On Jul 30, 2012, at 17:22 , Michael Thomas <[email protected]> wrote:
> On 07/30/2012 05:21 PM, james woodyatt wrote:
>> On Jul 30, 2012, at 11:08 , Michael Thomas <[email protected]> wrote:
>>> 
>>> If we believe that ipv6 is ready to go for mass deployment, why do we
>>> not pressure home router vendors to default to sending router advertisements
>>> with ULA addresses that, if necessary, get NAT'd at the border just like
>>> 192.168 space does today.
>>> 
>>> I mean, nothing bad would happen, right?
>> 
>> What does the conditional phrase "if necessary" mean in your mind?  Under 
>> what circumstances do you imagine this would not be "necessary" for 
>> operational continuity?
> 
> When you have native v6 connectivity.

On Jul 30, 2012, at 17:23 , Michael Thomas <[email protected]> wrote:
> 
> I should add that by native v6 connectivity, you'd also be getting a provider 
> public v6 as well where rfc3484 saves the day.

I thought as much.  Here's why I considered that idea and rejected it when I 
was thinking about this problem in the context of a certain home router 
product.  What happens when I have a persistent connection from my ULA address 
through the NAT64 path to a remote IPv4 address when the router acquires IPv6 
connectivity?

+ If we stop translating all packets to and from the ULA address, then that 
connection is broken.

+ If we only stop translating packets for new flows, then we effectively break 
3rd-party referral for the 64:ff9b::/64 prefix.

+ If we continue translating every packet, potentially in perpetuity, then we 
break locality of uniqueness by leaking our ULA routability into the global 
IPv4 Internet.

It seemed clear to me that all of these options collectively exhausted the 
available alternatives, and they all came with damaging outcomes without any 
real benefit visible to anyone that I could identify, so I decided not to 
pursue the idea very vigorously.


--
james woodyatt <[email protected]>
member of technical staff, core os networking



_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to