Re: [v6ops] The meaning of "IPv6-only" -- Re: sense of draft-vf-v6ops-ipv6-deployment

[Brian E Carpenter <[e-mail]>]

On 31-Mar-21 12:07, Fred Baker wrote:
> 
>> An "IPv6-only host" seems quite unambiguous to me, as does "IPv6-only CPE". An "IPv6-only ISP" seems like a recipe for bankruptcy, however: the only question for an ISP is *how* to provide dual stack service, surely?
> 
> Didn’t the speaker from Reliance JIO say, at IETF 109, that most of their services were in fact IPv6-only? The only reason they had for implementing IPv4 at all was certain services that they marketed to enterprises, which appear to be composed of Luddites.
> 
> I guess that part of the question is whether the phrase “IPv6-only ISP” says what the ISP implements, or whether it says how they communicate  with neighboring IPv6-incapable networks. I might describe T-Mobile as IPv6-only in the sense that they only implement or manage IPv6; they run 464XLAT (which is to say, translation) when a neighboring ISP is IPv6-incapable, but that is invariably in another network. An IPv6-incapable device in a T-Mobile network is about as communicative as a rock.
> 
> No, I don’t think dual stack is the end of the road. It is where we are at the moment.

I think this conversation shows that we're *only* talking about terminology. I don't care *how* an ISP connects retail customers to IPv4-only services, or *how* it connects IPv4-only enterprise customers to their users. But it will do so, unless it wants to leave a lot of money on the table.

Believe me, I'd love to see IPv4 down to 1% of total traffic, but I doubt that I will live long enough. Meanwhile, I don't expect AF_INET to vanish from the socket API.

   Brian