[184821] in North American Network Operators' Group
Re: IPv6 Irony.
daemon@ATHENA.MIT.EDU (Masataka Ohta)
Thu Oct 22 02:42:50 2015
X-Original-To: nanog@nanog.org
To: Mark Andrews <marka@isc.org>
From: Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>
Date: Thu, 22 Oct 2015 15:42:37 +0900
In-Reply-To: <20151020220820.A18133AD2B3D@rock.dv.isc.org>
Cc: nanog@nanog.org
Errors-To: nanog-bounces@nanog.org
Mark Andrews wrote:
>>> Customer support, especially network troubleshootings and so on...
>>
>> Customer support for IPv6 costs a lot, at least because of:
>>
>> 1) Unnecessarily lengthy IP addresses, not recognized by most, if not
>> all, customers
>>
>> 2) Lack of so promised automatic renumbering
>
> Upgrade the vendors. Nodes already renumber themselves automatically
> when a new prefix appears.
Can the nodes treat multiple prefixes on multiple (virtual) interfaces
for smooth ISP handover?
> Nodes can update their addresses in the DNS if the want to securely
> using DNS UPDATE and TSIG / SIG(0).
How much is the customer support cost for the service?
> This isn't rocket science. Firewall vendors could supply tools to
> allow nodes to update their addresses in the firewall. They could
> even co-ordinate through a standards body. It isn't that hard to
> take names, turn them into addresses and push out new firewall rules
> on demand as address associated with those names change.
As I and my colleague developed protocol suites to automatically
renumber multihomed hosts and routers
The Basic Procedures of Hierarchical Automatic Locator
Number Allocation Protocol HANA
http://delivery.acm.org/10.1145/2090000/2089037/p124-kenji.pdf?ip=131.112.32.134&id=2089037&acc=ACTIVE%20SERVICE&key=D2341B890AD12BFE.E857D5F645C75AE5.4D4702B0C3E38B35.4D4702B0C3E38B35&CFID=723424660&CFTOKEN=36506659&__acm__=1445495785_e3533480d8843be13ab34593a1faf194
which is now extended for DNS update including glue, I know it is
doable.
But, as it is a lot more simpler to do so with IPv4 with
NAT, 48 bit address space by NAT is large enough and NAT can
enjoy end to end transparency, I see no point to use IPv6 here.
Automatic renumbering of IPv6 *WAS* promising, because it was
not necessary to replace existing IPv4-only boxes.
Masataka Ohta
