In this situation you can use different solution. If it's new connection between different MPLS network you can use two variant.
Your situation it's very often in the VPNv4 prefix exchange with MP-BGP implementation.
About RD, RD using by VRF import process to import routing information to the isolated VRF instance. Take a look, this sample (knows as «route leaking»):
ip vrf test
route-target import 1:1
route-target import 1:2
route-target export 1:1
Ok we have one vrf (it's name test) and two different RD value, first it's 1:1 and second it's 1:2. ISP to define for this vrf RD 1:1, but this client should be need another service from another vpn (like VoIP… it does not matter).
First we import all routing update information and routers from MP-BGP, this operation operate by VRF-import process, this process knows from what RD import and install routes. As result from this configuration we can see at routing table for this vpn, many different routes from one VRF (VPN) and another VRF (VPN). But you should be controlling in this situation address space between different vpn's (for this situation you can controlling import process with route-map).
Information about route leaking, you can find here - http://www.cisco.com/en/US/tech/tk436/tk832/technologies_configuration_example09186a0080231a3e.shtml
Another solutions it's rewrite RD value in the MP-BGP updates. About this solution and how can configure at Cisco IOS, you can find here - http://www.cisco.com/en/US/docs/ios/12_0s/feature/guide/fsrtrw.html
And finally you can use "inter-vrf connect" best practice without VPNv4 NLRI MP-BGP exchange.. You can configure logical sub-interface or using Tunnel's interface to connect between PE routers different ISP with our Layer 2 domain. (Like one VLAN ID between trunk or access physical connection between different PE routers and place this interface with needed vrf instance).
PS: I think you can use this three solution's: route leaking, RD rewrite and inter-vrf connect. That your solutions? I don't know it's your chouse...
