Julien Laganier wrote: > Hi Vlad, > > On Wednesday 25 October 2006 18:22, Vlad Yasevich > wrote: >>> So, yes, there is a reason to prefer a configured >>> address over a stateless autoconf one. Same >>> argument applies with DHCPv6 configured addresses. >> [...] >> >> Also, this preference really depends on the usage >> cases. I can see scenarios, where I would rather use >> autoconfigured address over the DHCP or manually >> assigned one. >> >> [...] >> >> The concern I have is someone else may find the >> above preference backwards and may want to always >> prefer autoconf addresses over DHCP or statically >> configured. This is a policy matter and you are >> proposing to encode it into the rules without >> ability to change. > > Would privacy be the reason why such someone would > prefer autoconf addresses over DHCP or statically > configured addresses?
In a sense, yes. I was thinking of a well known server with a manually configured address and autoconf. A potential policy would be to prefer outgoing connections to use autoconfigured address. > > Would adding to the proposed new rule a statement > similar to those of Source Address Selection Rule 4 > (Prefer home addresses) and 7 (Prefer public > addresses) as below solve your concern? > > Implementations SHOULD/MUST provide a mechanism > allowing an application to reverse the sense of this > preference [...] (e.g., via appropriate API > extensions). Use of the mechanism should only affect > the selection rules for the invoking application [...] > Implementations for which [insert reason here] MAY > reverse the sense of this rule [...] That would certainly address this concern. This is all rather hypothetical though. The concept that a DHCP address is more stable then EUI64 base address is flawed in my opinion. Both depend on a piece of hardware that can fail or be changed. I guess manually configured addresses are a bit more stable. Additionally, the address preference would completely depend on a deployment scenario. In Alain's case, he would prefer DHCP/manual addresses over Autoconf. In someone else's deployment, that many not be the case. The rules as specified now tend to be agnostic more or less. They would work no matter how things are set up. (there are exceptions, such as ULA). Of course, implementations may override Rule 8 (longest prefix match) with something better/different. I wouldn't object as strongly to something like this: Rule 8 may be superseded if the implementation has other means of choosing among source addresses. For example, if the implementation somehow knows which source address will result in the "best" communications performance or knows relative stability of addresses and wants to select a more stable one. Regards -vlad -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------