Papers by Author | in ( Plenary | Parallel | Poster | Summary ) Sessions

MOOSE : a strategy for the future 

           Christian Arnault

  LAL Orsay - arnault@lalcls.in2p3.fr
   MOOSE (RD41) collaboration
             CERN
  • Paper (Postscript)
  • Paper (PDF)

  • Slides 
                    Abstract
    The problems that HEP software environments must address in the next few years are:

  • to manage the wide geographical dispersion of software providers and users,
  • to provide for a smooth cooperation between the increasingly sophisticated software technology and the trends for applications to become more and more user oriented.
  • to allow the long life time of HEP software tools and products to take advantage of the rapid evolutions of software technology.
  • to cope with the sociological and professionnal impact of the technology choices made by the HEP world as compared to the ones made by the industry.

    The CERN's R&D project MOOSE (RD41) has been formed to study the technologies, the methodologies and the human behaviours that could help provide solutions to these questions considering as key issues the cultural, technogical and projects constraints and requirements.

    This presentation will discuss some of the concepts that have been addressed so far by this R&D project such as :

  • The team and group behaviour in cooperative actions for sharing information and managing shared analysis phases or common evaluation of software production.
  • The methodology for software engineering dealing with the meaning and impact of the object oriented concepts, the conceptual methods for analysis and design, the process of the sofware production and the techniques for specifying a problem or for reverse engineering.
  • The technology for software production such as the tools that implement the various conceptual methods or the impact of using various programming languages(OO or not).

    The difficulties encountered during this study happen to be also the critical points that must be solved for a succesfull evolution of our software environments such as preserving the vast amount of cultural investment provided in the existing physics code, or dealing with the increasing influence of industry, where the constraints or requirements may not be the same as ours.The preliminary results that emerge from this first year study show that even more than the concept of object orientation (which nevertheless isthe most mature and coherent approach in the domain of SE methods) the demand for an engineering approach to software production seems to be the principle criterion for a quality improvement process. In particular, we have shown and experienced that the following actions are essential in this process :

  • establishing an important training and teaching program in our community.
  • sharing the software production process between all members of the team, from the early specification collection down to the final design phases.
  • introducing explicit quality criteria that can be tested upon at every stage of the sofware life cycle, through conventions, code reviews or management tools,
  • using standards (industrial or not) as much as possible in the underlying technology.
  • organizing the software architecture around the notion of a framework of reusable components that are responsible for providing services to each other. 
    Christian Arnault 
Laboratoire de l'accelerateur Lineaire - Orsay
Faculte des sciences - 91405 ORSAY CEDEX (France) 
arnault@lalcls.in2p3.fr
16 (1) 64 46 84 17 
16 (1) 69 07 94 04