Plural repositories do not produce a plural knowledge infrastructure by aggregation alone. Baglioni and colleagues compare FAIRsharing, re3data, OpenDOAR and ROAR, examining overlapping coverage, divergent descriptive properties and incompatible terminologies. Their operative device is the metadata crosswalk: a comparative alignment of registry data models from which they propose and evaluate a common descriptive model. The article makes a valuable distinction between preserving multiple registries and requiring that their information remain untranslatable to one another. Interoperability is presented as pragmatic coordination, not the eradication of institutional plurality. Yet the practical obstacles they identify, including differences in subject classifications and information not exposed through interfaces, also reveal that a shared field name may conceal unlike underlying concepts. The paper connects digital libraries, semantic modelling, research infrastructures and open-science governance. It provides an exacting technical counterpart to philosophical discussions of difference: the political question of what may be connected acquires material form in mappings between attributes, persistent identifiers, missing values and curatorial decisions.