[address-policy-wg] 2009-03 New Draft Document Published (Run Out Fairly)

On 25/09/2009 13:42, Alex Le Heux wrote:
> 2008-10 3266560 /11,/12,/16,/17,/18,/20,/21
> 2008-11 3047424 /11,/13,/14,/15,/17
> 2008-12 3274752 /11,/12,/16,/17,/18,/19,/20,/21
> 2009-01 6502400 /10,/11,/15,/16,/19,/20,/21
> 2009-02 4182016 /11,/12,/13,/14,/15,/16,/17,/18,/20
> 2009-03 2643968 /11,/13,/18,/20,/21
> 2009-04 2478080 /11,/14,/16,/17,/18,/20
> 2009-05 3629056 /11,/12,/14,/15,/16,/18,/19
> 2009-06 4540416 /10,/14,/16,/18,/21
> 2009-07 2844672 /11,/13,/15,/16,/18,/19,/21
> 2009-08 7272448 /10,/11,/13,/14,/15,/17,/18,/19,/20,/21
> 2009-09 1277952 /12,/15,/16,/17

taking this and http://www.potaroo.net/tools/ipv4/ into account, the run 
rate is between 2.5 and 3 /8s per annum; alternatively, a /8 will be 
gobbled up in about 4 months under normal circumstances.  Turning this 
around, the "what do we do with the last /8" discussions really revolve 
around what to do with 4 months worth of address space.  So, if people are 
proposing radical changes to the assignment policy, does it really make 
sense to do so for a mere 4 months of run-time.  Or probably a lot less, if 
address stock-piling and panic sets in?  I can't help seeing this as a 
knee-jerk reaction to the unknown.

The current date-based suggestion in 2009-03 currently works out as 
starting the "Run Out Fairly" proposal about a year before IANA depletion, 
and 2 years before RIR depletion (according to Geoff Huston's estimates), 
with the 3 month window kicking at around the IANA depletion date and about 
year before RIR depletion, or given RIPE NCC's burn rate and proposal 
2008-03, between 4 and 13 months before RIPE depletion, depending on how 
things pan out.  I.e. if the RIPE NCC were to receive 2 x /8 immediately 
before IANA exhaustion, then there would be about 4.2*3 = 13 months space 
available.  At the other end of the scale, if the RIPE NCC were almost 
exhausted at the time that IANA gets to its last 5 /8s, then there would be 
only a little more than 4 months address space available.

Limiting allocations / assignments to 3 months worth of address space would 
be fine if we knew that this policy were to last around 12 months. 
However, if it were only to operate on 4 months worth of space, that would 
be a bad thing, because it would introduce new guidelines at a time of 
market consternation / panic.  I'd like to propose that last minute tweaks 
are objectively bad - and 4 months is very last-minute.

Please excuse the thinking-out-loud thing going on here. It just seems that 
we need to model RIPE's run-out process carefully so that we have a good 
idea of when IANA will assign the RIPE NCC its last /8s, and what policies 
are in place to deal with them (whether 2009-03 or "last /8" type).  Has 
any modelling of this form been done within the NCC?

Using naive projection, I'd suggest that the RIPE NCC may be in line to get 
2x/8 in 2010-05, 2011-01 and possibly 2011-09 before it receives the last 
/8.  Going by this finger-in-the-air estimate and Geoff Huston's 
prognostications, the RIPE NCC will either be at the end of its previous 
2x/8 or will just be starting off on a new 2x/8 when it receives the last 
/8 disbursement.  This would suggest that there is a significant risk that 
2009-03 will end up running its 3 month requirement window period at a time 
when the RIPE NCC only has enough space for 4 months worth of assignments. 
   I find that alarming.

Nick