Hi,

* Arifumi Matsumoto

> Off the top of my head, I thought about a new rule to the RFC 3484.
> It's really a simple rule, and seems to solve several problematic
> cases related to address selection.
> The rule is:
> 
>   Prefer to select an address as the source address that is
>   assigned by the selected next-hop.
> 
> This rule assumes that the next-hop selection (rule 5) is already
> performed before visiting this rule. Also, it assumes the address
> assignment and routing are correlated. The second assumption does
> not apply to every environment, but typically this is true for
> a site that uses RA. 

This would perhaps have solved a problem we just experienced when
bringing native IPv6 service to a network that earlier was IPv4-only and
had big problems with rogue 6to4 routers spamming RAs onto the access LAN:

Before deploying IPv6, RFC 3484-compliant hosts like Windows and Linux
would de-pref the 6to4 addresses that was assigned by the rogue RAs and
use IPv4, and things would work just nicely.

After deploying IPv6, the 3484-compliant hosts have begun having
difficulties connecting to dual-stack destinations.  I believe what is
going wrong is that the presence of a native IPv6 address makes IPv6
preferred but that the next-hop selection is wrong and the outbound
packets are sent to one of the rogue 6to4 routers (probably with the
native IPv6 source address).

However, if you're determining the source address based on an
already-selected next-hop instead of the other way around, won't you
risk ending up preferring a 6to4 source address (and its associated
router) over a native source address and its associated router?  That
would be very undesireable if so.

If not, I support the rule.

Best regards,
-- 
Tore Anderson
Redpill Linpro AS - http://www.redpill-linpro.com/
Tel: +47 21 54 41 27
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to