IPv6 traffic data in Asian networks?

Mohsen Souissi mohsen.souissi at nic.fr
Thu Mar 22 18:23:56 CET 2007


 On 22 Mar, Rémi Denis-Courmont wrote:
 | On Thursday 22 March 2007 18:05:35 Kevin Loch wrote:
 | > With the exception of the ARIN website itself,
[...]
 | > I have not seen "much 
 | > more" transient reachability problems on IPv6.  I have seen IPv6 enabled
 | > on commercial websites without any problems.  I'm not saying
 | > it's perfect but it's alot better than "NO NO NO".
 | 
 | It will definitely cause problems to some users, adds extra cost (HW/SW 
 | updates, maintenance). And it does not bring any advantage to the other ones, 
 | because HTTP works fine with NATs and proxies.
 | 
 | I am a bit bored with the "If only Google advertised IPv6 on their websites" 
 | statements that show up every now and then, every here and there. *HTTP* is 
 | simply NOT a good use-case for switching to IPv6 at the moment. Or well, 
 | IPv6-only may make sense if you cannot afford an IPv4 address, but dual-stack 
 | HTTP server really looks useless to me from a business perspective.
 | 
 | Fortunately, there are other application-layer protocols where IPv6 makes a 
                          ^^^^^^^^^^^^^^^^^^^^^^^
 | lot more sense.
   ^^^^^^^^^^^^^^^

Please, let me say that I'm also a "bit bored" to see "every here and
there" peopole opposing two alternatives as if we can do only one of
them, either "a) deploy IPv6 where we cannot do with IPv4", or "b)
integrate IPv6 in classical IP world (for well known services such as
web, ftp, DNS, ...)".

IMHA, a better strategy would ask everybody to try their best on both
a) AND b) as far as they are concerned (in terms of network services
they provide) and as far as they have the skills.

As an example (and I know that its also the case for other TLD
registries), fortunately AFNIC didn't wait that French domain name
holders and registrars support IPv6 on their DNS servers to do so for
.fr servers. AFNIC DNS service has been officially supporting IPv6
since 2001 and that didn't cost us millions of Euros... Our web server
(www.afnic.fr) has been supporting IPv6 for a long time and the very
poor IPv6 httpand DNS requests that arrive at our servers nowadays
wont't make us shut IPv6 down on our servers. What we are doing
instead is humbly educating and raising consciousness among French and
European network actors as far as we can get in touch with them.

There are certainly still operational (and cost) problems with
integrating IPv6 in the classical IP world (connectivity, stability,
performance) but there will certainly other bigger problems if we do
nothing and if we keep waiting for the "magic killer IPv6-only"
application or for a sudden ROI increase, because meanwhile, Network
engineering (with NATS and NAT Traversals, P2P, poor transtion
mechanisms such as 6to4 and Teredo) will become a nightmare, sooner or
later!

Mohsen.


 | 
 | > As for transition mechanisms, sites will find that having their own
 | > local 6to4 and teredo relays will help alot.
 | 
 | If you do RFC3484, I think it does not matter, IPv4 will be favored over 
 | 6to4<->native or Teredo<->native connections. They really only matter if you 
 | do not have IPv4 at all on one side (or if the client is legacy non-RFC3484).
 | 
 | -- 
 | Rémi Denis-Courmont


More information about the ipv6-ops mailing list