multiple prefixes

Brian E Carpenter brian.e.carpenter at
Wed Feb 13 09:14:50 CET 2013


On 12/02/2013 18:00, Matthew Huff wrote:
> Tore,
> Do you know of any knowledge base of mainstream commercial network equipment (routers, switches, firewalls, load balancers, etc...)
> that supports NPTv6? There seems to be very few equipment on the market currently that supports NPTv6.
> IMHO, the push to force end-to-end connectivity (no-nat) that pervades the IPv6 community has slowed the growth of IPv6
> considerably. There are a lot of corporate entities (us included) that have no interest in having p2p connectivity. In fact, we do
> everything possible to block it and will continue to do so in IPv6. 

Indeed. I hate to repeat myself, but that's *exactly* why we wrote RFC4864.
Neither NAT66 nor NPTv6 are necessary for protection. There is a case for
NPTv6 for multihoming, but there's also a case against it in the RFC queue:


> Imagine if there were wide support for some sort of NAT46
> solution so that internal networks could stay ipv4 but have ipv6 connectivity while they slowly migrate to ipv6. Perhaps they would
> never move internally from ipv4, but externally they could be 100% ipv6. This would be fine with the vast majority of corporate
> networks. 
> FYI, we have a PI IPv6 address space and advertise it via BGP, and are slowly migrating internally to dual-stack systems. However,
> we have run into a lot of application issues so far - mostly applications that don’t expect multi-homing and legacy equipment that
> will never support ipv6.
> ----
> Matthew Huff             | 1 Manhattanville Rd
> Director of Operations   | Purchase, NY 10577
> OTA Management LLC       | Phone: 914-460-4039
> aim: matthewbhuff        | Fax:   914-460-4139
>> -----Original Message-----
>> From: at [mailto:ipv6-ops-
>> at] On Behalf Of Tore Anderson
>> Sent: Tuesday, February 12, 2013 12:41 PM
>> To: David Barak
>> Cc: ipv6-ops at
>> Subject: Re: multiple prefixes
>> * David Barak
>>> --- On Tue, 2/12/13, Tore Anderson <tore at> wrote:
>>>> You'll need [non-free] hardware to run the [free] software
>>>> implementation on, even at a small scale.
>>> Many non-ISP organizations view hardware capital costs (i.e.
>>> non-recurring) as vastly preferable to monthly recurring charges,
>>> especially when the term length for the MRC is "forever", even if the
>>> MRC is low.  Whether or not this is a good idea is a separate
>>> question, but it's a well-known behavior of many in the enterprise
>>> world.
>> Sure. But hardware also comes with recurring charges. Electricity,
>> hosting, support contracts, sysadmin hours, and so forth.
>>> I do think the attitude presented by many of the folks who encourage
>>> IPv6 adoption is not conducive to getting the enterprises to adopt v6
>>> - we as a community come across as dismissive of the level of effort
>>> required for the enterprises to learn a whole new way of doing things
>>> for benefits which are neither obvious not quantifiable in the
>>> near-term.  Modifying our tone and presentation, and spending more
>>> time listening to the concerns of the enterprise IT folks, would be
>>> worth it.
>> I like facts. If someone has considered the facts, and concluded that
>> NPTv6 is the best approach for them, then great, that's what they should
>> be using. Same thing goes for PI or PA or anything else.
>> My domain name notwithstanding, I don't like FUD. FUD may lead folks to
>> make the wrong decisions, and that's certainly not, to borrow your
>> words, conducive to getting the enterprises to adopt v6.
>> So. Some facts:
>> - NPTv6 isn't free. It might be cheap, or expensive. It Depends.
>> - PI isn't free. In the RIPE region, it's cheap.
>> - NPTv6 has adverse effects on certain applications.
>> - PI doesn't cause any application breakage.
>> - NPTv6 doesn't cause DFZ growth.
>> - PI causes DFZ growth.
>> - PI doesn't require the end user to run BGP.
>> - PI and NPTv6 aren't even mutually exclusive.
>> Best regards,
>> --
>> Tore Anderson

More information about the ipv6-ops mailing list