[ppml] Policy Proposal: IPv4 Soft Landing

[Stephen Sprunk]

Thus spake "David Conrad" <drc at virtualized.org>
> On May 15, 2007, at 1:19 PM, Stephen Sprunk wrote:
>> IANA could release some historically reserved space such as
>> 0/8 and 1/8 to the RIRs,
>
> I am skeptical we'd be able to "unreserve" 0/8, 127/8, and the
> RFC 1918 space.  All other reserved addresses (including 1/8 and 14/8 
> which the IANA Number Liaison, Leo Vegoda, has
> been doing a fantastic  job tracking registrants down) should
> consider their days of wine, song, and kicking back to watch
> the world go by near its end.

1/8 is not usable for human reasons.  10/8 and 192.168/16 are not 
recoverable because that'd give people nothing left to NAT with, and NAT is 
about to become even more prevalent than it already is as people desperately 
try to avoid migrating to v6.  172.16/12 could probably be reduced to 
172.16/16 without much grief, but pulling 172.16/16 would be tough (though 
not as bad as 192.168/16).

>> and there's a side discussion about opening up 240/4,
>> however IANA expects the IETF to publish an RFC on such
>> things first,
>
> Yep.  Some folks are working on a draft to do just this.

255/8 is tough to unreserve in practice for the same reasons as 0/8 and 
127/8, so it's probably not 240/4 but rather 240-254/8.  If we're going to 
do that, we might as well switch 225-238/8 to unicast too -- only 224/8 and 
239/8 see much use in the (miniscule) multicast world, and the exceedingly 
few oddballs using 225-238/8 could migrate to the remaining two /8s with 
minimal hassle.

So, we could theoretically open up 29 new /8s just by writing an RFC.  In 
reality, that's what, a couple more years of v4 space in exchange for 
billions of dollars in upgrades and time?  That's on par with the cost of 
just doing v6 now -- and we'll still have to pay that v6 cost later anyways.

(Also, the idea of opening those /8s up only for private use is IMHO daft 
and merits RFC publication on 1 Apr.  Anyone who is looking to deploy a new 
private network that'll be incompatible with the rest of the world would use 
v6 and ULAs -- and that solution already exists today without any need for 
new RFCs or product upgrades.)

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov