Re: draft-kawamura-ipv6-isp-listings-00.txt

[Chuck Anderson <cra@WPI.EDU>]

On Sat, Jul 24, 2010 at 05:21:06PM -1000, Antonio Querubin wrote:
> On Wed, 21 Jul 2010, Lindqvist Kurt Erik wrote:
>
>> So I suggest that minumum is
>>
>> - PA Block
>> - Customer Assignments
>> - DNS resolvers IPv6 capable
>> - Native (v6 only or dual-stack) upstream connection
>
> If you by native you mean non-tunneled, I disagree with this.  
> Operationally how does a user even verify that their connectivity does or 
> does not transit through one or more tunnels on their way to whatever  
> they're trying to reach even if their own connectivity is native?  If the 
> ISPs peer(s) or upstream(s) have their act together, you may not even see 
> much of a difference.  And consider that the upstream(s) could very well  
> be tunneling internally all over their backbone.
>
> So for the stated purposes of this draft why is requiring the ISP to have 
> native upstream connectivity relevant at all?  Focus needs to be kept on  
> the services visible to and directly accessed by users.

Agreed.  I think there is a bias against tunnels because they often 
imply:

- poor interconnectivity
- poor locality of interconnectivity
- high latency
- lower MTUs and the problems that brings

So the focus should be on improving those aspects, rather than 
avoiding tunnels explicitly.