I concur that we need to be realistic about this. Having had discussions with
operators who are trying to deploy IPv6, the reality is that even if IPv6 were
enabled universally tomorrow - a subset of subscribers, subscribers'
devices/applications, and content providers will only support IPv4 until key
pieces of consumer gear are replaced (many of which do not yet support IPv6,
even if replaced). CGN facilitates transition, as the move to IPv6 happens.
ISPs have already indicated (a few times) that RFC1918 space is not practical
behind the CGN due to the (real) possibility of overlap with customer
addressing. Thus, having a shared /10 minimizes the use of public globally
assigned addresses and avoids address squatting.
(as an individual contributor)
[mailto:ietf-bounces(_at_)ietf(_dot_)org] On Behalf Of Ida Leung
To: ietf(_at_)ietf(_dot_)org; iesg(_at_)ietf(_dot_)org
Subject: Re: Consensus Call: draft-weil-shared-transition-space-request
I read a lot of emails today regarding this subject. I would like to express
my personal thought on it. I support the allocation of the /10 for this
purpose as laid out in "draft-weil-shared-transition-space-request"
In organizations like the one I work for, we have solid IPv6 rollout plans
which include the necessity to support some ongoing IPv4 connectivity beyond
run out. This is related to the fact that too mach IPv4-Only equipment remains
in the network (and still on retail store selves, selling daily just in time
for Christmas demand) which cannot be feasibility removed in a short period of
We have worked tirelessly with vendors to move forward, but reality is king.
IPv4 (with address sharing in some form) will need to accompany the IPv6
deployment for a period of time, of which CGN plays a vital role (in the form
of CGN and later potential in the form of DS-Lite or the such technologies).
To facilitate this functionality, non-RFC1918 space will need to be used such
that we can offer a working service to customers. Using a pre-defined
allocation helps us and other operators achieve a deterministic approach
without the variances of needing to find other, less legitimate space for such
purpose. The alternative to the /10 is likely squat space. Worse yet.. Many
operators choosing many pools of RIR space which in aggregate will be much
greater then a single /10 (with no guidance as to what and how it's used).
Ietf mailing list