This article, based on an ethnographic survey conducted over five years in a free software developers’ « community », focuses on the research relationship. Since this type of group develops its activities using the digital resources of the Internet, it does not have a place in a physical territory or field. A classical way to take this characteristic into account is to conduct surveys from a distance and undercover, by collecting electronic traces and sometimes complementing them with direct inquiries with the participants (interviews or questionnaires, notably by electronic mail). This method has significant limits since most developers’ groups are not very institutionalized and function in a poorly codified way. Analyzing collective action and studying free software production mechanisms thus involves the identification of rules that are only revealed during action, since they mostly remain informal. The in-the-field ethnographic approach, focused on the group in activity, is then the most appropriate approach. However, this ethnographic approach can prove tricky to implement, since beyond the fact that they reside within the Internet’s digital cloud, these groups appear very elusive: their participants are scattered, authority is diffuse, boundaries are unclear, their organization is not at all formalized, affiliations are fragile, etc. These difficulties have direct consequences on the research relationship, which is malleable and uncertain, and takes forms that are described and discussed as levers for the ethnographic approach. In conclusion, such an ethnographic approach is appropriate to analyze organizational work, beyond the specific case of free software, in poorly formalized organizations.