[arin-ppml] v4 to v6 obstacles

Joe Maimon jmaimon at chl.com
Wed Oct 28 16:32:32 EDT 2009



Lee Dilkie wrote:
> My comment on the subject, repeated from last year.
>
> The only proper way forward is dual stack and the faster we achieve some
> magic number (80%?) of dual stack penetration, the faster we can roll
> out v6 only.
>
>   
Its not the proper way forward. It is the theoretically ideal way 
forward (albeit at 100%). It is also the way forward that hasnt gotten 
enough momentum yet and is uncertain that it will in time, unaided. 
Waiting for 80% penetration before depletion is very likely overly 
optimistic, probably because real uptake of IPv6 depends on depletion. 
Even then, it is hardly likely it will occur in any meaningfully quick 
fashion.

Here is the oft-quoted chicken and egg problem in its expanded form.

Why would any existing user of IPv4 need to add dual stack to IPv6?

To access the IPv6 only users.

Why would anybody ever be publicly accessible only via IPv6?

Because they cant get any IPv4.

Who is going to put up with that?

Only people who dont mind waiting for large percentages of the internet 
to decide it is worth getting IPv6 to talk to them.

What will the rest do?

Beg borrow and steal to get IPv4.

Rinse, Repeat.

The  waiting for dual stack to reach critical mass plan is proceeding 
too slowly, calling into doubt whether it is smart to continue waiting 
on it.

Due to address shortage, continuing with waiting for dual stacking to 
reach critical is going to require more and more NAT, and more and more 
wrangling over past inefficiencies. Which is bad, even as it may become 
more and more necessary.

Furthermore, as a plan formula it sucks. We have to invest 80% of the 
effort to get the 20% payoff? The 80 20 rule is supposed to work in the 
other way.

If the plan was to wait until 20% of the internet was dual stack and 
then the rest would "automatically" follow and cause IPv6 only to be 
practical, now thats more achievable, but still unlikely.

> One good way for organizations like ARIN to help with dual stack is to
> simply give out v6 addresses, free, to all current v4 address holders
>   
Which they do for the most part. To date the only significant complaints 
I have seen have been regarding those who arent IPv4 holders or with the 
single prefix policy.
> and encourage them to roll it out. Or add a stick like "here's your free
> v6, you have x years to roll out v6 to your customers or you loose your
> v4"... not that I favour the stick, just tossing it out.
>   
ARIN is not in the position to be waving sticks at people for things 
like this, and I think they know that.

And where is the carrot in your plan?
> {flame proof underwear on}
>
> -lee
>
>
>   
Joe



More information about the ARIN-PPML mailing list