[arin-ppml] Encouraging IPv6 Transition (was: Clarify /29 assignment identification requirement)
owen at delong.com
Sat May 12 12:06:34 EDT 2012
>> - Actively promote the establishment and maintenance of 6to4 gateways by
>> all present IPv4 allocation holders above a sensibly arbitrary size,
> 6to4 gateways are not necessarily an advisable migration strategy.
> 4to4 / CGN has significant advantages.
6to4 gateways aren't so much a migration strategy as a way to connect IPv6 islands across an IPv4 ocean.
4to4/CGN offer an alleged solution to a completely different problem -- How to multiplex more IPv4 connections onto a single IPv4 address. (Also not a migration strategy, but instead a strategy for attempting to avoid migration.) Frankly, anyone who has looked at the details of CGN and truly considered the problems of CGN would prefer to do native IPv6 with CGN providing only the minimal amount of connectivity to unfortunate IPv4 only end-points and would be actively seeking to encourage popular IPv4 only end-points to add IPv6 capabilities as quickly as possible.
The alternatives which could be called a migration strategy are NAT64/DNS64 which provides a mechanism for an IPv6-only host to make an IPv6 connection to a proxy host which "NATs" that connection (it's really more like a proxy) to an IPv4 connection to the true destination and DS-LITE which tunnels IPv4 private addresses over the providers IPv6 network to a CGN NAT44 box.
All of these so-called solutions have tradeoffs and the only one which does not provide a degraded user experience is dual-stacked content providers during the transition period with the eventual goal of native IPv6 everywhere.
>> - Actually bother to pronounce an IPv4 deprecation date. Only some weak
> It would be inappropriate at this time for ARIN to announce a
> deprecation date for
> IPv4 or IPv4 addressing. It's not a RIR's place to do that;
> it's IETF's.
Good luck getting the IETF to come to consensus on that one.
> If ARIN were to decide they wanted to get out of IPv4 addressing prematurely,
> it would be time at that point for them to be replaced as RIR for IPv4.
>> Indeed. I'd revisit his suggestions. A market, when left to its
>> devices, solves these problems with remarkable speed and little
> It's amazing that despite the enormous amount of evidence to the contrary,
> people keep regurgitating the totally unfounded myth that markets
> provide the good
> solution to all problems.
> If this were true, TCP/IP would not be deployed very much,
> because it would
> never have caught on; the market would have recognized this problem,
> and prevented
> TCP/IP from succeeding. Instead most networks would be using the
> best closed proprietary (and therefore most lucrative) replacement for
> TCP/IP that the market had brought us, which would have no IP
> address resource exhaustion problem from the beginning.
Here, I think you drive off the rails a bit. Open Source can be the most
efficient and most lucrative solution to a problem and a market can and
has recognized that efficiency many times (Linux, Apache, SSH).
Just because markets are often wrong (X11, Motif, Windows, Enron,
Tulips, Housing) does not mean that they always are.
More information about the ARIN-PPML