[arin-ppml] Against 2013-4

[William Herrin]

On Tue, Jun 4, 2013 at 12:17 PM, Jason Schiller <jschiller at google.com> wrote:
> I do value your input and would like to hear specifically
> which concepts that are popping up that have died.

Hi Jason,

The obvious one is heirarchy. There's just one more nail to put in
that coffin: move NRPM 4.2.3.6 over to 4.3.2.2. We still talk about
aggregation a lot but nobody pretends that insisting end users always
get their IPs from their ISP is still credible. Quite the contrary,
the industry is steadily shifting towards resilient, redundant,
multi-homed service.


Others on my hit list include:

Aggregation. Aggregation is an artifact of BGP. Critical while BGP
reigns, it should die quickly when a more scalable protocol rises.
IMO, principles should not die quickly as a result of foreseeable
technology evolution.


Conservation. Conservation where? We maintain three number resources.

Conservation for IPv4 is done. They're gone now. Our concern is making
sure that high value uses aren't starved for addresses while low value
uses hoard them. That means redistribution; conservation has nothing
to do with it.

Conservation for IPv6 is not yet a going concern. The protocol has to
be deployed and if throwing wastefully large blocks of addresses at it
helps, that's exactly what we should do. To an extent, that's what we
*are* doing.

Conservation for AS numbers is not credible. With the change in the
protocol to 32-bit AS numbers and the unrestrained demand curve,
they'll last longer than recorded human history. By which time we can,
if we're still using BGP, make another relatively trivial change to
the protocol.

So just what the heck is this conservation that you speak of as a core
principle of an Internet registry? It's not a core principle! It's
circumstantial and we happen to be in a lull where the circumstance
fits *none* of the number resources we're charged with maintaining.


Justified need. Justified need yields absurd results.

I work a $10M/yr project inside a multi-billion dollar defense
contractor. The project maintains three operational sites 5,000 miles
apart with a couple hundred clustered machines sharing IP addresses
using BGP. Because of the labyrinth created by justified need, I
couldn't just say, "Hey, here's the set up. Sell me IPs." As the
project lead, I can't talk to ARIN at all. I'd have to fight through
dysfunctional corporate heirarchy to even talk to ARIN and then ARIN
couldn't provide the obviously justified addresses to my project
without first examining resource use by the entire multi-billion
dollar company.

So every year my project spends around $4000 of taxpayer money on a
contract to maintain a separate corporation whose sole output is the
payment of a $100 ARIN end-user fee. Because that's the sanest and
*least expensive* path to getting the analysis of justified need to
work out. Ridiculous!

Is a pure market better? That's not yet proven. But for sure it
wouldn't have yielded the absurd result above.



Really, the only thing in draft 2013-4 that I don't immediately
disagree with is uniqueness. Even there I can't help but think of the
work we did here with the CGN /10. Granted it was appropriate to
handle part of the discussion at the IETF, but I'd hate to have seen
the discussion cut off at the knees because of our principle of
uniqueness.

Regards,
Bill Herrin





-- 
William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004