Hi Brian, On 2008/02/08, at 7:06, Brian E Carpenter wrote:
Hello again, On 2008-02-07 16:34, Hiroshi MIYATA wrote:Thanks Brian for your comment. On 2008/02/07, at 11:05, Brian E Carpenter wrote:Miyata-san, I haven't completely studied this yet, but I have two quick comments.The prefix PREFIX::/96 is advertised in the IPv6 network by theNAT-PT gateway, and packets addressed to this PREFIX will be routedto the NAT-PT gateway. The pre-configured PREFIX only needs to beroutable within the IPv6 network and as such it can be any routableprefix that the network administrator chooses.This seems to assume that the NAT-PT is at the boundary between a closed IPv6 routing realm (where it's OK to advertise a /96) and the IPv4 network. What happens if someone needs to put a NAT-PT between a closed IPv4 routing realm and the open IPv6 network?I think advertising /64 is enough. Prefix for routing should be 64-bit.And following 32-bit is indicator used by host to detect the translator.Sorry, my description maybe confusable.But you can't advertise a /64 to the Internet either...
Aggregated prefix is also fine. If the packet arrived to the gateway, it works. I think it is same as router who has stab network below.
The following 32-bit, represented as IDENT MUST be an value assignedby IANA to indicate that translator would exist in the path. More requirement for IDENT is described in Chapter 13.Why is this useful?IDENT can help end-node to know the existence of translator. And it must be unique. Then the end-node can have some options. Put dialog to user, quit the session, search other address, etc...I don't understand the value of these options. If you wanted the upper layer protocol to calculate "correct" checksums, you could play tricks with these "IDENT" bits to achieve that, I suppose.
Yes, above story supposes upper layer action. ;-) Thanks, ....miyata
BrianIf the end-node does not have function to detect the translator, like existing implementation, it would connect to the address. Regards, ...miyataBrian