Society and ACM. supported cooperative work, and interorganizational business processes. Contact him at [email protected]; http://hometown.aol.com/ He holds an MSc in computing science and a PhD in mathematics, both chbussler/index.html. from the Technische Universiteit Eindhoven. He is a fellow and manage- ment team member of the research institute BETA. Contact him at Eind- Alexander Maedche is the head of the Knowl- hoven Univ. of Technology, P.O. Box 513, NL-5600 MB Eindhoven, edge Management research department at the Netherlands; [email protected]. FZI Research Center for Information Technolo- gies at the University of Karlsruhe. His research V. Richard Benjamins is the director of R&D interests include knowledge discovery in data at Intelligent Software Components (iSOCO) in and text, ontology engineering, learning and Madrid, Spain. His research interests include application of ontologies, and the Semantic areas such as knowledge technologies, artificial Web. He has taught courses at the University of intelligence, knowledge management, the Seman- Karlsruhe in the areas of knowledge discovery tic Web, and ontologies. He received his MS and in databases, data and text mining, ontology engineering, and applied PhD in cognitive science from the University of computer science and has given several tutorials on ontologies. He Amsterdam. He serves on many international received a diploma in industrial engineering and his PhD in applied program committees and has been co-chair of informatics, both from the University of Karlsruhe. He is a member of many international workshops (including at IJCAI and ECAI) and con- the IEEE and German Society of Computer Science. Contact him at FZI, ferences (EKAW). He is member of the IEEE Intelligent Systems editor- Univ. of Karlsruhe, 76131 Karlsruhe, Germany; [email protected]; www. ial board. Contact him at [email protected]. fzi.de/wim.

Amit Sheth is a professor of computer science Dieter Fensel works at the University of Inns- and the director of the Large Scale Distributed bruck, Austria. His research interests include Information Systems (LSDIS) Lab, at the Uni- ontologies, the Semantic Web, Web services, versity of Georgia. He also founded Taalee, an knowledge management, enterprise application enterprise software and Semantic Web technol- integration, and electronic commerce. He is an ogy startup based on the research at the LSDIS editor of Knowledge and Information Systems: lab. He received his BE from B.I.T.S., Pilani, An International Journal (KAIS), IEEE Intelli- India, and his MS and PhD from Ohio State Uni- gent Systems, the Electronic Transactions on versity. He is on the editorial board of six jour- Artificial Intelligence (ETAI), and Web Intelli- nals, has served on over 70 program and organization committees, and gence and Agent Systems (WIAS). He has been involved in several has chaired or cochaired six international conferences or workshops in national and internal research projects—for example, the IST projects the areas of the Semantic Web, digital libraries, multidatabase systems, COG, Esperonto, H-Techsight, IBROW, Multiple, Ontoknowledge, and parallel and distributed information systems. Contact him at the Ontoweb, SWAP, SWWS, and Wonderweb. He has been the project Dept. of Computer Science, 415 Graduate Studies Research Center, coordinator of Ontoknowledge, Ontoweb, and SWWS. Contact him at Univ. of Georgia, Athens, GA 30602-7404; [email protected]; lsdis.cs. [email protected]. uga.edu/~amit. Dennis Gannon is a professor in and chairs the John A. Miller is a professor of computer sci- Computer Science Dept. at Indiana University. ence at the University of Georgia and is also the His current work is on designing software com- Graduate Coordinator for the Computer Science ponent architectures for distributed scientific department. His research interests include data- applications and studying the architecture of grid base systems, simulation and workflow, and systems. He received a BS from the Univerisity parallel and distributed systems. He received of California, Davis, a PhD in computer science his BS in applied mathematics from Northwest- from the University of Illinois, and a PhD in ern University and his MS and PhD in informa- mathematics from the University of California, tion and computer science from the Georgia Davis. He is one of the cofounders of the Common Component Archi- Institute of Technology. He is an associate editor for ACM Transactions tecture project (now supported by the DOE Center for Component Tech- on Modeling and Computer Simulation and IEEE Transactions on Sys- nology for Terascale Simulation Software), the Java Grande Forum, and tems. Contact him at the Dept. of Computer Science, 415 Graduate Stud- the Global Grid Forum. He is also the Science Director for the new Indi- ies Research Center, Univ. of Georgia, Athens, GA 30602-7404; jam@ ana Pervasive Technologies Labs and the Chief Computer Scientist for ca.uga.edu. the NCSA Alliance. Contact him at [email protected].

In addition to OGSI, there is a GGF vices that are needed for a new generation service messaging’s document model is Open Grid Service Architecture working of grid applications. Other working groups that security can be end-to-end rather group, which is exploring grid uses cases are looking at specific topics such as secu- than link level. In other words, we can and defining the specific standard ser- rity. Another important advantage of Web encrypt and sign a message that might

84 computer.org/intelligent IEEE INTELLIGENT SYSTEMS pass through several intermediaries before Like any programming language, References reaching its destination. In standard RPC BPEL4WS has a rich control structure to models, the security is often at the level sequence operations. While it is not yet 1. I. Foster et al., “The Physiology of the Grid: An Open Grid Services Architecture for Dis- of the supporting stream protocols, and known if BPEL4WS will also prove to tributed Systems Integration,” Grid Comput- the contents of method arguments are be the appropriate composition tool for ing: Making the Global Infrastructure a Real- exposed at any intermediate steps in the grid services, it does seem to be a strong ity, F. Berman, A.J.G. Hey, and G. Fox, eds., processing. candidate. John Wiley, New York, 2003. Any component architecture must have a A final area of concern might be the rela- 2. R.T. Fielding, Architectural Styles and the concept of component composition. The Web tion of Web services and grid services to Design of Network-Based Software Architec- services community has produced a new the Semantic Web. Others have observed tures, PhD dissertation, Dept. of Computer Sci- draft standard, called the BPEL4WS, which that Semantic Web markup languages can ence, Univ. of California, Irvine, Calif., 2000. is one candidate.3 A BPEL4WS process play an important role in allowing agent 3. F. Curbera et al., “Business Process Execution defines a composite Web service described technology to automatically use Web ser- Language for Web Services, Version 1.0,” IBM the way you might build a flowchart for an vices.4 Within the Grid Forum, there is a developerWorks, 13 July 2002, www. algorithm. It is composed of a set of activities research group arising out of the Semantic ibm.com/developerworks/library/ws-bpel. that include invoking its composite services, Grid of David De Roure and his colleagues 4. S.A. McIlraith, T.C. Son, and H. Zeng, “Seman- waiting for input messages, copying data, and (see www.sematnicgrid.org) that is actively tic Web Services,” IEEE Intelligent Systems, throwing exceptions. considering this question. vol. 16, no. 2, Mar./Apr. 2001, pp. 46–53.

PURPOSE The IEEE Computer Society is EXECUTIVE COMMITTEE the world’s largest association of computing President: professionals, and is the leading provider of STEPHEN L. DIAMOND* technical information in the field. Picosoft, Inc. MEMBERSHIP Members receive the P.O.Box 5032 monthly magazine COMPUTER, discounts, San Mateo, CA 94402 and opportunities to serve (all activities are Phone: +1 650 570 6060 led by volunteer members). Membership is Fax: +1 650 345 1254 open to all IEEE members, affiliate society members, and others interested in the com- [email protected] puter field. President-Elect: CARL K. CHANG* COMPUTER SOCIETY WEB SITE Past President: WILLIS. K. KING* The IEEE Computer Society’s Web site, at http://computer.org, offers information VP, Educational Activities: DEBORAH K. SCHERRER and samples from the society’s publications (1ST VP)* and conferences, as well as a broad range of VP, Conferences and Tutorials: CHRISTINA information about technical committees, SCHOBER* standards, student activities, and more. COMPUTER SOCIETY OFFICES VP, Chapters Activities: MURALI VARANASI† BOARD OF GOVERNORS Headquarters Office VP, Publications: RANGACHAR KASTURI † 1730 Massachusetts Ave. NW VP, Standards Activities: JAMES W. MOORE† Term Expiring 2003: Fiorenza C. Albert- Howard, Manfred Broy, Alan Clements, Richard A. Washington, DC 20036-1992 VP, Technical Activities: YERVANT ZORIAN† Kemmerer, Susan A. Mengel, James W. Moore, Phone: +1 202 371 0101 • Fax: +1 202 728 9614 Secretary: OSCAR N. GARCIA* Christina M. Schober Treasurer:WOLFGANG K. GILOI* (2ND VP) Term Expiring 2004: Jean M. Bacon, Ricardo E-mail: [email protected] Baeza-Yates, Deborah M. Cooper, George V. Cybenko, 2002–2003 IEEE Division VIII Director: JAMES D. Haruhisha Ichikawa, Lowell G. Johnson, Thomas W. Publications Office ISAAK† Williams 10662 Los Vaqueros Cir., PO Box 3014 2003–2004 IEEE Division V Director: GUYLAINE M. Term Expiring 2005: Oscar N. Garcia, Mark A Grant, Michel Israel, Stephen B. Seidman, Kathleen Los Alamitos, CA 90720-1314 POLLOCK† M. Swigger, Makoto Takizawa, Michael R. Williams Phone:+1 714 8218380 Computer Editor in Chief:DORIS L. CARVER† Executive Director: DAVID W. HENNAGE† Next Board Meeting: 22 Feb. 2003, San Diego, CA E-mail: [email protected] * voting member of the Board of Governors Membership and Publication Orders: † nonvoting member of the Board of Governors IEEE OFFICERS Phone: +1 800 272 6657 Fax: +1 714 821 4641 President: MICHAEL S. ADLER E-mail: [email protected] President-Elect: ARTHUR W. WINSTON EXECUTIVE STAFF Asia/Pacific Office Past President: RAYMOND D. FINDLAY Executive Director: DAVID W. HENNAGE Executive Director: DANIEL J. SENESE Watanabe Building Assoc. Executive Director: Secretary: LEVENT ONURAL 1-4-2 Minami-Aoyama,Minato-ku, ANNE MARIE KELLY Treasurer: PEDRO A. RAY Tokyo107-0062, Japan Publisher: ANGELA BURGESS VP, Educational Activities: JAMES M. TIEN Phone: +81 3 3408 3118 • Fax: +81 3 3408 3553 Assistant Publisher: DICK PRICE VP, Publications Activities:MICHAEL R. LIGHTNER E-mail: [email protected] Director, Finance & Administration: VIOLET S. VP, Regional Activities: W. CLEON ANDERSON DOAN VP, Standards Association: GERALD H. PETERSON Director, Information Technology & Services: VP, Technical Activities: RALPH W. WYNDRUM JR. ROBERT CARE IEEE Division VIII Director JAMES D. ISAAK Manager, Research & Planning: JOHN C. KEATON President, IEEE-USA: JAMES V. LEONARD

JANUARY/FEBRUARY 2003 computer.org/intelligent 85