Re: ocean: do not boil

[itojun@iijlab.net]

>Shouldn't we also be considering how any of these choices effect the
>overall IPv6 transition of the Internet? Gradually we'd like to see IPv6
>replace IPv4, right? So, if we marry the existence of an IPv6 instance
>to existance of a matching IPv4 instance, how are we going to see
>Internet's transition to IPv6...  If we pick the 3rd choice, we cannot
>really
>have IPv6-only hosts. When can we have one? Maybe as soon as
>IPv6 has a critical mass and some of those hosts declare IPv4 Internet
>is not worth the access...?
>
>This choice also keeps the pressure on the IPv4 address demand and
>the complexity to handle them..
>
>I think deployment of IPv6->IPv4 (IPv6 connecting to IPv4) will
>expedite the IPv6 transition of the Internet... "Co-existence" of IPv6 in
>the current Internet is the must-have first step, but "transition" of
>Internet
>to IPv6 is the ultimate goal.. (unless we want to live with IPv4 forever,
>and create an alternate/parallel universe of IPv6..)

	it's just like we have transitioned from closed TSS systems like
	compuserve to IP.  we will eventually see services available only
	over IPv6, IPv4 services (as a collective means) become less attractive,
	and people will less motivated to maintain their IPv4 connection.
	then IPv4 will gradually cease.
	so yes, IPv4/v6 dual stack age could be very long.  it will take
	a long time until all IPv4-capable-and-configured devices go away.

	(speaking of codebase, IPv4 code will stay forever in the kernel.
	in fact, we still have ISO/CCITT protocol stack in NetBSD and such...)

itojun