Dissertation, Otto-Friedrich-Universität Bamberg, 2023

How can we facilitate the exchange of datasets and other data assets across heterogeneous technical and social environments? This is a classical problem that is more relevant today than ever before, relevant both on a global scale as well as the scale of organizations. We need to distinguish between two notions of data sharing and exchange: one based on the integration of data at the schema level, and the other, which is our focus here, is based on the description, publication, discovery, compilation of data as an asset (similar to how a book is managed in a library).

The state-of-the art of the data sharing and exchange problem is polarized. On one end we have emerging 'platforms' which can be susceptible to become walled-gardens that do not interoperate with each other. On the other, we have recommendations such as FAIR guidelines, or architectural abstractions such as data lake and data catalog that help orchestrate vendors in developing interoperable technologies; taking us in the right direction but not too far, due to being too high-level and more normative/prescriptive than constructive. This work fills a gap by making a (1) 'medium-level' proposal, (2) in the form of a constructive recommendation (an architectural pattern), (3) which incorporates several major recommendations.

We propose the Basin Network, a distributed architectural pattern which revolves around two novel abstractions: the Offering (message exchange format) and the Basin (node). We adopt an indirection approach: a data asset is managed and exchanged via a surrogate (intermediate representation) which identifies, represents, describes, and effectively 'stands in' for it. The Offering is a model for such surrogates; built on the URI-Resource framework for hypermedia authoring but avoiding some of its pitfalls. The Basin is structure to author, catalog, and manage Offerings and exchange them with other Basins (publish/subscribe), resulting in a Basin Network (akin to a network of data lakes).

We demonstrate the applicability of the proposal by applying it to use-cases in three domains: a computer-aided manufacturing project, an IoT project in smart agriculture, and one in crowd data management.