> -----Original Message----- > From: Bob Hinden [mailto:bob.hin...@gmail.com] > Sent: Monday, August 12, 2013 12:59 PM > To: Mark ZZZ Smith > Cc: Bob Hinden; Ole Troan; Brian Carpenter; IPv6 IPv6 List; Dave Thaler > Subject: Re: draft-ietf-6man-stable-privacy-addresses: Document title > > Mark, > > > Why does this method have to be limited to SLAAC, or rather, wouldn't > describing it as SLAAC related imply that this or a similar technique can't be > used with DHCPv6 or any other address configuration methods? > > > > While DHCPv6 assigned addresses are naturally per-subnet, I'd think the IID > generation techniques described should be an option for DHCPv6 provided > addresses e.g., different IIDs are generated for different subnets by the > DHCPv6 server even when the client provides the same DUID. > > > > It's starting to seem to me that more clearly separating address generation > methods from address configuration methods would be worth while.
I agree with all of the above discussion. > If I > statically configure an address that complies with this algorithm (perhaps > because my host doesn't yet implement the SLAAC method of configuring > them), is it a "stable privacy address" (or what ever name it ends up with)? > In > practice I'd probably still call it a "stable privacy address". > > That's a good point, the same issues apply to SLAAC and DHCPv6 generated > addresses. The next version of Privacy Considerations for IPv6 Address > Generation Mechanisms will hopefully address the issues in DHCP. > > Regarding the title, of this draft, how about: > > Stable Temporary Per-Subnet Interface Identifiers "Temporary" definitely doesn't belong in the title because they're not nearly as temporary as the ones in RFC 4941. > > My thinking behind this suggestion is that the draft is about a new type of > Interface Identifiers, addresses are somewhat separate. The are temporary > as they are not tied to a hardware token like an IEEE MAC address. They are > stable as they intended to be used for a long while on the same subnet and > don't change per connection/application. It is a bit of an oxymoron :-) > > Another choice: > > Pseudo-Random Per-Subnet Interface Identifiers > > They are pseudo-random and intended to be used on a single subnet. I have no problem with the above. > Personally, I would prefer something like these titles because "privacy" is > once of several characteristics and the privacy characteristic depend a lot on > how they are used and other factors like if they are put in the DNS, shared at > the application layer, the prefix remains constant, etc. I agree it would be good to keep the word "Privacy" out of the name of the type of interface identifiers. It should only be in the title if used in a way that doesn't modify the term "address" or "interface identifier". (By comparison the title of RFC 4941 has Privacy used in a way that doesn't, and is fine.) -Dave > > Bob -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
