International audienceDrawing on an ethnographic study of the installation and maintenance of Paris subway wayfinding system, this article attempts to discuss and specify previous claims that highlight stability and immutability as crucial aspects of material ordering processes. Though in designers' productions (such as guidelines or graphic manuals), subway signs have been standardized and their consistency has been invested in to stabilize riders' environment, they appear as fragile and transforming entities in the hands of maintenance workers. These two situated accounts are neither opposite nor paradoxical: they enact different versions of subway signs, the stabilization of which goes through the acknowledgment of their vulnerability. Practices that deal with material fragility are at the center of what we propose, following Annemarie Mol and Maria Puig de la Bellacasa, to term a care of things. Foregrounding such a care of things is a way to surface a largely overlooked dimension of material ordering and to renew how maintainability issues are generally tackled. Keyword

Drawing on a two-year ethnographic study within several French administrations involved in open data programs, this article aims to investigate the conditions of the release of government data - the rawness of which open data policies require. This article describes two sets of phenomena. First, far from being taken for granted, open data emerge in administrations through a progressive process that entails uncertain collective inquiries and extraction work. Second, the opening process draws on a series of transformations, as data are modified to satisfy an important criterion of open data policies: the need for both human and technical intelligibility. There are organizational consequences of these two points, which can notably lead to the visibilization or the invisibilization of data labour. Finally, the article invites us to reconsider the apparent contradiction between the process of data release and the existence of raw data. Echoing the vocabulary of one of the interviewees, the multiple operations can be seen as a 'rawification' process by which open government data are carefully generated. Such a notion notably helps to build a relational model of what counts as data and what counts as work.