Centre de Recherche INRIA Nancy-Grand Est
615, rue du Jardin Botanique
Meeting will take place in Room B-013
Monday, November 29th
- Meeting from 10:30am to 6:00pm
- Dinner at 8:00pm
- Brasserie Excelsior
50 Rue Henri Poincaré, 54000 Nancy
- Brasserie Excelsior
Tuesday, November 30th
- Meeting from 9:00am to 5:00pm
SCORE: Claudia-Lavinia Ignat, Gérald Oster, Hien Thi Thu Truong, Pascal Urso
REGAL: Marc Shapiro, Marek Zawirski
CASSIS: Asma Berregba, Abdessamad Imine, Michael Rusinowitch
ASAP: Achour Mostefaoui
XWiki SAS: Fabio Mancinelli
GDD: Pascal Molli
ANR: Gaëll Guibert
Invited people: Stéphane Weiss
Sessions (preliminary program)
|1:30pm||3:30pm||Opening Session / Task 2: Requirements and Infrastructure|
|4:00pm||6:00pm||Task 4: Security and Privacy|
|11:00am||12:30pm||Task 3: Efficient Replication and Consistency Maintenance|
|1:30pm||3:30pm||Task 5: Experimentations and Performance Analysis / Discussions|
The STREAMS Project: Overview
Presenter: Gérald Oster Documents: PDF
Abstract. In this talk, we will review the STREAMS project objectives and organisation. Then, we will discuss the proposed description of work.
Distributed Data Structures
Presenter: Achour Mostefaoui Documents: PDF
Abstract. Distributed systems are characterised by their high degree of non determinism implied by asynchronism and potential failures of systems components. Nowadays distributed systems are also faced to new sources of non determinism such as untimely connection/disconnections (churn) and system components compositions (size, structure, etc.). In peer-to-peer systems, problems are even harder since the system size, generally unknown, is very high. This is the main reason why proposed solutions must scale. This presentation is about distributed data structures such as queues, stacks, bags, wikis, etc. These data structures grely on synchronisation services named agreement services. The goal of this presenation is to give quick overview of agreement problems, to show which of them are equivalents, which of them are impossible, under which conditions they become impossible.
A Log-Auditing Approach for Peer-to-Peer Trust Management
Presenter: Hien Thi Thu Truong Documents: PDF
Abstract. Most collaborative distributed systems including social software rely on a central authority and place personal information in the hands of a single large corporation which is a perceived privacy threat. In a peer-to-peer collaboration rather than having a central authority which has access to all users data, control over data is given back to users. Users define their network of trust containing people that they trust and with whom they wish to share their data. In this setting, data management policies are more difficult to ensure. Peers may misbehave and violate data privacy policies by not respecting obligations associated with them. Traditional access control models require an explicit security mechanism where obligations are checked by an authority before data is used and therefore are not suitable for a peer-to-peer collaboration. We propose a novel trust management mechanism based on log auditing where we log modifications done by users as well as the obligations related to usage policy of shared data. Peers have assigned trust values and a peer collaborates only with trusted peers. By using logs we detect mismatching between user actions and the received obligations and then adjust trust values for the users who misbehaved.
Optimistic Access Control Model for Collaborative Editing Systems
Presenter: Asma Berregba Documents: PDF
Abstract. Security represents a real challenge in collaborative applications. In this presentation, we focus on collaborative editors which represent a famous application of collaborative work allowing many dispersed users to edit a shared document simultanously. It is very important to add a security layer on the top of such an application to control the access to shared documents. However, it is difficult to manage both document and security layer updates while maintaining the convergence of the document. Security may also degrade the performance of the distributed collaborative editor if it is managed in a centralised way. To overcome these problems, we propose an optmistic, distributed and replicated access control model allowing users to administrate their documents. Moreover, we will discuss an extension of a distributed collaborative editing model with a garbage collection mechanism in order to apply the model in mobile devices.
The XWiki Platform: An overview
Abstract. In this talk I will present the XWiki platform from a technical point of view, its components, and the XWiki ecosystem (usages, community, etc.)
Messages and Expectations from ANR
Abstract. This presentation will give an overview of the selection process used by ANR. Then, it will present the expectations of ANR (from an administrative and also technical point of view).
The ConcoRDant Project (ANR-10-BLAN 0208)
Presenter: Marc Shapiro
Abstract. A short presentation of the ConcoRDant project (http://concordant.lip6.fr/) which aims to investigate in depth Commutative Replicated Data Types (CRDT).
The Kolflow Project (ANR-10-CORD 021)
Presenter: Pascal Molli
Abstract. A short discussion about Kolflow project (http://kolflow.univ-nantes.fr/) which will investigate human-computer collaboration in continuous process of knowledge building.
The Wiki3.0 Project
Abstract. The objective of Wiki3.0 project (http://wiki30.xwikisas.com/) is the development of an open-source platform based on XWiki that addresses the three major evolution axes of collaborative Web: real-time collaboration, social interaction integrated into the production (chat, micro-blogging, etc) and on demand scalability (cloud computing). This platform should be competitive with major editors of collaborative Web developed by Google such as Google Wave, IBM and Microsoft.
The Compatible One Project
Abstract. Compatible One is a research and development project funded under the competitiveness cluster System@tic. This project aims to develop a complete "Cloudware" solution under a free license. The different software products from the project partners, among others, will be adapted and integrated. Compatible One has the ambition to enable everyone to create, deploy and administer private, public or low-cost hybrid Clouds providing services of all kinds (IaaS, PaaS and SaaS).
Logoot: A Scalable Optimistic Replication Algorithm for Collaborative Editing on Peer-to-Peer Networks
Presenter: Stéphane Weiss Documents: PDF
Abstract. With the arrival of Web 2.0, collaborative editing becomes massive. This scale change is undermining the existing approaches that were not designed for such a charge. To distribute the load, and thereby achieve greater scalability, many systems use an architecture known as peer-to-peer. In these systems, data is replicated on several peers and it is then necessary to define optimistic replication algorithms adapted to the characteristics of peer-to-peer : dynamicity, symmetry and of course the massive number of users and data. In this talk, we propose Logoot a commutative replicated data type (CRDT) for text documents. We also present experimental results of the performance evaluation of Logoot that we analysed on a testbed of changes produced over 2000 pages of Wikipedia.
TreeDoc: A Commutative Replicated DataType Designed for Cooperative Text Editing - Issues and Future Work
Presenter: Marek Zawirski Documents: PDF
Abstract. We recall the design and issues of Treedoc, a CRDT design for cooperative text editing, based on an extended binary tree. We discuss future engineering work designed to minimise storage overhead: leveraging the equivalence between a balanced tree and an array, and a mixed (binary-Nary) tree encoding.
A Library of Commutative Replicated DataTypes - Work in Progress
Presenter: Marc Shapiro Documents: PDF
Abstract. We examine the basic properties required for a replicated data type to be a CRDT, i.e., to converge without concurrency control. We distinguish the state-based and operation-based approaches. In the former, a sufficient condition is that object histories form a semi-lattice (convergent RDT, or CvRDT). In the latter, a sufficient condition is that concurrent operations commute commutative RDT, or CmRDT). We will discuss how to build a CvRDT. The two models appear equivalent in a broad sense, but the details differ. Finally, we will discuss some basic ideas/approaches to design CRDTs: unique identifiers, UID-encoded order, tombstones, and propose a number of CRDT constructions, generally speaking variations over sets and maps.