Star and Ruhleder’s “Steps Towards an Ecology of Infrastructure” argues that infrastructure should not be understood as a neutral substrate beneath social or technical activity, but as a relational, contextual and politically saturated ecology that becomes visible when systems fail, exclude or demand translation across communities. Through the case of the Worm Community System, a collaborative software environment designed for geographically dispersed geneticists studying C. elegans, the article shows that even technically successful systems can encounter deep problems of access, uptake and use when designers mistake installation, connectivity and training for merely technical matters. Its central contribution is the claim that infrastructure is always relational: what appears simple to developers, such as downloading files, using UNIX or configuring X Windows, may become a complex organisational problem for biologists whose laboratories, funding, equipment, disciplinary habits and support networks differ radically. The authors develop Bateson’s levels of learning to distinguish between first-level resource problems, second-level contextual clashes and third-level political or epistemic disputes involving standards, trust, collaboration, competition and disciplinary identity. The Worm Community System becomes a case study in infrastructural double binds: a tool intended to democratise access may reinforce inequality when richer laboratories, stronger technical cultures and better institutional support make participation easier. Ultimately, Star and Ruhleder conclude that collaborative systems must be designed through an ecological understanding of work, where technical architecture, local practice, tacit knowledge, institutional power and community norms are treated as inseparable. Their argument remains decisive because it shifts infrastructure from background to foreground, revealing it as the negotiated condition through which knowledge work becomes possible, fragile or inaccessible.