Open science" today is a boundary object : invoked across policy, publishing, technology, and academia, it means different things to different communities, and yet it helps to coordinate their activities. But this vagueness is not neutral -it creates frictions whenever the values and practices underpinning "open" come into focus. These frictions are particularly salient in the infrastructures built to support open science, which do more than host and distribute research. They embody particular epistemologies: ways of knowing, valuing, and governing scholarly work. Part of my ongoing research is to explore this issue through the case of OpenEdition, the French public infrastructure for open access in the social sciences and humanities (SSH). My study helps us see how infrastructures are not neutral backdrops but actors that organize scholarly practices, and how studying them requires specific methodological and ethical commitments.