Theorem 5. The PST of a TTG G can be computed in time linear in the number of edges of G.
only be obtained by arbitrating between overlapping fragments. Our definition is fur- ther justified by Prop. 2, i.e., by the fact that all fragments and hence all parses can be derived from the PST in a simple way.
Acknowledgments. The work published in this article was partially supported by the SUPER project (http://www.ip-super.org/) under the EU 6th Framework Programme Information Society Technologies Objective (contract no. FP6-026850).
References
1. Ayesta, J.V.: Parsing flowcharts and series-parallel graphs. PhD thesis, Stanford University, CA, USA (1978)
2. Gutwenger, C., Mutzel, P.: A linear time implementation of SPQR-trees. In: Marks, J. (ed.) GD 2000. LNCS, vol. 1984, pp. 77–90. Springer, Heidelberg (2001)
3. Hopcroft, J., Tarjan, R.E.: Dividing a graph into triconnected components. SIAM J. Com- put. 2, 135–158 (1973)
4. Johnson, R., Pearson, D., Pingali, K.: The program structure tree: Computing control regions in linear time. In: Proceedings of the ACM SIGPLAN 1994 Conference on Programming Language Design and Implementation (PLDI), pp. 171–185 (1994)
5. Johnson, R.C.: Efficient program analysis using dependence flow graphs. PhD thesis, Cornell University, Ithaca, NY, USA (1995)
6. K¨uster, J., Gerth, C., F¨orster, A., Engels, G.: Detecting and resolving process model differ- ences in the absence of a change log. In: Dumas, M., Reichert, M., and Shan, M.-C., (eds.) BPM 2008. LNCS, vol. 5240, pp. 244–260. Springer, Heidelberg (2008)
7. Lassen, K.B., van der Aalst, W.M.P.: WorkflowNet2BPEL4WS: A tool for translating un- structured workflow processes to readable BPEL. In: Meersman, R., Tari, Z. (eds.) OTM 2006. LNCS, vol. 4275, pp. 127–144. Springer, Heidelberg (2006)
8. Lohmann, N., Kleine, J.: Fully-automatic translation of open workflow net models into human-readable abstract BPEL processes. In: Modellierung 2008, GI, March 2008. Lecture Notes in Informatics (LNI), vol. P-127, pp. 57–72 (2008)
9. Ouyang, C., Dumas, M., ter Hofstede, A.H.M., van der Aalst, W.M.P.: From BPMN pro- cess models to BPEL web services. In: ICWS, pp. 285–292. IEEE Computer Society, Los Alamitos (2006)
10. Sadiq, W., Orlowska, M.E.: Analyzing process models using graph reduction techniques. Inf.
Syst. 25(2), 117–134 (2000)
11. Tarjan, R.E., Valdes, J.: Prime subprogram parsing of a program. In: POPL 1980: Proceed- ings of the 7th ACM SIGPLAN-SIGACT Symposium on Principles of Programming Lan- guages, pp. 95–105. ACM, New York (1980)
12. Vanhatalo, J.: Structural analysis of business process models using the process structure trees (to appear)
13. Vanhatalo, J., V¨olzer, H., Koehler, J.: The refined process structure tree. IBM Research Re- port RZ 3712 (2008)
14. Vanhatalo, J., V¨olzer, H., Leymann, F.: Faster and more focused control-flow analysis for business process models though SESE decomposition. In: Kr¨amer, B.J., Lin, K.-J., Narasimhan, P. (eds.) ICSOC 2007. LNCS, vol. 4749, pp. 43–55. Springer, Heidelberg (2007)
Christian Stahl1 and Karsten Wolf2
1 Humboldt-Universit¨at zu Berlin, Institut f¨ur Informatik Unter den Linden 6, 10099 Berlin, Germany
2 Universit¨at Rostock, Institut f¨ur Informatik 18051 Rostock, Germany
Abstract. We present a finite representation of all services M where the composition with a given serviceN is deadlock-free, and a given set of activities ofN can becovered (i.e. is not dead). Our representation is an extension of the existing notion of an operating guideline which only cared about deadlock freedom. We further present an algorithm to decide whether a serviceMmatches with the extended operating guideline ofN.
Keywords: process modeling and analysis, SOA, Petri nets, operating guidelines.
1 Introduction
One of the objectives of service-oriented computing (SOC) [1] is the modular structuring and loose coupling of interorganisational business processes. In this aspect, SOC meets the area of modeling and analysing workflows [2]. While SOC aims at composing complex business activities from more elementary ones (services), workflow modeling is (among others) concerned with the study of well-designed workflows and business processes. Central to the wellformedness of workflows is the concept of soundness. This property basically states that every process instance will terminate in a well-defined final state while there are no useless (dead) activities. In the intersection of SOC and workflow modeling, we are thus interested in mechanisms for service composition (and related tasks such as discovery) which assure soundness in the overall system (e.g. a service orchestration).
Current approaches for matching and discovering services are incapable of asserting soundness in service discovery scenarios. Some approaches propose to compute and publish a public view P of a provided service P [3,4]. Then, a service requesterRcan check its compositionR⊕Pto decide proper interaction.
However, public view approaches do not explicitly state whether soundness of P⊕Rimplies soundness ofP⊕R. Thus, existing public view approaches cannot be applied to obtain a globally sound system.
Other approaches suggest to compute anoperating guidelineOGP for a given service P which represents all correctly interacting partners of P [5]. Then, a
M. Dumas, M. Reichert, and M.-C. Shan (Eds.): BPM 2008, LNCS 5240, pp. 116–131, 2008.
c Springer-Verlag Berlin Heidelberg 2008
matching procedure betweenRandOGPcan be used for deciding whetherP⊕R would interact correctly. Here, correctness refers to deadlock freedom so far.
Deadlock freedom is a necessary but insufficient condition for soundness. In this paper, we extend the operating guideline approach by asserting—in addi- tion to deadlock freedom—the absence of dead activities in the composed sys- tem. This is another necessary condition for soundness. The only remaining gap between the new approach and soundness is the possible existence of livelocks.
For acyclic services, our approach already establishes soundness in the composed system since acyclic services cannot contain livelocks.
Another motivating scenario for our approach is inspired by [6]. In this arti- cle, all partners of a given service, whichenforce or exclude certain behavioral patterns such as occurrences of activities, are characterized. This approach can be used, among others, for
– filtering of service registries for services that fit specific specifications (“en- force book”: I want to get a book selling service; “exclude credit card”: I do not want to pay by credit card),
– validating services by checking whether there exist partners that access cer- tain features
Sometimes, enforcing some behavior is too strict. Consider an application for a credit with an online bank service. Of course, the user (service requester) wishes to have the activity “credit approved” executed in the service. However, there is hardly an online bank service where “credit approved” can be enforced by the user (which would mean that the user can always obtain a credit by just following a suitable communication pattern). There will rather be an internal decision based on which a credit is either approved or denied. In typical service models, the decision appears to the user as a nondeterministic choice. Thus, we need a weaker criterion that rules out at least all those services where “credit approved” is completely impossible. That is,Rshould match withP if and only if it is at leastpossibleto execute activity “credit approved” in the composition of the online bank service and the requester.
Formally, we want to compute a finite representation of the (generally infinite) set of all those partnersRof a given serviceP where the composition P⊕Rof both services is deadlock-free, and a certain setX of activities is not dead. For establishing soundness, this setX would be the set of all activities ofP. In the online banking example,X would consist only of activity “credit approved”. We achieve this goal by extending the existing operating guideline approach with deadlock-free interaction.
The paper is structured as follows. In Sect. 2 we recall open nets and operating guidelines. Next, in Sect. 3, we extend our notion of partnersRfor P to those partners R where a certain set of places and transitions in P ⊕R is covered (i.e. each place can be marked and each transition is not dead). We show how to calculate a finite representation of all these partners by extending our notion of an operating guideline with a global constraint. Section 4 presents related work and finally conclusions are drawn in Sect. 5.