[ppml] Policy Proposal 2003-15: IPv4 Allocation Policy for the Africa Portion of the ARIN Region

[jlewis at lewis.org]

On Wed, 24 Sep 2003, Owen DeLong wrote:

> > In principle I support 2002-3, but..
> >
> > * It may have a much bigger impact on the size of the global routing table
> 
> Likely it will not, since, many of the effected providers are announcing
> more specific holes of PA space anyway.

And it may actually have the opposite effect people seem to be worried 
about.  Take a client of mine as an example.  He currently has multiple 
T1's to 3 "Tier 1's".  I'll just call them A, B, and C.  He's had his C 
connection the longest, so he's using one of C's /24's for his public IPs.  
This is announced via BGP to A, B, and C.  

Things have changed with C, and he knows that connection will be
terminated at some point in the not too distant future.  In fact it
probably would have been already if not for the need to renumber...so he's
effectively keeping that connection for the IP space he's using.

Not knowing which of A or B will be more stable, he's requested and
received /24's from each of them.  So, for the time being, he's announcing
3 PA /24's.  If he could qualify for a micro-allocation from ARIN, those 3
routes would be replaced with 1.

I don't know how many other networks are in similar situations, but I bet
the number is >0.

It might be an interesting exercise to look at a current show ip bgp
snapshot and see how many AS's announce multiple small PA blocks,
especially from multiple P's.  What if all those multiple blocks could be
traded in for single ARIN blocks?  Wouldn't that actually be good for the 
routing table...and for those AS's?

----------------------------------------------------------------------
 Jon Lewis *jlewis at lewis.org*|  I route
 Senior Network Engineer     |  therefore you are
 Atlantic Net                |  
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________