[arin-ppml] Draft Policy 2010-2: /24 End User Minimum Assignment Unit - Correct Title

[William Herrin]

On Fri, Jan 29, 2010 at 7:46 PM, Jim McBurnett <jim at tgasolutions.com> wrote:
> 1. If an end user gets a /24 IP address space from an ISP via the /24 MH rule--
>        Why such a big difference in PA from ARIN?
>        BUT does the ISP check usage?

Hi Jim,

As a thought experiment: if you buy a $20 Verizon residential DSL
account, and ask Verizon to give you a /24 because you also have a Cox
cable modem, what sort of response do you expect? Bear in mind that a
/24 for multihoming means they have to reprogram their border routers
not to reject packets originating from that /24.

If you ask ARIN for a /24 justified because you have Internet
contracts with Verizon and Cox Business Services, what sort of
response do you expect?

That's one reason why letting the ISP decide whether or not to give
you a /24 suppresses abuse.



> 2. If I am going to get a /24 to multi-home for an end user, they will use at least 64.
>         BUT do we have a formal definition of usage?  See below.
> 4. Usage--
>        I have never seen a clear definition of usage.

Usage is host count, network structure overhead and anything else you
can convince the well educated application evaluator at ARIN to accept
as legitimate. It's intentionally informal so that it doesn't
accidentally exclude reasonable uses.


> 3. Fraud--
>        Please correct me if I am wrong here.. BUT isn't
> there a requirement now for a C Level or VP To put their
> John Hancock on the request?  For Fraud to be a reason
> to vote this down is wrong. IMHO. Attempts to make rules
> to account for fraud can never include all the variables to
> be worth the time.

Process issues can encourage or discourage fraud. I personally think
this proposal is neutral on the issue of fraud, but I can't agree that
its inappropriate to consider a policy's potential impact on fraud.



> 5. De-aggregation--  If I am an ISP today and I have a /16.
>And I have to de-aggregate a block of say /19 to give a /24 for an end user..
>        How much growth do I add to the global route table ?

If he's multihomed, you announces two routes in the routing table. You
announce the original /19 and you also propagate the customer's /24
announcement "cut out" from the /19.

In the case of a /19 for most of your customers and an ARIN /24 for
the one multihomed guy, you also announce two routes: Your /19 and you
propagate the customer's /24 announcement.

So it's a wash. Either way the same number of routes are carried in
the DFZ. In neither case is the /24 aggregable with the /19 at any
point in the DFZ although either one can be aggregated with a default
route once you reach the DFZ border.

However, the renumbering requirement in the proposal adds a kicker.
You don't have to renumber out of an ISP /24 to add a second. So, when
you have two ISP /24's, you have to announce two /24 cutouts of the
ISP's /19. When it's time for a second ARIN /24 you're required to
renumber into a /23. This /23 can be announced as just one route in
addition to the ISP's /19.

So at least in theory this proposal -reduces- the routing table load
due to small multihomed organizations.

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