Re: [nznog] NZIX OpenFlow adventures phase 2

On 1/9/2013 1:00 PM, Lloyd Parkes wrote:

On 9/01/2013, at 12:53 PM, Kris Price wrote:

...

So forwarding is done based on MACs?

My understanding is that forwarding can be done on anything. You could forward based on ethertype, IPv6 payload length, or IPv4 header checksum if you really wanted. The idea is you get to decide what you want and Dean et al. are scratching their heads trying to work what it is they want.

Yep, OpenFlow is flexible with what it can match and what actions it can take (dependent on hardware support of course). Only thing lacking in the OpenFlow specification I really want to see is push/pop of IP headers in order to do IP based encapsulation (limited to MAC, MPLS, and PBB encap at the moment). But, yes, I'm curious how the data plane does/will work in terms of the IX. Maybe I misunderstood and am leaping too far ahead, but I thought it was already pushing flow entries via openflow on this trial switch.