Abstract. The continual improvement in computer performance together with the prevalence of high-speed network connections having high throughput and moderate latencies enables the deployment of multimedia applications, such as collaborative virtual environments, over wide area networks. These applications can serve as simulated environments in scenarios such as emergency response training to catastrophic disasters, military training, and entertainment. Many of these systems use 3D graphics for display and may be required to distribute geometric models on demand between participants. Progressive meshes provide an attractive mechanism for such distribution. Previous uses of progressive meshes have sent data using reliable protocols (TCP). However, such protocols have disadvantages in on-demand settings, in that they: (1) use flow control, which limits performance in wide area networks; (2) add additional bandwidth when there is loss; (3) treat all loss as an indication of congestion; and (4) require feature-rich multicast support, which is not always available. In this paper, we modify progressive mesh models to allow reconstruction even in the event of packet loss. We use these modifications in two transmission schemes, a hybrid transmission that uses TCP and UDP to send packets and a forward error-correcting transmission scheme that uses redundancy to decode the information sent. We assess the performance of these transmission schemes when deployed on network testbeds that simulate wide-area and wireless characteristics.