This paper addresses the design and verification of infrastructure for mobile computation. In particular, we study language primitives for communication between mobile agents. They can be classified into two groups. At a low level there are location dependent primitives that require a programmer to know the current site of a mobile agent in order to communicate with it. At a high level there are location independent primitives that allow communication with a mobile agent irrespective of any migrations. Implementation of the high level requires delicate distributed infrastructure algorithms. In earlier work with Wojciechowski and Pierce we made the two levels precise as process calculi, allowing such algorithms to be expressed as encodings of the high level into the low level; we built Nomadic Pict, a distributed programming language for experimenting with such encodings. In this paper we turn to semantics, giving a definition of the core language and proving correctness of an example infrastructure. This requires novel techniques: we develop equivalences that take migration into account, and reasoning principles for agents that are temporarily immobile (eg. waiting on a lock elsewhere in the system).

In this paper we present sofar, a versatile multi-agent framework designed for Distributed Information Management tasks. sofar embraces the notion of proactivity as the opportunistic reuse of the services provided by other agents, and provides the means to enable agents to locate suitable service providers. The contribution of sofar is to combine some ideas from the distributed computing community with the performative-based communications used in other agent systems: communications in sofar are based on the startpoint/endpoint paradigm, which is the foundation of Nexus, the communication layer at the heart of the Computational Grid. We explain the rationale behind our design decisions, and describe the predefined set of agents which make up the core of the system. Two distributed information management applications have been written, a general query architecture and an open hypermedia application, and we recount their design and operations.

We study the distributed infrastructures required for location-independent communication between mobile agents. These infrastructures are problematic: different applications may have very different patterns of migration and communication, and require different performance and robustness properties. Some applications also demand disconnected operation (on laptop computers). Algorithms must be designed with these mind. In this paper we describe simple algorithms and techniques such as a central server, forwarding pointers, broadcast, group communication, and hierarchical location directory, and use Nomadic Pict to develop and implement an example infrastructure. The infrastructure can tolerate site disconnection; a user can disconnect the computer from the network, work in a disconnected mode for extended periods, and later reconnect. All messages that cannot be delivered to a laptop or sent out from the laptop due to disconnection will be transparently delivered upon reconnection. 1

In this paper, we propose a global tracking service for mobile agents, which is scalable to the Internet and accounts for security issues as well as the particularities of mobile agents (frequent changes in locations). The protocols we propose address agent impersonation, malicious location updates, as well as security issues that arise from profiling location servers, and threaten the privacy of agent owners. We also describe the general framework of our tracking service, and some evaluation results of the reference implementation we made.

A reliable communication layer is an essential component of a mobile agent system. We present a new fault-tolerant directory service for mobile agents, which can be used to route messages to them. The directory service, based on a terhn~que of forwarding pointers, introduces some redun-dancy in order to ensure resilience to stopping failures of nodes contaln~-g forwarding pointers; in addition, it avoids cyclic routing of messages, and it supports a technique to collapse chains of pointers that allows direct communica-tions between agents. We have formalised the algorithm and derived a]uil ~ mechanical proof of its correctness using the proof assistant Coq; we report on our experience of design-Lug the algorithm and deriving its proof of correctness. The complete source code of the proof is made aveglable f~om the WWW. 1.

Mobile Agent systems have attracted considerable attention as means of exploring and manipulating distributed information sources. However, many existing multi-agent platforms present limitations in terms of adaptability and scalability, indicating difficulties when trying to replicate these results on a large scale. We describe the core of a novel mobile agent toolkit known as DIET, (Decentralised Information Ecosystem Technologies), which addresses some of these limitations and provides a foundation for an open, robust, adaptive and scalable agent ecosystem. We introduce DIET core features and describe how they support basic mobile agent capabilities such as migration and real-time interaction. We then illustrate how an ecosysteminspired design approach differs from conventional design approaches. Finally, we experiment with a simple information retrieval scenario, demonstrating the emergence of agent communities through the evolution of environmental preferences. In this way we hope to clarify how applications built on this foundation could be used to tackle problems in adaptable and open real-world scenarios.

In this article we investigate secure global name services which are scalable to the Internet and account for particularities of mobile agents (frequent changes in locations). Name services are required for instance for message routing between mobile agents and for tracking agents. In this article we propose first protocols targeted at providing such name services. We focus in particular on security issues.

We have recently defined a new algorithm for distributed garbage collection based on reference-counting [20, 24]. At the heart of the algorithm, we find tree rerooting, a mechanism able to reduce third-party dependencies by reorganising diffusion trees. In reality, the algorithm describes a spectrum of algorithms according to the policy used to manage messages. In this paper, we present the implementation of the algorithm and evaluate its performance. We have implemented two policies, which are extremes of the spectrum, respectively using and not using tree rerooting. In addition, two different strategies for managing action queues have been implemented. The conclusions of our experimentations are the following. Tree rerooting offers more parallelism during distributed gc activity; we explain this phenomenon by the length reduction of causality chains in the distributed gc. Grouping messages per destination dramatically reduces the number of messages, but requires a more complex implementation as messages have to be sorted per destination. Speed up of 100% has been observed on some benchmarks.

We emphasise and briefly review existing infrastructure required to realise the Computational Grid, and define such Grid with reference to Knowledge and Information Grids. We then propose an agent-based approach for the Computational Grid, which is centered on providing "services" for managing resources.

Mobile Objects in Java provides support for object mobility in Java. Similarly to the RMI technique, a notion of client-side stub, called startpoint, is used to communicate transparently with a server-side stub, called endpoint. Objects and associated endpoints are allowed to migrate. Our approach takes care of routing method calls using an algorithm that we studied in [22]. The purpose of this paper is to present and evaluate the implementation of this algorithm in Java. In particular, two different strategies for routing method invocations are investigated, namely call forwarding and referrals. The result of our experimentation shows that the latter can be more efficient by up to 19%. 1