We manage a large network of OpenVPN clients that connect into a central location (think Hub & Spoke design). We have a need to distribute RIP routes to these remote machines - but we've run into a design limitation of RIP and was hoping for some guidance. RIPd will not send out announcements through an interface that it knows it learned the routes from (which I believe is 100% in keeping with the RFC). The problem for us is these remote hosts connect into a single tap interface, which OpenVPN uses to maintain their tunnels. This means the Hub can receive routes from the remote hosts, but it will not send out route announcements to the tap that it learns from the remote hosts so that the remotes can then reach other. We have experimented with using the RIP neighbor statement to try to force RIPd to override this restriction, but it looks like this is not honored. That RIPd still will refuse to send out routes to a tap interface from which it sees it learned the route from. We know we could define a series of point-to-point gre tunnels between each remote host and the RIPd hub, and then configure RIPd to distribute routes over each individual gre interface, but this is pretty messy, and we were hoping someone else might have a better idea. Thanks Jim C.