PIRATES Project

Méthodes et Outils pour la Programmation Répartie Transparente et Sûre

(Methods and Tools for Dependable Transparent Distributed Programming)

Convention 9713540 de la Région Wallonne

Peter Van Roy
Department of Computing Science and Engineering
Université catholique de Louvain

In collaboration with:

Programming Systems Laboratory

Swedish Institute of Computer Science

Project members

Researchers:
- Donatien Grolaux
- Valentin Aurelian Mesaros
- Luis Quesada
Associated researchers and members:
- Raphael Collet
- Almetwally Mostafa
- Luc Onana Alima
- Freddy Gridelet
- Gustavo Ospina

The Mozart Programming System

Public release: Jan. 25, 1999 (current version 1.1.0 released Feb. 7, 2000)
Systems supported: Many Unix platforms, Windows 95/98/NT, planned MacOS X port
Developers:
- DFKI, UdS, and SFB 378 (German Research Center for Artificial Intelligence and Universit\144t des Saarlandes)
- SICS (Swedish Institute of Computer Science)
- UCL (Université catholique de Louvain)
Mozart 1.0 is an extensive development platform for open fault-tolerant distributed applications, constraint programming applications, and logic programming applications. The platform includes a complete development environment with many tools and full documentation including tutorials. Applications developed include a collaborative graphic editor (TransDraw), an extended version of ICQ (MIM), a corpus browser, a real-time bus scheduler, a configuration tool, and much more. The platform is released with a very liberal license (X11-style) that allows both commercial and non-commercial use. Full source code is available.
Mozart 1.0 implements the Oz 3 language, which is almost completely upward-compatible with its predecessor Oz 2 (see DFKI Oz 2.0). The main extensions are functors, which provide first-class software components and specifications, and futures, for secure dataflow behavior. Oz 2 is itself a successor to the original Oz 1 language, whose implementation was first released publicly in 1995.

Mozart announcement and comparison with Java platform

Vision and current status

Distribution is a major problem in computer science today. The goal of PIRATES is to provide a platform for general-purpose application development that will let us program networks of computers as easily as single computers. The major research issues to be resolved are network transparency, network awareness, fault tolerance, security, and resource management. Other important issues are openness and network scalability.

First results of PIRATES are the release of the Mozart Programming System and the TransDraw collaborative graphic editor. The current Mozart release provides complete network transparency and network awareness, first-class software components, and basic primitives for fault tolerance and open computing. It implements a conservative extension to Oz 2, a concurrent object-oriented language that is state-aware and has dataflow synchronization. In Mozart, all network operations are invoked implicitly by the system as an incidental result of using particular language constructs. Extending the implementation of these constructs to a distributed setting is practical because Oz 2 has a simple formal semantics.

In the current release, writing applications is done in four parts, more or less according to the following schema. The first two parts are fully implemented and trivial to use for the application developer; the last two are ongoing research and currently require a bit more thought, although the basic primitives exist. Future releases will provide more functionality that is easier to use, and will include security.

First, write the application's functionality ignoring the network, the fault tolerant behavior, and the open connectivity. Network transparency guarantees that the same program will run on one site or multiple sites, differing only in performance.
Second, to achieve efficient distribution, specify the distribution structure and the resource usage of the application's entities. This is a key idea to provide control over network communication patterns in a predictable manner. For example, stationary servers, mobile objects, and object caches are identical from the programmer's perspective except that they have different (but well-defined) network behaviors.
Third, to achieve the desired robustness, specify the fault-tolerant behavior of the distributed entities. We are currently building abstractions to separate this as much as possible from the application's functionality.
Fourth, add openness to allow dynamic connections between parts of the application that belong to different users.

The goal is to concentrate most of an application's development effort in realizing just its functionality. The other characteristics (geographical distribution, fault tolerance, etc.) are realized by adding small specifications that are almost independent of the functionality. This goal is currently fully realized for distribution and openness, and is well on its way to be realized for fault tolerance.

Project proposal

PIRATES: Méthodes et Outils pour la Programmation Répartie Transparente et Sûre (slides) (in French). English translation.

Papers and talks

For more Mozart-related papers see the Mozart publications page. For more Oz-related papers see the Programming Systems Lab publications page at the University of Saarbrucken.

QTk -- A Mixed Declarative/Procedural Approach for Designing Executable User Interfaces (50K) (pdf version), short paper, 8th IFIP Working Conference on Engineering for Human-Computer Interaction (EHCI'01), May 2001.
Fault-Tolerant Mobile Agents in Mozart (80K), poster, ASA/MA 2000, Zurich, September 2000.
Path Redundancy in a Mobile-State Protocol as a Primitive for Language-Based Fault Tolerance (100K), INGI research report RR2000-01, January 2000.
Logic programming in the context of multiparadigm programming: the Oz experience (DRAFT) (600K), extended version of tutorial given at International Conference on Logic Programming, November 1999. More than just a tutorial on logic programming, this article explains true multiparadigm programming as realized in Oz, gives the history of Oz, and summarizes the main lessons learned in the Oz project.
Mozart: A Programming System for Agent Applications (120K) (slides), January 2000. An introduction to open, robust distributed programming in Mozart. The article gives many small code examples and the slides give an overview of the main concepts.
A Lightweight Reliable Object Migration Protocol (290K), in Springer Lecture Notes on Computer Science, Vol. 1686, October 1999. Gives an intuitive yet precise presentation and justification of Mozart's mobile state protocol.
A New Mobile State Protocol for Distributed Oz (360K) (slides of talk), PDCS 99, Ft. Lauderdale, Florida, August 1999. Gives an extension of the mobile state protocol to improve behavior on hierarchical networks.
On the separation of concerns in distributed programming: Application to distribution structure and fault tolerance in Mozart (180K), International Workshop on Parallel and Distributed Computing for Symbolic and Irregular Applications (PDSIA 99), Sendai, Japan, July 1999. Explains the Mozart design principles and applies them to distribution and fault tolerance of mutable pointers.
Efficient Logic Variables for Distributed Computing (600K), ACM Transactions on Programming Languages and Systems, May 1999. Explains how to use and implement dataflow execution in a distributed system. Defines a practical distributed unification algorithm and proves its correctness.
Le Protocole Réparti de Distributed Oz (190K) (in French), Colloque Francophone sur l'Ingénierie des Protocoles (CFIP99), Nancy, France, April 1999. Explains how Mozart's different distributed algorithms are integrated together.
A Collaborative Graphic Editor Based on Transactions (140K). Overview of the design of the TransDraw graphic editor. Talk given at Oz/Mozart workshop, SICS, Sweden, June 1998.
Programming Languages for Distributed Applications (400K) (pdf version), Journal of New Generation Computing, May 1998. Second published overview of the Mozart approach. Much more complete than the first published overview.
Mobile Objects in Distributed Oz (900K), ACM Transactions on Programming Languages and Systems, September 1997. Explains the mobile state protocol, proves it correct, and shows its use for network-transparent distribution.
Making distributed programming easy, extended version of article that appeared in FSA Research, Université catholique de Louvain, 1997. Very brief summary of the problems that are addressed by Mozart.
An Overview of the Design of Distributed Oz, PASCO '97, July 1997 (early slides). First published overview of the Mozart approach.
Networks from the programming language point of view, or using and implementing mobile objects (slides), talk given at Alcatel-Bell Symposium In search of a Perfect Marriage between IP and ATM, May 23, 1997.
La Disparition de l'Ordinateur (in French) (transparents de l'exposé), leçon inaugurale, Faculté des Sciences Appliquées, UCL, 21 octobre 1996.
Internal discussions. A chronological and open-ended list of working documents on various aspects of Distributed Oz.
PERDIO project proposal, Programming Systems Lab, Universität des Saarlandes. The original project proposal with which the distribution part of the Mozart project started.
Distribution at SICS.