Section: New Results
Other results
Programming models and calculi
Participant : P. Fradet.
We have been interested for a long time in formal calculi in order to study programming language issues in the simplest possible setting. We present here work within the calculus (compilation of higherorder sequential languages) and the calculus (higherorder parallel and nondeterministic programming).
calculus and the Krivine abstract machine
The Krivine machine is a simple and natural implementation of the callbyname calculus. While its original description has remained unpublished, this machine has served as a basis for many variants, extensions and theoretical studies. We have presented the Krivine machine and some wellknown variants in a common framework [Oops!] . We have characterized the essence of the Krivine machine and have located it in the design space of functional language implementations. This work is based on the framework that we had previously developed for the systematic study of functional language implementations [49] .
This is joint work with Rémi Douence from the Obasco project team (École des Mines de Nantes).
calculus and higherorder chemical programming
The chemical reaction metaphor describes computation in terms of a chemical solution in which molecules (representing data) interact freely according to reaction rules. Formally, chemical programs can be represented as associativecommutative rewritings (reactions) of multisets (chemical solutions).
This model of computation is wellsuited to the specification of complex computing infrastructures. In particular, the orderless interactions between elements that occur in large parallel or open systems are naturally expressed as reaction rules.
We have been working on the application of Hocl to the programming of distributed applications, in particular to autonomic systems [Oops!] . We have shown that autonomicity features ( e.g. , selfhealing, selfprotection, selfoptimization, etc.) are naturally expressed as reaction rules.
This work is conducted in collaboration with JeanPierre Banâtre and Yann Radenac from the Paris project team at Irisa . It was the central topic of Yann Radenac's PhD thesis. This line of research is related to the AutoCHEM project (see section 8.2.1 ) starting this year.
Componentbased modeling and analysis of genetic networks
Participants : G. Goessler [ contact person ] , A. Richard.
Genetic regulatory networks usually encompass a large number of genes, proteins, and metabolites. Being able to model and analyze its behavior is crucial for understanding the interactions between the proteins, and their functions. There has been a wide variety of modeling approaches, including the influential early work of [76] based on logical equations, and [54] based on differential equations. However, simulation and verification of the continuous model are expensive, and many properties are not even decidable in this framework. The approach of [48] based on the approximation of nonlinear models by piecewise linear differential inclusions, uses a discrete abstraction preserving the qualitative dynamics of networks. As [48] approximates the continuous behavior with a monolithic discrete transition system, it still suffers from state space explosion. This problem has been addressed with the componentbased approach of [55] where the discrete abstraction is constructed and analyzed modularly, allowing to deal with complex, highdimensional systems. We have further improved this technique by allowing for a more precise, conservative abstraction.
Using the same approach, we are currently studying, in cooperation with H. de Jong (Helix) and G. Batt (Contraintes), the definition of a symbolic representation of the network behavior as a compact exchange format between the Genetic Network Analyzer (GNA) developed in the Helix group, and the model cheker CADP developed by Vasy.
The lack of numerical values for the parameters characterizing the interactions of a genetic regulatory network makes classical numerical analysis techniques difficult to apply. The approach of [48] defines a discrete abstraction preserving the qualitative dynamics of networks for wide ranges of parameter values. We have developed, in cooperation with H. de Jong, a novel algorithm to enumerate all classes of parameter values of an incompletely specified network. This technique has been implemented and applied to the analysis of a model of a network controlling the stress response of bacteria, and has allowed to uncover a shortcoming in the model.
Interactions Between Law and Information and Communication Sciences
Participant : D. Le Métayer [ contact person ] .
Daniel Le Métayer is initiating a new activity (which is to become an independent action in the short term) on the interactions between ICT (Information and Communication Technologies) and law. The motivation for this activity is the observation that the impact of ICT on the every day life of most individuals raises new challenges which cannot be tackled by a purely technological approach. Our position is that the first step for a fruitful and useful exploration of the relationship between the legal and technical dimensions is the definition of a formal framework for expressing the notions at hand, understanding them without ambiguity, and eventually relating or combining them. The first application of this approach, which is conducted within the PRIAM ARC, concerns privacy protection in the “ambient intelligence” context. Privacy is a complex issue, especially in the context of ambient intelligence, both from the legal and the technical perspective:

The very definition of privacy is far from obvious since it is by essence subjective and based on a fuzzy notion of boundary or “private sphere”. In particular, these boundaries are blurred in the ambient intelligence landscape. In addition, the legal framework, which has to reflect the social expectations, needs to be revisited to account for the new possibilities offered by the technology.

An ambient computing infrastructure is by nature heterogeneous and dynamic, with new nodes, of different natures – and belonging to potentially unknown mistrusting users – connecting to the network and being able to communicate in a spontaneous way. Last but not least, the smart objects can be tiny, inexpensive, devices with limited resources (chips on clothes, banknotes, etc.). It is thus difficult to rely on these to implement complex privacy policies.
The PRIAM project emphasizes the design of privacy policies that are amenable to both a formal description and a realistic implementation in the ambient world. The techniques under study are based on a combination of a priori controls ( e.g. , access controls), which are the most conservative, and a posteriori controls ( e.g. , audits) which may be easier to enforce on resourceconstrained devices.
Control for dataparallel systems
Participant : E. Rutten.
Data intensive computing is increasingly getting high importance in a wide range of scientific and engineering domains. Such systems manipulate large amounts of data; so high performance, scalability and throughput are important requirements. Reconfigurability is another interesting feature because it makes the systems flexible enough to be adapted to various environment and resource constraints. The Gaspard2 (http://www.lifl.fr/west/gaspard )development framework aims at proposing a solution to the design of data intensive applications in general, and highperformance embedded systemonchip (SoCs) in particular.
We have proposed a synchronous model of Gaspard2 , in order to bridge the gap between Gaspard2 and analysis and verification tools of the synchronous technology so that formal validation is favored [Oops!] .
The automation of the transformations is implemented within an MDE framework [Oops!] .
We extend Gaspard2 , by adding reactive control features based on finite state machines [Oops!] , [Oops!] , and are integrating this extension in the synchronous model.
This work is conducted in cooperation with the DaRT project at UR Futurs in Lille.