El 11/05/2006, a las 1:21, Perry Lorier escribió:

Rémi Denis-Courmont wrote:
Le Mercredi 10 Mai 2006 10:35, marcelo bagnulo braun a écrit :

ulas and private address can be used to reach a global destiantion
address heavily depending on the local setup, hence local
configuration per case is needed in the general case...


This assumes that one may use an ULA (IPv6) to reach a globally routable
(IPv6) address. In other words, that someone has introduced some kind
of NAT or transparent proxy in the middle. I'm not so gullible as to
believe that nobody will ever try to implement this, but... if some
network administrator really wants to provide “transparent” IP-layer
access to the globally routable IPv6 Internet, shouldn't (s)he hands
out public IPv6 addresses/prefix(es) to his/her users, possibly in
addition to ULAs?

A more probable solution is to have all the client PC's on ULA, and all
the servers have a globally routeable IPs. clients should still be able
to reach the servers, and visa versa, they just can't reach other hosts
out on the internet.

What would be wrong with the IPv6 stack throwing an unreachable error
whenever an application tries to send a packet/connect to a global IPv6
address while its socket is bound to an ULA or while the host has no
global (or at least has no non-ULA-nor-link-local) addresses?

The Correct Solution(tm) for this entire problem (IMHO) is for Pekka to
make sure that the gateway box for the client PC's throws an unreachable
error back to the hosts when it realises it can't forward it out to the
Internet.  The client boxes should detect the unreachable and give up
immediately and fall back to IPv4.


as mentioned to Remi my understanding is that the routability of the ULAs can be extended beyond the local site using agreements, so probably this approach would not support this case (maybe it is not a big problem though...)

regards, marcelo


Also, Pekka should use /etc/gai.conf to modify glibc's preferences as to
which one it prefers.  I however can't find this documented anywhere,
however it does seem to be available in the most recent fedora packages.


--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------



--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to