Re: ULA scope [draft-ietf-6man-rfc3484-revise-05.txt]

[Brian E Carpenter <[e-mail]>]

On 2012-03-17 14:37, Hemant Singh (shemant) wrote:
> Arifumi,
> 
>  
> 
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
> Arifumi Matsumoto
> Sent: Friday, March 16, 2012 8:30 PM
> To: Brian E Carpenter
> Cc: [e-mail]; Brian Haberman; Bob Hinden; Dave Thaler
> Subject: Re: ULA scope [draft-ietf-6man-rfc3484-revise-05.txt]
> 
>  
> 
>> ULA-to-ULA will not be preferred, because ULAs are assigned lower
> 
>> precedence value in the policy table than those of IPv4 and IPv6
> 
>> global addresses.
> 
>  
> 
> I certainly understand if a host has an IPv4 and a ULA assigned, the
> policy table prefers the IPv4 over the ULA otherwise Internet
> connectivity for the host breaks down.  Likewise if the host has an IPv6
> global and a ULA (which is also globally scoped), the IPv6 global is
> preferred or Internet connectivity breaks down.   However, the issue
> that Brian raised needs some thought.  The host has an IPv4, IPv6
> global, and a ULA address. What if the packet destination is a ULA for
> local communication, so why not use a longest-prefix match and use a ULA
> source?  How does a longest-prefix match relate to the policy table?  Do
> we have a mode where the policy table is the only entity used for source
> address selection (SAS)? If yes, then how does the ULA to ULA
> communication above work?

Let me be clear. If a local service has (for some reason) both a ULA
and a non-ULA global address, and the host has both, I think the
correct default behaviour is for the ULA address pair to be used.
This is a matter of address-pair selection, however, not of
source or destination selection alone.

   Brian