Re: [v6ops] The need for local-ipv4 socket transition solutions -- NAT64/DNS64 remains insufficient

On Fri, Apr 3, 2015 at 2:15 PM, James Woodyatt <jhw@nestlabs.com> wrote:

> On Fri, Apr 3, 2015 at 3:57 PM, Lorenzo Colitti <lorenzo@google.com>
> wrote:
>
>> what do you purpose as a better alternative?
>>
> I propose that IETF decline the invitation to publish a BCP to the effect
> that host OS implementations should include a CLAT function.
>
> Aren't we just changing the outcome from "we'll never get away from
>> IPv4-only socket calls" to "we'll never get away from running IPv4"?
>
>
> I think the former entails the latter. I think we get away from running
> IPv4 by deliberately and carefully removing support for IPv4 from
> application developers and users in a way that continuously makes forward
> progress at a manageable and predictable expense. If we deploy CLAT
> everywhere, then we will be making IPv4 more difficult to displace by
> removing support for it in the network.
>
> You could well argue that for as long as IPv4 is around, we'll never get
>> rid of IPv4 socket calls either, because it's impossible for anyone to
>> impose a "thou shalt not write an IPv4-only app" policy without placing
>> themselves at a competitive disadvantage.
>
>
> Is it really impossible? Because that's exactly what Facebook did— impose
> a "thou shalt not use IPv4" policy on its internal developers, and it
> doesn't look to me like it put them at a competitive disadvantage. Have you
> looked at Google+ lately?
>
> In fact, I suspect they might have found it more difficult if the majority
> of their developers were using workstations that had CLAT implementations
> in their operating systems. Taking away IPv4 was a critical step for them,
> according to what their delegation to the World IPv6 Congress reported
> earlier this month. If their developers could have still written IPv4
> applications and relied on the system to translate it into IPv6 on their
> interior networks, then I think it would have greatly complicated
> transition.
>
>
>

The fundamental issue is that you have no control over these on the
internet, https://sites.google.com/site/tmoipv6/ipv4literals

You, and i, also have no control over what is implemented in signalling
channels. Your sockets my be AF agnostic, but that does not stop the data
structures for peer to peer apps being ipv4 addesses like Bittorent or
Skype use.

In fact, Skype works on iOS today on IPv6-only.... why?  Because they built
a CLAT into the App.  It is a horrible idea of ask all the app vendors to
invent their of unique CLAT.  Because they have already reacted to the
threat of IPv6-only... and this is what they do.

Regarding the Flash analogy, that was one ecosystem acting unilaterally.
The ecosystem is well invited to uniformly and unilaterally move forward
with IPv6-only universally.  They are not invited to say choose IPv6 or
IPv4, and those that choose IPv6 only get 99% of the internet while the
IPv4 guy gets 100%

CB

> --
> james woodyatt <jhw@nestlabs.com>
> Nest Labs, Communications Engineering
>
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops
>
>