[ppml] IPv6 Assignment Guidelines, Straw Man #2
david.kessens at nsn.com
Tue Aug 21 14:25:54 EDT 2007
On Mon, Aug 20, 2007 at 07:00:50PM -0400, Leo Bicknell wrote:
> In a message written on Mon, Aug 20, 2007 at 04:02:00PM -0500, Stephen Sprunk wrote:
> > ... and most ARIN members don't have that many customers. Those that do
> > mainly provide residential service and will be handing out /56's at most to
> > those customers, giving them just shy of 17M assignments before they need
> > more space. I'd expect those folks are smart enough to ask for larger
> > allocations up front to save hassle later.
> You mean, like these smart folks...
> inet6num: 2003::/19
> netname: DE-TELEKOM-20050113
> descr: Deutsche Telekom AG
> country: DE
> inet6num: 2a01:c000::/19
> netname: FR-TELECOM-20051230
> descr: France Telecom
> country: FR
> For the record, a /19 is 536,870,912 /48 assignments. There's 64
> million people in France, 83 million in Germany. All of those
> people could get 5 /48's FROM BOTH PROVIDERS (10 total) with that
> much space.
I don't see what the number of people in France or Germany has
anything to do with the size of their allocations. Both companies are
multinational operations. I am sure you are familiar with the concept
that allocations are based on the number of customers they have, not
on the number of people in a particular country.
For the record, I actually agree that /48's for residential customers
are a bit at the lavish side ...
Also, I am actually not convinced that these kind of superallocations
are actually supported by the current IPv6 policies. The policy is
quite detailed about /32 allocations but has few words on what to do
when there is the perception that somebody might need more. That was
done on purpose at the time as people felt that it was more important
to get a policy out the door so that people could finally receive IPv6
allocations and to allow the policy to get refined when more
experience is gained with LIRs that actually manage to sign up more
than 200 or so ipv6 customers.
I for one am quite surprised that it has been interpreted by at least
one RIR that one can get these very large allocations even without
first starting with a /32 and not even having 200 IPv6 customers
signed up and/or a running native ipv6 network that spans more than a
couple of links. On the other hand, some of these very large
enterprises tend to roll out services to all their customers at once
in a particular region and I can see that they might have a need to
grow beyond a /32 if they really get started. However, it seems that a
sparse allocation scheme could do the job in such cases. At the same
time, what is really the (routing system) cost if only the biggest
providers end up with a few prefixes ? In fact, they might actually
have to deaggretate their space anyways if they got one very big block.
More information about the ARIN-PPML