EVALUATION OF INTER-ORGANIZATIONAL BUSINESS PROCESS SOLUTIONS: A CONCEPTUAL MODEL-BASED APPROACH Johny Ghattas and Pnina Soffer Management Information Systems University of Haifa, Carmel Mountain 31905, Haifa, Israel Email: GhattasJohny@gmail.com, SPnina@is.haifa.ac.il ABSTRACT Collaboration and coordination between organizations are necessary in today's business environment, and are enabled by inter-organizational processes. Many approaches for the construction of such processes have been proposed in recent years. However, due to the lack of standard terminology it is hard to evaluate and select a solution that fits a specific business scenario. The paper proposes a conceptual model which depicts the nature of interaction between organizations through business processes under specific business requirements that emphasize the privacy and autonomy of the participating organizations. The model is generic, and relies on the Generic Process Model (GPM) framework and on Bunge's ontology. Being generic and theory-based, we propose to use the model as a basis for comparing and evaluating design and implementation-level approaches for interorganizational processes. We demonstrate the evaluation procedure by applying it to three existing approaches. Keywords: Inter-organizational processes, Business process design, Conceptual model, RosettaNet, P2P, CWM, Privacy, Autonomy, Flexibility. 1. Introduction Collaboration and coordination between organizations are considered necessary in a business environment, where organizations focus on their competitive advantage and complement their offering through partners and suppliers. Inter-organizational processes are the enabler of such business environments. Various mechanisms at various levels of detail have been proposed for achieving interoperability or shared business processes among organizations. Most of them focus on implementation details (e.g., [1]). A key issue discussed in the literature with respect to inter-organizational processes is the required balance between trust and control, visibility and privacy. Addressing this delicate balance, solutions vary from complete central control to pure distribution. Various models [2][3][4][5][6] are proposed, where an inter-organizational workflow is defined as part or as a result of a contract between organizations. The organizations are then contractually committed to the defined workflow (or to a partial definition). Different levels of visibility of a partner’s internal process by the other partners at run time are also proposed and supported. In general, a high degree of central control and required visibility imposes constraints on the internal operations of an organization, thus reduces its flexibility. Given this variety of approaches, it is difficult to assess and evaluate which one is better suited for a given situation. This difficulty is particularly severe considering the absence of a standard vocabulary or set of concepts in this domain, so each solution has its own terminology. Furthermore, going into implementation details and solutions, many of these approaches do not rely on a theoretical basis or explicit requirements specified in business terms. We claim that an implementation solution needs to rely on a solid conceptual model, depicting the essence of shared business processes and needs the solution must realize. Such model should enable the assessment of a proposed solution in terms of its completeness and compliance with the behavior required in a specific situation. This paper proposes a conceptual model of inter-organizational processes, which aims at identifying the minimal definition required to enable a smooth operation of shared processes, while allowing the partners a maximal degree of privacy and flexibility. The model, based on the formal Generic Process Model (GPM) and Bunge’s ontology, is at a rather high level of abstraction, and does not address implementation details. It includes a set of concepts which are well-defined and generic, relating to a set of business-oriented requirements whose focus is autonomy and privacy. As such, it can form a basis for evaluating implementation models and solutions when such business requirements are due. The paper proposes such evaluation 2 procedure and demonstrates it with respect to three existing inter-organizational process frameworks. The remainder of the paper is structured as follows: The next section provides a literature review to motivate this current work. Then a brief introduction to the main concepts of GPM is made, as a basis for the presentation of inter-organizational process model. The following section presents the use of our model as a benchmark for evaluating inter-organizational process solution approaches, and demonstrates it with respect to three well-known approaches. This evaluation is discussed and finally, conclusions are drawn. 2. Related work The literature about business process modeling offers a rich variety of solutions for implementing business processes within organizations and between organizations. The proposed solutions differ from each other in the implementation details, in the scope of issues addressed, and in the level of privacy and autonomy granted to the participating organizations. For example, [7] addresses interaction among workflows (both within and between organizations). They suggest a mechanism for exchanging events between interacting workflows. Their work provides a technical mechanism that allows workflow interactions, and a means for specifying them in workflow models. Privacy and autonomy of the organizations are enabled since the workflows interchange events and do not expose further information. Another example is [8], who propose to model inter-organizational processes using UML, combined with a translation of the model into a designated formal language (COALA). Their approach, which enables specification of the interaction among roles (organizations), management of shared resources, and exception handling, is focused on dependability and error recovery at run time. Infrastructures, such as XRL/flower [1], are designed to support the exchange of information required by inter-organizational business processes. Means for defining the process are provided. However, the support of flexibility and autonomy of the participating organizations is not explicitly addressed, and can be assured only within a specific process. A business contract language (XLBC) was introduced in [9] to formally link the component definition language (CDL) specification of workflow systems based on business objects. The authors briefly discuss the object visibility specified through contracts, but do not go into details. Infrastructural issues also include the work on process specification standards (BPMI [10]; ebXML BPSS [11], BPMN [12]; W3C [13]; OASIS [14]). E-business standards, traditionally focused on simple message exchange between organizations, have been expanding to 3 consider structured sequences of messages and the process implications behind such sequences. Considering privacy and autonomy issues, the Workflow Management Coalition (WFMC) proposed an interface (WF-XML) that allows variable degrees of visibility between organizations [15]. A different direction for approaching the inter-organizational process is by addressing the relationships between organizations. [5][4] deal with legal contracts between organizations, and show how a detailed workflow can be derived from a contract. They also provide rules for matching the contract-based workflow with the existing organizational business processes. The contract-based workflow relates to the entire process, thus it allows no flexibility to a single organization for changing the process or deviating from it in specific cases. In [6] the focus is on the privacy of organizational processes, which is maintained at run time, limiting the visibility of the internal process. However, the entire workflow has to be defined at build time, when a contract between the parties is established. Being contractually obligated to the workflow model, the flexibility of the parties is limited. In [2], the contract specifies the internal process of the “service provider” party and the interface between the organizations, and allows a limited visibility of the provider process. Here autonomy and flexibility are reduced only for the “provider” party, and not for the other side. The PRODNET project [16] is aimed at facilitating autonomy and heterogeneity of organizations participating in a virtual organization. However, their solution is based on a central mandating and coordinating party, to whom all parties must report. This requirement forms a limitation on the privacy and autonomy of the participants. A model that includes a means for process verification, both between and within the organizations, is proposed by [17]. Steps in defining the inter-organizational process are proposed, facilitating some privacy and autonomy of partners through workflow views and inheritance. Autonomy of the partners is also facilitated by the ebXML BPSS model [11]. However, its evaluation on the basis of Bunge’s ontology [18] indicates a lack of ontological completeness as well as clarity, which may lead to modeling and interpretation difficulties. Clearly, when more constraints are imposed by the inter-organizational process, less flexibility can exist in the internal processes of an organization [19][20]. Summarizing the above review of related literature, we observe that the issue of interorganizational process modeling and support has been extensively addressed. Yet, while the proposed solutions strive to enable the operation of an inter-organizational process, no explicit consideration of specific business requirements is made to relate specific solutions to specific business scenarios. 4 3. Deriving Inter-organizational Business process requirements For any formal conceptual model to be complete and valid, the ontological assumptions of the model need to be made explicit. In this paper we have set our major objective to establish a model that would maximize the interacting parties' autonomy and privacy while not losing the run time governance of the inter-organizational process. Our model is designed to satisfy the following set of requirements, which emphasize the privacy and autonomy of the participating organizations, as well as the operational feasibility of the process. (1) No central governance of the over-all process. There is no entity that designs, implements, executes, and monitors the end-to-end process. This requirement follows the assumption that central governance reduces the autonomy of the parties and may require visibility to details that are not necessarily visible in a purely distributed process. Hence, organizations’ privacy and autonomy is expected to be higher in purely distributed processes as compared to centrally governed ones. (2) Partners' business privacy conservation. Each partner shares the minimum needed information for enabling the over-all process design, implementation, and enactment. (3) Partners' business autonomy conservation. Each partner has a full autonomy to design, implement, execute and monitor its internal processes, provided they comply with the partner's obligations toward the other partners. (4) Process run time governance assurance. All collaborating partners need to have governance of the overall process status at all time. Since visibility is limited, such governance can be based on mutual commitments, which reduce the uncertainty and foment the trust among the partners. Note that these are among the most restrictive requirements that may be set for a valid interorganizational process definition. When privacy and autonomy are not at the top priority of the organizations, these requirements may not be necessarily relevant and some of them may need to be relaxed or even eliminated. Once we have established a conceptual model that complies with these requirements, we would be able to evaluate business process (BP) design and implementation models on the basis of the conceptual model. 5 4. The conceptual model The proposed model builds upon the Generic Process Model (GPM) and extends it to incorporate the relevant concepts for modeling inter-organizational business processes. GPM does not explicitly address inter-organizational processes. We will first briefly present GPM and to extend it to a conceptual model that addresses inter-organizational processes. 4.1 The Generic Process Model This section introduces the Generic process Model (GPM), which is the basis of our interorganizational process conceptual model. GPM is based on Bunge’s ontology [2321][22], as adapted for information systems modeling (e.g.,[23][24]), for conceptual modeling, and for modeling business process concepts. According to the ontological framework, the world is made of things that possess properties. Properties are perceived by humans in terms of attributes, which can be represented as functions on time. The state of a thing is the set of values of all its attribute functions (also termed state variables). When properties of things change, these changes are manifested as state changes or events. State changes can happen either due to internal transformations in things (self action of a thing) or due to interactions among things. Properties that establish relations between things are mutual properties, while properties that relate only to one thing are intrinsic to that thing. The rules governing possible states and state changes are termed state laws and transformation laws, respectively. States can be classified as being stable or unstable, where an unstable state is a state that must change by law, and a stable state is a state that can only change as a result of an event external to the thing or the domain. A domain is a part of the world, namely, a set of things and their interactions. It is represented by a set of state variables, whose values represent the state of the domain at a moment in time. A sub-domain is a part of the domain, represented by a subset of the domain state variables. A process is a sequence of unstable states, transforming by law until a stable state is reached. A process is defined over a domain, which sets the boundaries of what is in a stable or an unstable state. Events that occur outside the domain are external events and they can activate the domain when it is in a stable state. A process model in GPM is a three-tuple <L, I, G>, where L is the law, specified as mapping between subsets of states; I is a subset of unstable states, which are the initial states of the process after a triggering external event has occurred; G is a subset of stable states, which are the goal of the process. Subsets of states are specified by conditions over the state variables of the domain. Hence, a process starts when a certain condition on the state of the domain holds, 6 and ends when its goal is reached, i.e., when another condition specified on the state of the domain holds. It is possible to define the projection of a process over a sub-domain, where the set of state variables addressed by the law is a subset of domain state variables. Then all transformations outside the sub-domain are considered external events, and the sub-domain may be in a stable state while the process is active in other parts of the domain. The process goal as addressed by GPM is the state achieved by the process. However, the goal concept is sometimes used also to describe business objectives. GPM distinguishes process goals from soft-goals, which are defined as an order relation on goal states [25]. In other words, soft-goals relate to the desirability of possible states in the goal set (all meeting the condition that terminates the process) according to defined business objectives. For example, assume the goal of a sales process is a set of states where an order has been fulfilled, but achieving it in two days is considered "better" than in two months. Finally, GPM entails criteria for assessing the validity of a process, namely, its ability to achieve its goal [26]. It enables the analysis of a process to identify causes for invalidity, and suggests appropriate redesign actions to eliminate these causes. 4.2 The Inter-organizational process model The notion of domain sets the scope of control over the process in a GPM process model. Consider an inter-organizational process modeled from the point of view of one of the partners (that is, defined over the domain of that organization). The process reaches a state where some request was made for another partner organization (e.g., supplier) to operate. Since actions taken by other partner organizations are outside the scope of the domain, the process domain remains in a stable state, waiting for the results of the other partner’s action. These results are conceived as an external event, and are not within the control of the requesting organization. At the occurrence of the expected external event, the state of the domain transforms from stable to unstable and the process is reactivated. This kind of situation is termed a discontinuity point [26], since the process (as defined within the organization) depends on an external event in order to proceed and achieve its goal1. When looking at the overall process, the process domain is extended beyond the organizational boundaries of each of the participants. Definition 1: The inter-organizational process domain is a domain including all the participating organizations. 1 Note that synchronous actions in both organizations are possible according to the model. Even so, at the synchronization point the process might be waiting for an external event and will not proceed without it. 7 For the sake of brevity, we herein refer to the inter-organizational process domain simply as the process domain. The participating organizations are all sub-domains of the process domain, and shall hence on be denoted as the partner sub-domains. The process domain is not centrally mandated, but it can be captured as an aggregation of all its partner sub-domains, which are autonomous and private. Definition 2: A sub-domain is said to be private if its states, goals, and law cannot be controlled or viewed from outside the sub-domain. The partner sub-domains interact with each other. In ontological terms, this interaction takes place through properties that are mutual to at least two partners. When the value of such property (reflected by a state variable value) is changed by one partner, this event affects the state of the other partner too and triggers a transformation in it. Definition 3: The sub-domain that includes all the state variables representing mutual properties of the partners shall be termed the shared sub-domain. Note that if the collaboration spans more than two parties then a shared sub-domain may exist with respect to any sub-set of parties which includes more than two members. For simplicity, we herein refer to a single shared sub-domain. Since the process domain is composed of private sub-domains, the process cannot be viewed or defined as a whole, but it is known to exist. Nevertheless, the internal processes of the partners, which are well defined and known within each partner sub-domain, are projections of the inter-organizational process over the partner sub-domain. Considering the agreements that bind the partners, they concern the process parts that are not private, namely the shared sub-domain. It appears as if the agreements define the projection of the inter-organizational process over the shared sub-domain. However, since the shared sub-domain includes only mutual properties, its state cannot transform by itself. Rather, every transformation in the shared sub-domain is a result of an event which is external to that subdomain, be it in a partner sub-domain or external to the entire process domain (e.g., time). This argument leads to the following lemma. Lemma: All the states of the shared sub-domain are stable. A state can only be regarded stable or unstable with respect to a (sub)domain. Since all the transformations in the shared sub-domain are a result of external events, all its states are stable. Hence, no meaningful projection of the law exists over the shared sub-domain. Rather, all its states can be considered as discontinuity points, where the (sub)domain is “waiting” for an external event for its state to be transformed. For the inter-organizational process to progress towards its goal, these events should be generated by the partner organizations. 8 Consider a set of states SI in the shared sub-domain, defined by a condition over some of the state variables. A state in this set is stable, and it waits for a partner P to transform it into a state in the next set of states, SF. Definition 4: An obligation is a three-tuple <SI, SF, P>, where, given a state in SI, partner P will transform the shared sub-domain to a state in SF. As an example, consider an inter-organizational process in which once an order has been made by a customer, the supplier must commit to a delivery schedule. In this case SI = {s| Order_status=inserted}, SF = {s| Order_status = delivery schedule committed}, and the partner P who should transform a state sSI to sSF is the supplier. In other words, an obligation of a partner to a state (or a set of states) forms a commitment that the law defined in its private sub-domain assures the achievement of that state. In the above example, the law in the supplier’s private sub-domain is designed so as to plan, provide, and commit to a delivery schedule when an order is received. The details of how this is done are not known to other participants in the process. The outcome of an obligation is that although an observer is not familiar with the projection of the process over a private subdomain, he knows it is designed so that the required state will be achieved. The set of obligations can be viewed as equivalent to a law mapping between sets of states in the shared sub-domain. Yet, the transformations are to be brought about by the partner organizations. The implications of the obligations on the process design in each of the partner organizations are in the form of given sets of states that the process must reach. These may be the goal of the (local) process or some mandatory states within the process that constrain its course. The validity of the process can be assured by: (a) ensuring that the law of every private process meets the defined obligations (to be performed by each partner). (b) Verifying that each of the defined shared states is on a "path" to the goal, based on the mappings established by the obligations. Another element in the process model is related to Quality of Service (QoS) parameters or soft-goals in GPM terms. Constraints on performance measures related to the generation of the events in the shared sub-domain can be agreed upon. For example, response to a request (event) should be made within a given time limit (performance measure). Then this constraint becomes a part of the state definition in the obligation. To summarize this discussion, the construction of a process should address the states in the shared sub-domain to which the partners will be obligated, and the constraints on QoS parameters that are implied in these states. It may be possible that a partner organization already has existing (local) business processes that should be matched to the defined obligations. Due to the obligation definition, this matching becomes rather simple, verifying that the required states are reached by the existing process. 9 The obligation mechanism allows maximal flexibility and autonomy to the partner organizations. The partners' organizations are free to change their internal processes or to deviate from them as long as they fulfill their obligations. The obligations are defined with respect to the mutual properties of the organizations, which are clearly distinguished from intrinsic properties. Hence, the requirements posed by the obligations do not concern issues beyond those necessary for the interaction. Since only the shared sub-domain is visible to all partners (both at design time and at run time), no unnecessary information is revealed and the privacy of the organizations is maintained. The concepts of the model and their relationships are summarized in Figure 1. The interorganizational process operates over a process domain, composed of the shared domain and many partner domains. The shared domain includes mutual state variables of the partner domains. Private processes, which are part of the inter-organizational process, operate over the partner domains, and have obligations, which relate to the mutual state variables. The obligations include QoS constraints and shared states, which hold values of shared state variables. Note that these are concepts that reside as an upper level on top of GPM and Bunge's concepts. Figure 1. The overall process conceptual model 5. Evaluating Potential Inter-organizational Process Solutions 10 As already discussed, the conceptual model, whose set of requirements is clearly defined in business terms, can serve for evaluating potential solutions in situations where these requirements hold. lists the evaluation criteria, which are derived from the conceptual model’s Table 1 requirements. In addition, the criteria relate to the possibility of assessing the validity of the process. For each criterion, Table 1 presents evaluation indicators related to structural elements that should exist in a model in order to meet the criterion. The evaluation indicators are expressed using the concepts of the conceptual model, and can be evaluated on the basis of a mapping established between the conceptual model’s concepts and those of the solution under evaluation. The evaluation procedure includes a step of establishing a concept mapping between the two models, and an evaluation step, as described below. 5.1 Evaluation procedure We denote our conceptual model as CM, while the evaluated model shall be denoted M. Step 1. Concept mapping (1) Identify the concepts of model M, based on a conceptual map of model M. (2) Map each M concept to CM concepts. Several cases may occur: a. The mapping may be exactly one-to-one, b. In some cases an M concept may correspond to a set of CM concepts, c. In other cases a given M concept may be partially mapped to a CM concept. (3) Map CM concepts to model M concepts (“reverse mapping”), to assure the mapping completeness. (4) Once steps (2) and (3) completed, we have a table with three columns: CM concept, M concept, and the degree of equivalence (complete, partial, no equivalence). Several cases may occur: a. A CM concept does not have a corresponding M concept, b. An M concept does not have a corresponding CM concept c. It may also be possible that several M concepts are fully or partially mapped to a single CM concept. Step 2. Evaluation Based on the concept mapping established in step 1, assess the level of each criterion evaluation indicator ( Table 1). Note that the evaluation is qualitative and its result is on a discrete scale (Compliance, Partial compliance, Non Compliance). Table 1. Evaluation criteria Criterion Criterion evaluation indicators (in relation to the concept mapping) (1) Non Central governance No party should have full visibility or some control of the domains at any moment while designing, implementing, executing or monitoring the overall process Non existence of transformations outside of private domains. (non existence of shared sub-domain transformations) Separation between shared sub-domain state variables and private state variables (2) BP run time governance assurance Ability of partners to Identify the overall process Shared sub-domain state definition + association state at any moment in time. between obligations and sets of shared subdomains states Partners' access to mutual state variables value. Partner’s knowledge of expected run time Definition of obligations for each shared subbehavior. domain state. Identification of soft-goals (QoS parameters) Definition of performance constraints for each shared state Mapping of the shared state performance constraint to internal performance constraints Association of events and transformations in the Inclusion of QoS state variables in obligation model with QoS constraint incompliance. definition (3) Domain Privacy conservation Distinction of necessary from optional private Distinction between internal and mutual process visibility needs (Intrinsic and mutual state variables. properties mapping). Privacy of transformations execution Privacy of process laws No shared sub-domain transformationsonly private transformations No shared process law Shared domain state transition only through Shared domain state transition only obligations fulfillment. through obligations fulfillment. (4) Domain Autonomy conservation (private process design & run time) Minimal constraints on private process design Minimal constraints on private process execution Design constraints imposed only on mutual properties through obligation definition. Run-time constraints imposed only on mutual properties through obligation definition. (5) Validity of the overall process Private Processes validity establishment Shared domain process validity establishment Private process model includes mutual properties in addition to intrinsic ones. Goal state identification (shared and unshared ones) Inclusion in the obligation definition of the relevant external event definition Existence of a Validity criteria for the private process For each state of the shared sub-domain, there must be at least one obligation associated with it. 5.2 Applying the procedure to existing solutions We proceed to the evaluation of three potential inter-organizational process design and implementation solutions: (1) the Public to Private (P2P) approach for inter-organizational 12 workflow [7], (2) the Contract Workflow Model (CWM) [3][4] [5] [27] and (3) the RosettaNet approach [28][29]. These three approaches were selected as examples of models whose aim and focus are different. In addition, the documentation of these approaches is detailed enough to enable their analysis. A brief introduction to these models is presented below, followed by their evaluation. The Public to Private (P2P) model The P2P approach [17] uses models expressed as workflow (WF) nets, which are a specific form of Petri-nets. The following steps are defined for setting the inter-organizational workflow: Step (1): Following the contract specification, the partners form an end to end workflow, named "the public workflow". Step (2): The public workflow is partitioned onto domains: each interacting entity (e.g. customer, supplier, partner, etc.) is associated with a domain. Public parts, while related to each domain, form the definition of the inter-organizational workflow net (IOWF net). An IOWF-net consists of a set of WF-nets, a set of channels, a set of methods, and a set of channel flow relations. Methods represent the public view of domain tasks, and are associated to channels through the channel-flow relations. Step (3): Each domain creates its private workflow, which, for preserving the overall workflow, must be a subclass of its public part. The model focuses mainly on the control-flow view of the process, whilst other aspects of the process are not dealt with (Data flows, time constraints). Workflow nets are based upon Petrinets that are constrained to a single-token per place. A WF net must have one initial place and one final place, and can be verified for soundness. A sound WF net has to be connected, safe, deadlock free, and free of dead transitions. Moreover, sound WF nets always terminate properly (i.e., reach an end state which contains one token in the final place and no tokens in other places of the net). Aalst and Basten [130] defined the projection inheritance criterion, which establishes equivalence of WF-nets, preserving soundness under a specific list of change operations. The criterion ensures the equivalence of WF nets without accessing WF instance data. They also provided a diagnosis tool called Woflan to automatically check inheritance equivalence rules for two given schemas. The projection inheritance is intended to render schema change impact as local as possible. Through projection inheritance soundness of the process can be maintained, while preserving the privacy of the internal processes implemented by each 13 organization. However, it would be difficult to deal with information beyond the basic control flow of the process (e.g., data flow). Excluding data tokens from the model (that is, addressing control flow only), the problem is largely simplified. Moreover, there is no inheritance relation that allows state order change, parallelization of otherwise sequential transitions, etc, thus limitations are formed on the potential internal processes. The model does not deal with time or performance constraints. Time constraints affect sets of process paths and might cause difficulties in retaining inheritance equivalence. The mapping of P2P concepts to CM concepts (including the relevant basic concepts taken from GPM and Bunge's ontology) is presented in Error! Reference source not found.. The equivalence in the table can be full (+), partial (P), or none (-). Table 2. Mapping P2P concepts to CM concepts CM concept P2P concept Process domain Overall workflow Shared domain IOWF/public workflow/public domain Private domain Private domain/workflow State The set of marked places Transformation transition/task --------------------------- Method Event Internal event A change in some place markings (transition). External event A place marking -result of a method’s task firing. Stable state Sink place Unstable state Some place markings (other than the sink) Shared sub-domain state The set of all enabled channels (IOWF net state). mutual state variable Channel/place Process Goal Sink place Law Arcs Obligations Channel-flow relation/arcs/transition input or output. Soft goal constraint ------------------------Process validity criteria Workflow net soundness + projection inheritance criteria ------------ stands for no equivalent concept Equivalence of concepts + P + P + P P P P P P P P P P Detailed explanation of the mapping is given in Appendix 1. In summary, P2P model distinguishes private from shared domains. However, this distinction is not based on identification of mutual properties. Hence, some intrinsic properties of the organizations may be exposed in the public domain. For example, the goal of the overall process, which may be in one of the private domains, is public according to the P2P model. In addition, P2P specifies transitions (methods) in the public domain, whereas in CM the shared (public) domain holds states only, and all transformations are performed (privately) by the partners. Finally, P2P does not address soft-goals or performance constraints. 14 The Contract Workflow Model (CWM) model Within a multi-tier Contract Ontology (MTCO), [5] established a 3 tier ontology for modeling business contracts. The goal was to integrate the contractual terms and conditions into a regular business process workflow, in order to improve the alignment of the business workflow with the expected behavior pattern of the contractual obligation. MTCO integrates several ontologies in order to offer a model that is open and flexible enough to represent many contract types: - Upper level core contract ontology - Specific domain level contract ontology - Template level contract ontology The Upper level ontology is the one that, according to the authors, defines all the required and necessary components of a business contract in order to be legally valid. The Specific Domain level ontology relates to specific contract types and the Template level ontology groups pre-defined contract models, using standard and/or established contract model forms such as the ICC. For the sake of evaluation we are interested in the basic concepts upon which the model is built, which are modeled within the Upper Level Core Contract Ontology. The MTCO main concepts are: Contract, Role, Consideration, Obligation, Performance, Timeframe, Non-performance, Rights and Prohibitions. A contract is an abstract concept representing the contract document. A Role is the responsibility an actor undertakes to perform under the period and scope of the particular contract. A Consideration describes the objective or purpose for which the agreement is being made. An Obligation is a promise made by the contracting parties and the duties and responsibilities that accompany them. Different levels and kinds of obligations are distinguished in the model. Performance is a concept used for defining the expected behavior or actual execution of the contract. It may be a business event or activity like the delivery of ordered goods. Every performance is related to some obligation. A group of performances may be required to fulfill the same primary obligation. Non-performance relates to situations where performance is not achieved as promised. Time-frame is a specified time period after which the contract is either terminated or renewed or a new business contract may be renegotiated. Rights are powers and options awarded by one party to the other empowering the receiver of the rights to certain entitlements or guarantees subject to certain conditions and situations. Finally, Prohibitions specify conditions that must not be met. An occurrence of prohibited acts would result in a contract violation. In Table 3 we establish a mapping of the CWM model to CM (again, including relevant GPM and Bunge's concepts). 15 Table 3: Mapping CWM concepts to CM CM concepts CWM concepts Process domain Shared sub-domain Private sub-domain State Transformation Contract + actors + roles. --------------------------------------Obligation states + performance state. Performance/ performance event, Nonperformance/Non-performance event Event Business events/obligation state change Internal event -------------------External event -------------------Stable state -------------------Unstable state -------------------Shared sub-domain State Obligation states+ performance states Mutual state variable obligation states + performance states Process Goal Consideration, Timeframe Transformation Law Obligation State transformation diagrams, prohibitions. Obligation Obligation (primary, reciprocal, conditional, secondary). Obligation Permissions/Rights Process validity criteria Valid contract criteria Soft-goal ------------------------Annotations legend:(+)= equivalent; (P)= partially equivalent; (-) = non equivalent. Equivalence of concepts P P P P P P P P P P P - Detailed explanation of the mapping is given in Appendix 1. In summary, CWM addresses the entire process domain without establishing private and public domains. Hence, no mapping is found regarding the different domains, the external/internal events, and stable/unstable states. CWM does not address soft-goals and validity criteria either. On the other hand, it distinguishes different kinds of obligations as well as permissions and rights, all mapped to a single CM concept. We note that these distinctions are useful in the context of contract establishment and monitoring, but can all be generalized to a single concept. RosettaNet RosettaNet [28] is consortium of more than 500 companies within the electronic component and information technology industry, which have joined forces to address the need for more accurate information throughout the global supply chain. This goal is achieved by the standardization of RosettaNet Partner Interface Processes (PIPs) to define business processes between trading partners. Each PIP defines how two specific processes, running in two different partners’ organizations, will be standardized and interfaced across the entire supply chain. A PIP includes all business logic, message flow, and message contents to enable alignment of the two processes. RosettaNet relates to an eBusiness Process, which comprises a set of PIPs. Each PIP is a template of a specific process, including documented specifications of its services, 16 transactions, messages (including dictionary properties), class and sequence diagrams in UML, validation tools, and an implementation guide. A PIP depicts the activities, decisions and partner role interactions that fulfill a business transaction between two partners in the supply chain. Each partner participating in the partner interface process must fulfill the obligations specified in a PIP. If any partner fails to perform a service as specified in the PIP implementation guide then the business transaction is null and void. PIP Specifications include three major sections. These three specifications represent three different levels of abstractions. The Business Operational View (BOV) describes the business goal that the Partner Interface Process is supposed to achieve. The Functional Service View (FSV) describes the transaction dialog. The Implementation Framework View (IFV) includes message guidelines and protocols for communications between software components. In this paper we focus on the BOV aspects of the PIP, partly complemented by FSV. RosettaNet PIP blueprints contain the following sections: (a) Business process definition; (b) PIP purpose; (c) PIP business process flow diagram (figure); (d) PIP start state and end state (specified as pre-and post-conditions); (e) A table specifying partner role descriptions; (f) A table specifying business activity descriptions, including its responsible role and pre- and post-conditions; (g) A table specifying business activity performance controls, which specifies controls such as time limits for specific events (e.g., acknowledgement) and authorizations; (h) A table specifying PIP business documents and its data entities; (i) Business data entity security requirements. Table 4: Mapping RosettaNet concepts to CM CM concept Rosetta Net solution Process domain Shared domain Private domain e-Business Process PIP Private process (including PIP integration) PIP state Business activity descriptions (table) State Transformation Event Internal event External event Equivalence of concepts P + implementation P Part of the Business activity Pre and post conditions Part of the Pre and Post-conditions; Messages defined within the Functional service view Stable state PIP state, PIP start state, PIP end state Unstable state ------Shared sub-domain state PIP state mutual state variable Business documents Process Goal PIP end state Law PIP business process flow diagram Obligations Business activity descriptions (table) + Business activity performance controls Soft goal constraint Included in the Business activity performance controls) P P P + + + + P P + P 17 CM concept Rosetta Net solution Equivalence of concepts Process validity criteria PIP definition conditions P A detailed explanation of the mapping is provided in Appendix 1. Criteria Evaluation Based upon the previous mapping, we proceed in the following to the evaluation of the investigated models, presented in Table 5. Table 5. Models evaluation Criterion/criterion evaluation Rosetta Net P2P evaluation CWM evaluation P (P2P Methods are shared domain transformations, done by one of the partners) P (Channels represent mutual state variables, but there is no structured distinction of mutual from intrinsic variables) (All transformations are visible to all domains) P (through the method concept the IOWF is related to internal Petri-nets; yet no QoS variables + (the view of a partner includes its private workflow and the public one) P (Obligation states, like all states, are shared. No QoS variables) P ( methods mapping, partial because of missing QoS variables) (performance constraints are not identifiable in P2P) P (Obligations and performance states represent partially shared domain states, missing QoS variables) (performance constraints are not identifiable in CWM) (performance constraints are not identifiable in P2P (performance constraints are not identifiable in CWM) Non Central governance Non existence of transformations outside of private domains. Separation between shared sub-domain state variables and private state variables + (PIP activities are done by one of the partners, although all of them are visible) + (RosettaNet deals only with documents to be shared) (No distinction of mutual from intrinsic variables) BP run time governance assurance Shared sub-domain state definition and association between obligations and sets of shared sub-domains states Partners' Access to mutual state variable values. Definition of obligations for each shared sub-domain state. Definition of performance constraints for each shared state Mapping of the shared state performance constraint to internal performance + (PIP state, business activity, and business activity performance controls) + (only shared documents (mutual variables) are modeled and all partners have access to them) + (Business activity performance controls) P (as part of the PIP business activity performance controls; limited to specific ones defined) (RosettaNet focuses on the shared domain; the private domain + (access to all variables, not only mutual) 18 constraints (on private interaction states) Inclusion of QoS state variables in obligation definition details are out of scope) + (the business activity performance associated to the PIP ) model) (performance constraints are not identifiable in P2P) (performance constraints are not identifiable in CWM) P (Channels represent mutual state variables, but P2P offers no method to distinguish mutual from intrinsic properties) P (Method are shared transformations visible to external domains, but there are private transitions as well) (Process law exist in the IOWF) (In CWM all process variables are shared) Domain Privacy conservation Distinction between internal and mutual state variables. + (only shared documents which are mutual variables are modeled in RosettaNet) No shared sub-domain transformations P (PIP activities are done by one of the partners in his internal domain, although they are publicly visible) No shared process law (PIP blueprint, shared between partners, includes the process law) + (PIP state is governed by the set of business activity descriptions and business activity performance controls) Shared domain state transition only through obligations fulfillment. P (The shared domain has a process law by itself. However state changes are related to channel-flow relations, partially mapped to obligations) (All performance/ performance events, Non- performance events are shared in CWM) (All Obligation State transformation diagrams, prohibitions are visible to all partners) P (No shared domain identification, but the obligations and performance states are the mechanisms of process state transition) Domain Autonomy conservation (private process design & run time) Design constraints imposed only on mutual properties through obligation definition. Run-time constraints imposed only on mutual properties through obligation definition. + (Only the shared documents are modeled; each partner implements its internal processes based upon the desired PIP's for attaining its transaction goals + (Rosettanet addresses and constrains only exchanged documents & data) P (Projection inheritance constraints on internal processes) (No distinction of mutual from intrinsic properties, process design addresses all properties) P (Methods are visible and constrained in run time, internal transitions are not) ( no distinction between mutual and intrinsic properties exist ) P (Channels are mapped to internal places) P ( no distinction between mutual and intrinsic properties exists) P (P2P has one (shared) sink place) P (CWM's considerations and contract timeframe Validity of the overall process Private process model includes mutual properties in addition to intrinsic ones. Goal state identification (shared and unshared ones) (only mutual properties are modeled; private ones are out of scope) P (only shared goal states are modeled 19 Inclusion of the relevant external event in the obligation definition Existence of a Validity criteria for the private process through the PIP end state; the goal of the overall process is not necessarily modeled) + (Pre- and postconditions of business activities) (only for the interorganizational process is validated in RosettaNet) are all shared). P (P2P channels and channel-flow relations provide the relation between methods of different domains) + (although there is no distinction of external and internal events, obligation state transition diagrams specify events for each transition) (CWM provides no mechanism to assure the overall process validity) P (WF net Soundness + projection criteria. However, they address process structure rather than goal reachability) For each state of the + + + shared sub-domain, (PIP validation (Soundness criteria + (Obligations states there must be at least criteria) projection criteria diagrams ) one obligation ensure that each associated with it. channel is associated with methods through channel flow relations) Annotations legend: (+) = compliant; (P) = partial compliant; (-) = non-compliant. Evaluation Summary The P2P solution establishes an extensive implementation model through the concept of projection inheritance, and the adoption of Petri-nets at the basis of its workflow net. This provides a solid run time governance of the process state. However, the model does not support performance constraints (time and quality constraints over process paths), nor does it deal with data flow modeling, which may lead to a risk of incomplete transformation and law definition. Its focus is on the construction of sound private workflows that maintain the soundness of the public IOWF. To this end, the P2P approach models the shared domain using IOWF-net, mapped through projection inheritance into the private workflows. The existence of shared transformations and laws and the lack of an explicit distinction of mutual properties from intrinsic domain properties may imply the external visibility of intrinsic private domain properties; hence domain privacy may be compromised. The privacy of the parties can be improved if it is explicitly checked that no intrinsic properties are exposed. As well, P2P includes the process goal in the public workflow, although it may be within some private domain. When the exposure of the goal state is not necessary, it is recommended to avoid it, thus gain an improved privacy. P2P process validity verification relates to its soundness, which contributes to the goal reachability in terms of the process logical flow, but not, e.g., the completeness of its 20 definition. As well, validity of data flows and transformations are out of the scope of the model. The soundness of the private processes can be established wherever the projection inheritance criterion is applicable. Hence, it provides some flexibility to the partners in designing and modifying their internal processes as long as projection inheritance is kept. However, projection inheritance is applicable under specific conditions, and is not applicable whenever the necessary modification in the private domain implies changing state order, parallelization of otherwise sequential tasks, etc. Performance (soft goal) constraints mapping may result in design modifications of internal process domains (state laws, transformation laws, adding state variables). As well, process path selection may be notoriously changed by soft-goal related conditions. P2P provides no mechanism for such modeling. The CWM solution establishes a contract workflow model based upon an extensive ontology of contracts. Herein we consider only the generic layer of the MTCO ontology. CWM defines the end-to-end process in a completely visible manner. Its focus is supporting the transition from a business legal contract to an operational workflow model. During both process design and process implementation all domains are fully visible. Hence, the domains' autonomy and privacy are compromised. Although this provides a high level governance of the process state at run time, it disables the capability of the partners to modify their internal processes or to deviate from their definition, thus decreases their flexibility. To gain some autonomy and flexibility, it may be recommended to model the internal activities of the organizations at a low level of granularity and keep their details hidden. CWM offers an exhaustive model for obligations and distinguishes different types of these. It manages and monitors the life-cycle of obligations. It also addresses goals (referred to as considerations). However, it does not enable soft-goals and performance constraints mapping and handling. Soft-goal constraints mapping may result in design modifications of internal process domains (state laws, transformation laws, adding state variables). As well, process path selection may be notoriously changed by soft-goals related conditions. CWM provides no mechanism for such modeling. Finally, CWM postulates that the contract document which is agreed upon between partners allows the specification of a valid overall process. However, process validity criteria are not established. Note that, as mentioned, our analysis relates to the upper-level ontology of CWM, while disregarding the lower-level ones, which provide additional concepts, whose value is in the context of contracts between organizations. However, the lower-level ontologies rely on the upper-level one and extend it, thus they preserve its assumptions. 21 The Rosettanet approach is focused on facilitating the interaction within an interorganizational business process. As such, it seems to provide a high level of compliance regarding the autonomy and privacy requirements. PIP's provide a solid framework for defining the building blocks of possible transactions between partners, with a well defined set of validity criteria for the PIP, including obligations and performance constraints definition at the PIP level. However, RosettaNet PIP's are only the building blocks for an overall process, which is not modeled in RosettaNet, leaving the internal process design to other methodologies or approaches, thus avoiding a view of an overall process whose validity needs to be established. As compared to, e.g., P2P, RosettaNet preserves privacy by relating only to information that needs to be shared. However, while P2P supports the construction of compatible and sound private processes, this aspect is not addressed at all in RosettaNet. Moreover, each PIP in RosettaNet is designed for a specific use. If processes that are not predefined are to be implemented, they first need to be modeled, thus the RosettaNet would need to be extended for that. RosettaNet is designed for a specific market segment, and not a generic one (EC and IT markets specific). Although RosettaNet provides specifications that encompass many aspects of an automated business relationship, it does not address problems of aligning legacy systems and internal business processes towards the specified application integration scenarios. RosettaNet can beneficiate from the conceptual model established in this paper to link its PIP building blocks as elements within the overall business process. This would provide also for overall BP validity criteria, essential for linking the PIP's goals to the overall process goals. 6. Discussion We have established a minimal but consistent set of requirements which are related to the required balance between the control over the process and the autonomy and privacy of the partners organizations involved in business transactions. The completeness of such model can be assessed with respect to: (a) the set of requirements it is aimed to fulfill. Our model was designed based on a defined set of requirements and derived from these requirements. Considering possible requirements which are not in this set, the model may need to be extended. (b) The expressive power of the model in representing the phenomena under consideration. Our model is based on Bunge’s ontology, which has been widely used as a basis for evaluating the expressive power of models. This basis provides for completeness in terms of the model’s expressive power. From these requirements we derive a formal, ontology based Inter-organizational BP conceptual model, based upon a formal generic process model, the GPM. 22 Establishing a CM prior to selecting potential implementation solutions provides a formal and valid framework for analyzing potential solutions, while remaining at a high level of abstraction and allowing traceability between the solution implementation and the business requirements. Key concepts in the model are the domains, whose demarcation is clear. Defining the shared sub-domain as a sub-domain whose states are all stable clarifies the responsibility of the organizations to transforming the state of this sub-domain and makes their obligations concrete. The concept of obligations as it is defined in the model minimizes the constraints posed on the internal operation of the participating organizations. The concept of state variables is highly generic, and may bear a large variety of forms. This can be considered both as a strength and as a weakness. It is a strength because it enables the model to address diverse issues, such as status attributes, data flows, and performance constraints. It is a weakness when implementation is considered, because these issues require different implementation mechanisms, which are not addressed by the model. Hence, such mechanisms should be supported when an implementation solution is designed. Based upon the CM, we presented an evaluation procedure that relates to our underlying set of requirements, and applied it to three potential BP implementation solutions in light of the established set of BP requirements: the P2P approach [17], the CWM approach [4], and RosettaNet [28]. The findings provide a clear assessment of the strength and weaknesses of these approaches with respect to our requirements, and support the selection of an appropriate solution. The evaluation itself is comparable to ontological evaluation of modeling languages [24]. Green and Rosemann [18] discuss weaknesses of such evaluations, including lack of completeness, lack of objectivity, lack of guidance, and others. They propose a reference methodology for conducting ontological analysis, based on meta-models. The evaluation procedure proposed in this paper follows this approach. The first step of concept mapping relates to meta-model concepts; cross-mapping is intended to ensure completeness, and the evaluation criteria indicators establish a direct relation between the concept mapping and the evaluation, in order to provide a strong guidance when performing the evaluation and to avoid a lack of objectivity. The evaluation procedure can be applied to any other inter-organizational process solution providing a clear mapping of possible adaptation for satisfying the above list of requirements. However, our model is based on a specific set of requirements. It may need to be extended or modified to satisfy different sets of requirements. In fact, we recognize that in many cases not all our requirements are required. Such is the case, for example, of associations of organizations which have a central governance committee, where the non-central governance 23 assumption is not relevant. Another example is the case of a supply chain established between partners, where powerful partners may impose visibility of details on other partners, less powerful. Even in this case the work need not be redone from scratch, as the model is at at a high level of abstractions and traceability of the requirements to conceptual model components is straight forward. Finally, we highlight the importance of a sound requirement based conceptual model as an underlying basis for BP implementation solutions. 7. Conclusion The importance of inter-organizational processes has been widely recognized, leading to a variety of approaches and proposed solutions to their design and implementation. While most of the approaches focus on a specific aspect of these processes and on their implementation details, this paper focuses on a high-level conceptual model, aimed at achieving an understanding of the interaction among organizations in situations where particular requirements are made. Proposing this model, the paper makes two main contributions. First, the model is sound and theory-based, and provides a sharp understanding of how organizations can interact in ontological terms. It satisfies a set of restrictive requirements regarding the governance of the overall process and the conservation of privacy and autonomy of the participating organizations. Furthermore, it can be used for designing an inter-organizational process, as shown by 1.32]. Second, we propose the CM as a platform for evaluating potential solutions for inter-organizational process establishment. This proposition relies on the generic nature of the model, on its sound theoretical basis, and on its explicit relation to specific business requirements. By using the established conceptual model, we have illustrated the ability to map different implementation models and compare their compliance with our preset model requirements. We have also demonstrated how possible improvements can be made to the examined approaches based on our model. Future research may address other sets of requirements and identify the difference required in the underlying conceptual model when such different assumptions are made. We believe that the current model provides a basis that can, with minor modifications, be valid for other sets of requirements too. Another research direction is to develop an implementation solution based on the model, which will address the entire scope of issues represented by state variables and allow control as well as privacy and autonomy. 24 References 1. Aalst, W. M. P and Kumar, A. (2003) “XML-Based Schema Definition for Support of Inter-organizational Workflow”, Information Systems Research 14(1), p. 23-46. 2. Grefen P., Abere K., Hoffner Y. And Ludwig H. (2000), “CrossFlow: CrossOrganizational Workflow Management in Dynamic Virtual Enterprises“, International Journal of Computer Systems Science & Engineering, 15 (5), p. 277-290. 3. Kabilan V. and Johannesson P. (2003), “Semantic Representation of Contract Knowledge using Multi Tier Ontology“, Proceedings of Semantic Web and Databases Workshop, (SWDB 2003). 4. Kabilan V. (2005), “Contract Workflow Model Patterns Using BPMN“, EMMSAD’05, Proceedings of CAiSE’05 workshops Vol. 1, Porto, Portugal, p. 557-568. 5. Kabilan V., Johannesson P. and Rugaimukammu, D. (2003), “Business Contract Obligation Monitoring through Use of Multi-tier contract ontology“, Proceedings of Workshop on Regulatory Ontologies (Worm Core 2003), Italy, (LNCS 2889), SpringerVerlag, Berlin 6. Kafeza E., Chiu D. K.W. and Kafeza I. (2001), “View-Based Contracts in an E-Service Cross-Organizational Workflow Environment“, Proceedings of TES 2001 (LNCS 2193), Springer-Verlag, Berlin, p. 74-88. 7. Cassati F. and A. Discenza (2001) “Modeling and Managing Interactions among Business processes”, Journal of Systems Integration (10), pp. 145-168. 8. Guelfi N., G. le Cousin and B. Ries (2004), “Engineering of Dependable Complex Business Processes Using UML and Coordinated Atomic Actions“, Proceedings of OTM workshops (LNCS 3292), Berlin: Springer-Verlag, pp. 468-482. 9. Weigand, H. and Van den Heuvel, W.J, (2002), “Cross-organizational workflow integration using contracts“, Decision Support Systems 33(3), p. 247-265. 10. Business Process Management Initiative (2006), http://www.bpmi.org. 11. ebXML BPSS (Electronic Business using eXtensible Markup Language) (2006), ebXML BPSS specification, website:http://www.ebXML.org/. 12. OMG (Object Management group) (2006), Business Process Modeling Notation (BPMN), http://www.bpmn.org/ 13. W3C (World Wide Web Consortium) (2006), http://www.w3.org. 14. OASIS (Organization for the Advancement of Structured Information Standards) (2006), http://www.oasis-open.org. 15. WFMC (Workflow Management Coalition) (2006), “Workflow Management Coalition Interface: Process Definition Interchange Process Model“. http://www.wfmc.org. 25 16. Camarinha-Matos, L.M and Afsarmanesh, H. (Eds) (1999), “Infrastructures for virtual enterprises –Networking industrial enterprises“, Kluwer Academic Publishers. 17. Aalst, W. M. P and Weske, M. (2001) “The P2P Approach to Inter-organizational Workflows“, Proceedings of CAiSE’01 (LNCS 2068), Springer Berlin p. 140-156. 18. Green, P. F., Rosemann, M. and Indulska, M. (2005), “Ontological Evaluation of Enterprise Systems Interoperability Using ebXML“, IEEE Transactions on Knowledge and Data Engineering 17(5), p. 713-725. 19. Soffer, P. (2005), “On the Notion of Flexibility in Business Processes“, Proceedings of the CAiSE’05 Workshops, p. 35 – 42. 20. Soffer P. and Ghattas J. (2006), “Business Process Flexibility in Virtual Organizations”, Workshop on Business Process Modeling, Design and Support (BPMDS’06), Proceedings of CAiSE’06 Workshops, p. 188-197 21. Bunge, M. (ed.) (1977), “Treatise on Basic Philosophy: Volume 3: Ontology 1: The furniture of the world“. Reidel, Boston. 22. Bunge. M. (ed) (1979), “Treatise on Basic Philosophy: Vol. 4, Ontology II: A World of Systems, Reidel, Boston. 23. Wand, Y. and. Weber, R. (1990), “An Ontological Model of an Information System“, IEEE Transactions on Software Engineering, Vol. 16, No. 11, pp. 1282-1292. 24. Wand Y, and Weber R. (1993), “On the ontological expressiveness of information systems analysis and design grammars“. Journal of Information Systems 3 pp. 217–237 25. Soffer, P., and Wand, Y. (2005), “On the Notion of Soft Goals in Business Process Modeling“, Business Process Management Journal 11(6), p. 663-679. 26. Soffer P. and Wand Y. (2004), “Goal-driven Analysis of Process Model Validity“, Advanced Information Systems Engineering (CAiSE’04) (LNCS 3084), p. 521-535 27. Zdravkovic J. and Kabilan V. (2005), “Enabling Business Process Interoperability Using Contract Workflow Models“, OTM 2005: CoopIS, DOA, and ODBASE: OTM Con. International Conferences, Cyprus (LNCS 3760) Springer-Verlag, Berlin, pp. 77-93. 28. Rosettanet Organization & Standards, website: http://www.Rosettanet.org/, 2007. 29. Damodaran. S (2004) “B2B integration over the Internet with XML: RosettaNet successes and challenges”, Proceedings of the 13th international World Wide Web conference, pp. 188-195 30. Aalst, W. M. P and Basten, T (2002), “Inheritance of Workflows: An approach to tackling problems related to change“, Theoretical Computer Science 270(1-2), pp. 125-203. 31. Green, P. and Rosemann, M. (2005) "Ontological analysis of business systems analysis techniques: experiences and proposals for an enhanced methodology" in P. Green and M. Rosemann (eds.) Business Systems Analysis with Ontologies, IDEA, Hershey, pp. 1-27. 26 32. Ghattas J. and Soffer P., (2007) “Facilitating Flexibility in Inter-Organizational Processes: A Conceptual Model”, International Journal of Business Process Integration and Management (forthcoming) APPENDIX 1: Models detailed mapping Detailed mapping of P2P CM : P2P concept Explanation Process domain (CM); Overall workflow (P2P) The overall workflow in P2P includes the public and private workflows, as the process domain in CM unites all the partner domains (including the shared sub-domain). Shared domain (CM); A P2P Inter-organizational net (IOWF net) represents the public part of the IOWF/ public workflow/ workflow. While the CM shared domain includes no process (only stable public domain (P2P) states), the IOWF net describes a complete process model, where methods ate transformations generated by partners, defined in the shared domain; hence the mapping is partial. Private domain (CM); P2P offers a complete process model for the private domain based upon the Private workflow (P2P) WF-net concept. CM private domains set the scope the private processes. Transformation (CM); P2P has transitions/tasks in the private workflows, and method, which are transition/task (P2P) transformations that are visible-to all domains. CM transformations are all private. CM postulates that the minimal requirement for a valid inter(no equivalent in CM); organizational process definition is to define the shared states (mutual Method (P2P) properties, their associated semantics and possible values transitions) and obligations (partners' commitment to implement the needed transformation of shared states within their domains). Hence there is no need for transformation visibility outside the private domains. The transformation public visibility inevitably exposes intrinsic domain properties to the public. Internal event (CM); CM establishes the concepts of internal/external events and stable/unstable Place markings change states with respect to a particular domain. In P2P transitions in the private (P2P) workflows cause internal events to these domains, while methods fired by External event (CM); other partners cause external events Place marking after method firing (P2P) Stable state (CM); Sink The stability of a state in a private domain in P2P cannot be seen. Hence, place (P2P) only the sink place in P2P is identified as stable, while the others are Unstable state (CM); considered unstable Place marking (P2P) Shared state (CM); The A channel is mapped into places within the private workflow. A token may set of all enabled arrive to this place whenever a transition (representing a method) is fired. channels (P2P) Hence an enabled channel's place may be compared to the activation of a Mutual state variable shared state. P2P channels represent mutual properties' values. Hence they (CM); Channel/place are mapped to mutual state variables. However, not all mutual state variables relate to channels (e.g., QoS), thus the mapping is partial (P2P) Process Goal (CM); Sink In CM there may be multiple process goal states for each domain (process, Place (P2P) private and shared). While private domains have a set of defines goal states, the overall process domain goals may be located in part of the domains and be invisible to other partner domains. P2P IOWF net sink places represent shared goal states, while WF-nets may represent private goal states in addition to the mapping of shared goals into the private domain. Hence there may be cases where CM unshared goal states are shared within the P2P model. Law (CM); Arcs (P2P) CM law is private by definition. Arcs in P2P may represent public laws, which have no equivalent in CM. Obligations (CM); A channel flow relation is a unidirectional association from a specific method Channel-flow; to a specific channel. A method generates a token that is passed to another 27 relation/arcs/input or output of a transition (P2P) Soft goal (CM); (none in P2P) Process validity criteria (CM); WF net soundness criteria+ projection inheritance criteria (P2P) domain through a channel. A method of domain A is mapped to a method of domain B through channels and channel-flow relations. Hence channel flow relations are state mappings between domains. This is somehow equivalent to CM obligations concept. P2P establishes no Soft goal (Performance) constraints in general. CM performance constraints are associated with obligations; these obligations are mapped into the different domains, as state laws that ultimately are constraints/conditions set upon state variables. IOWF nets are ultimately mapped into WF-nets, for which soundness criteria exist (no deadlock, live-locks, free of dead transitions). The projection inheritance assures for some cases that soundness of the process is maintained through the construction of the private workflow-net. Unlike CM process validity, soundness relates to structural properties of the process rather than to its goal. Detailed mapping of CWM CM : CWM concepts Process domain (CM); Contract + actors + roles (CWM) Shared sub-domain (CM); (none in CWM) Private sub-domain; (CM); (none in CWM) State (CM); Obligation states; performance state (CWM) Transformation (CM); Performance/ performance event, Non-performance/ Non-performance event (CWM) Explanation CWM provides a model for the end-to-end process, through the contract definition, actors and roles mapping. Although it may be argued that actors represent domains, their full process visibility is not consistent with CM domain definition. All the actors' processes are completely visible and there are no demarcations of domains. Domain goals, states, laws are not distinguished from overall process domain goals, states and laws; hence CWM establishes neither private domains nor shared domains. As all the actors' processes are public, all the properties are shared by definition. Hence, CWM does not identify mutual properties, although obligation states and performance states may represent them implicitly. CWM defines for each obligation a state transformation diagram, where possible obligation states are: inactive, active, pending, fulfilled or terminated/ cancelled. Each obligation is associated with a set of performance and non-performance events, which have their own states. Hence the process state is described through the obligation state and its performance/non-performance events' states. CWM performance events and non-performance events are transformations that are visible to all interacting actors. CM transformations visibility is limited to a single specific private process domain. While CM established a single concept, CWM offers two concepts: one for expressing transformations that handle desirable process flow and the other for handling undesirable situations. Such distinction highlights exception handling, important in the context of the contract between the parties. However, on the conceptual level they are of the same kind. Event (CM); Business events/ obligation state change (CWM) Internal / External event; Stable / As all the domains are shared, all states are shared and there are no private states; Unstable state (CM) hence there is neither distinction between stable/unstable states nor Shared State (CM) external/internal events, except for the end-to-end process domain, where external Obligation states+ events may be originated by third parties. performance states (CWM) Shared state variable (CM); obligation states + performance states (CWM) 28 CM : CWM concepts Explanation Process Goal (CM); CM enables defining different goals for different domains (process, private and Consideration, shared). If the process goal is in a private domain it is not visible to the others. Timeframe (CWM) CWM's considerations and contract timeframe are all shared; there are neither private goals, nor process goals visible to one side only. Law (CM); CWM distinguishes between lawful and unlawful transformations using two Obligation State different concepts: lawful transformations are expressed within the obligation state transformation transformation diagrams, while unlawful transformations are represented though diagrams, the prohibitions concept. prohibitions (CWM). Obligation (CM); (1) CM obligations form part of the shared state definitions and represent a Obligation (primary, partner's commitment to transform shared domain states. CWM obligations and reciprocal, roles define the commitment between actors. conditional, (2) CWM obligations establish publicly visible tasks (performance/nonsecondary) (CWM) performance events), associated with actors through roles. CM defines only mutual properties and their required transformations and leaves the task definition to the private domains. (3) CWM obligations types include primary, secondary, reciprocal, and conditional. CWM obligations may be nested (primary obligations may include secondary obligations), while CM obligations are not. CM obligations may be mapped to CWM conditional (triggered in specific conditions) obligations; CM obligations represent both primary and reciprocal obligations as inherent in the definition of CM obligation, i.e., commitment of the domains to implement a set of transformation of shared state variables for a given shared state. Obligation (CM); CWM distinguishes obligatory obligations ("obligations") from optional ones Permissions/Rights ("rights"). CM obligations are synchronous (triggered by shared events), as (CWM) opposed to rights which are asynchronous. While rights may be enacted by any actor, it is not clear what mechanism helps other actors to identify their triggering. This may cause ambiguity in the state of the overall process as perceived by other actors, and consequently a risk of loosing control, or misinterpretation of obligations. However, with a more accurate definition these concepts can be partially mapped to obligations. Process validity CM process validity criteria apply to both the inter-organizational process and the criteria (CM) Valid private processes. CM obligations assure the overall process validity, while private contract criteria process validity is assured by each organization, based on goal reachability. CWM (CWM) claims that a valid contract is established between two contracting parties/actors, but it provides no mechanism to assure the overall process validity. Soft-goal (CM); CWM establishes no equivalent for CM Soft goal (performance) constraints. CM (none in CWM) performance constraints are associated with obligations mapped into the different domains as constraints set upon state variables. Detailed mapping of RosettaNet CM : RosettaNet concepts Process domain (CM); eBusiness process (RosettaNet) Shared sub-domain (CM); PIP blueprint (RosettaNet) Private sub-domain; (CM); Private process (RosettaNet) State (CM); PIP start state, PIP end State (RosettaNet) Explanation RosettaNet relates to an overall process termed the e-Business process for B2B transactions. However, it defines only the inter-organizations aspects of the transactions the partner should implement between them, which would be a subset of all RosettaNet PIP's. Each Partner needs to model its own private processes, but RosettaNet does not provide any model to be used internally. The details of the implementation are completely private and out of scope of RosettaNet. Rosettanet does not address the overall process state. RosettaNet PIP state represents the shared state. Hence the CM state has no complete equivalent in RosettaNet. 29 CM : RosettaNet concepts Transformation (CM); Business activity descriptions Event (CM); Business activity Pre and post conditions, Messages within the Implementation view (RosettaNet) Internal / External event (CM); Pre/post conditions (RosettaNet) Stable / Unstable state (CM); Business activity (RosettaNet) Shared State (CM) ; PIP state (start, end) (RosettaNet) Shared state variable (CM); Business data document (RosettaNet) Process Goal (CM); No equivalent (RosettaNet) Explanation RosettaNet defines the transformations required of each partner within the business activity descriptions, which provides also a definition of the pre and post conditions of the transformations. RosettaNet does not provide the overall process model- hence only B2B transactions are modeled – which provide a partial equivalent to CM transformation concept. RosettaNet Business activity Pre and post conditions are the triggers for the PIP state change; However as RosettaNet models only the inter-org. aspects of the private process these concepts are only partially equivalent to CM event concept; RosettaNet Pre/Post conditions that trigger another activity in the same domain are internal events to that domain. RosettaNet Pre/Post conditions that trigger another activity in the same domain are internal events to that domain. All the rest of pre-post conditions would be equivalent to external events. RosettaNet models only internal events related to the inter-org. process; in PIP's where several related transformations need to be occurring in a sequence there may be unstable states. These states would be invisible to the external partners- however their effect upon shared documents (that are mutual properties) would be visible to other partners. RosettaNet provides a complete model of the PIP state, which is based upon the shared business document transformations. PIP business documents diagrams (table) included in the PIP blueprint are shared state variables. CM enables defining different goals for different domains (process, private and shared). If the process goal is in a private domain it is not visible to the others. RosettaNet 's PIP end state represents the goal of executing the PIP. However the goal of the overall process is not defined. Law (CM); PIP business process flow diagram (RosettaNet). RosettaNet PIP business process flow diagram represents the PIP subprocess law. However the overall process law is not modeled in RosettaNet. Obligation (CM); PIP Business activity + Business activity performance control (RosettaNet) CM obligations form part of the shared state definitions and represent a partner's commitment to transform shared domain states. RosettaNet PIP defines the commitment between partners. RosettaNet obligations establish publicly visible tasks (business activity diagrams), associated with actors through roles, though these tasks are applied only to shared documents (that are the equivalent of CM mutual properties) CM process validity criteria apply to both the inter-organizational process and the private processes. CM obligations assure the overall process validity, while private process validity is assured by each organization, based on goal reachability. RosettaNet defines PIP validity criteria, however these do not ensure goal reachability. RosettaNet establishes PIP soft-goals within the Business activity performance control included in the PIP blueprint. CM performance constraints are associated with obligations mapped into the different domains as constraints set upon state variables. The equivalency between the two is only partial as the overall process soft-goals are not modeled. In addition, only specific pre-defined soft-goals can be addressed, while others cannot be added. Process validity criteria (CM) Valid contract criteria (RosettaNet) Soft-goal (CM); Business activity performance control (RosettaNet) 30