Report from the visiting committee

Size: px

Start display at page:

Download "Report from the visiting committee"

Rose Doyle
5 years ago
Views:

1 Section des Unités de recherche Report from the visiting committee Research unit : Preuves, Programmes, Systèmes (PPS) UMR 7126 University Paris 7 Denis Diderot February 2008

2 Section des Unités de recherche Report from the visiting committee Research unit : Preuves, Programmes, Systèmes (PPS) UMR 7126 University Paris 7 Denis Diderot February 2008

3 Report from the visiting committee The research unit : Name of the research unit : Preuves, Programmes, Systèmes (PPS) Requested label : UMR N in case of renewal : 7126 Head of the research unit : Mr Pierre-Louis CURIEN University or school : University of Paris 7 Denis Diderot Other institutions and research organization: CNRS Date(s) of the visit : 18 janvier

4 Members of the visiting committee Chairman of the commitee : Mr Jean-Pierre JOUANNAUD, Prof. Université Paris Sud et Ecole Polytechnique Other committee members : Mr Samson ABRAMSKY, Prof. Oxford Mr Martin HOFMANN, Prof. Munich Mr Xavier LEROY, DR INRIA-Rocquencourt Mr Davide SANGIORGI, Prof. Bologne Mr Patrick DEHORNOY, Prof. à Caen CNU, CoNRS, CSS INSERM, représentant INRA, INRIA, IRD representatives : Mr Bruno DURAND, Prof. à Marseille, pour le Comité National de la Recherche Scientifique Mr Jean-François DUFOURD, Prof. Strasbourg pour le CNU Observers AERES scientific representative: Mr Luis FARIÑAS DEL CERRO, DR CNRS à l Université Paul Sabatier, pour l AERES University or school representative: Research organization representative (s) : Mrs Véronique DONZEAU-GOUGE, Prof. au CNAM, pour le CNRS 3

5 Report from the visiting committee 1 Short presentation of the research unit Numbers of lab members including researchers with 19 teaching duties, 12 full time researchers, 1 ingeneers, 23 Phd students, 1 technician and administrative assistant Numbers of HDR : 12 and of HDR who are PhD students avisors Numbers of PhD students who have obtained their PhD and average length of a PhD during the past 4 years : 16 Numbers of lab members who have been granted a PEDR : 5 Numbers of publishing lab members : 26/31 EC, C PPS is a laboratory whose aim is to develop the mathematics of computation. To ease the collaborations between its various members, whose backgrounds are variously in mathematics or computer science, the laboratory is not divided into teams, but into four themes, most researchers working in at least two themes. The laboratory is hosted by the University Paris Diderot (P7), as is LIAFA, the other computer science laboratory. Both laboratories have an excellent relationship, in particular within the hiring committee. Both are active members of the RTRA Sciences Mathématiques de Paris, which is run by mathematicians who are sympathetic to computer science activities. Membership in this RTRA is an important source of funding and dynamism. Personnel has more than doubled since PPS was created on January 1st, 1999, now reaching 56 people (among which are 31 people in permanent positions and 23 PhDs). This is the result of the hiring of 2 new professors since 2005, 4 MdC since 2004, 1 CR CNRS and the arrival of 2 DR CNRS. On the other hand, 2 senior leading researchers left the laboratory this year, to take up positions in Lyon and Edinburgh respectively. 2 Preparation and execution of the visit The agenda of the meeting and the review was well organized by PPS. The presentation of the structures of the lab and the scientific projects was very comprehensive. 3 Specific appreciation team by team and/or project by project The first two themes reflect the original emphasis of PPS, which has now broadened, as reflected by the other two themes. 3.1 Mathematical structures of programming This theme involves a large number of researchers 25 in all, including 11 established (permanent) researchers. The work in this theme remains extremely active, and indeed this group represents an exceptional critical mass of strong researchers working in the mathematical foundations of computation, emphasizing the structural aspects (logic and semantics) rather than the algorithms and complexity side (although there is some activity on implicit complexity). 4

6 Given the large number of active researchers working in this theme, it is perhaps inevitable that there are a large number of research topics being pursued, some of them interacting with each other and with work in other themes, some rather separate from any other activity in PPS. We briefly summarize the main subthemes : - There is a major activity in Game Semantics. The main thrust here is to explore the mathematical deep structure of the subject, rather than, for example, the connections with and applications to verification and program analysis, which has been a theme of much recent work in this area in the U.K. Some of the ideas have been inspired by ideas of true concurrency, and lead on to work in which ideas from homotopy and higher-dimensional categories play an increasingly prominent role. - Another leading current activity is toward developing a Curry-Howard isomorphism for concurrency. It has been shown that a large fragment of the pi-calculus can be encoded faithfully into the interaction net representation of proofs in Differential Linear Logic. There is an ANR-funded project on this topic which has just started. This work represents a pleasing interaction with the work in Theme 3. - There is a substantial activity on models of lambda calculus and functional programming languages. Some of this concerns the structure theory of the lattice of lambda theories, and some studies specific models with interesting properties. This work, in its several different lines, mostly stands apart from other activities within PPS, although some of it involves international collaborations. - There is also general work on algebraic and homotopic methods, which provide foundations for diagrammatic notations, which are of increasing interest in computer science, and also have some promise to provide a more algebraic and conceptual foundation for game semantics. - There is work on studying denotational semantics of bounded complexity variants of linear logic, notably ELL and LLL on an internationally competitive level. If PPS intends to pursue this strand further, we would recommend that goals beyond the mere definition of a semantics be set, and that the cooperation with Paris XIII on this topic should be intensified. All of this work is of high technical quality, as shown by the excellent record of publication in leading conferences, and the very good record of publication in leading journals. We strongly encourage maintaining and even increasing the efforts to publish detailed versions of the major contributions in journals; the extended-abstract format of conferences often does not allow full justice to be done to the ideas. The section on Perspectives offers some unexpected but welcome ideas towards linking the foundational work on game semantics with more applied topics, notably on compilation and abstract machines, and on using game semantics to formalize APIs. This reflects in part some of the work done in Theme 2, and also the influence of interactions with collaborators outside PPS. 3.2 Proof and rewriting Twelve people in permanent positions are involved in this theme, including the research engineer and an active professor emeritus. Despite the low level of activity by some members, the overall quality of the research is very good and has a strong reputation, as witnessed by the many excellent publications and the Ackerman prize in This theme appears as a collection of many different topics in proof theory, programming languages and proof systems, such as typed languages with patterns, calculi with explicit resources, proof terms for non-necessarily intuitionistic logics, proof search for intuitionistic logics, classical realazibility and proof extraction. Most research projects are carried out with outside colleagues of international reputation. All these diverse topics are however closely related by the tools employed and the goals aimed at. The tools are the Curry-Howard correspondence, typed lambda calculi, Gentzen systems and linear logic. The goals are: to design and understand programming languages with ever richer type systems; to design proof systems for intuitionistic or classical logic in which proof search can be (semi-) automated; to implement these languages and proof systems efficiently. These activities are partly supported by software developments allowing for experiments, carried out in relation to theme 4. 5

7 There are many Perspectives. A quite original one in our view is the program for classical realizability, although it is still early to conclude that this will eventually lead to tools challenging those developed for functional programming and intuitionistic logics. Another very interesting and original shorter term program is to reconcile sequent-based proof systems, which are very good at proof search, with natural deduction-like proof systems, which are very good at proof checking. 3.3 Concurrency This theme covers 3 main topics: models of concurrent programming (including type systems for languages with concurrency), probabilistic systems, and models of biology based on process calculi. The topics are closely related; for instance, the use of probability is central in the models for biology and its applications. The concurrency theme has become more and more important during the last 4 years, due to recruitment specifically aimed towards it, and the new orientation of the activities of some strong researchers in the lab. At present, 6 permanents are involved in this theme 3, all of them being very active researchers. The overall quality of the research is excellent. This is well witnessed by the quantity and the quality of the publications produced. This research is indeed very well known and highly appreciated by the international scientific community. The lab has weekly meetings on concurrency that are well attended (with an average of 10 to 15 people each time). The fact is that concurrency has become very relevant also for the other themes of the lab, which was not the case during the early years of the lab. For instance, several works in theme 1 are published at the CONCUR conference, while the work on web services in theme 4 concerns certain special forms of concurrent systems. A concern for the future is the departure of the main actor in the biology topic. Bioinformatics is considered by the director of the lab (as well as the president of the university) a very important topic. This departure will involve a general re-organisation whose consequences are difficult to predict now, with a transition period that might last well over a year. However the funding situation of the lab at present is very solid indeed. The transition period resulting from this departure should be monitored closely by the lab. 3.4 Software tools This theme is devoted to applying principled approaches to practical problems arising in software development and in systems. It is characterized by major software development and experimental validation efforts. Most of the software developed is distributed as free software; some of it has significant communities of users outside the lab. A unifying feature of these developments is the emphasis on solid formal foundations, drawn from the common background in semantics and logics of PPS s members. This theme involves approximately 8 permanent members of PPS but only 3 to 4 PhD students. It brings more than 50% of PPS s contractual funding. Two projects stand out in terms of originality, technical contributions, visibility outside the lab, and publications. The first is the work on the CDuce and XDuce domain-specific languages for type-safe, high-level processing of XML semi-structured data. Initially represented by mostly theoretical work on XDuce, this line of work moved to a much larger scale with a new appointment in 2006, bringing to PPS the tasteful design and high-quality implementation of the CDuce language. This work continues with a reorientation towards specification and coordination of Web services, a hot area where PPS s expertise in process calculi should help. The second project that stands out is the work on static analysis of the large collections of software packages that are typically found in Linux distributions (e.g. the EDOS and Mancoosi European projects). This work scores high on originality and practical impact among distributors and users of free and open source software. Strong connections with SMEs working on free software were established during these projects, resulting in PPS leading a proposal for a Pôle de Compétitivité on free software, which was eventually created as a branch of the System@tic Pole. 6

8 In its future objectives, PPS proposes a third software project: the Ocsigen Web server and infrastructure for programming of Web applications. A considerable amount of design and implementation work was performed during the last 4 years but did not result in major scientific advances so far. In particular, only one paper was published, in a workshop. The committee is also worried about some apparent duplication of effort with competing projects such as Links and HOP. The committee recommends that PPS, before investing more in this project, performs a careful analysis of the expected development costs vs. expected scientific results, and tries harder to collaborate and join forces with competing projects. 4 Appreciation of resources and of the life of the research unit The laboratory council meets regularly. The discussion with the different categories of personnel did not uncover any problem. People are clearly quite satisfied, despite the lack of office space. The candidate director for the next contract is a strong CNRS researcher. He will compensate for his lack of experience by co-leading the laboratory during the last year of the present contract. 5 Recommendations and advice Strong Points : PPS is a medium size research laboratory, whose excellence is in large part the result of the vision, action and outstanding quality of the original team led by the founding director. Because the founding goal was to favour interactions between mathematicians and computer scientists, PPS is organised as a laboratory without teams, although four research themes can be distinguished in its activities. With a few exceptions, people contribute to at least two of the themes. All themes need mathematical tools and expertise, but clearly belong to computer science. The publication activity is very good, with a high participation in the most prestigious conferences in logic and computer science. Recognition is excellent, with participation in the editorial boards of 6 journals, PCs of 30 conferences of the highest reputation, and the jury of the Gödel Prize. Journal publications could be slightly increased. The scientific environment at PPS is excellent, with well-attended seminars, including one jointly organised by the students of PPS and LIAFA for their own sake, and an important flow of visitors, who usually stand at the whiteboard near the coffee machine, since they can hardly sit in a designated office. Thanks to ANR (7) and European contracts (2), the financial situation is quite good. What needs to be improved : Despite the overall quality, a small proportion of the staff does not contribute much to the research activity. This led us to regard the overall activity as somewhat less outstanding than would otherwise have been the case. Besides the few heavily engaged in teaching activities, this includes a talented CNRS researcher with an international reputation. Recommendations : Assuming a solution to the problem of office space, the university plans to increase its computer science activities by diversifying the spectrum at the interface with other sciences, mostly biology this has started but should expand, medicine, and physics. The current situation of the university makes it difficult to meet this objective before the delivery, in late 2012 at best, of the buildings that will then host mathematics and computer science activities. The new university president, a former member of PPS, has recently proposed a temporary solution which would result into relocating part of the lab in a remote building. This proposal should be considered very seriously and elaborated by the partners until it becomes compatible with the laboratory s strong scientific life. As for the expansion plans, both PPS and LIAFA have a strong scientific identity: any diversification of the computer science activities requires their collaboration, as well as that of the hiring commission. Discussions should start as soon as possible within these different bodies and with the university to plan this diversification. 7

9 Relocating part of the laboratory a kilometre away won t be possible without some increase in the support staff, for both secretarial support and network management. CNRS and P7 should join their forces to satisfy this need, which is indeed already strong (in both PPS and LIAFA), because of the large increase in personnel during the present contract. Although the number of past and present PhDs at PPS is reasonable, we think that the number of PhDs could be slightly increased, and more evenly distributed among the members, while still ensuring the same excellent quality of supervision. This is feasible given the current funding situation in France. We note that 2 theses have been terminated before completion, both for specific reasons. Six theses are carried out with a co-supervisor in the home country of the student. Several internships at the master level are also carried out abroad with joint supervision; this is supported by the newly created RTRA Sciences Mathématiques de Paris. PPS has been very successful in establishing a distinctive and internationally leading activity in some important foundational aspects of programming languages. However, more could be done to relate the work being done in PPS to the work of researchers outside, and to relate the objectives of the research to questions seen as of wider interest in the community at large. This would further enhance the international visibility and impact of PPS. One very nice example where this does occur is in the work applying game semantics to the issue of equational characterizations of type isomorphisms. This has led both to interesting new developments in the game semantics of polymorphism, and to a striking demonstration of the utility of game semantics to a question already of independent interest. One would like to see more examples of this kind. The activity report of PPS lists a number of software developments other than those already mentioned performed during the evaluation period, ranging from routing in IPv6 networks to a virtual machine for the π-calculus. The report does not clearly distinguish between research prototypes, software products, and personal hobbies. While such a prolific activity in software development is healthy for a lab that has software as one of its topics of study, the committee fears that PPS is spreading its manpower too thinly on too many developments. The committee recommends that PPS analyses the expected results and costs of these and future software developments, and concentrates its efforts on a limited number of them. 8

REPORT ON THE RESEARCH UNIT: UNDER THE SUPERVISION OF THE FOLLOWING INSTITUTIONS AND RESEARCH BODIES:

REPORT ON THE RESEARCH UNIT: UNDER THE SUPERVISION OF THE FOLLOWING INSTITUTIONS AND RESEARCH BODIES: Research evaluation REPORT ON THE RESEARCH UNIT: Institut of Ecology and Environmental Sciences (iees) UNDER THE SUPERVISION OF THE FOLLOWING INSTITUTIONS AND RESEARCH BODIES: Université Pierre et Marie