> I just took another look at RFC 2765. I believe that for SIIT to work, > when an IPv6-only node sends to an IPv4-mapped address (so it is sending > an IPv6 packet with an IPv4-mapped source address, as opposed to using the > IPv4-mapped address in the API and sending an IPv4 packet), then it must > use an IPv4-translatable source address. Anything else will just not work > when the packet gets to the SIIT box.
Correct. > So to my mind, this feels more like a section 3 requirement. It's not a > matter of the IPv6-only node *preferring* to use an IPv4-translatable > source address. Instead, if there is no IPv4-translatable source address > available, then the IPv6-only node should *fail* to send to the > IPv4-mapped destination address. I agree that it should fail. > > But I don't understand Erik's comment that this only applies to nodes that > implement SIIT. Why does it not apply to all IPv6-only nodes sitting > behind a SIIT box? Perhaps there is some terminology confusion. When I said something like "nodes that implement SIIT" I meant nodes that have mechanisms to acquire the IPv4-translatable addresses (which isn't specified in the SIIT spec). This is quite different than the nodes that run the SIIT translation rules. So "all IPv6 nodes capable of sitting behind a SIIT translator" catches what I tried to say. Erik -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
