Are pseudowires a simple and efficient way to build mobile broadband networks?

There is a topic that I have been discussing often with colleagues and vendors, but none of the views that I have heard have caused me to form a firm opinion yet.

I understand perfectly the advantages of using pseudo wires (PWE3) within the IP/MPLS backhaul network. Pseudo wires can be differentiated by QoS and used to backhaul the mobile traffic (voice, data, and signalling) generated at the NobeB and RNC. Besides that, it is possible to carry legacy mobile services through the same infrastructure, in a very simple and effective way.

But, what I’m not sure is if using PWE3 in a completely new backhaul network, without legacy traffic, is the best solution. Some vendors says that yes, PWE3 is the best and simplest way to backhaul even 3G packets. But I am afraid they think this way only because they are used to evolve networks from GSM to UMTS (or CDMAto CDMA2000).

I wonder if using L3VPN or L2VPN (H-VPLS), or another method that I don’t know, wouldn’t be better, but it is very hard to get a conclusion, even because all the networks that I have seen migrated from 2G to 3G, and had to adapt the legacy network to transport 3G and legacy traffic.