I would still like to encourage adoption of this proposal, and I thank
the many people who have expressed their support for it so far.
- APNIC policy still formally requires demonstration of multihoming in
order to be allocated a portable or provider independent IPv6 allocation;
- This means that currently a company with a singly-homed and
statically-configured configured network would need to number their
network from their ISP's allocation, and changing ISP would require
renumbering the entire network. This could deter them from implementing
IPv6 in their enterprise. Removing APNIC's multihoming requirement would
remove this deterrent by allowing them access to portable address
allocations.
- APNIC is the only RIR remaining that requires demonstration of
multihoming in order to allocate portable IPv6 addresses.
- It is therefore proposed that multihoming be removed as a requirement
for portable IPv6 allocations, and be replaced instead with requirements
aimed to limit allocations to organisations truly requiring portable
addressing for reasons of difficulty of renumbering or criticality of
network services.
The only particular concern that has been raised about this draft, and I
believe only explicitly by one member of this list, is that the removal
of the multihoming requirement could lead to a potential explosion of
the IPv6 global routing table because of a large number of requests for
portable address allocations, especially as the the draft only requires
"reasonable technical justifications" rather than mandatory and
explicitly detailed criteria (although two examples of "reasonable
justifications" are included, based on complexity of renumbering or
importance of civic services).
The draft itself does provide some reasons suggesting why the number of
additional allocations arising should only have a relatively small or
manageable impact on the global routing table, and some provisions are
proposed in the draft for progressive reporting of the number of
portable allocations to provide timely warning to the community should
the resulting number of allocations prove more significant than expected.
I would welcome any further comment on this draft before the Phnom Penh
meeting in a few weeks, especially if there are any remaining concerns
about the draft, as well as any additional indications of support or
opposition.
Can we remind you all that a revised version of this proposal,
available at http://www.apnic.net/policy/proposals/prop-101, is up for
consideration at our next meeting. Of course, you should consider the
whole proposal before deciding to support or oppose this proposal.
The change from the previous version is the *removal* of the following clause:
(e) The first Policy SIG meeting of 2014 (expected to be APNIC
Meeting 35) will as an agenda item consider the observed rate of IPv6
portable assignments and potential 10-year forecasts of growth of
portable assignments prepared by the APNIC Secretariat extrapolated on
the observed data, and by consensus consider the question "Should the
IPv6 portable assignment criteria revert to requiring multihoming?"
--
This clause was a major point of debate in the debate at the last
policy SIG meeting and while some people now are happy to support the
proposal there has been no other discussion. We encourage you express
your views on the list so that those who won't be able to attend the
meeting in person have a chance to take part in the discussions.
Andy, Masato, Skeeve
prop-101-v004: Removing multihoming requirement for IPv6 portable
assignments
- Introduction
This a proposal to change the "IPv6 address allocation and assignment
policy" to allow portable (that is, provider independent or PI)
assignments of IPv6 address blocks to be made by APNIC to any
organization with due justification and payment of standard fees,
removing the current requirement that the requestor is or plans to be
multihomed.
- Summary of the current problem
Current APNIC policy only permits portable assignments of IPv6
addresses to be made to an organization "if it is currently multihomed
or plans to be multihomed within three months." [1] This requirement may
unnecessarily complicate the implementation of IPv6 in some networks
that are large or complex and use static assignment of addresses. It is
therefore proposed to remove this requirement.
IPv6 models tend to assume widespread assignment of registered IPv6
addresses to equipment throughout a network; so if provider assigned
IPv6 addresses have been used in an organization's network, then any
change of ISP would require a renumbering of the entire network. Such
renumbering may be feasible if the network is small or dynamically
assigned (for example, through use of prefix-delegation), but
renumbering a large, statically-assigned network would be a significant
operational challenge, and may not be practically possible.
Although it is likely that many large networks would be multihomed,
there will be technical or commercial reasons why some will not be;
currently those networks cannot obtain portable IPv6 assignments from
APNIC, and would need to use assignments from their ISPs, and accept the
associated difficulties of future renumbering if they do so. This
consideration and complexity could significantly delay IPv6 use by the
affected organisations, which is not desirable.
There is a risk that removing the multihoming requirement could cause
a significant increase in demand for portable assignments, which in turn
could cause the Internet routing tables to grow beyond manageable
levels. It is not feasible to quickly generate any realistic model of
likely demand increase which would arise from the proposed policy
change, but it is argued that any such increase would only be of a scale
to produce a manageable impact on global routing, for reasons including:
- Organizations would only be likely to seek portable addressing if
they believed it were essential for their operations, as provider
assigned IPv6 addressing would be likely to be offered
automatically and at no additional cost with their Internet
services from their ISP;
- APNIC membership fees would be expected to naturally discourage
unnecessary requests, as these would be a far greater cost than
that for provider assigned addressing;
- Many or most organizations that require portable addressing will
be multihomed, so the demand increase caused by removing the
multihomed requirement should be small;
- Only a limited set of an ISP's products is likely to allow
customers to use portable assignments if they are singly-homed.
3.Situation in other RIRs
APNIC is now the only RIR remaining with an absolute requirement for
multihoming for portable address assignments.
AfriNIC: The "Policy for IPv6 ProviderIndependent (PI) Assignment for
End-Sites" [2] does not mention any requirement for multihoming;
ARIN: Section 6.5.8 of the "ARIN Number Resource Policy Manual" [3]
only identifies multihoming as one of several alternative criteria for
direct IPv6 assignment to end-user organizations;
LACNIC: There is no mention of multihoming anywhere in the IPv6
section (Section 4) of the current LACNIC Policy Manual (v1.8 -
07/12/2011) [4].
RIPE: The latest version (RIPE-545 [5]) published in January 2012 of
the "IPv6 Address Allocation and Assignment Policy" does not mention
multihoming, removing the requirement that existed in previous versions
of the document.
4.Details
It is proposed that section 5.9.1 of APNIC's "IPv6 address allocation
and assignment policy" (apnic-089-v010) is rewritten to remove the
absolute multihoming requirement for portable assignments, and to
incorporate the following conditions:
A. Portable IPv6 assignments are to be made only to organizations
that have either joined APNIC as members or have signed the
non-member agreement, under the standard terms & conditions and
paying the standard fees applicable for their respective category.
B. An organization will be automatically eligible for a minimum IPv6
portable assignment if they have previously justified an IPv4
portable assignment from APNIC.
C. Requests by organizations that have not previously received an
IPv4 portable assignment will need to be accompanied by:
(a) a reasonable technical justification indicating why IPv6
addresses from an ISP or other LIR are unsuitable - examples of
suitable technical justifications may include (but are not
limited to):
(i) Demonstration that the relevant network is statically
addressed and of a size or complexity that would make IPv6
renumbering operationally impractical within an acceptable
business period, together with evidence that dynamic or
multiple addressing options are either not available from
the relevant ISP or are unsuitable for use by the
organization;
(ii) Demonstration that any future renumbering of the relevant
network could potentially interfere with services of a
critical medical or civic nature;
(b) A detailed plan of intended usage of the proposed address block
over at least the 12 months following allocation.
D. The minimum IPv6 portable assignment to any organization is to be
an address block of /48. A portable assignment of a larger block
(that is, a block with a prefix mask less than /48) may be made:
(a) If it is needed to ensure that the HD-ratio for the planned
network assignments from the block remains below the applied
HD-ratio threshold specified in Section 5.3.1 of the APNIC IPv6
policy [6], or;
(b) If addressing is required for 2 or more of the organization's
sites operating distinct and unconnected networks.
Any requests for address blocks larger than the minimum size will
need to be accompanied by a detailed plan of the intended usage of
the proposed assignment over at least the following 12 months.
E. In order to minimise routing table impacts:
(a) Only one IPv6 address block is to be given to an organization
upon an initial request for a portable assignment; subnets of
this block may be assigned by the organization to its different
sites if needed;
(b) It is recommended that the APNIC Secretariat applies sparse
allocation methodologies so that any subsequent requests from an
organization for additional portable addressing would be
accommodated where possible through a change of prefix mask of a
previous assignment (for example, 2001:db8:1000::/48 -> ]
2001:db8:1000::/44), rather than through allocation of a new
prefix. An additional prefix should only be allocated where it
is not possible to simply change the prefix mask.
(c) Any subsequent request for an additional portable assignment to
an organization must be accompanied by information
demonstrating:
(i) Why an additional portable assignment is required, and why
an assignment from from an ISP or other LIR cannot be used
for this purpose instead;
(ii) That the use of previous portable IPv6 allocations
generated the minimum possible number of global routing
announcements and the maximum aggregation of that block;
(iii) How the additional assignment would be managed to minimise
the growth of the global IPv6 routing table.
(d) The APNIC Secretariat will produce reports of the number of
portable IPv6 assignments requested, preferably as an
automatically-generated daily graph of the number of cumulative
IPv6 portable assignments published publically on the APNIC
website, or else as regular (at a minimum, quarterly) reports
sent to the sig-policy mailing list detailing the incremental
assignments of new IPv6 portable assignments made since the last
report, plus the cumulative total of IPv6 portable assignments.
5.Pros/Cons
Advantages:
- This proposal would provide access to portable IPv6 addresses
for all organizations with valid needs, removing a potential
impediment to industry standard IPv6 addressing for large
singly-homed networks
- This change would align APNIC with the policies of all other RIRs
on portable assignments
Disadvantages:
- There would be a risk of an unmanageably large increase in
global IPv6 routing table size and APNIC workload if there were to
be a substantial and widespread increase in demand for portable
assignments arising from the removal of the multihoming
requirement
- But demand is expected to be limited by the requirements specified
in section 4 for justifications and APNIC standard fees, as well
as other industry factors such as the capability of Internet
services to support portable addressing.
6.Effect on APNIC
The impact of this proposal on the APNIC Secretariat would depend on
the increase of demand for portable assignments. Even if demand is
eventually large, it is unlikely that there will be an significant
change in hostmaster workloads for a long time because of the slow rate
of take up of IPv6, and so there should be sufficient time to identify
and take steps to modify policies and processes if necessary to manage
the increase.
7.Effect on NIRs
This proposal specifically applies to portable assignments made by
APNIC. It would be the choice of each NIR as to whether they would adopt
a similar policy.
References:
[1] Section 5.9.1, IPv6 address allocation and assignment policy,
http://www.apnic.net/policy/ipv6-address-policy#5.9
[2] http://www.afrinic.net/docs/policies/AFPUB-2007-v6-001.htm
[3] https://www.arin.net/policy/nrpm.html#six58
[4] http://www.lacnic.net/en/politicas/manual5.html
[5] http://www.ripe.net/ripe/docs/ripe-545 [6]Section 5.3.1, IPv6
address allocation and assignment policy,
http://www.apnic.net/policy/ipv6-address-policy#5.3
_______________________________________________
sig-policy mailing list
sig-policy@lists.apnic.net
http://mailman.apnic.net/mailman/listinfo/sig-policy
