I usually file things like this under "duh", but a recent conversation with a network architect made me think that it's worth noting in public space.
I recently had a problem related to dynamic route distribution in an MPLS environment. Several remote sites were served via a Paetec MPLS network, all sites falling under a common major net, 192.168.0.0/16, with DMZ and edge networks built on another major private net. In a frame-relay or PTP network, routing would easily be handled by EIGRP or OSPF, or even static routes, and moving subnets from one site to another would be as simple as referencing them in the routing table of a core switch or router at the target site. In an MPLS environment, however, the routers connecting the LAN to the MPLS network are controlled by the provider, and the nature of MPLS dictates that these routers have explicit routes that are propagated among the participating MPLS routers via BGP.
The problem was that in order to be able to move subnets around between sites, for instance in order to accommodate the failure of an inbound Internet circuit at one site by moving VPN subnets to another site, the provider would need to be in on the configuration, and that can hardly be considered a workable proposition in an emergency. So, I built an OSPF area at each site that was redistributing connected and static subnets, and had the MPLS provider configure their router to participate in that area, and redistribute routes learned via OSPF into BGP. With this in place, adding a route statement to any router at any site causes that route to be propagated throughout the MPLS network, and then into the OSPF area at each site, including defroute information. Turning the routing around now is absurdly simple and doesn't require the involvement of the provider at any level. Prior to this, all routes were static, with subnet of the major net assigned to each site, such as 192.168.0.0/20, 192.168.64.0/20 and so on. Now, it's as granular as necessary.
If you're going to undertake this configuration, be sure that you properly assign loopback interfaces on the core routers and switches with the highest IP in the subnet assigned to that site to ensure that it becomes the DR/BDR interface and the status of any physical interface doesn't interfere with proper OSPF operation.