This is great news! During my studies, I have learned to really like EIGRP. But, I would never recommend EIGRP, soley because it is proprietary. To me, it has always been like the VHS vs Beta war and anything else similar that has a proprietary product involved. This is likely going to rock the networking world. The options have really opened up.
I vote against EIGRP for three reasons:
1: Challenging to control the overall topology metrics, especially with QoS in place (as QoS often uses the configured bandwidth to derive its values). In my world, bandwidth is of no use in setting link metrics.
2: Unequal-cost load-splitting can (and will) lead to truly excess hops (imagine a triangle of routers R1/R2/R3, links are all metric=1, variance >=2. R1 can send to R3 direct or via R2. R1 sends 2/3 of traffic direct, and 1/3 to R2. R2 sends 2/3 of the indirect traffic direct to R3, and 1/3 of the indirect traffic BACK to R1. At this point, 2/3 of the traffic went direct to R3, 2/9 of the traffic went indirect to R3, and 1/9 of the traffic is back where it started!)
3: Lack of support for the necessary inner workings for MPLS Traffic Engineering.