Invited Talks at P2PKM

Invited Talks

Bernard Traversat, SUN Microsystems, USA

Title: "JXTA™: Beyond P2P File Sharing, the Emergence of Knowledge Addressable Networks"

Abstract: In the last couple of years, P2P applications such as music swapping, instant messaging, and media distribution have emerged as the most popular applications on the Internet. This P2P revolution has led the research community and computer industry to explore this novel way to communicate and distribute content on the Internet. P2P is fundamentally extending the traditional centralized content indexing model with a decentralized model where indexes are distributed, and used to dynamically route discovery queries to content provider peers, similarly to IP packets in an IP network. P2P is also fundamentally changing the way we address and locate contents on the Internet. From a physical IP addressing model, the Internet is reconstructed into a logical knowledge addressable network, where IP addresses are replaced by content IDs. Each P2P application has the ability to define its own logical network overlay providing its own ID representation, meta-data information, and query routing infrastructure allowing peers to self-organize into protected peer group domains.

The talk will introduce the fundamental advantages of shifting from a physical IP addressing model to a logical knowledge addressing model for building content delivery network. We will cover some of the key issues that need to be addressed when building a P2P knowledge addressable network: content ID representation, meta-data ontology, peer churn rate, super-peer hierarchy, cost of maintaining a distributed index on an unstructured network, quality of services, and security and trust. We will illustrate our presentation using the JXTA platform. JXTA defines a standard set of protocols for ad hoc, pervasive, peer-to-peer (P2P) network that provides a generic platform for developing a variety of knowledge addressable network applications.

Dr. Bernard Traversat is the Senior Engineering Manager for the RFID and JXTA technology at Sun Microsystems Inc. Bernard is leading the development of the Sun Java RFID product and JXTA technology working with customers and partners. Previously he led Sun's effort in pervasive computing for small consumer devices, and was lead developer on the SunCluster product. Prior to that, he worked at the NASA Ames Research Center on distributed-memory operating systems for massively parallel supercomputers. He has numerous technical publications in distributed-memory operating systems and resource management systems.
Bernard co-authored the "JXTA in a Nutshell" book by O'Reilly Press.
He received his Ph.D. and M.S. degree in Computer Sciences from the Florida State University.

Matteo Bonifacio, ITC-Irst, Italy

Title: "Centralized vs. Distributed Knowledge Management: Is it the right question?"

Abstract: Until a few years ago the dominant conceptual model in knowledge management was the one of centralization. Organizations, consultants and software vendors where focused on providing solutions able to codify, store, and disseminate knowledge in terms of content. Driving concepts where those of corporate knowledge taxonomies and portals. In parallel, many research efforts have been made in order to define methods and tools able to support this paradigm focusing in particular on the creation of shared semantics whereby the driving concept was the one of ontology. More recently, the disappointing outcomes of KM experiences and the emergence of new technological paradigms (such as P2P) led to an opposite approach that views knowledge as an intrinsically distributed matter localized in social and cognitive contexts. Accordingly, both researchers and practitioners are now looking at Distributed Knowledge Management as a new possible answer, focusing on new driving concepts such as the one of semantic interoperability rather than standardization, and localized communities of practice rather than monolithic organizations. Still, weak signals are evidencing some potential limitations of this view. In this talk I will first explore the evolution from KM to DKM, and show a P2P solution that exemplifies such evolution (KEEx). Second, based on both the commercial experience of KEEx and some interdisciplinary research reflections, I will explore some of the weak signals that in my opinion should alert both DKM researchers and practitioners not to consider DKM as an alternative to centralized solutions. Each view represents an extreme of a trade-off that poses to researchers and practitioners intriguing as much as challenging questions.

Matteo Bonifacio is currently Researcher in Organization Sciences at the University of Trento and Program Manager at ITC-IRST in the area of Knowledge Management. He teaches Knowledge Management and Learning Organizations at the Faculty of Economics of the University of Trento. He’s involved in several research project such as Knowledge Web (6th Program Framework Network of Excellence), VIKEF (6th Program Framework Integrated Project), Innovanet (5th Program Framework Startegic Action), and EDMOK (Regional Funds of the Trento Local Government). He’s Program Chair of the First Workshop on Peer-to-Peer Knowledge Management (P2PKM), and Program Committee Member of Context 01, 03, The I-Know Conference 03, 04, 05, Mobiquitous 04, the Workshop on Peer-to-Peer and Agent Infrastructures for Knowledge Management WM2005. Among he’s research interests are the topics of Distributed Knowledge Management, Organizational Learning, Communities of Practice, social and economic aspects of Semantic Evolution and Meaning Negotiation, Innovation, Decision Making under ambiguity, Collaborative and Peer to Peer technologies for KM. He’s also involved in several industrial projects on Knowledge Management related issues with organizations such as Trenitalia, and Telecom Italia.