Re: draft-arkko-ipv6-transition-guidelines WGLC

[Brian E Carpenter <brian.e.carpenter@gmail.com>]

On 2010-09-01 05:51, Gert Doering wrote:
> Hi,
> 
> On Tue, Aug 31, 2010 at 07:34:48PM +0200, Mohacsi Janos wrote:
>> On Tue, 31 Aug 2010, Gunter Van de Velde (gvandeve) wrote:
>>> Can be mananaged... but... if you use 6to4, then do you know the person 
>>> running the relays? Do you even know who is running the relays? And why 
>>> should the people running the relays care about you if you are not there 
>>> direct customer?
>> If a provider is encouraging to use 6to4, it will provide 6to4 relay for 
>> their customers: announcing anycast 6to4 relay address to them (probably 
>> only for them). Provider is monitoring operational status of 6to4 relay, 
> 
> And the response traffic takes this relay, because...?

Actually, it doesn't. It takes whatever relay 2002::/16 happens to be
routed to at the remote host's location. It's when there is no
such route to a "willing" relay that 6to4 fails. That is the analysis
that's missing in draft-vandevelde-v6ops-harmful-tunnels.

>> traffic volume etc. plus help debugging MTU problems... Yes I know, this 
>> is can be done only for outgoing direction.... But if every 6to4 relay 
>> provider would be doing the same....
> 
> "If".  But this is not so, and therefore, 6to4 with anycast relay is
> just not something that makes sense for global traffic.

Well, the problem is usually not the anycast relay, but the return path.
However, I agree; the problem cases seem to be caused by RFC 3068.

     Brian

> 
>> For example I used to know who is running the 6to4 relay used by me...(me 
>> and network operation team).  For 1.5 years I don't care much about 6to4 
>> relay anymore since I am using native IPv6 both at home and at work.
> 
> Our (well-maintained!!) 6to4 relay works very well for traffic between
> our users and our IPv6-enabled machines - but as soon as the users have 
> native IPv6, no more need for the relay...
> 
> Gert Doering
>         -- NetMaster