[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: tiny v6 operational issue



Randy Bush wrote:

> >> i wonder if doc needs to be updated to check this.
> > It should be documented here in v6ops since it is not a dns 
> problem but
> > general problems among services.  I.e. if you put AAAA 
> record to a node,
> > all the services at the node should be served also via 
> IPv6.  Otherwise,
> > the FQDN should be splitted between the services which 
> served only via
> > IPv4 and served via IPv6/IPv4.
> 
> it is obvious to me that one must provide all services on all
> addresses.  given
> 
>    foo.bar.  A    1.2.3.4
> 	     A	  5.6.7.8
> 	     AAAA 2001:666:777::42
> 
> we do not currently require that all services provided on 1.2.3.4
> also be provided on 5.6.7.8.  so how can we explain doing so for
> 2001:666:777::42?

"We" don't have too. IPv6 capable software _should_ use getaddrinfo()
This call nicely returns a set of host/port combo's to connect to and
the software should walk this list till it is able to fulfill it's job.
Check out, the 4 year old doc: http://www.kame.net/newsletter/19980604/
And one can instruct getaddrinfo() to only return say AF_INET6 etc.

Greets,
 Jeroen

PS: Yeah yeah I know PuTTY doesn't fall back to IPv4 if the IPv6
connection fails, don't kick me now I am down, I admit, I admit :)