[arin-ppml] Rationale for /22

[David Farmer]

On 30 Jul 2009 William Herrin wrote:

> On Thu, Jul 30, 2009 at 6:46 PM, Leo Bicknell<bicknell at ufp.org> wrote:
> > In a message written on Thu, Jul 30, 2009 at 05:48:06PM -0400, William Herrin wrote:
> >> If I may draw it back to the question I started with: Can you offer a
> >> well grounded reason to believe that changing the minimum allocation
> >> size for *multihomed* systems is likely to affect the size of the
> >> global table?
> >
> > I believe it is reasonable to assume each decrease in prefix size
> > allows more orgs to qualify, and thus increases the routing table.
> 
> Hi Leo,
> 
> How do you figure that "more ARIN qualifying orgs = increased routing
> table" is a reasonable assumption?
> 
> Let me rephrase the question by expanding your assumption a little bit:
> 
> 1. Multivendor multihomed user connects to the Internet.
> 2. User announces his ISP /24 via BGP.
> 3. User qualifies for an ARIN /24
> 4. ...
> 5. ...
> 6. User announces more than just one /24.
> 
> What are steps 4 and 5 that get you from qualifying for an ARIN /24 to
> consuming extra BGP slots?

You are assuming that the number of multi-homed users will remain a 
constant, if you allow a multi-home user with a PA based /24 from a provider 
to switch to a PI based /24 from ARIN.  It is entirely possible that the 
availability of PI /24s from ARIN for multi-homing could create an inducement 
for more users to multi-home, just so they can get PI space.  

Personally, I'm not sure this would actually be a bad thing.  But, the 
inducement of PI space is the kind of thing that could cause Leo's tipping.  
I'm not sure that a PI /24s are enough inducement to cause a flood of users 
to want to become multi-homed.  But, I would discount it out of hand either, 
and with an AC hat on it is enough to make me worry at least a little bit.

None of us can prove it will be a problem, but none of us can prove it won't 
be a problem either.

That said, there are ways we could mitigate the possibility of a flood, we 
could require someone who wants to get a PI /24 for multi-homing to have 
multi-homed for a year using PA space.  This wouldn't totally prevent 
someone from multi-homing solely to get PI space, but it would prevent 
instant gratification for those that would try it, and maybe make them think 
twice about it.

As a general concept I support making PI space available to end users and 
smaller providers too, but we can break the Internet doing it.




===============================================
David Farmer                                      Email:farmer at umn.edu
Office of Information Technology
Networking & Telecomunication Services
University of Minnesota		       Phone: 612-626-0815
2218 University Ave SE		       Cell: 612-812-9952
Minneapolis, MN 55414-3029	       FAX: 612-626-1818
===============================================