Even Web Services Can Socialize: A New Service-Oriented Social Networking Model Zakaria Maamar Zayed University Dubai, U.A.E [email protected]

Leandro Krug Wives Instituto de Informatica UFRGS Porto Alegre, Brazil [email protected]

Abstract The incessant growth of the number of Web services (WSs) makes their discovery more difficult. Standard discovery registries, such as UDDI and ebXML, have their own inherent limitations as they only describe the functionality aspect of each WS and not how it relates to others. Capturing the relationships between WSs as they interact with each other can be useful in many ways. In this paper, we present a novel model that captures such relationships using social networks. We describe how these WSs social networks can be initiated and how they evolve. We also discuss the types of associations (edges) that can exist among WSs (nodes). Traversing a WS social network makes it possible to identify a community of homogeneous WSs that are functionally similar. It can also enable us to use the model as a recommender system in case we need to replace a faulty WS, find a partner WS, or an add-on WS that may enrich the current business scenario. Keywords. Discovery, Metric, Social Network, Web service.

1. Introduction Web services are paving the way for a new generation of loosely-coupled and cross-enterprise business applications. This can be noticed from the large number of standards and initiatives related to Web services [21], [25], [32], which tackle a variety of issues such as security and fault tolerance. These issues hinder, to a certain extent, the automatic composition of Web services, which is a cornerstone of these business applications. Service composition, which is one of Web services’ selling points, handles users’ requests that cannot be satisfied by any single (atomic) Web service available, whereas a composite Web service obtained by combining available Web services may be used. Despite the tremendous capabilities that Web services have for the development of business applications, they still lack some capabilities (e.g., self-assessment, self-healing) that would propel them to a higher level of adoption by the IT community and make them compete with other traditional

Youakim Badr Universit´e de Lyon INSA-Lyon, F-69621, France [email protected]

Said Elnaffar UAE university Al-Ain, U.A.E [email protected]

integration middleware such as CORBA, DCOM, and JBI. Due to this shortage of capabilities, the adoption of Web services could be slowed down if some persistent issues like the complexity of these Web services discoveries are not properly addressed [16]. For the particular issue of Web services discovery, in this paper we examine the use of recommendation-based techniques with focus on social networks [8], [31]. Conventional social networks establish among people different kinds of relationships such as friendship, kinship, collegiality, and conflict. These relationships are dynamic and hence, adjusted over time depending on different factors like outcomes of previous interaction experiences, penalties incurred following the execution of prohibitive actions, etc. Within a social network, the person who questions the trustworthiness of a colleague could either break or review the trust relationship that she has with this colleague. Replacing people with Web services is doable since Web services are constantly engaged in different types of interaction sessions with users and peers as well [18], [20]. For example, Manuel Serra da Cruz et al. identified interaction patterns between Web services and users with the purpose of offering users better interaction patterns in the future [20]. The purpose of the discovery process is to find a suitable Web service for a given consumer’s request. A consumer could be a user or another Web service. This process relies considerably on WSDL documents that Web service providers post on registries like UDDI and ebXML. Unfortunately, current UDDI and ebXML registries still grapple with a number of inherent shortcomings (e.g., lack of semantics, security) despite multiple extensions and improvements that are reported in the literature [23], [30]. We trust that Web services social-networks can help address some of these limitations by letting Web services take advantage of the previous composition scenarios in which they participated so they can establish relationships with peers that also participated in these composition scenarios. The followings are our contributions: (i) define social networks in the context of Web services, (ii) support Web services as they build, use, and maintain their respective social networks, and (iii) adopt these social networks to

discover Web services. The rest of this paper is organized as follows. Section 2 give an overview of social networks. Section 3 discusses the use of social networks for Web services discovery. Conclusions are outlined in Section 4.

2. Overview of social networks Social networks have been used in different domains like social and political sciences, Artificial Intelligence (AI), and e-business [13], [26], [27], [33]. According to Ethier, “the study of social networks is important since it helps us to better understand how and why we interact with each other, as well as how technology can alter this interaction. The field of social network theory has grown considerably during the past few years as advanced computing technology has opened the door for new research” [8]. Generally, a network consists of nodes and edges. Nodes refer to any type of object (or entity) like individuals or organizations, and edges refer to relationships (or associations) between these nodes such as the degree of friendship between two persons or the distance between two cities. Relationships are sometimes directional, bidirectional, with weight, or a mixture of all of these. Research in a number of academic fields has revealed that social networks operate on many levels, from families up to the level of nations, and play a critical role in determining the way problems are solved, organizations are run, and the degree to which individuals succeed in achieving their goals [10], [22]. In the field of Distributed AI (DAI), social networks have been widely used in the specification of coordination, cooperation, and negotiation mechanisms of software agents. According to Castelfranchi, “an agent can be helped or damaged, favored or threatened, it can compete or cooperate” [6]. The idea of using social networks in DAI stems from the fact that an agent does not form an independent world, rather it is part of a society. Agents can engage in interactions with peers from the same society or different societies, which requires specific mechanisms to support the progress of these interactions. An example of the use of social networks is discussed in [26] where the reputation of agents are assessed with respect to the position they hold in a society or community. In the field of recommender systems, Donovan and Smyth propose two computational models of trust and show how these models could be incorporated into collaborative-based recommender systems [24]. The authors report that users tend to ask friends (i.e., persons they normally trust and are part of their social networks) for advice prior to taking actions. To address the reliability problem that could undermine provided recommendations, Donovan and Smyth consider historical evaluations that users give to these recommendations. If a user is the source of a good number of accurate recommendations over time, his or her level of

trust as a reliable source increases compared to the one who makes poor or misleading recommendations. Though the aforementioned paragraphs offer a good overview of some initiatives on social networks, there is a lack of initiatives germane to Web services. A good number of questions such as how Web services can build their social networks in relation to composition scenarios, how Web services are discovered using social networks, how they evolve or shrink over time, and what kind of economical dimension can be added to social networks, are left unanswered and, thus, responses are provided on a case by case basis.

3. Social networks of Web services Web services discovery is critical to fulfilling users’ requests. A plethora of Web services exist on the Internet, which makes identifying the right ones not always a straightforward task. Additionally, independent providers continue to deliver almost the same set of Web services with different non-functional properties, which further increases the complexity of the discovery process. Bui and Gacher point out that despite the heterogeneity of Web services, their functionalities are sufficiently well-defined and homogeneous enough to allow for market competition to happen [5]. For example, both the US National Digital Forecast Database XML Web Service (www.weather.gov/xml) and the US National Weather Service Forecast Office (www.srh.noaa.gov/mfl) propose weather forecast-related Web services. In this section, we discuss the development of social networks to discover Web services by answering how to build, use, and maintain social networks. Social networks of WSs are different from classical type of social networks of people (e.g., LinkedIn, Plaxo, Facebook). The latter type is based on the absolute collaboration and mutual assistance between their members (i.e., no competition). Contrarily, in WSs social networks1 , members are mainly competing as each WS wishes to be (i) a part of compositions, (ii) a replacement for a faulty WS, and (iii) nominated as an add-on WS to augment the currently invoked one (e.g., a WS to arrange for conference participation, and an add-on WS to arrange for excursion). Building a WSs social network is an incremental and continual process that begins when a Web service is selected for the first time to participate, either as a partner in a composition scenario or as a replacement in case of a peer failure (partnership and replacement are explained later). Dependency interactions2 between a Web service and other 1. We adopt social networks idea to model and explain some types of interactions between WSs and treat these interactions as “social” relationships. 2. These interactions are part of the business logic that underpins the specification of a composite Web service. For example, a person could buy her air ticket subject to hotel confirmation.

peers in a given composition scenario permits to at least constitute the initial edges of this social network. Future participations of this Web service in additional composition scenarios would permit to extend its social network, but at the same time would call for a possible review of alreadyestablished edges as new edges might have to be added. Another condition that may induce reviewing the status of a social network is when a Web service ceases functioning for various reasons such as withdrawal. If that happens, the node and incoming and outgoing edges associated with this Web service are to be checked out, which may affect the overall topology of the social network. In this context, we can understand that, as Web services interact with each others, new relationships can be formed and existing ones may vanish (i.e., become inactive) or get modified. This is measured and expressed by specific weights that are dynamically adjusted. The following sections describe how Web services social networks are thus, built and maintained.

3.1. Social networks building Building a social network means identifying types of nodes and edges that constitute this network. In our context, Web services are the sole constituents and they designate nodes. In terms of edges, we suggest three types of association between Web services namely Recommendation (R), Similarity (S), and Collaboration (C). Due to lack of space, R and S associations are presented only. Formally, a Social N etwork SN of a Web Service WS is a couple: SNWS = (N , E) where N and E are the set of N odes and Edges, respectively. Each edge e ∈ E is a tuple < WS i , t, w, WS j >, where the edge is directed from WS i to WS j , t denotes the type (or name) of association between WS i and WS j , and w denotes the weight of the association. This weight is a calculated numerical value between 0 and 1. Recommendation-based association (R). In [14], we discuss the combination of recommender systems and Web services in terms of what recommender systems can do for Web services and what Web services can do for recommender systems. We, hereafter, rely on the first part of this combination to show recommendation cases that could stem from Web services interactions. We identify two cases, namely partnership3 and robustness that promote the recommendation-based association. 1) Partnership case (Rp): a component Web service that participates in a composition scenario could propose that new peers should be part of this composition as well. Though the proposed peers are not ultimately required to satisfy the request of a user, they could 3. Partnership is seen from an “add-on” perspective, where the use of a Web service is optional and subject to the user approval.

yield to extra business opportunities with the user. For example, a delegate attending an overseas conference could be interested in sightseeing activities according to his or her profile albeit this delegate did not explicitly express this interest. The new peers (i.e., component Web services) are subject to the delegate’s approval prior to their execution. • Example: : if RoomBookingWS is part of a composition scenario, then it will recommend that SightSeeingWS could be part of this scenario subject to checking and seeking the requestor’s profile and approval, respectively. • Properties: Rp is asymmetric and transitive (transitivity may be limited in terms of transition cycles/paths by a threshold set by the Web service provider4 ). • Association weight for recommendation based on partnership wtRp is given by the following equation: wtRp(W Si ,W Sj ) =

|W Sj selection| |W Si participation|

(1)

where |W Si participation| and |W Sj selection| stand for the number of times that WSi participated in composition scenarios and the number of times that WSj was nominated by WSi to participate in these composition scenarios. The higher the weight is, the better it is for WSj . Remark: |W Sj selection| tends to be smaller than |W Si participation| for two reasons: W Sj participation in compositions could be rejected by users or composition designers, and W Sj could refuse to participate in compositions. In the recommendation-based association, a question that arises is what happens if all weights drop to zero. This issue is common in the recommender system field, and is known as the “cold start problem”, i.e., when new items arrive and the system has no information about them (i.e., evaluations) to start recommending. The solutions for this problem has been reported in many papers [1], [15], [29]. We plan to adopt similar approaches. 2) Robustness case (Rr): a Web service could propose peers that substitute it in case of failure. These peers are identified with respect to functional and nonfunctional characteristics they have in common with the Web service in question. Reasons for substitution 4. Another way to limit transitivity would be the application of a function that controls the propagation of recommendation. Such a function should introduce a minimization rate per transition performed, somehow similar to what Google’s PageRank [4] algorithm does. It propagates the “reputation” of a page to the rest of pages that it refers to.

are multiple including Web service failure to respond to a client’s requests or inability to meet a certain level of Quality of Services (QoS). • Example: : if RoomBookingWS fails at run-time, then HotelReservationWS will substitute for this Web service subject to the approval of the user. • Properties: Rr is asymmetric and transitive (like Rp , Rr is limited in terms of transition cycles/paths by a threshold). • Association weight for recommendation based on robustness wtRr is given by the following equation: wtRr(W Si ,W Sj ) =

|W Sj selection| |W Si f ailure|

(2)

where |W Si f ailure| and |W Sj selection| stand for number of times that WSi failed in composition scenarios and number of times that WSj was requested to substitute for WSi in these composition scenarios upon the recommendation of WSi , and obviously did not fail. The higher the weight is, the better is for WSj . Remark: Similar to Equation 1, |W Sj selection| tends to be smaller than |W Si f ailure|. Similarity-based association (S). In [2], we describe communities as a structure that allows gathering Web services with the same functionality independently of many factors such as origin, location, non-functional properties, and performance. A community has a dynamic behavior by which new Web services join, other Web services leave, some Web services become temporarily unavailable, and other Web services resume operation after suspension. By letting Web services join communities, it is possible to form similarity associations between them, which could help build social networks. Similarity could reduce the search space of Web services in case the first discovered Web service does not accept to fulfill a user’s request for various reasons that we report in [19]. As a result, the discovery continues with the Web services that are similar to the discovered Web service in the community without screening registries (e.g., UDDI) from scratch. •

Example: :

•

RoomBookingWS and RoomReservationWS have the same functionality. However this does not guarantee that both Web services have similar non-functional properties, which is a prerequisite to their substitution in the robustness case. Properties: S is symmetric and transitive.

•

Association weight for similarity wtS is given by the following equation: wtS(W Si ,W Sj ) =

|W Sj selection| |W Si similar|

(3)

where |W Si similar| and |W Sj selection| stand for number of Web services that are similar to WSi and number of times that WSj was selected out of the set of Web services that are similar to WSi . The higher the weight is the better for WSj . Remark: |W Sj selection| is set to zero when the number of times that WSj has been selected becomes greater to |W Si similar|. By doing so, the opportunity of participation to other Web services=j that are similar to WSi is given. One of the concerns that similarity-based associates may have is that only a limited number of Web services end up being frequently selected without giving any room to other peers to compete. This could discourage Web services’ engineers from developing new Web services, leading to diminishing the number of Web services in the market. To mitigate these problems, “serendipity” approaches [11], [17], [28] could be used.

3.2. Social networks use Among the many goals of having WS social networks is ?rstly to help in identifying additional Web services based on the Web services that are already recognized and secondly, reinforcing (or measuring) associations that connect all these Web services together in terms of recommendation, similarity-, and collaboration-based associations (Section 3.1). The principle here is to combine traditional Web services discovery mechanisms (like UDDI) with various details that social networks carry on. Any Web service that is discovered using these mechanisms constitutes an entry point to its own social network and probably to other peers’ social networks if navigation (or graph traversal) rights are granted to this Web service. We identify, hereafter, two cases that show how social networks could be used. Composition scenario. The recommendation-based association with focus on partnership could enrich a composition scenario with additional Web services that were not initially taken into consideration during the specification of this scenario. Starting from a Web service that is already part of this scenario, this Web service could recommend a number of other Web services that have the highest weights wtRp (equation 1) in the social network that this Web service manages. This number is usually set by the user or composition designer. It should be noted that sometimes neither the user nor the designer would be interested in expanding or reviewing the specification of the established composition scenario. The rejection of expansion within a given composition scenario may be induced by the additional

financial charges that need to be bore by either user of the designer. Failure situation. Recommendation-based association with focus on robustness could enable a direct (and probably automatic) selection of a Web service that will smoothly substitute a faulty one. Furthermore, to guarantee the success of this selection, similarity-based association would ensure that this Web service is compatible with peers (especially subsequent ones) that were expected to interact with the failed Web service. In these cases (similarity and compatibility check), the selection of the new Web service would depend on the weights of similarity (wtS , equation 3) and robustness (wtRr , equation 2), respectively (i.e., so upon a failure, one looks for a replacement in the “superset” of “similar” Web services and afterwards looks for the top reliable (robust) ones within that set.

3.3. Social networks maintenance The initial establishment of social networks falls within the responsibilities of providers of Web services who set up associations between their Web services prior to letting some (semi-) automated techniques take over the maintenance responsibilities. A provider could use the time of posting a Web service on a registry (or making this Web service join a certain community) to establish some associations. For instance, the provider could screen a registry for Web services that are similar to its Web service. Section 3.1 mentions three types of associations. Some associations could be established at the time of announcing Web services, while others could be established at the time of composing and invoking Web services. Announcement time. Similarity-based associations could be established to maintain a social network. Different techniques that assess the similarity between Web services are reported in the literature [3], [9] and could be adopted in our work. However, because these techniques primarily focus on the similarity at the functionality level, we judge some of them inappropriate for the recommendation-based association with focus on partnership. When Web services participate in the same composition scenario, they basically complement each other so they cannot be treated as similar. To address this lack of techniques, previous composition scenarios that Web services participated in could be used. This makes a provider establish the recommendation-based association with focus on partnership at composition and not announcement times. Composition time. In addition to the recommendationbased association with focus on partnership, collaborationbased association could be established during the time of maintaining a social network. Because a composite Web service could be either proactively or reactively constructed [7], the way associations can be established may vary. In the proactive case, the designer knows the, beforehand, the Web

services that will collaborate with each other. In the reactive case, the designer will have to wait until the composite Web service is formed and executed to ?nd out more about the Web services that interacted with its Web service at run-time. In Fig. 1 we illustrate how the three types of associations can be connected to each other. To this end, we split functionality into two types: similar, which makes Web services compete to participate in composition scenarios, and complementary5 , which makes Web services have a stake in the same composition scenarios. In the same figure, we show how a similarity-based social network supports a recommendation-based social network with focus on robustness by identifying the Web services that would be compatible with the rest of Web services that are already in the composition scenario. On top of the announcement and composition times that illustrate when social networks are affected, these networks could prove useful in the specification of new composite Web services. On the one hand, recommendation- and similarity-based social networks are used at design time by identifying extra Web services and not screening registries. On the other hand, collaboration-based social networks are used and updated at run-time by keeping track of all the interactions between Web services. p

4. Conclusion In this paper we provided a conceptual description of our research on Web services and social networks with focus on the discovery of these Web services. Our current efforts are put into completing our testbed and validating experimental results. Coupling Web services and social networks is an original contribution and has a potential of further research in different directions: 1) The current work can be extended by defining “topics” or “communities” to gather inter-related Web services (i.e., Weather, Hotels, Travel). The “topic” concept can be considered as relationship between Web services without any particular weight. In addition, a Web service can belong to different topics. This feature will reduce the search space when looking for Web services, facilitate the automatic or manual assignment of new Web services to its topics and find out that all Web services relevant to a certain topic. By doing so, it becomes easy to make relationships with other web services. 2) What will the situation be when a user request has to take into consideration multiple relationship types at 5. Complementary notion is reported in [12] where Jureta et al. gather Web services into groups called service centers that are dedicated to specific types of functionalities and hence, facilitate the development of composite Web services.

Web services' functionalities Viewed as

Similar

Complementary

Creates

Similarity SN

Supports

Creates

Recommendation SN (robustness)

Recommendation SN (partnership)

Supports

Collaboration SN

Figure 1. Social networks in interaction

the same time? In this case, do we have to establish a rank, or propose a minimal distance metric? Acknowledgements. L. K. Wives is partially supported by CNPq, Conselho Nacional de Desenvolvimento Cientfico e Tecnolgico, Brazil, and the Brazilian National Institute of Science and Technology for the Web.

References [1] G. Asela and M. Christopher. Tied Boltzmann Machines for Cold Start Recommendations. In Proceedings of the 2008 ACM conference on Recommender Systems (RecSys’2008), 2008. [2] J. Bentahar, Z. Maamar, W. Wan, D. Benslimane, Ph. Thiran, and S. Sattanathan. Agent-based Communities of Web Services: an Argumentation-driven Approach. Service Oriented Computing and Applications, 2(4), 2008. [3] L. Bordeaux, G. Sala¨un, D. Berardi, and M. Mecella. When are Two Web Services Compatible? In Proceedings of the 5th International Workshop on Technologies for EServices (TES’2004), Toronto, Canada, 2004. [4] S. Brin and L. Page. The Anatomy of a Large-Scale Hypertextual Web Search Engine. Computer Networks and ISDN Systems, 30(1-7), 1998. [5] T. Bui and A. Gacher. Web Services for Negotiation and Bargaining in Electronic Markets: Design Requirements and Implementation Framework. In Proceedings of the 38th Hawaii International Conference on System Sciences (HICSS’2005), Big Island, Hawaii, USA, 2005. [6] C. Castelfranchi. Commitments: From Individual Intentions to Groups and Organizations. In Proceedings of the International Conference on Multiagent Systems (ICMAS’1995), San Francisco, California, US, 1995. [7] D. Chakraborty and A. Joshi. Dynamic Service Composition: State-of-the-Art and Research Directions. Technical report, TR-CS-01-19, Department of Computer Science and Electrical Engineering, University of Maryland, Baltimore County, Maryland, USA, 2001. [8] J. Ethier. Current Research in Social Network Theory, Visited September 2008. http://www.ccs.neu.edu/home/perrolle/archive/EthierSocialNetworks.html.

[9] N. Hameurlain. Flexible Behavioural Compatibility and Substitutability for Component Protocols: A Formal Specification. In Proceedings of the Fifth IEEE International Conference on Software Engineering and Formal Methods (SEFM’2007), London, UK, 2007. [10] M. O. Jackson and A. Wolinsky. A Strategic Model of Social and Economic Networks. Journal of Economic Theory, 71(44):44–74, October 1996. [11] H. Jonathan L., K. Joseph A., T. Loren G., and R. John T. Evaluating Collaborative Filtering Recommender Systems. ACM Transactions on Information Systenms, 22(1), 2004. [12] I. Jureta, S. Faulkner, Y. Achbany, and M. Saerens. Dynamic Web Service Composition within a Service-Oriented Architecture. In Proceedings of the IEEE International Conference on Web Services (ICWS’2007), Salt Lake City, Utah, USA, 2007. [13] C. Kadushin. A Short Introduction to Social Networks: A Non-Technical Elementary Primer, Visited September 2008. http://construct.haifa.ac.il/ cerpe/papers/kadushin.html. [14] L. Krug Wives, Z. Maamar, S. Tata, M. Sellami, B. Defude, and J. P. Moreira de Oliveira. Towards A Marraige Between Recommender Systems & Web Services. Technical report, College of Information Technology, Zayed University, Dubai, U.A.E, September 2008. [15] Xuan Nhat Lam, Thuc Vu, Trong Duc Le, and Anh Duc Duong. Addressing cold-start problem in recommendation systems. In ICUIMC ’08: Proceedings of the 2nd international conference on Ubiquitous information management and communication, pages 208–211, New York, NY, USA, 2008. ACM. [16] C. Langdom. The State of Web Services. IEEE Computer, 36(7):93–94, July 2003. [17] Tuck W Leong, Frank Vetere, and Steve Howard. The serendipity shuffle. In OZCHI ’05: Proceedings of the 17th Australia conference on Computer-Human Interaction, pages 1–4, Narrabundah, Australia, Australia, 2005. ComputerHuman Interaction Special Interest Group (CHISIG) of Australia. [18] Z. Maamar, S. Kouadri Most´efaoui, and M. Lahkim. Web Services Composition Using Software Agents and Conversations. Ing´enierie des Syst`emes d’Information, 10(3), 2005.

[19] Z. Maamar, S. Kouadri Most´efaoui, and H. Yahyaoui. Towards an Agent-based and Context-oriented Approach for Web Services Composition. IEEE Transactions on Knowledge and Data Engineering, 17(5):686–697, May 2005. [20] S. Manuel Serra da Cruz, L. Maris Campos, M. Luiza Machado Campos, and P. F. Pires. A Data Mart Approach for Monitoring Web Services Usage and Evaluating Quality of Service. In Proceedings of the XVIII Brazilian Symposium on Databases (SBBD’2003), Manaus, Brazil, 2003. [21] T. Margaria. Service is in the Eyes of the Beholder. IEEE Computer, 40(11):33–37, November 2007. [22] J. Moody and D. R. White. Structural Cohesion and Embeddedness: A Hierarchical Concept of Social Groups. American Sociological Review, 68(1), February 2003. [23] P. Nasirifard. Web Services Security Overview and Security Proposal for UDDI Framework. In Proceedings of the International Conference on Security and Management (SAM’2003), Las Vegas, Nevada, USA, 2003. [24] J. O’Donovan and B. Smyth. Trust in recommender systems. In IUI ’05: Proceedings of the 10th international conference on Intelligent user interfaces, pages 167–174, New York, NY, USA, 2005. ACM. [25] M. Papazoglou, P. Traverso, S. Dustdar, and F. Leymann. Service-Oriented Computing: State of the Art and Research Challenges. IEEE Computer, 40(11):38–45, November 2007. [26] J. M. Pujol, R. Sang¨uesa, and J. Delgado. Extracting Reputation in Multi Agent Systems by Means of Social Network Topology. In Proceedings of the First International Joint Conference on Autonomous Agents & Multiagent Systems (AAMAS’2002), Bologna, Italy, 2002. [27] J. Raab and H. B. Milward. Dark Networks as Problems. Journal of Public Administration Research and Theory, 13(4), 2003. [28] B. Ranieri, L. Claudio, O. Salvatore, S. Massimo, and S. Fabrizio. A Privacy Preserving Web Recommender System. In Proceedings of the 2006 ACM symposium on Applied Computing (SAC’2006), 2006. [29] P. Seung-Taek, P. David, M. Omid, G. Nathan, and D. Dennis. Na¨ıve Filterbots for Robust Cold-Start Recommendations. In Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD’2006), 2006. [30] C. Sun, Y. Lin, and B. Kemme. Comparison of UDDI Registry Replication Strategies. In Proceedings of The IEEE International Conference on Web Services (ICWS’2004), SanDiego, California, USA, 2004. [31] S. Wasserman and J. Glaskiewics. Advances in Social Networks Analysis. Sage Publications, 1994. [32] Q. Yu, A. Bouguettaya, and B. Medjahed. Deploying and Managing Web Services: Issues, Solutions, and Directions. The VLDB Journal, 17(3):537–572, 2008.

[33] J. Zhang and M. S. Ackerman. Searching for Eexpertise in Social Networks: a Simulation of Potential Strategies. In Proceedings of the 2005 International ACM SIGGROUP Conference on Supporting Group Work (GROUP’2005), Sanibel Island, Florida, USA, 2005.