Abstract
Nowadays, several different actors are involved in acquiring, distributing and updating spatial data. Such a situation naturally leads to a multiplication of heterogeneous data of different types, formats, abstraction levels and qualities. Blind integration of this data can undermine the consistency of the database and degrade the dataset's quality. Within this broad issue, we limit our scope to the updating of geographic data by different actors spread out over a communications network – with some data liable to change while the actor is disconnected from the network. Data is replicated on different sites and there is no server holding centralized information. Moreover, updating is conducted simultaneously by the different actors, sometimes in disconnected mode. The evolutions are therefore not necessarily relevant to a particular user, sometimes contain errors, can be conflicting and consistency issues can arise during their integration into different datasets. In this context, our objective is to propose solutions to allow coherent integration, as far as possible automatically, of spatial-data updates in a multi-master and asynchronous optimistic replication environment. To achieve this, we propose a global integration strategy, based on standardized metadata, to provide solutions to improve the quality of a dataset during updates by multi-source evolutions. We will have to, on the one hand, deal with the relevance of the proposed evolutions and, on the other, ensure dataset consistency commensurate with the final user's requirements. This strategy depends on the prior establishment of a spatial data infrastructure in which a communications network has been defined, the users and their different roles are known, and updates can be exchanged using a common strategy [Pierkot et al., 2006]. In this infrastructure, a metadata model that allows formal relationships between data, actors and the evolutions can be set up. This paper is structured as follows: We start by defining consistency and quality in spatial databases. Then, we present the different parts of the metadata model which we have defined in the infrastructure and we cover in some detail the quality metadata relating to the evolutions. Subsequently, we present the integration strategy and, in particular, the modules for verifying the relevance and consistency. Then, we analyze the results obtained for the consistency-checking process. Finally, we conclude and provide some perspectives for this work.