picking up an old thread. [...]
> - DHCP and stateless autoconf. This document is probably not the > right place to discuss the M&O bits, but IMO this document should > say more about DHCP vs. stateless and the issues surrounding when > to implement one or the other. Not to mandate them. Actually, that > raises an intersting point. This document (and RFC 4294) mandate > (MUST) that hosts implement stateless autoconfiguration. This > despite that this document is only informational, and no where in > standards track RFCs is stateless autoconf mandated. This takes us > back to the question of what the scope of this document should be. the draft currently says that stateful address assignement aka DHCP MAY be supported. I think that is to weak and that this document should recommend that all nodes support DHCP for address assignment just like it does for SLAAC. there are a number of shared access networks (EPON, DSL N:1...) which currently depend on option 82 in DHCPv4 to initiate state in the access network. if the assumption is going to be that nodes do not implement DHCPv6 address assignment then this functionality may have to be replicated in ND (e.g Line-id options inserted by access node). a stateless protocol is ill-suited for this purpose and I suggest we just bite the bullet and accept that there are networks which require a stateful protocol for address assignment. cheers, Ole -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------