[ppml] IPv6 Assignment Guidelines, Straw Man #2

[David Schwartz]

> In a message written on Mon, Aug 20, 2007 at 12:09:54PM -0700,
> David Schwartz wrote:
> > 1) It is largely senseless to have policies for small
> > allocations from an
> > ISP to their single-homed end users. There is simply no reason
> > to care. I
> > think we pretty much all already agree that this is so at the /48 level.

> So, when Randy Bush gets his /32 from ARIN, and chooses to give two
> /33's to residental DSL customers, and the next day asks for more
> space that's ok?

No. That's my point. What we don't want is people coming back to the
registries too quickly. That's the cost we need to minimize.

> Since we give most residential users a /32 in IPv4, and we're running
> out, I would posit letting someone give out /33's to residential
> users in IPv6 will lead to the exhaustion of the IPv6 address space
> in rather short order.

Right, that's why I am saying the complete sub-assignment freedom model is
not reasonable. That's the argument I was responding to and that's the part
of it I disagree with.

> Which is why, in version 2 I put the same stake in the ground as
> RFC 3177.  /48-/128, we don't care, do what you want.  /1-/48, we
> care, there are rules.  Is that not the right break point?

That's not the only question. We also need to decide what the time interval
for forecast need is going to be for assignments to multi-homed end users
and LIRs. Is a decade unreasonable?

As for whether /48 is the right choice or not, I honestly do not know. I
certainly do not see any reason to increase it. How many end users on the
planet could justify a /47?

If your point was to convince me that I already have what I say I want, I
think you're basically right.

DS