On Sun, Jan 16, 2011 at 7:12 PM, Happy-melon <happy-melon(a)live.com> wrote:
I don't entirely understand the point of this.
The plan seems to be """get
a large enough fraction of 'the internet' to make a change which breaks for
some people all at the same time, so that those people get angry with the
ISPs that haven't got off their arses to fix said breakage, rather than
angry with the broken sites""", which is fair enough.
No, the point is to test what happens if IPv6 is supported on a large
scale. It's known from small-scale testing that this will break
things for some small percentage of users, but no one's sure what the
consequences are of switching this on fully for everyone.
But AFAICT, the
breakage won't occur if your connection can't 'do' IPv6, but only if
connection can't 'do' both IPv4 *and* IPv6 on the same site at the same
time. Surely that's not actually the problem that we need to solve if we're
to be able to migrate smoothly onto IPv6? When the IPv4 addresses run out,
we need to be able to start setting up websites which are *only* v6, surely?
There are many more clients in the world than servers, and servers
have always been able to get dedicated IPv4 addresses much more easily
than clients. A server Internet connection in America will typically
come with as many IPv4 addresses as you need, while you usually can't
get a dedicated residential IP address unless you pay extra. (And
America has more IP addresses allocated per capita than anywhere else
in the world, since it originally developed the Internet.)
So as IPv4 addresses become scarcer, the pressure to use IPv6 only
will fall mostly on residential users. Clients with only an IPv6
address will only be able to get direct connections to IPv6-enabled
servers. The way servers are supposed to do this is serve both A and
AAAA records for the same domain, so IPv4 clients use the A record and
IPv6 clients use the AAAA record.
Unfortunately, someone at some point decided that if the client
supports both IPv4 and IPv6, and the server publishes both A and AAAA
records, the client should connect via IPv6. In practice, almost no
sites use IPv6, so the infrastructure is much less well-tested.
Clients that think they have IPv6 connections might actually have the
connection eaten by a middlebox, or just be slower or less reliable.
So sites don't turn on the AAAA records in practice because it
degrades service for clients with IPv6 connections, which means the
servers aren't accessible to IPv6-only clients without workarounds.
IPv6 day is an attempt to see what happens if major sites publish AAAA
records for a while. Stuff will break, but hopefully not too
horribly, and it will give both site operators and ISPs the chance to
analyze what's wrong with their IPv6 support and what they can do to
fix it. This is a step toward major sites publishing AAAA records all
the time, which is necessary to support IPv6-only clients.
Something like that, anyway. I'm hardly an expert on these things.