[arin-ppml] ARIN Multiple Discrete Networks Policy

[Richard A Steenbergen]

Hello,

I've recently found myself at odds with ARIN's interpretation of the 
Multiple Discrete Networks policy, and thought it might be useful to 
bounce to issue off of the community to see what they think.

https://www.arin.net/policy/proposals/2001_6.html

As a quick background, this policy is used by entities who operate 
multiple discrete networks (for any of a variety of reasons, many of 
which are cited in the policy), and who wish to have their allocations 
and utilization calculated separately for each discrete network. An 
example of where this makes sense is when one discrete network is 
growing much faster than another, and the organization wishes to receive 
new space for use by only that discrete network, without having to "play 
games" like creating multiple maintainer IDs to receive such treatment.

The problem is that ARIN doesn't believe "promoting routing aggregation 
on the Internet" is a stated goal of this policy, and thus anyone who 
admits "well yes technically I could always just break up my single ARIN 
allocation into a bunch of individual /24s for use by each discrete 
network" does not "NEED" the policy, and therefore does not qualify, 
even if every other stated justification and requirement of the policy 
is met.

In my mind, the ENTIRE PURPOSE of letting a single entity apply for 
space with each discrete network standing on its own merits is really 
for aggregation purposes. Otherwise, almost any network would be able to 
take their single ARIN allocation and break it up into a bunch of 
smaller deaggregates for use by each discrete network, thus being able 
to achieve the normal overall utilization goals. At the end of the day, 
I can't think of anything else that is actually being accomplished by 
this policy EXCEPT maintaining reasonable aggregation.

To me, ARINs policy interpretation seems not only wrong given the stated 
goals of the existing policy, but intentionally harmful to the shared 
global resources of the Internet. Increased deaggregation carries 
significant costs in terms of routing hardware upgrades necessary to 
support larger FIB sizes and increased memory and CPU loads, as well as 
slower convergence times, especially within large networks who must 
carry many views/copies of the routing table (think route reflectors, 
l3vpn providers, etc). Wasteful announcements impact every BGP speaking 
router in the global DFZ, and while ARIN's policy has traditionally (and 
sensibly) been "routing agnostic", I don't believe intentionally making 
the problem worse for no good reason is the correct solution.

I'd very much like to see something done about this, either through the 
community voicing its opinion to ARIN about their interpretation of the 
existing policy, or through a policy proposal which amends this to make 
maintaining reasonable route aggregation a stated goal of the MDN 
policy.

Thoughts?

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)