[arin-ppml] Initial ISP Allocation Policy

[Randy Carpenter]

----- Original Message -----
> On Mon, Jul 22, 2013 at 1:56 PM, Randy Carpenter <rcarpen at network1.net>
> wrote:
> > I strongly disagree with your disagreement :-) and fully support the
> > proposed changes.
> >
> > I have worked with several small rural ISPs who really only have one option
> > for
> > upstream connectivity. In most cases, the upstream refuses to give out any
> > more
> > IP addresses. Some of these ISPs would be fine with a /21, while others
> > would
> > need a /20, but cannot get a aggregate /20 from the upstream.
> 
> Hi Randy,
> 
> BS. You can add satellite coverage anywhere, and I do mean anywhere.

Satellite coverage? Has there been some crazy new development that let's us do low-latency GigE+ links over satellite?

> More, nearly all rural ISPs can contract a private point to point line
> to the nearest city with a carrier-neutral data center and pick up
> another ISP there.

Not for anything even close to reasonable cost. We're talking 100+ miles of fiber that would have to be newly installed.

> What you mean is that they can't get a second upstream at a price
> that's viable for their customer base.

Of course that is what I mean. If the companies I am talking about had billions of dollars, then there would not be an issue. But, they would also no longer be the companies I am talking about.

> And I'll bet the same thing is
> true with IP addresses -- wave money and documentation at the upstream
> and see how fast they find more IPs for you. 

If it were as simple as waving (a reasonable amount of) money, everything would be fine, and we would not be having this conversation.

> That's life in a
> post-IPv4 exhaustion world. If you can't afford it, use carrier NAT.

CGN is not exactly a good|affordable solution.

I guess I am confused a bit... If the remaining addresses in the ARIN pool shouldn't go to entities that most need them, then who are they for?

> At any rate, if Daniel ties the two proposals together, I'd bet he'll
> sink them both. Since the disucssion seems to focus on unifying
> registrant classes rather than reducing minimum allocations, I'd
> recommend he split the latter out.

I agree that separate proposals may be better.

If we are really worried about routing table size, then we should never have reduced the needs requirement to 3-months. I have seen many cases where people now have several smaller blocks, rather than once larger, contiguous block because of that policy. I really don't think there are enough entities like I am talking about to make a huge impact. Certainly not as much impact as getting the really big ISPs with stupid deaggregated announcements to fix their policies.

-Randy