[arin-ppml] A modest proposal for IPv6 address allocations
joelja at bogus.com
Fri Jun 5 02:46:12 EDT 2009
James Hess wrote:
> IPv6 is abundant, but it's not abundant enough that the same scarcity
> problem can't accidentally be created as with V4.
Good stewardship is why we're on this mailing list. We know that if you
give out a longer prefix than the number of subnets that someone will in
fact need over the sourse of their use of that prefix, that they have to
come back for more. That is the world we live in today. We pay a
signficant price for it in the size of both our interior, and dfz
routing tables, and in the onerousness of the allocation regime that we
live under. The same scarcity problem as v4 is rather hard to create, it
hard after-all to create a circumstance where you have to renumber out
of subnet because you ran out of addresses, all the other reasons to do
so still apply but 10^19 is frankly enough for now.
The IPNG working group saw clearly that while cidr would alleviate the
immediate pressure that there are costs associated with small (and
fragmented allocations) that are both local (I pay them when I manage
space) and collective (we all pay them due to fragementation).
> It's a salient and justified fear, given the experience with IPv4.
> Consider what happens if the expoential internet rate of growth over
> the next 30 years expands in the manner it has in the past 30 years.
> It's conceivable there could be a quintillion internet hosts before 2020.
10^18 is a lot of hosts . sticking to orders of magnitude, that's 10^8
per person in 10 years...
> And with V6, many of those would consume an end site. If 1% of those
> quintillion is a logically end site that gets a /56, that's
> 13% of the IP space.
1% of 10^18 is 10^16 so hey only a million end sites per person no big deal.
> Also, a forgeone conclusion is major new uses of IP addresses; it's
> the reason end sites get a /56, so they can have multiple subnets.
If you consider nat translation the semantic equivalent if subdividing
your assigned v4 /32 and imposing a layer-3 boundry then in fact they
all subnet today so you don't exactly have to looking very far for an
application for subnetting.
> It's conceivable V6 users could want more, some may even be giving
> end sites a /48 instead of a /56 or /64.
> There must be some specialized app that really burns IPs for this
> Well, there aren't enough IPs in the IPv6 address space to assign
> 1% of those 10000000000000000 end sites each a /48, so you have
> exhaustion in that case.
since your premise is flawed your conclusion is kind of hard to refute.
> You are receiving this message because you are subscribed to
> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> Please contact info at arin.net if you experience any issues.
More information about the ARIN-PPML