[arin-ppml] Q1 - ARIN address transferpolicy: whythetriggerdate?

[Paul Vixie]

> From: "Robert E. Seastrom" <ppml at rs.seastrom.com>
> 
> >> As with many other technologies, there is a substantial 
> >> last-mover advantage to going dual-stack or single-v6.
> >
> > On what do you base this opinion?
> 
> I base it on the fact that technologies mature, that the cost to support
> goes way way down as the average joe becomes more familiar with it and it
> becomes less dodgy and idiosyncratic.  Remember the joy of supporting
> customers who were running Trumpet Winsock or MacSLIP?  If you want to
> relive those days, chase your customers into single-stack v6.

that analogy doesn't hold, since a macslip customer back in the day would
still be a customer in the second month, whereas, not so an ipv6-only customer
today.

> These days, in the home gateway department for instance, we are in the flint
> knives and bearskins era when it comes to IPv6 support.  Some of us choose
> to run v6 and offer it to our customers anyway, but the guy who leaves the
> business on the table for now and picks it up after others have blazed the
> table may be making a smarter business decision.  Of course, we get a bunch
> of intangibles like early experience etc.  The stockholders probably don't
> care.

the biggest reason why a last-mover advantage exists for ipv6 is the "network
effect".  when i was at paix we opened some new facilities after palo alto's
runaway success, and in each city we had to set the IX port fee to $0 until
there were ten ISPs connected to the switch in that city, at which point the
normal pricing would kick in.  nobody wanted to pay for a switch port when
there was effectively no benefit to it since nobody else was connected yet.

the benefit to connecting to a network is a function of the size of that
network.  the size of the ipv6-reachable public internet is very very small
today, whereas for the last-mover it will be as large or larger than ipv4 is
today.  so, equipment costs and quality, and moore's law, all enter into it,
but the dominant term of the "last-mover advantage" equation is the "network
effect".

which is why i've been saying for the last couple of years that ARIN needs to
investigate a policy whereby at some date before some runout milestone, new
ipv4 is only available to those who can demonstrate that they will use it on
dual-stack public-facing networks.  i don't LIKE this idea, but i don't see
any way to soften the impact between the internet community and our brick wall
otherwise.  (note, as before, this is a personal position, not an ARIN BoT
position, and in particular, counsel has not weighed in on its advisability.)