On Wed, Aug 10, 2005 at 10:23:46AM +0200, Stig Venaas wrote: > On Wed, Aug 10, 2005 at 05:24:23PM +1000, Greg Daley wrote: > [...] > > I'd suggest that if preferences are all that's needed, then the > > function matches that for which the options are used now. > > I agree that if only prefix preference is needed (possibly also > v4 vs v6), then it seems obvious to learn this together with the > prefixes themselves. I.e. if you use slaac, then also get the > preference that way (similarly with dhcp or other mechanisms). > The preference is for destination address selection, right? > > So a question then is whether that is enough or if there are > many cases where the full policy (including source address > selection) is needed. If the full policy is needed in some cases, > then we have to consider whether it's worth having two solutions. > > I don't know myself, I think it requires further study.
I myself think that full policy is needed, of course. But, even if you define a option only for prefix *preference* for this framework of policy table, I wonder what the label value should be. As for IPv4 and IPv6 selection, even if you choose an arbitrary label value, it doesn't have any effects on source address selection. But, once that option is used for multiple IPv6 prefixes, the label value has some effects on source address selection also. So, as far as you use the policy table framework as it is, IMHO, you have to distribute prefix preference with label value. -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
