>> We are already 90% of the way here: Make IA_PD work for hosts, not >> just for routers. That way Android handsets can have as many addresses >> as they want. > > DHCPv6 PD is one of the means suggested by RFC 7934, yes. I'm not sure that > the folks asking for IA_NA would be happy with IA_PD though. The reason > most often cited for wanting DHCPv6 is that it fits well with tracking > practices and systems that are built to support on-request addressing and > networks assigning individual IP address(es) to devices. DHCPv6 PD provides > request-based addressing, but it wouldn't do much to interoperate with > those tracking systems because they deal with addresses, not subnets. From > that perspective, ND snooping might be more likely to interoperate well.
Well, I work for one othe ISPs who would *love* to use DHCPv6 PD. Yes, we'd have to make some modest changes to systems to handle prefixes instead of individual addresses. But we'd be happy to just track IPv6 prefixes and *not* have to track individual addresses. Our estimates is that the work to incorporate such a change (handling IPv6 prefixes) would be significantly less than doing ND snooping and similar in all relevant boxes and systems. YMMV. Steinar Haug, AS2116
