Thoughts about ipv6 white listing

Brian E Carpenter brian.e.carpenter at
Sun Dec 5 03:52:01 CET 2010


On 2010-12-05 08:44, Erik Kline wrote:
> On 4 December 2010 11:09, Brian E Carpenter <brian.e.carpenter at> wrote:
>> On 2010-12-05 07:57, Nick Hilliard wrote:
>>> On 04/12/2010 18:42, Doug Barton wrote:
>>>> A content provider is not going to knock 470,000 users off line, that
>>>> just
>>>> isn't going to happen.
>>> Think again.
>> Except that it's encouraging the wrong solution (turn off IPv6
>> and 6to4 at the subscriber end if they don't work).
>> The right solution is: make them work, which is mainly the
>> responsibility of the ISPs at the content provider end.
> I'm really not sure how to parse this last sentence.  Can you clarify?

I think my previous reply to Tore explained what I mean, and
I need to write a draft, but now I will be out for a week, so
I will try to work on this before the end of the year.


> Past experience shows that users apparently suffer in silence, and
> they do not call their ISP if only 1 destination is down.  They just
> assume that destination is having some problem.  Hell, I would
> probably do the same thing and just suck it up, until several hours
> had passed (assuming I /really/ needed to reach that destination such
> that I kept trying; otherwise I'd probably wait a day or even more).
> Statistically speaking, a network like Comcast's will have an upper
> limit of O(10,000) broken users/NAT'd endpoints (see Jason Livingood's
> presentation at IETF79).  [To be clear, for the record: this is not in
> any way Comcast's fault.]  Let us assume, for the sake of argument,
> that Comcast were willing to fund the contacting of and upgrading of
> or CPE replacement for these users/NAT'd endpoints.  Super nice guys!
> So, how do they find them?  How does a content provider help with
> finding them without painting a privacy-concern-seeking missile target
> on itself?
> Ron Broersma has repeatedly said that folks shouldn't be afraid to
> "break some glass".  But everybody has their own notion of what "some"
> means.

More information about the ipv6-ops mailing list