[policy-wg] Policy proposal: IPv4 allocation to End Users

[Bill Woodcock]

      On Mon, 25 Apr 2005, Gregory Massel wrote:
    > /24 ?
    > This seems to be a bit out of line with the other RIRs and given the
    > current exponential growth of the global routing table, may need
    > reconsideration.

This is an issue we've been discussing a bit within ARIN recently, with 
regard to IPv6, but it's the same issue at heart...  

There are a bunch of independently-routable /24s out there, and they were 
allocated or assigned early on, as "Class C" "swamp" addresses.  I mean, 
they were called "Class C" at the time, and we later designated the 
portion of the IP spectrum which they came out of as "the swamp" because 
it was made up of lots of little blocks.

Now, there's something of an inequity in that those /24s are typically 
held by people who arrived early, who weren't in Africa by and large, or 
those who have enough connections or money to make a transaction to get 
one from someone who got one that way.  Again, probably not someone in 
Africa.

So on the one hand, we have a "do as I say, not as I do" inequity being 
handed from the early-arriving North American and European operators to 
the later-arriving African and Latin American operators.  On the other 
hand, we do have a very real need to conserve memory and CPU in routers, 
and that need is in fact greater in areas like Africa and Latin America, 
where operators can't afford to upgrade equipment as frequently or spend 
as much when they to, to accommodate routing table growth.

The approach we've been talking about in ARIN is to find a reasonable 
"cap" number.  That is, to begin with a liberal policy (like /24 minimum 
allocation) but only up to a certain number of allocations (like 250 or 
500 of them) which is not enough, even if replicated in other regions, to 
cause routers to fall over in great numbers.  This would have the effect 
of creating another, African-specific, "swamp" of /24s.  They would be 
filtered out of the routing table by some people, which would make them 
less valuable than larger allocations, but they might meet some people's 
needs, and they wouldn't become a big problem.  Moreover, by being liberal 
up-front, you might find that you address a strong need of a small number 
of organizations, without creating any difficulty for anyone else.  For 
instance, you might allocate a block of 512 /24s, and get a "run" on them 
initially, allocating 100 in the first few months, and then five years 
later, you might find that you'd only allocated another 100...  No 
problem, it takes care of those who need it, without creating any big 
issue for anyone else.

One approach to think about.

                                -Bill