One of the challenges of open science today remains the interoperability of reliable data. Both the exchange and exposure of data within scientific communities, as well as the ability to search resources within a given field, face significant gaps, not only in standardizing datasets and querying across them, but also in the shared infrastructures and governance mechanisms that would let communities take collective responsibility for FAIR data exchange. The LUMEN project addresses these issues by building a federation of four scientific communities, Social Sciences and Humanities, Mathematics, Earth System Sciences, and Molecular Dynamics at both the governance and technological levels, with a particular focus, here, on the shared PaaS (Platform as a Service) infrastructure that hosts its tools and, we argue, acts as a governance instrument in its own right. This contribution describes the current implementation of LUMEN infrastructure, combining conceptual argumentation with an infrastructure case study.

In this context, the concept of infrastructure as governance instrument is understood in an operational sense. The PaaS layer not only hosts applications, but actively structures how data can be shared, searched, accessed, and reused. Governance is therefore embedded in technical capabilities: for example, the requirement to provide metadata in machine-readable form for indexing or the requirement to exist in interoperable formats.

On the one hand, LUMEN is working to create a shared governance model across communities. On the other hand, LUMEN proposes a set of technological services covering the full chain, from data exposure and metadata enrichment to cross-platform discovery and semantic querying, over FAIR data.

Communities are thus federated while retaining autonomy over their data (a decentralized approach to data ownership). Each community has access to LUMEN’s technological services – from the Discovery Platform and its customization layer, through the advanced Meta-Search engine, to metrics and visualization tools. A community can also join the federation with its own data management platform. All these tools are embedded within a single PaaS infrastructure, lowering the technical barrier for each community and freeing resources for what matters most to them: curation, editorial work, and engagement with their researchers.

A shared PaaS infrastructure, used by federated communities to deploy and maintain their applications and services, can effectively support FAIR data exchange, provided it is designed as a community infrastructure, not just as a hosting layer.

PaaS thus functions as a common platform for the federated communities, offering, for example, a uniform application execution environment and common security, scaling, and integration mechanisms. Communities can run their own applications without having to build the entire technological infrastructure from scratch. They retain full control over the data and application logic. Crucially, in the PaaS model the application owner, the data steward and the infrastructure provider are deliberately distinct roles, a separation that turns the infrastructure into a device of shared responsibility.

Additionally, in the PaaS model, community exchange does not rely on manual data transfers, but rather on shared access to infrastructure layers.

Each community publishes its data as “data products” (self-contained, governed, and contract-bound units), exposes it through common mechanisms of the given platform, and decides the scope and context of sharing. Through appropriate interfaces, other users are able to discover this data while preserving context and accountability. For instance, in the LUMEN environment, a dataset published by a community in Earth System Sciences can be exposed through its own discovery platform (as a data product), indexed by the Meta-Search engine, and then discovered by a researcher in Social Sciences and Humanities. The discovery process relies on shared metadata standards, while access is controlled by each community.

PaaS therefore presents an interesting contrast to other models, such as IaaS (Infrastructure as a Service), which leaves too much technical work to communities without dedicated infrastructure teams, and SaaS (Software as a Service), which hands control to a single provider and erodes community autonomy over data and policies. PaaS standardises what must be shared across communities by providing common services and rules for how data are described, accessed, and discovered, thus ensuring a consistent baseline for collaboration. At the same time, it preserves flexibility where disciplinary diversity is essential, allowing each community to maintain its own data practices and governance. This combination enables federation without centralization, supporting both interoperability and community autonomy within a shared framework.

A PaaS‑based shared infrastructure enables community exchange not by centralising data, but by providing common technical services, discovery, interoperability, access control and provenance, that allow FAIR data to circulate across community applications while remaining governed by their respective communities. The LUMEN’s approach demonstrates that a shared PaaS infrastructure can function as a governance tool not through central control, but by standardizing interfaces, exposure mechanisms, and interoperability requirements. In this sense, the infrastructure itself becomes an act of community governance.