Questions for an ISP

[Mark Townsley]

On May 25, 2012, at 11:17 AM, Jeroen Massar wrote:

>> ·         How will IPv6 be implemented (Native IPv6 with Dual-Stack, 6rd
>> or whatever)
> 
> If you are a company you should demand NATIVE IPv6, anything else is
> just a transition method and should be avoided as much as possible.

Tunnels that trombone traffic around and cause latency, brokenness, or MTU issues are no fun for anyone, but shouldn't a company be more worried about the SLA itself vs. how it is encapsulated by the ISP once it leaves the company premises? 

e.g., for business services, much of the dual-stack service you will find today is actually based on 6PE, which takes an IPv4 MPLS network modified with the minimum necessary to transport IPv6 packets in a way that gives the user production-quality service that the ISP can operate. For residential, 6rd isn't all that different in that it takes an IPv4 access network modified with the minimum necessary to deliver IPv6 in a way that gives the user production-quality service that the ISP can operate. 6PE uses a 4 (or perhaps 8) byte encap, setup with IPv4-based protocols, in order to route IPv6 in a mesh fashion across a network. 6rd other uses a 20-byte encap, setup via IPv4-based protocols, in order to route IPv6 in a mesh fashion across a network. Does the actual presence or make-up of the shim sitting between the IPv6 header and Data-Link matter here?

I'm the first one to try and help ISPs that deploy 6rd have all options on the table to sunset it when the time comes (draft-townsley-v6ops-6rd-sunsetting), but I see that more as a decision for the ISP to make about how to operate its network. As long as the service is on par with or better than IPv4 (including a 1500 byte MTU, comparable RTT and bandwidth, turnaround time for problems or outages, etc), why would an average everyday company or end-user care how their IPv6 packets are encapsulated?

- Mark