[ppml] Policy Proposal 2006-7: Changes to IPv6 initial allocation criteria - revised text

[Owen DeLong]

On Feb 20, 2007, at 1:52 PM, Stephen Sprunk wrote:

> Thus spake "JORDI PALET MARTINEZ" <jordi.palet at consulintel.es>
>> Basically my intend with this proposal is to support those  
>> organizations
>> that aren't "knoww ISPs" (as required by section 6.5.1.1.d). This  
>> can fall
>> into two categories (but may be there are others that I don't see  
>> right
>> now):
>> 1) An organization which want to provide IPv4 and IPv6 services.
>>
>> 2) An organization which only want to provide IPv6 services.
>>
>> In both cases, seems to me unnecessary to ask for a slow start  
>> with *only*
>> IPv4 to be "known" or have a plan for 200/48 (there may be cases  
>> which a
>> smaller number of customers and there is no need to force that
>> organization
>> to use a single upstream and get allocated the address space from  
>> that
>> upstream, being forced to stay with that upstream on order to avoid
>> renumbering of its network and customer ones).
>
> It's hardly "slow start" when we're giving every LIR a /32 to start  
> with.
> That's a lot of addresses.
>
> I also don't think that mere intent (even if documented) is sufficient
> grounds to give an org a globally-routable prefix.  If you wish to  
> debate
> the 200 customer requirement, I'm sure many folks would be open to  
> lowering
> it, but I cannot in any way support a policy change that allows  
> anyone to
> get a /32 just because they buy a couple routers and sign up one or  
> two
> customers.  To do so completely undermines the entire the  
> allocation policy;
> we might as well stop asking for justification entirely and just  
> hand out
> /32s to everyone who fills out the paperwork.
>
This is another case of trying to have our cake and eat it too.

ARIN doesn't control or have anything to do with routability or global
routability of prefixes.

ARIN only provides uniqueness and registration.

So, giving someone a globally unique prefix does not necessarily mean
the same thing as a globally routable prefix.  Routability is  
determined on
a case-by-case basis by each AS to which the prefix is advertised or
attempted to be advertised.  It's not ARIN's business or responsibility
to get involved in routing decisions, and, the more these types of
discussions continue, the more I become convinced that we really need
to stop tying address policy decisions to the current set of routing
assumptions.

Jordi pointed out a number of scenarios in which globally unique  
addresses
are very desirable, yet, will not be globally routed.  There's no  
reason we should
avoid doing this in IPv6 where numbers are not scarce.  In IPv4, we  
had a
situation where there simplly weren't enough numbers to address the
publicly connected needs of the internet in the long run, so, there had
to be a priority give to making those unique.  As near as I can tell,  
the ONLY
problem IPv6 still solves out of the entire list of problems it was  
originally
intended to solve IS the not enough numbers problem.  As such, it does
not make sense to me to act as if that problem is not solved in IPv6.

> I also question the commercial viability of an ISP that does not  
> offer IPv4
> services and plans to subsist for 5+ years on less than 200  
> customers who
> are so small they do not qualify for PI assignments.  Asking the  
> ISP to make
> do with PA space (for the year or two it takes to deplete their VC  
> money) is
> not unreasonable.  And, if they do somehow survive, they will  
> qualify as
> "known" and be able to get an LIR allocation anyways.
>
Again, I don't see how that is relevant to ARIN policy.  If the ISP  
is not commercially
viable, then, they will stop renewing their ARIN resources and the  
addresses
can be reclaimed (or, they will be absorbed by the acquiring  
entity).  Either
way, I don't see how ARIN should be passing judgment on peoples business
plans.

Owen