Re: why I'm not implementing IPv6

[Tim Chown <tjc@ecs.soton.ac.uk>]

On Sat, Nov 23, 2002 at 08:08:17PM -0500, Margaret Wasserman wrote:
> 
> Specifically, what problem does this cause?  Are there many hosts
> that have IPv6 AAAA records in the DNS, but no connectivity via
> IPv6?

Maybe you missed the 6bone thread on this :)   If you take the IETF meeting
example my IPv4 RTT home was 100ms, but my IPv6 RTT home was 400-500ms.  I was
trying to run IPv6-only (including DNS transport) at the site, but ended up
using IPv4 (except when due to the L2/L3 local WLAN issues we had this year 
IPv6 was present when IPv4 was not :)

Note in some cases it is not a problem, e.g. between the UK and Abilene
west coast via 6NET and SURFnet I have a 20+ hop path for IPv4 and IPv6, 
and both perform equally well.  But those paths are dual-stack and similar,
whereas "in the wild" your connectivity is not predictable or reliable enough
to use IPv6 day-to-day (or at least enough that you wish to use IPv4 ahead
of IPv6 to reach a dual-stack service).  

While it was great of Itojun et al to get IIJ connectivity for IETF, it was 
ironic that we were a few blocks from an Abilene gigapop, but that traffic 
still went via Japan to Abilene sites (and this path also doubled the RTT 
to the UK research networks).  I think Itojun fixed this before the end of
the event, which was good :)

I appreciate the intent of preferring v6 ahead of v4 where a host has A and
AAAA records, but the reality is dictated by the 6bone-mess as written down
by Pekka in draft-savola-v6ops-6bone-mess-01 and discussed at the Atlanta
6bone meeting.   It is a problem we are trying to tackle between 6NET, Euro6IX
and Abilene (and WIDE) by getting some sane policy and community tagging
running, to get a stable and well-connected international research network
that may connect some 6bone sites but that is relatively shielded from the
"6bone-mess".   We had a meeting about this in Atlanta, and will be trying 
some new policies soon.

In short, we need to unravel the 6bone mess or, more realistically, build
some good policies between connected networks and transit providers, to
begin to make IPv6 usable internationally.   Note I am talking here about 
Pekka's 6bone-mess thesis, and not Dan's ipv6-mess document which started
this thread and which I largely don't agree with.

Cheers,
Tim