With the emergence of video on community media sites such as YouTube or TED, there is a need to understand interactions by community participants around video in order to maximise the potential of these systems to support communities and create meaningful interactions online. In this paper, we illustrate how ethnomethodologically-informed studies of the social interactions among a small group of co-located participants around video may be used to inform the design of an online video interface which may support interactions by a large distributed group of participants. Our point of view is grounded in the idea that the interactional accomplishments of a co-located group remain relevant at a distributed level, thereby allowing an ethnomethodologically-informed approach to arrive at effective implications for design of online systems. A strength of this approach is the potential to create implications which are properly grounded in a set of observations of the precise ways in which people interact to accomplish social interaction around video. For the purpose of illustration, we perform an analysis of a fragment of video data collected from a quasi-naturalistic experiment of a co-located group collaboratively annotating a video, from which we propose a number of design implications for a video annotation interface.