Even i am interested in knowing the exact issue with
using IGPs? What is the most common CE-PE route
exchange behaviour now ... Static routes OR IGP OR
BGP??
Dave, are you referring the CE loopback address also
to be local?
cheers,
Elwin.
Oops... I answered this privately, assuming NANOG at large would
not be interested. The issue here is that Cisco will tend to
add IGP routes to the default table, not the VPN table. Bad
things ensue.
I was actually referring to the CE interface address; peering
with the CE's loopback is, IMHO, more trouble than its worth
unless you have multiple connections on the same router. But
as long as the address you are peering with is in the private
routing table, you're fine, regardless of whether or not it is
also in the default table.
I cannot speak to the "most common;" I think it is too early
to tell. But we are tending towards static routes (nice and
stable, without the chance of the other guy breaking you)
and BGP (which is already designed to handle trans-border
communication.)
I have not tried it, but I would assume the OSPF area "repair"
toys would work nicely over this, if you want an IGP running
across your CE routers. (I'm more of an ISIS guy than an
OSPF guy... anybody know why this would blow up in your face?)
-Dave
not be interested. The issue here is that Cisco will tend to
add IGP routes to the default table, not the VPN table. Bad
things ensue.
What ... ? For protocols that have been vrf aware routes go into vrfs
and not global RIB. Those IGPs would be in shipping images: RIP, OSPF.
Very soon also EIGRP & ISIS. Just watch your CPU before using IGPs on a
wide scale with tons of customer routes flappoing :).
Just not be understood that I recommend the above IGPs 
I am also
seeing worldwide the following set of protocols on the PE-CE in order or
preference: static, BGP & RIP.
R.
That's the problem: not all the IGP's (specifically, not ours) are
aware, and handling for a non-aware IGP on a VRF interface is
more dangerous than one might expect.
-Dave