Re: [pcp] WGLC: draft-ietf-pcp-anycast-02 comments due by NOV 10

[Sebastian Kiesel <[e-mail]>]

Hi Med,

On Thu, Oct 30, 2014 at 10:11:11AM +0000, [e-mail] wrote:
> Dear all,
> 
> I have a comment about this text:
> 
>    The PCP client SHOULD select the PCP anycast address to be of the
>    same IP address family as its requested PCP mapping, i.e., the
>    address family of the Requested Internal IP Address.
> 
> This assumes the PCP client is always dual-stack. 

This is not true. Or at least not what we wanted to say.

What we want to say is:  Whatever address family you decide to use on
your side of the NAT for the actual user data traffic, you should use
the same address family for the related PCP anycast signaling to the
NAT.  It doesn't matter whether the NAT will translate the user data
traffic to the same or another address family on its other side.

If a host is not dual stack, i.e., IPv4-only or IPv6-only, it has no
other choice than sending both the user data and the PCP anycast
signaling using the one address family it supports.
This is completely fine and inline with this requirement.

Does this make sense?

> More importantly, this is a case that is covered in general by this behavior : https://tools.ietf.org/html/draft-ietf-pcp-server-selection-06#section-3.
> 
> I suggest removing the text above. Instead add a reference to the server selection document.

I must say that I have difficulties finding this special case in
draft-ietf-pcp-server-selection-06#section-3 
Can you narrow down which statement includes our case?

Thanks
Sebastian